Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Black sheep or devil protection?: the effects of cooperative and competitive intergroup goal structures on group members reactions to an offensive act when committed by an ingroup or an outgroup ...
(USC Thesis Other)
Black sheep or devil protection?: the effects of cooperative and competitive intergroup goal structures on group members reactions to an offensive act when committed by an ingroup or an outgroup ...
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
BLACK SHEEP OR DEVIL PROTECTION?:
THE EFFECTS OF COOPERATIVE AND COMPETITIVE INTERGROUP GOAL
STRUCTURES ON GROUP MEMBERS REACTIONS TO AN OFFENSIVE ACT
WHEN COMMITTED BY AN INGROUP OR AN OUTGROUP MEMBER
By
Mayuko Onuki
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PSYCHOLOGY)
December 2014
2
ACKNOWLEDGEMENTS
I would like to acknowledge my advisor, Norman Miller, for his utmost dedication to
research and teaching. His uncompromised approaches to the pursuit of scientific knowledge and
his meticulous articulation of them taught me how to become a better researcher. I would also
like to acknowledge my former advisor, Brian Lickel, for continuing to provide guidance and
support throughout my graduate studies. I also appreciate the scientific support from my other
committee members Stephen Read, Wendy Wood, Richard John, and Scott Wiltermuth.
I would like to thank my senior colleague, Marija Spanovic, for providing me with
numerous hands-on guidance and support. Thanks also to Gurveen Chopra for kindly
proofreading my earlier drafts.
3
TABLE OF CONTENTS
List of Tables 4
List of Figures 5
Abstract 6
Chapter 1: Introduction 7
Chapter 2: Experiment 1 21
Chapter 3: Experiment 2 34
Chapter 4: General Discussion 55
References 63
Appendix A: Texts of the Newspaper Articles for Experiment 1: Manipulation 73
of Intergroup Goal Structure (Dornsife Participants’ Version)
Appendix B: An Opinion Sheet under the Guise of Facilitating the Face-to-Face 76
Discussion
Appendix C: Instructions for Recalling Incidents of Ingroup and Outgroup 77
Offenders in a Cooperative or Competitive Goal Structure
4
List of Tables
Table 1: Descriptive Statistics and Intercorrelations among Variables Examined 28
in Experiment 1
Table 2: Descriptive Statistics and Intercorrelations among Variables Examined 44
in Experiment 2
Table 3: Simple Slopes Representing the Relationship between level of Group Identity 47
and Action Tendencies, Group-Based Emotions, and Perceived Threats in
Cooperative and Competitive Intergroup Goal Structures for Ingroup and
Outgroup Offenders
5
List of Figures
Figure 1: Mean group-based emotions (+/- SE) toward an ingroup or outgroup offender 31
as a function of intergroup goal structure, Experiment 1
Figure 2: A 3-way interaction between offender’s group membership, intergroup goal 48
structure, and level of group identification (at 1 SDs above and below its mean)
on punishment
Figure 3: A 3-way interaction between offender’s group membership, intergroup goal 49
structure, and level of group identification (at 1 SDs above and below its mean)
on protection
Figure 4: Intergroup-threat and group-based shame and anger as serial mediators of 52
the black sheep effect under cooperative intergroup condition. The model
reports unstandardized regression weights
Figure 5: Intergroup-threat and group-based shame and anger as serial mediators of 53
the devil protection effect under competitive intergroup condition
6
Abstract
Ingroup justification and outgroup punitiveness are pervasive intergroup processes that
augment intergroup conflict. This dissertation examines whether a cooperative intergroup goal
structure reduces these destructive processes. It does so within the theoretical framework of the
black sheep effect wherein people, at times, form more negative attitudes toward an offensive
ingroup member than an equally offensive outgroup member. This enables them to maintain
positive subjective evaluations of their ingroup (Marques, Yzerbyt, & Leyens, 1988). In two
experiments an offender’s group membership (ingroup vs. outgroup) and the intergroup goal
structure (cooperative vs. competitive) were orthogonally manipulated -- via bogus newspaper
articles in Experiment 1 and the guided recall of past incidents in Experiment 2. Under
intergroup cooperation, Experiment 1 found greater shame and anger toward an offensive
ingroup member than an equally offensive outgroup offender, thereby supporting a black sheep
effect. Under competition, the opposite pattern, occurred -- a devil protection effect. In
Experiment 2, level of group identification further moderated these black sheep and devil
protection effects. Under intergroup cooperation, stronger group identification elicited
punishment of an ingroup offender. Harsher punishment was mediated by the intergroup-threat
to the ingroup’s moral reputation that was caused by the offender’s act and by ensuing feelings
of group-based shame and anger. Under competition, by contrast, stronger group identification
elicited protection of an ingroup offender by suppressing group-based shame and anger and
thereby promoting leniency. In the context of the black sheep and intergroup goal structure
literatures, the implications of group-based emotions are further discussed.
7
Chapter 1: Introduction
People readily perceive, feel, evaluate, and act in ways that favor those with whom they
share an important identity and discriminate against others with whom they do not. Such ingroup
favoritism and outgroup derogation are well studied and pervasively found to maintain and
further intergroup conflict (Brewer, 1999). Moreover, denial and justification of an ingroup
wrongdoing is rather widespread (e.g., Branscombe & Miron, 2004) and dehumanization of
outgroup members often serves to justify offensive actions toward them (e.g., Castano & Giner-
Sorolla, 2006). Hence, there is continued need for study of alternative routes and boundary
conditions wherein these destructive psychological tendencies are overcome in a way that fosters
positive intergroup relationships.
In the face of wrong-doing, people have a strong tendency to support their group and
blame outgroups. In some instances blame is directed inward towards one’s own ingroup and
induces vicarious self-conscious emotions such as shame, guilt, and anger (Lickel, Steele, &
Schmader, 2010). As a result, people may come to admit the fault and punish culpable members
of their group, at times, even more strongly than they do so for an offender with whom they have
no association. This dissertation attempts to augment our understanding of how individuals react
to a fellow ingroup member who commits a wrongdoing toward an outgroup, and to examine
how the goal structure that links the ingroup and outgroup – whether it is cooperative or
competitive – moderates retributive reactions to the wrongdoer.
The Black Sheep Effect
Despite the pervasiveness of intergroup biases, research on the black sheep effect shows
that individuals emit harsher treatment to ingroup offenders or deviants in comparison to
similarly deviant outgroup members. Marques, Yzerbyt, and Leyens (1988) argued that the
8
concomitant emergence of favorable ingroup bias toward desirable members (viz. ingroup
favoritism) and ingroup derogation of undesirable members (viz. the black sheep effect) is a
strategy to preserve the group’s sense of positivity as a whole. This proposition was later
incorporated into Abrams and Marques model of subjective group dynamics as reflecting a
propensity for maximizing intergroup differentiation and normative intragroup differentiation
(Abrams, Marques, Bown, & Henson, 2000; Marques, Abrams, Paez, & Martinez-Taboada,
1998; Marques, Abrams, & Serodio, 2001; Marques et al., 1988).
Subjective group dynamics theory postulates that people are motivated to sustain a
psychological representation of a cohesive, well-defined, and normatively legitimated group, and
hence reject deviant ingroup members because they threaten the definition of their group. This
postulate goes back to Festinger (1950)’s theory that people choose to affiliate with similar
others in order to reduce uncertainty in life. Accordingly, the majority of black sheep studies
have examined deviants who violate group specific norms that qualitatively differentiate the
ingroup from the outgroup. In these studies, deviancy varied on the basis of expressed values or
opinions (e.g., Study 2 of Abrams, Marques, Brown & Dougill, 2002; Study 2 of Abrams et al.,
2000; Begue, 2001; Chan, Luis & Hornsey, 2009; Eidelman & Biernat, 2003; Eidelman, Silvia,
Biernat, 2006; Marques et al., 1998; Study 1 of Marques et al., 2001; Matheson et al., 2003;
Morton, Postmes, & Jetten, 2007; Pinto, Marques, Levin, & Abrams, 2010), personality traits
(Study 1 of Abrams et al, 2000), attitudes toward the ingroup or outgroup (Abrams, Rutland, &
Cameron, 2003; Abrams, Rutland, & Cameron, & Ferrel, 2007; Marques et al., 1988), to skills
and performances (Coull, Yzerbyt, Castano, Paladino, & Leemans, 2001; Hutchison & Abrams,
2003; Hutchison, Abrams, Gutierrez, & Viki, 2008; Khan & Lambert, 1998; Lewis & Sherman,
2003; Marques & Yzerbyt, 1988). However, Marques et al. (2001) also showed that a black
9
sheep effect occurs for a deviant who violated generic prescriptive norms shared by both the
ingroup and outgroup (e.g., equality for homosexuals). They argued that people care more about
the moral integrity of their ingroup members than that of their outgroup members, and thus
general norm violation by an ingroup member (i.e., an ingroup offender) can be more threatening
than the same violation by an outgroup counterpart.
In addition to social identity theory and subjective group dynamics theory, expectancy
violation theory also supports the occurrence of black sheep effects. Expectancy violation theory
(e.g., Jussim, Coleman, & Learch, 1987) suggests that people form stereotypes about the
dispositions, behaviors, and physical traits to be expected from group members. An individual
who violates the stereotype provokes an evaluation in the direction of that violation, either
positive or negative. An ingroup offender will provoke stronger negative reaction because people
expect ingroup members, by comparison with outgroup members, to behave less negatively, or
more fully act in accord with ingroup norms. The extent to which ingroup deviants or offenders
threaten the value of the ingroup, either by violating specific group norms or universal norms,
potentiates this regulatory mechanism.
Social Identity Theory (Tajfel & Turner, 1976) proposed that, in the face of group
identity threats, people engage in one of three strategies to maintain a sense of positive social
identity: social mobility (distancing one self or group from a tainted ingroup), social creativity
(thinking flexibly, such as blacks promoting the notion that "black is beautiful"), and social
change (e.g., blacks engaging in civil protest to change the status quo so as to eliminate
segregated public transportation). The black sheep effect is an instance of a social change
strategy in which people form a negative attitude toward their ingroup offender as reflected in
actions that punish, exclude, or derogate him or her and thereby reestablish the moral integrity of
10
the ingroup. In the face of an offensive ingroup member, an alternative to the black sheep effect
is a social mobility strategy wherein people distance themselves from the ingroup when it is
tainted by the offensive ingroup member’s act (Eidelman & Biernat, 2003; Eidelman et al., 2006).
For instance, a black sheep effect was found only when a judgment about an offender was
measured immediately after reading about his or her offense (Eidelman & Biernat, 2003). When,
instead, group identification was measured immediately after reading about the ingroup offender
and before the judgment of the offender, participants showed less identification with their group
by comparison with those who read about a similarly offensive outgroup member. In the latter
condition, wherein participants were given an opportunity to distance themselves from the
culpable group, their evaluation of the offensive person did not differ as function of that
offender’s group membership.
Retributive Attitudes toward an Ingroup Versus Outgroup Offender
Contrary to the notion of the black sheep effect, in the criminal justice arena, racial
ingroup favoritism of juries is well documented and ranges from being relatively blatant to being
subtle in its magnitude (Feather & Oberdan, 2000; Gordon, 1993; Gordon, Bindrim, McNicholas,
& Walden, 1988; Kemmelmeier, 2005; Pfeifer & Ogloff, 2003; see Sommers & Ellsworth, 2001;
Sargent & Bradfield, 2004; Sweeney & Haney, 1992). For example, Whites are more likely to
vote to convict Black defendants than White defendants (e.g., Foley & Chamblin, 1982; Klein &
Creech, 1982) and they give longer sentences to Black defendants (e.g., Gray & Ashmore, 1976;
Sweeney & Haney, 1992). Other studies show no evidence of juries’ racial bias in non-laboratory
court room settings, possibly due to the buffering effects of judicial instructions (Pfeifer &
Ogloff, 1991) and jury deliberation (Bernard, 1979; Kerr, Hymes, Anderson, & Weathers, 1995) .
In over 418 closed felony cases, Taylor and Hosch (2004) coded archival data for defendant
11
ethnicity, jury ethnic composition, and strength of evidence against the defendant, to test the
black sheep effect vs. an ingroup–leniency effect. They found no evidence, however, of either
effect, finding instead that the strength of evidence had the most impact on juries’ decisions.
Black sheep research concerned with retributive attitudes toward ingroup vs. outgroup
offenders, has provided insights into the mixed findings reported in the criminal justice arena.
Several factors determine whether an ingroup offender is protected or derogated. Van Prooijen
(2006) found that people show stronger retributive reactions to a suspected ingroup offender
(relative to a similarly offensive outgroup member) when their guilt is certain, thereby
augmenting the black sheep effect. When guilt was uncertain, however, an ingroup offender was
favored in comparison to an outgroup offender. Thus, there was a cross-over interaction between
an offender’s group membership and the certainty of his or her guilt. Similarly, the certainty of
maliciousness and the criminal history of offenders moderate the black sheep effect (Gollwitzer
& Keller, 2010). Greater anger, social concern, and punishment severity were reported for
repeated ingroup offenders than for outgroup counterparts, whereas the reverse was true when
offenders were first-timers. This suggests that individuals protect ingroup offenders to the point
at which justification and positive change becomes unlikely. There is also evidence that people
shift their standard of justice to defend culpable ingroup action. People who are highly identified
with their group reported less collective guilt and set higher standards for admitting that
questionable past actions of their ingroup are wrong and harmful than did those with low group
identity (Miron, Branscombe, and Biernat, 2010). That is, the greater their identification with the
ingroup, the more likely that people will justify and defend an ingroup offender and avoid feeling
guilty.
Another factor known to moderate a black sheep effect is group status. People are more
12
punitive towards an ingroup offender than an outgroup counterpart when an ingroup's status
exceeds that of the outgroup; by contrast, no differentiation occurred when ingroup status was
lower than that of the outgroup (van Prooijen & Lam, 2007). Further examination of the high vs.
low group status conditions showed greater retribution toward a low status outgroup offender
than a high status outgroup offender, whereas there was no status differentiation effect for an
ingroup offender. This suggests that observed black sheep effects under high ingroup status
conditions were attributable to leniency towards an offender in a lower status group. Similarly,
White Australians report less anger and punitive attitudes, and attributed less responsibility to an
offender when he is an Aboriginal Australian (a low status outgroup member) than when he is a
White Australian (a high status ingroup member; Feather & Souter, 2002). Similarly, in the
Netherlands, Dutchmen, -- a relatively high status group, showed less ingroup leniency toward
an ingroup member who aggressed against the outgroup than did lower status Turks, suggesting
that members of a high status group are more willing to show retributive attitudes toward an
offending ingroup member than are low status group members (Schruijer & Lemmers,1996).
There is additional indication that the black sheep effect may at times be driven by
leniency toward outgroup offenders. Prior black sheep research typically does not include a
control group that can be compared with the ingroup and outgroup offender conditions. Thus, it
is not clear whether effects reflect harsher treatment of offensive ingroup members or more
lenient treatment of outgroup counterparts. But when Braun and Gollwitzer (2012) addressed this
issue by including a vignette of an offender with no reference to a group membership, they found
that outgroup offenders are punished more leniently than are either ingroup or individual
offenders of unknown group membership. Furthermore, the leniency effect vanished when
participants were given an opportunity to show that they were not prejudiced against the
13
outgroup by expressing their political support for the outgroup. Presumably this occurred
because they had established their own moral credentials before having judged the outgroup
offender. They call this effect “patronizing leniency” because “positive behavior that looks like
'true' sympathy and 'true' compassion for members of an out-group actually reflects an ideology
of inequality: out-group members are treated positively because they are seen as inferior,
incompetent, and passive” (Braun & Gollwitzer, 2012, p. 884). As the authors indicated,
particular outgroups were selected for the purpose of the study. In experiment 1, participants
were Germans and the outgroup was South East Asians in Germany who previously had been
found to elicit less threat compared to other ethnic outgroups such as Turks, Russians, and
Eastern Europeans who are major ethnic groups in the country (Asbrock, 2010; Asbrock,
Lemmer, Wagner, Becker, & Koller, 2009). In experiment 2, West Germans also showed this
patronizing leniency effect toward East Germans toward whom, for reasons of political
correctness, they try to avoid acting in a prejudiced or discriminatory manner. Note that in these
studies, too, the status of the ingroup exceeded that of the outgroups, making them consistent
with the circumstances wherein the black sheep effect found in the aforementioned study by van
Prooijen and Lam (2007).
Attitudes Toward an Ingroup Versus Outgroup Offender as a Function
of the Intergroup Goal Structure
In summary, extant research on attitudes toward ingroup vs. outgroup offenders shows
that harsher treatment of an ingroup offender (i.e. a black sheep effect) is relatively rare, and that
leniency toward an offending member of a lower status outgroup may explain some of the black
sheep effects seen in the literature. The leniency found toward a low status outgroup was
presumably due to the lack of need for positive distinctiveness and a heightened self-
14
presentational concern for intergroup fairness (Braun & Gollwitzer, 2012; van Prooijen & Lam,
2007). The findings of these past studies, however, are rather inconclusive because they neither
systematically examine nor discuss the target person or group that was harmed by the offender’s
action. For example, some studies examined circumstances wherein the offender’s actions solely
harm the ingroup (Study 3 by van Prooijen, 2006); in others, both the ingroup and outgroup or a
mutual resource was the target of harm (Gollwitzer & Keller, 2010; van Prooijen, Study 2,
2006); in another, the harmed target’s group membership was crossed with the offender’s group
membership (Schruijer & Lemmers, 1996); and finally, in some studies the target of harm was
ambiguous or unknown (e.g., Braun & Gollwitzer, 2010). Thus, it is unclear whether the prior
findings are due to the offenders’ group membership, the target of harm, or a specific
combination of the two. Particularly when the harm befalls an ingroup or one of its members,
one might display a strong punitive attitude, not because the offender is a member of one’s group,
but instead, as a result of ones empathy for the victim and a vicarious tit-for-tat reaction to the
harm itself. In sum, the group membership of the harmed person(s) needs to be taken into
account when examining attitudes toward an offender.
In considering the factors that foster intergroup conflict, reactions to an outgroup
offender who brought harm to an ingroup are important. Members of a harmed ingroup are
motivated to retaliate against the offender’s group because of their group pride and empathy for
their harmed fellow members (see “vicarious retribution” by Lickel, Miller, Stenstrom, Denson,
& Schmader, 2006). Therefore, when an outgroup member harms a member of an ingroup, it will
affect the judgments that other ingroup members make about the moral integrity of the
harmdoer’s group.
In the face of an ingroup member acting offensively toward an outgroup, fellow ingroup
15
members are not merely likely to be concerned about the consequences of whatever harm has
been done, but also, about the possibility that the moral credentials of their group have been
tainted as a consequence of the offender’s association with it. This is especially likely when their
group is working in cooperation with the outgroup. Were they to justify, minimize, or defend the
offending ingroup member and his or her wrongdoing, such actions are likely to be seen as
morally wrong. In turn, such actions will undermine a trusting positive relationship with the
outgroup. This specific threat of impugned moral integrity, which is even more strongly driven
when groups are in a cooperative intergroup goal structure, seems to provide a major condition
for the emergence of a black sheep effect. Showing retributive attitudes toward an ingroup
offender will not only serve as a strategy to cope with the threat to ingroup morality, but will also
function to better foster a positive intergroup relationship and achieve joint intergroup goals than
will attempts to minimize the wrong of the ingroup offender.
Furthermore, people are less concerned with positively differentiating themselves from
an outgroup under cooperation (e.g. Gaertner & Dovidio, 2000; Gaertner et al., 2000; Dovidio,
Gaertner, Isen, & Lowrance, 1995; Dovidio, Gaertner, & Validzic, 1998; Gaertner, Dovidio, Nier,
Ward, & Banler, 1999). Therefore, they may be more willing to admit the fault of their fellow
ingroup offender. Punishment of an ingroup offender by other ingroup members can further
signal to the outgroup that the ingroup does not tolerate the offense and is willing to correct the
wrong, thereby upholding their moral integrity.
When the intergroup goal structure is competitive, wherein the goal of one group is
achieved at the expense of the other’s loss, category differentiation between the groups becomes
salient and group members are more likely to want to positively differentiate themselves from
the outgroup. At the same time, they are likely to be less concerned about preserving their moral
16
reputation with the outgroup or improving intergroup relations with them. This seems to suggest
a reversed pattern from that of the black sheep effect, namely, a devil protection effect wherein
people justify and defend an ingroup offender to protect their group pride in spite of the
reputational harm done by the offending ingroup member’s actions. Normative influence also
explains why people may readily justify an ingroup offender’s offensive action toward the
enemy outgroup, rather than admitting the wrongdoing and punishing the ingroup offender. It is
normative to be antagonistic toward an enemy and when in competition with an outgroup, they
are an enemy. Moreover, it is likely that in the past there were some forms of offensive gestures
initiated by the outgroup that can be readily resurrected to justify application of a norm of
retaliation to justify the ingroup offense.
In sum, when in a cooperative relationship with an outgroup, people are likely to show
more lenient attitudes toward an outgroup offender as a sign of tolerance and generosity. This is
consistent with the patronizing leniency effect in that outgroup leniency serves to uphold the
egalitarian value of being unbiased and tolerant toward the outgroup (Braun & Gollwitzer, 2012).
As a result, a cooperative intergroup goal structure is expected to augment the black sheep effect
– the stronger derogation of an ingroup offender by comparison with a similarly offensive
outgroup member. On the other hand, when groups are in competition, outgroup derogation is
expected due to the decreased salience of egalitarian motives and the heightened motivation for
positively distinguishing ones own group from the outgroup.
Roles of Group-Based Emotions in Retributive Attitudes Toward an Ingroup Offender
The group emotion literature is another body of research that provides evidence wherein
people come to admit the fault of other ingroup members and feel group-based emotions such as
shame and guilt (Lickel, Steel, & Schmader, 2011). Group-based shame is particularly relevant
17
to the black sheep effect because of its conceptual link to group identity threats and it’s
regulatory function in repairing the tarnished image created by ingroup deviants (Iyer, Schmader,
& Lickel, 2007; Johns, Schmader, & Lickel, 2005; Lickel et al., 2005; Shephard, Spears, &
Manstead, 2013). The predominant motivation elicited by shame is distancing oneself from the
shame-inducing event (Tangney, Wagner, Hill-Barlow, Marschall, Gramzow, 1996). Shame is
often depicted as less effective in restoring equity by comparison with guilt, which is
traditionally conceptualized as eliciting prosocial motivation (Baumeister et al., 1994). Group-
based shame has also prompted motivations that are more adaptive than that of merely
psychologically distancing from the group. Examples include: an increased desire to withdraw
troops from Iraq in response to their countries occupation of it (Iyer et al., 2007), a motivation to
disconfirm a negative stereotype displayed by a fellow ingroup member (Schmader & Lickel,
2006), and support for reparations to harmed outgroups (Brown & Cehajic, 2008). Moreover,
anticipated group-based shame and anger elicited a desire to undertake collective action against a
proposed ingroup transgression (e.g., a military intervention in Iran) whereas anticipated group-
based guilt did not (Shepherd et al., 2013). Furthermore, members of a harmed group were more
likely to forgive and accept reparatory gestures from members of a culpable group when shame
rather than guilt was expressed (Giner-Sorolla, Castano, Espinosa, & Brown, 2008).
When anticipating group-based shame about future offensive actions by ingroup
members toward an outgroup, group members of a stable high status group were likely to inhibit
ingroup favoritism and even show outgroup favoritism by comparison with low status and
unstable high-status groups (Shephard et al., Study 2, in press). When the positive distinctiveness
of the group was not threatened (i.e., a stable high status), people suppressed ingroup favoritism,
presumably because they were able to engage in egalitarian behavior that reflected a concern
18
about moral principles relating to harm and fairness (Leidner & Castano, 2011). On the other
hand, when group status was threatened (i.e., low status or unstable high-status groups), group
members were presumably more concerned about ingroup loyalty, which prevented anticipated
group-based shame from inhibiting the identity-protective effects of ingroup favoritism.
The manner in which intergroup status differences moderate the effect of anticipated
group shame seems to conceptually parallel the effects of the type of goal structure that
characterizes an intergroup relationship. A lack of concern about the establishment of positive
distinctiveness from an outgroup allows expression of retributive attitudes toward an ingroup as
well as leniency toward the outgroup. Groups in cooperation are likely to exhibit the same
pattern as did groups with stable high status in the Shephard et al’s (in press) studies because, as
previously noted, the nature of cooperation lessens one’s concern for positive distinctiveness and
increases a concern for intergroup fairness. These same notions apply to reactions to an ingroup
offender and thereby bear on the occurrence of black sheep effects. Group members who are in a
cooperative goal structure with the outgroup are likely to feel group-based shame in response to
an ingroup member's offensive action, which in turn will elicit retributive attitudes toward the
ingroup offender – a black sheep effect. On the other hand, in a competitive goal structure,
wherein positive distinctiveness of one’s group is under threat, group members are less likely to
feel shame and more likely to defend an ingroup offender, and hence, exhibit a devil protection
effect. This is likely to be particularly true for group members who strongly identify with their
group. Avoidance of group-based shame and guilt by high group identifiers is well known as a
defensive mechanism that emerges in response to a heightened threat to the positive
distinctiveness of the group (see a review by Lickel et al., 2011).
In addition to group-based shame, in-group directed anger has been identified as possibly
19
an even more potent motivator for social change and the restoration of positive intergroup
relations. Anger is accompanied by a more intense level of arousal than shame and is known to
more powerfully trigger a variety of retributive actions (Thomas, McGarty, & Mavor, 2009; see
also Harth, Kessler, & Leach, 2008). Ingroup directed anger had the strongest impact, compared
to group-based shame and guilt, on support for a series of reparative actions in Iyer et al.’s
(2007) Iraq study: compensation to Iraq, confrontation of those responsible for Iraq, and
withdrawal from Iraq. In-group directed anger also predicted support for real political action to
rectify the situation, whereas guilt only predicted support for abstract reparation goals (Leach,
Wayne, Iyer, & Pedersen, 2006). In another study of Euro-Australian maltreatment of indigenous
Australians, ingroup directed anger predicted support for real political action to rectify the
situation whereas guilt only predicted support for abstract reparation goals (Leach, Iyer, &
Pedersen, 2006). Research has also shown that anger is strongly associated with retributive
reactions (e.g., Darley & Pittman, 2003; Miller & Vidmar, 1981; Tyler & Boeckmann, 1997;
Vidmar, 2002). Similarly, Gollwitzer and Keller (2010) found that anger and moral outrage
mediated the influence of criminal history on punishment of ingroup offenders.
Present Studies and Hypothesis
This dissertation examines both the effect of the intergroup goal structure – cooperative
versus competitive – on retributive attitudes toward an offender, and the moderation of such
effects as a function of the offender’s group membership. It is the first study of black sheep
effects that examines the effects of the intergroup goal structure on group-based emotions. The
following five hypotheses were examined.
First, under a cooperative intergroup condition, people will exhibit stronger retributive
attitudes as well as group-based emotions, particularly shame and anger, toward an ingroup than
20
an outgroup offender – a so-called black sheep effect. An opposite pattern is expected in the
comparative effects of an ingroup versus outgroup offender under a competitive intergroup
condition– a so-called devil protection effect (hypothesis 1).
Second, under a cooperative intergroup goal structure, people will have stronger
retributive attitudes and group-based emotions, particularly shame and anger, toward an ingroup
offender than under a competitive intergroup condition. The opposite expectation will hold for an
outgroup offender, thereby yielding a cross-over interaction between the offender’s group
membership and the intergroup goal structure (hypothesis 2).
Third, the preceding hypotheses are based on the assumption that one’s group
membership is sufficiently important and salient at the time of judgment. Therefore, hypothesis 1
and 2 will be particularly true for high group identifiers relative to low group identifiers, thereby
yielding a 3-way interaction between participants’ level of group identification, the intergroup
goal structure, and an offender’s group membership. In the past, black sheep studies concerned
with retributive justice have rarely examined the effects of level of group identification.
Although Miron, Branscombe, and Biernat (2010) is an exception, it did not have an outgroup
offender condition. Thus, a measure of group identification was included in order to test this
potential 3-way interaction (hypothesis 3).
Fourth, group-based emotions, particularly shame and anger, will mediate the interactive
effects of the offender’s group membership and the intergroup goal structure on retributive
attitudes (hypothesis 4). Group-based guilt was also examined, as it is generally well-known to
correlate with group-based shame in the context of ingroup wrongdoing.
Lastly, the threat to the ingroup of losing their moral integrity in the eyes of outgroup
perceivers was also expected to mediate the black sheep effect that is expected when groups are
21
linked by a cooperative, as opposed to a competitive goal structure (Hypothesis 5).
To test these hypotheses, two experiments were conducted. They orthogonally
manipulated the intergroup goal structure (cooperation vs. competition) and the group
membership of an offender (ingroup vs. outgroup). To test hypothesis 1 through 4, punitive
tendencies toward the offender and group-based emotions of shame, anger, and guilt were
measured. In order to test hypothesis 5, Experiment 2 measured the degree to which the
offender’s offense elicited perceptions of intergroup threat to the ingroup’s moral integrity. In
both studies, attitudes toward an ingroup offender who aggressed toward an outgroup member
were compared with those toward an outgroup offender who similarly aggressed toward another
outgroup member in order to control for the impact of the victim’s group membership on
retributive attitudes toward offenders.
Chapter 2: Experiment 1
In Experiment 1, USC Dornsife College of Letters, Arts, and Science students and
Marshall School of Business students participated in a study about “a campus resource issue that
relates to Dornsife and Marshall students.” They were led to anticipate partaking in a face-to-
face discussion about the issue with an outgroup member later in the study session. One well-
known strategy to deal with the threat associated with an ingroup offender is to psychologically
distance oneself from the group, thereby reducing the importance or salience of the group
membership associated with the offender (Eidelman & Biernat, 2003). Although individual
differences in strength of identification with one’s group were measured at the beginning of the
experiment, in order to eliminate the use of this strategy as a viable option, participants’ group
identity was made salient and important throughout the experiment.
Participants read a bogus campus newspaper article about an ongoing collaborative effort
22
by Dornsife and Business students to improve the campus environment. The intergroup goal
structure was manipulated by the contents of the article, which informed participants that
Dornsife and Business students are either in a competitive or cooperative situation with respect
to the acquisition of new resources. To manipulate the group membership of an offender, the
article then informed participants about a recent incident wherein either a Dornsife member or a
Marshall member committed a wrongdoing by stealing money from a fund that was set up for the
benefit of outgroup members. Dependent variables of interest were then assessed under the guise
that responses to them were going to be used to facilitate the upcoming discussion with an
outgroup member. It was expected that participants would express more punitive attitudes and
less leniency toward the ingroup offender, as opposed to an outgroup offender, accompanied by
stronger feelings of group-based shame and anger, when the intergroup goal was cooperative
rather than competitive.
Method
Participants and Design
Participants were 199 college students (135 female and 62 males)
1
, varying in age from
18 to 45, M = 20.21, SD = 3.36) and enrolled in the Psychology Department Subject Pool at the
University of Southern California. Most of the participants were either Caucasians (38.2%) or
Asian (37.2%), with an additional 10% Hispanic and 3% African American.
From this total, 162 self-identified themselves as Dornsife College of Letters, Arts, and
Science students and 37 self-identified themselves as Marshall School of Business students. For
Dornsife students, their ingroup is Dornsife College students and their outgroup is Marshall
School students, and vice versa for Marshall students. Participants were randomly assigned to a
1
There was no gender effect in the data.
23
2(Offender membership; ingroup vs. outgroup) X 2(Intergroup goal structure: cooperative vs.
competitive) between subject design.
Overview of Procedure
Upon arrival, participants were greeted and immediately escorted by an experimenter to
individual rooms in which an entire experimental session took place. Participants were asked to
wait in the room and read an information sheet on a computer screen while waiting for other
participants to show up (with a maximum four participants per session). A Qualtrics survey on
the computer was used to record most of the participants’ responses.
Upon the experimenter’s return, participants were informed that the study was
investigating attitudes of Dornsife and Marshall students about campus resource issues and that,
as part of the study, they were going to participate in a discussion with another participant who
was a member of the other group (i.e., an outgroup member who was a Marshall or Dornsife
student). Participants first completed a measure of identification with their ingroup. They then
answered a set of questions that probed for their awareness of the issues to be presented in the
article they were to subsequently read, and then read an excerpt from the “Daily Trojan,” the
campus newspaper. A competitive or a cooperative structure was induced via the bogus news
article, which was written so as to manipulate the nature of Marshall and Dornsife student inter-
relations (see Appendix A, 340 and 342 words). Crossed with these two types of intergroup goal
structure, the news article also presented a case of wrongdoing wherein a secretary of either the
Dornsife Student Council or the Global Business Student Association had embezzled money
from the fund that was to be used for the creation of new facilities and computer labs for
outgroup students. Participants then provided their initial reactions to the news article in an open-
ended questionnaire (see Appendix B), allegedly to facilitate the later discussion. They moved on
24
to fill out a questionnaire that contained measures of punitive and lenient attitudes toward the
offender, group-based feelings, and impressions of average ingroup members and outgroup
members. Next, participants answered questions concerning demographic information: their age,
sex, race and ethnicity, major, year in school, and whether English is their native language. They
then answered several questions that served as a manipulation check and assessed their
comprehension of the article. Finally, participants were asked to guess about the purpose of the
study, after which they read a debriefing form that explained the true nature of the study and the
deceptions involved. They were asked to bring their opinion sheet (see Appendix B) to the
experimenter at the place where they were originally greeted. The experimenter made sure that
participants understood the rationale behind the deception, answered questions they asked,
probed for participants’ suspicion and thought processes if time was left, apologized for the
deception, thanked participants for their participation, and asked them not to share the
information about the study with future potential participants (i.e., students) on campus.
Competitive intergroup goal structure. Participants in the Competitive goal structure
condition first answered two questions specifically designed to create a conflict between
Dornsife and Marshall students. When participants were Dornsife students, for example, one
asked if the participant was aware of the existence of an organization that provides benefits to
Marshall students at the expense of Dornsife students. These questions used five-point Likert-
type scales (1=not aware, 2= hardly, 3=somewhat, 4=very,5=fully aware). They then read about
a conflict between Dornsife and Marshall School students. The article, ostensibly extracted from
the Daily Trojan, described a shortage in computer labs and facilities faced by Dornsife and
Marshall students. Dornsife students had allegedly proposed creation of a fund that would aid in
fighting these economic difficulties but Marshall students were said to have rejected that same
25
plan due to their unwillingness to donate their own money. The article furthermore suggested
that University funding policies be altered to allocate more scholarships to Marshall students at
the expense of Dornsife students. For Marshall student participants, “Dornsife” was replaced
with “Marshal” and vice versa in the questions and the article.
Cooperative intergroup goal structure. Participants in the Cooperative goal structure
first answered two questions designed to create a cooperative attitude between Dornsife and
Business students. For example, one asked whether they were aware that relations between
Dornsife and Marshall students are friendly. Participants in the Cooperative goal structure then
read a news article about cooperation between the two student groups in order to improve the
economic difficulties on campus.
Manipulation of group membership of an offender. In the ingroup offender condition
the news article continued to read as follows for Dornsife and Marshall participants respectively
Recently a student secretary of the Dornsife Student Council (or Marshall Business
Student Government) was caught embezzling from the fund that was set up for the
creation of new facilities and computer labs for Marshall (or Dornsife) students. A total
of $3892 was deposited from the account under the guise of an expense for a Council
luncheon. The Dornsife student secretary (or Marshall student secretary) admitted the
guilt, yet remains silent about the motive behind the offense.
In the outgroup offender condition, italicized parts were replaced with the alternative in the
parenthesis for Dornsife and Marshall students respectively, making the outgroup both the
offender’s group membership and the target of the harm from the perspective of the participant’s
own group membership.
26
Measures
All responses were measured on seven point Likert-type scales that ranged from: 1 “not
at all” to 7 “very much”. The following items were used for Dornsife student participants. For
Marshall student participants, “Dornsife” was replaced with “Marshall” in all items.
Retributive attitudes: punitiveness and leniency toward the offender. Punitiveness
and leniency toward the offender were assessed with 3 items each: punitiveness, “This student
deserves a severe punishment”; “As a member of Dornsife community, I urge the student to be
sanctioned officially.”; and “the Office of the Dean of Dornsife should take this offense very
seriously” (α = .75 ); leniency, “This student deserves a lenient treatment.”; “As a Dornsife
student, I think that we should not make a big deal out of this offense.”; and “The Dornsife
student council should approach the issue with compassion” (α = .61).
As expected, principal components analysis with Oblimin rotation indicated that the
items intended to measure punitiveness and leniency loaded on the separate factors (explaining
37.92% and 24.94% of the variance respectively) with all item loadings exceeding .66.
Group-based emotions: shame, anger, and guilt. A feeling of group-based shame was
assessed with three items: “I feel that this student perpetrator is a disgrace to the Dornsife
community.”; “As a Dornsife student, I feel humiliated to face the fact that this student has stolen
money from the fund built upon the hard work by Marshall students”; and “I feel ashamed of this
student.” (α = .76). A feeling of group-based anger was assessed with three items: “As a Dornsife
student, I feel angry with what this student has done.”; “This student perpetrator upsets me.”; and
“As a member of Dornsife, I feel frustrated to learn that we have such a student who engages in
wrongdoing.” (α = .87). A feeling of group-based guilt was assessed with two items: “I feel
guilty of the harm done by this student on Marshall students.” and “I feel remorse for what this
27
student has done to Marshall community.” (α = .70).
Principal components analysis with Oblimin rotation indicated that the items intended to
measure shame and anger load on the same factor (explaining 54.33% of the variance) and the
items for guilt on the second factor (14.75%), with all item loadings exceeding .42. Hence, a
composite score of shame and anger (α = .89) was used in the analysis.
Group identification. An individual’s psychological inclusion in the group has been
shown to be the basis on which group-threat impacts group members’ identity and engenders
attitudes and feelings associated with the actions of another ingroup member (Leach et al., 2008).
Thus, the following four items were used to measure participants’ group identity: “Being a
Dornsife (Marshall) student is an important aspect of who I am.”; “I feel a bond with the
Dornsife (Marshall) community.”; “I am committed to my degree program which USC Dornsife
(Marshall) offers”; and “Being a Dornsife (Marshall) student has little to do with who I am
(reversed item)” (α = .77). Descriptive statistics and intercorrelations among all the variables are
reported in Table 1
2
.
Manipulation checks. At the end of the experiment, the intensity of the cooperation
between Dornsife and Business students were assessed using two items: “How much conflict is
there between Marshall and Dornsife? (a reverse item)” and “How much Dornsife and Marshall
students agree in fundraising strategies?” (α = .77). To check the manipulation of the offender’s
group membership, a short memory quiz was administered to participants to identify the group
membership of the offender.
2
Marshall students (M = 5.31, SD = 1.12) were higher on group identification than Dornsife
students (M = 4.22, SD = 1.02), t(167) = -5.20, p < .001. There were no other group differences
among other variables. When the group difference was entered as a covariate in the moderated
regression analyses, substantive results did not change.
28
Table 1
Descriptive Statistics and Intercorrelations among Variables Examined in Experiment 1
Mean SD 1 2 3 4
1. Punitiveness 4.36 1.32 -
2. Leniency 1.78 0.83 -.18* -
3. Shame and anger 3.71 1.43 .62** -.07 -
4. Guilt 2.78 1.46 .27** .16* .52** -
5. Level group identity 4.42 1.12 .09 -.05 .24** .33**
Note. **p < .01; *p < .05.
Results
Manipulation Checks
In a 2 X 2 ANOVA, the perceived cooperation measure administered at the end of the
experiment only showed a main effect of intergroup goal structure, F(1, 176) = 572.41, p < .0001.
Participants perceived greater cooperation between Dornsife College and Marshall School
students in the cooperative (M = 5.88, SD = 0.98) than the competitive condition (M = 2.59, SD =
0.85). There was neither a main effect of offender’s group membership nor an interaction (Fs <
1).
On the memory quiz toward the end of the experiment, eleven participants failed to
correctly recall the offender’s group membership and hence were excluded from the analyses.
Also excluded were the eighteen participants who failed to correctly recall that the harm was
done to their outgroup. As a result, 169 participants remained to be included in the analyses.
3
3
The percentage of participants who were discarded did not differ by experimental conditions,
χ
2
(3) = 1.33, p = .72.
29
Offender’s Group Membership, Intergroup Goal Structure, and Level of Group
Identification as Predictors of Retributive Attitudes and Group-Based Emotions
A series of moderated regression analyses was conducted on retributive attitudes and
group-based emotions, including an assessment of the 3-way interaction among offender’s group
membership (ingroup vs. outgroup), intergroup goal structure (cooperation vs. competition), and
level of group identification (continuous). A cross-over 2-way interaction between the offender’s
group membership and the intergroup goal structure was expected in support of the hypothesis 1
and 2. Theory suggests that this interaction be further qualified by a 3-way interaction, wherein
the level of group identification moderated the hypothesized black sheep and devil protection
effects (hypothesis 3).
Retributive attitudes: punitiveness and leniency. The moderated regression analysis
of punitiveness showed a marginally reliable interaction between the offender’s group
membership and level of group identification, β = -.21, p = .06, wherein the level of group
identification positively predicted punishment toward the outgroup offender, β = .23, p < .05,
whereas it had no relationship with that toward the ingroup offender, β = -.04, p = .73. An
analysis of leniency indicated a conceptually parallel interaction, β = .32, p < .01, wherein the
level of group identification negatively predicted leniency toward the outgroup offender, β = -.25,
p < .05, whereas it had no reliable relationship with leniency toward the ingroup offender, β
= .16, p = .15. Overall, regardless of the intergroup goal structure, stronger group identification
was associated with stronger retributive attitudes only toward the outgroup offender. The
predicted 3-way or 2-way interactions were not found.
Group-based emotions: shame, anger, and guilt. A similar moderated regression
analysis was conducted separately on the shame and anger composite as well as the guilt score.
30
The shame and anger composite score was positively predicted by level of group identification, β
= .25, p < .01, and showed an interaction between the offender’s group membership and the
intergroup goal structure, β = .31, p < .05. The pattern was the same for Guilt: level of group
identification, β = .41, p < .001; examination of the interaction between the offender’s group
membership and the intergroup goal structure yielded a reliable effect, β = .31, p < .05.
To probe the interactive effects of the offender’s group membership and the intergroup
goal structure on the shame and anger composite, as well as guilt, a series of post-hoc ANOVAs
were conducted
4
. These tests examined the effects of the offender’s group membership,
controlling for level of group identification separately for the cooperative and competitive
intergroup conditions. Under the cooperative condition, there was a simple effect of the
offender’s group membership on the shame and anger composite score, F(1, 81) = 4.28, p < .05,
but not on guilt, F(1, 81) = 2.40, p = .27. Specifically, participants felt greater shame and anger
toward the ingroup offender (M = 4.00, SD = 1.42) than the outgroup offender (M = 3.48, SD =
1.29). This is consistent with the predicted black sheep effect under a cooperative intergroup goal
structure. Under the competitive condition, on the other hand, there was a simple effect of the
offender’s group membership on guilt, F(1, 82) = 5.70, p < .05, but not on the shame and anger
composite, , F(1, 82) = 1.47, p = .23. Specifically, participants felt less guilt toward the ingroup
(M = 2.23, SD = 1.16) than the outgroup offender (M = 3.11, SD = 1.64). This is consistent with
the expectation of a devil protection effect under a competitive intergroup goal structure.
When the effect of the intergroup goal structure, controlling for the level of group
identification, was examined separately for the ingroup and outgroup offender conditions, a
4
Since there is no continuous variable in the post-hoc analyses, results based on ANOVAs are
reported instead of those based on regression analyses for ease of interpretation and their
consistency with the experimental design. Nonetheless, the substantive results are the same when
the regression approach is used.
31
marginal effect was found on both the shame and anger composite, along with a reliable guilt
effect for the ingroup offender condition, F(1,79) = 3.12, p = .08 and F(1, 79) = 5.59, p < .05
respectively, indicating that participants more strongly felt group-based emotions toward the
ingroup offender under the cooperative intergroup condition than the competitive intergroup
condition. This is consistent with the hypothesized expectation that intergroup cooperation will
elicit stronger group-based emotions in response to an ingroup offender’s act. For the outgroup
offender conditions, there was a marginal simple effect of intergroup goal structure on the shame
and anger composite, F(1,84) = 3.15, p = .08, but not on guilt, F(1, 84) = 1.78, p = .19, indicating
that participants felt less shame and anger toward the outgroup offender under the cooperative
intergroup condition than the competitive intergroup condition (see Figure 1).
Figure 1. Mean group-based emotions (+/- SE) toward an ingroup or outgroup offender as a
function of intergroup goal structure, Experiment 1.
Discussion
Experiment 1 was the first study to show that the induction of an intergroup goal
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
Cooperation Competition Cooperation Competition
Shame and Anger Guilt
Ingroup
Outgroup
32
structure of cooperation vs. competition moderates the occurrence of a black sheep effect.
Consistent with hypothesis 1, participants in a cooperative intergroup situation felt greater shame
and anger toward an ingroup member than an outgroup member, each of whom had acted
offensively toward the same outgroup. This is the result that is commonly termed the black sheep
effect. Consistent with hypothesis 2, the obtained cross-over interaction supports the notion that
intergroup cooperation, in comparison to intergroup competition, induces more negative
reactions to an ingroup offender, as well as more leniency toward an outgroup offender. The
latter result is conceptually consistent with the outgroup leniency observed among members of a
high status group, who presumably have less need for positive distinctiveness and more concern
about displaying fairness than do those of a low status group (Braun & Gollwitzer, 2012; Feather
& Souter, 2002; van Prooijen & Lam, 2007). The discovery of stronger derogation of an ingroup
offender, as compared with an outgroup offender when ones group is in a cooperative intergroup
structure with an outgroup is a novel contribution of the present study.
Experiment 1 also found that guilt yields some effects that parallel those of shame and
anger. Participants felt greater guilt toward an ingroup offender under a cooperative than a
competitive intergroup condition, whereas their guilt toward an ingroup offender was lower than
that toward an outgroup offender under a competitive intergroup condition. As seen in Figure 1,
these findings seem to be driven by the fact that under a competitive intergroup condition, guilt
for the ingroup offender was particularly low. This is consistent with the notion of a defense
mechanism in which group members avoid feeling group-based emotions in order to protect the
positive distinctiveness of their group identity.
The current experiment had participants anticipate interacting with an outgroup member
in order to eliminate distancing (e.g., disidentifying with their group or physically disavowing
33
group membership) as a means of regulating the threat that an ingroup offender might potentially
elicit. The presence of this constant feature of the experimental procedure may explain the failure
to find the predicted 3-way interaction (hypothesis 3) wherein the level of group identification
will moderate the cross-over 2-way interaction reported above. In retrospect, it seems likely that
because the dependent measures were obtained just before participants anticipated having a
discussion about the offender with an outgroup member, it made their group identity uniformly
salient across conditions and eliminated the influence of individual differences in the level of
participants’ group identification. That is, participants were temporarily made accountable for
their fellow group member’s individual actions in the eyes of an outgroup member, and were
more likely to think of themselves being part of that group regardless of their initial level of
group identification.
Note that a main effect of group identification was obtained on the group-based emotions.
Higher identifiers showed stronger emotional reactions to either offender, regardless of the
intergroup goal structure. Level of group identification interacted with the offender’s group
membership to affect retributive attitudes (i.e., punitiveness and leniency). The higher the
participant’s group identification, the stronger the retributive reactions to the outgroup offender.
These effects of group identification are likely due to the fact that the ingroup and the outgroup
in this experiment shared a fairly accessible superordinate identity (i.e., their University
affiliation) and that there is a significant overlap between the subgroup identification (i.e., their
college affiliation being either the Dornsife or Marshall school) and the common group
identification (USC). In other words, those who care about their school affiliation likely care
about their University affiliation, and stronger retributive responses were likely exhibited
because someone, regardless of which group member, engaged in an offense that harmed the
34
University.
Despite its major contribution, Experiment 1 failed to detect the hypothesized effects on
retributive attitudes. Hence, hypothesis 4, which is concerned with the underlying mediational
process, could not be examined. The only effect found was that, regardless of the nature of
intergroup condition, the higher the group identification, the more punitive and less lenient were
the attitudes expressed toward an outgroup offender. As previously noted, the lack of an effect
on attitudes might be attributable to the design of the experiment, wherein participants were
made to anticipate a face-to-face discussion session in which they shared their opinion about the
offender with an outgroup member. This may have eliminated the effect of a competitive
intergroup goal structure on shared attitudes by inducing agreeableness with an anticipated
outgroup member but not on reported emotions such as shame, anger, and guilt, perhaps because,
while these emotions were not expected to be shared with an outgroup during the discussion,
their attitudes were. Experiment 2 eliminated the possibility of inadvertently inducing
agreeableness and thereby vitiating the effects of the manipulated variables on retributive
attitudes. It did so by abandoning the experimental procedures of Experiment 1, using instead a
paradigm that examines actual behaviors and action tendencies that are recalled from the past.
Chapter 3: Experiment 2
Recall that the black sheep effect refers to the tendency of ingroup members to display a
harsher treatment of an ingroup offender, by comparison with their treatment of an outgroup
offender, when each offender has committed the identical offensive act. In the context of a
controlled laboratory experiment that confronted participants with an incident wherein an
ingroup or outgroup member committed an offensive action toward members of their outgroup,
experiment 1 provided initial evidence for the occurrence of a black sheep effect under a
35
cooperative intergroup goal structure, and conversely, a devil protection effect (the reverse of the
black sheep effect) under a competitive intergroup goal structure. In Experiment 2, in
conjunction with Mechanical Turk Amazon, an event sampling methodology was used with a
diverse sample of participants who were asked to recall actual incidents in their past wherein
others’ had engaged in an offensive act. Event sampling methodology is commonly used in
emotion research (e.g., Johns, Schmader, & Lickel, 2005) and has the advantage of allowing us
to examine felt emotions and actual behaviors in response to real and very intense situations (e.g.,
a fist fight) that may be hard to recreate in the lab. Accordingly, the black sheep effect was
assessed with measures that reflected actual punitive behaviors (i.e., whether the participant did
something to punish an offender) and punitive action tendencies (i.e., whether the participants
tried to or wanted to do something, if they could, to punish the offender). By contrast, to assess
the opposing devil protection effect, rather than inferring it from leniency or a lack of
punishment, it was measured directly with items that reflect justification of the offensive act,
minimization of its offensiveness, and protection of the offender.
The design of experiment 2 was conceptually parallel to that of experiment 1. Mechanical
Turk users were instructed to recall and write about an instance when they had observed or heard
about an ingroup member acting offensively toward a member (or members) of an outgroup with
whom they were either in a cooperative or competitive relationship. Their perceptions of the
level of threat associated with the offender; their emotional reactions such as shame, anger, and
guilt; their action tendencies to punish and defend the offender; and their level of identification
with their ingroup were measured.
An additional aim of the experiment 2 was to test the mediation hypotheses (Hypothesis 4
and 5). Besides assessing group-based emotions, perceptions of threat associated with an
36
offender were measured in the hope of better understanding the conditions that cause a black
sheep effect to occur under a cooperative intergroup goal structure and a devil protection effect
to occur under a competitive intergroup goal structure. As explained earlier, the occurrence of a
black sheep effect as well as group-based shame, is rooted in the notion that an ingroup offender
undermines a fellow group member’s group identity. By contrast, an ingroup offender who
harms an outgroup likely undermines the intergroup relationship by undermining the trust
between the groups. Therefore, when fellow ingroup members are concerned about having a
positive relationship with an outgroup, they are more likely to be susceptible to a threat of losing
their moral reputation in the eyes of that outgroup than when they are in a contentious
relationship. In addition to group-based shame and anger, this specific threat, henceforth called
“intergroup-threat”, was expected to mediate the relationship between group identification and
punishment of an ingroup offender under a cooperative intergroup condition. Moreover, a threat
associated with an ingroup image is known to be causally antecedent to group-based emotions in
reaction to an ingroup offender (e.g., Iyer et al., 2007). Therefore, the perception of the
intergroup threat was expected to precede the group-based shame and anger in the mediation
path.
Under a competitive intergroup goal structure, on the other hand, this intergroup-threat
should be low, regardless of one’s level of group identity. Therefore, when groups were in
competition, intergroup threat was not expected to mediate the relationship between group
identification and the protection of an ingroup offender (i.e., devil protection effect). Instead,
suppression or avoidance of group-based shame and anger were expected to mediate this
relationship. As previously indicated, the lack of a mediational effect of intergroup threat under
the competitive condition is expected to be countered by the presence of its effect under the
37
cooperation condition, wherein intergroup threat is expected to induce group-based shame and
anger and in turn suppress the protective tendency toward an ingroup offender – a tendency that
is implicit in the more general occurrence of ethnocentrism.
Method
Participants and Design
Participants were 185 Mechanical Turk users (109 males, 70 females, and 6 unknown)
varying in age approximately from 18 to 71 years old (37% of the sample was in their late 20s
and 19% in their early 30s). They were randomly assigned to the same 2(Offender membership;
ingroup vs. outgroup) X 2(Intergroup goal structure: cooperative vs. competitive) between
subject design employed in Experiment 1. In all conditions, participants recalled an incident
wherein either an ingroup member or an outgroup member had acted offensively toward an
outgroup member. Each participant was paid for 50 cents upon completion of the survey
5
. An
undergraduate student who was blind to the study hypotheses read participants’ stories and
flagged them to be discarded from the data set when 1) the offender’s group membership was not
consistent with an assigned experimental condition, 2) the victim was not an outgroup member,
or 3) the reported incident did not pertain to an offensive action. Data from 82 responses were
discarded, resulting in a final sample of 103 participants (63 male and 39 female, 1 unknown)
varying in age from 18 to 65 years old (36.9 % in their late 20s and 18.4% in their early 30s).
Most of the participants were Asians (57.3%), Caucasians (21.4%), or Native Americans (9.7%).
Procedure
Participants in the ingroup offender competitive intergroup conditions were instructed to:
5
For Mechanical Turk sample, the amount awarded is typically small (e.g., nickels and dimes for
5-10 minutes tasks) and participants are often internally motivated to complete the task
(Burhmester, Kwang, Gosling, 2011).
38
“think about an incident when someone in a group that you belong to committed an
offensive action toward a person who belongs to another group that has a competitive and
hostile relationship towards your group. That is, your ingroup and outgroup are in a
contentious relationship while someone in your group did something wrong to the
outgroup (see Appendix C for the full description of the instruction)”.
Those in the outgroup offender competitive intergroup conditions were instructed to:
“think about an incident when someone in a group that you do not belong to and which
has a competitive and hostile relationship towards your group committed an offensive
action toward a member of his or her own group. That is, your ingroup and outgroup are
in a contentious relationship while someone in your outgroup did something wrong to
his/her group (see Appendix C for the full description of the instruction)”.
In the cooperative conditions, the instruction was identical except that the italicized words above
were altered to be, cooperative, friendly, and harmonious respectively. Participants were also
told that they could report an offensive action that they had observed either first hand or had
learned about through another source. They were instructed to spend five minutes writing about
the event and to describe as clearly as possible what they were thinking and feeling at the time.
Next, they rated the perceived threat (to self, group, and the intergroup relationship) associated
with the offender’s act, their emotional reactions to the event, and their action tendencies to
either punish or defend and protect the offender. Finally, they provided further information about
the incident (e.g., the presence or absence of an audience, the presence or absence of
consequences to the offender, the nature of their own interpersonal relation to the offender and
the victim, how they knew about the incident, the competitiveness/cooperativeness between the
two groups, their level of identification with the ingroup, and the types of groups that were
39
involved, such as family, friends, teams, and social categories).
Measures
Perceived threats associated with the offender. Perceived self-threat caused by
association with the offender was assessed with the following four items on seven-point Likert-
type scales (1 = strongly disagree to 7 = strongly agree): “I was afraid that people would think
that I was like the offender,” “I wondered if people thought that I had something to do with the
offense,” “I felt that this incident made me look bad,” and “I was worried about myself being
held accountable for the offender” (α = .83). On the same scales, perceived general group-threat
by association was assessed using the following four items: “I was concerned that my group
would lose its reputation because of this incident,” “I was worried that people would make
negative judgments about the nature of my ingroup based on the event,” “I felt that this incident
made my ingroup look bad,” and “I was worried about my ingroup being held responsible for the
offensive action” (α = .84). Similarly, perceived group-threat by association for the specific
outgroup (henceforth, called intergroup-threat) was measured by the four items: “I was afraid
that people in the outgroup would think negatively of our group based on this incident,” “I
wondered whether we lost reputation among people in the outgroup because of this incident,” “I
felt this incident made the relationship between my ingroup and the outgroup bad,” and “I was
apprehensive whether the outgroup would blame my ingroup for the harm inflicted on their
victim(s)” (α = .82).
Principal components analysis with Oblimin rotation indicated that the items intended to
measure a general group-threat and an intergroup-threat load on the same factor (explaining
55.2% of the variance) and the items for self-threat loaded on a second factor (10.49%), with all
40
item loadings exceeding .48
6
. It is not surprising that the general group-threat and the intergroup-
threat items loaded on the same factor given that a general group-threat is conceptually inclusive
of the intergroup-threat (i.e., an inter-group threat is part of a general group-threat). The average
inter-item r among the two sets of items was .76. For hypothesis testing, however, effects
associated with these measures were examined separately in order to examine as directly and
specifically as is possible the hypothesis that intergroup-threat specifically induces black sheep
effects under a cooperative goal structure, but has no such mediational role in a competitive one.
Group-based emotions: shame, anger, and guilt. Participants rated the extent to which
they felt a series of emotions in response to the incident on nine-point Likert-type scales (1 = not
at all to 9 = very intensely). Items were constructed to measure shame, guilt, and anger based on
conceptual considerations. Anger was assessed with three items (α = .80): “I felt angry at the
offender”, “I was upset with the offender”, and “I was disappointed with the offender's action”.
Shame was assessed with three items (α = .75): “I felt disgraced by the offender”, “I felt ashamed
of the offender's action”, and “I felt embarrassed about the offender's action”. Guilt was assessed
with three items (α = .72): “I felt regretful of the offender's action”, “I felt remorseful for what
the offender had done”, and “I felt guilty for the offender's action”. To ensure that the effects of
these group-based emotions are not accounted for by general negative affect, 2 items that
assessed positive affect, happy and calm, were also included (r = .59). That is, the opposite
effects should be found for positive affect if group-based shame and anger were conceptually
isomorphic to a lack of positive affect. Similarly to Experiment 1, principal components analysis
with Oblimin rotation indicated that the items intended to measure shame and anger loaded on
the same factor (explaining 42.30% of the variance) and the items for guilt on a second factor
6
Principal components analysis with Varimax rotation yielded the same two factors.
41
(15.88%). The items for positive affect loaded on the last factor (8.29%), with all item loadings
exceeding .45. Hence, the composite score of shame and anger (α = .86) was used in the analysis.
Action tendencies: punishment and devil protection. Action tendency for punishment
was measured with five items, using seven-point Likert-type scales (1 = strongly disagree to 7 =
strongly agree): “I thought that the offender deserved severe consequences,” “I felt that the
offender needed to be punished severely,” “I thought that the offender had to be punished more
harshly,” “I reprimanded (or wanted to reprimand) the offender for his/her negative action,” and
“I showed (or wanted to show) my disagreement with the offensive attitude of the perpetrator” (α
= .82). On the same scales, action tendency for protection was assessed using five items: “I
defended (or wanted to defend) the offender for the reasons behind his/her act,” “I took sides
with the offender for his or her reasons for the action,” “I understood and sympathized with the
reasons why he or she engaged in such an offensive action,” “I personally agreed with the
offender even though some people might have taken his or her actions as offensive,” and “I
personally did not think that the offender did anything wrong” (α = .89).
Group identification. Participants’ identification with the group was measured using the
same four items used in Experiment 1, except that the group for which participants’
identification was assessed was the group that they recalled in their reported incident (α = .72).
Manipulation check. At the end of the survey, participants were asked to report the
usual level of cooperativeness/competitiveness of the ingroup and outgroup at other times in the
absence of the incident (1 = very cooperative to 9 = very competitive). Toward the end of the
survey participants were also asked to identify the group membership of the offender (ingroup,
outgroup, or other).
42
Results
Characteristics of the Incidents Recalled
An average of 124 words (Median = 104, SD= 98.44) were used to describe the incident.
Examination of participants’ narratives showed that offensive actions included: physical
violence; verbal insults and quarrel, and stealing, hiding, and manipulating organizational/team
information. Additional questions in the survey showed that 6 % of the recalled incidents took
place within the past week, 17% within the past month, 36% within the past year, 26% within the
past 5 years, and the rest longer than 5 years ago. The types of groups involved in the incidents
ranged from work related colleagues and teams (30%), friends (35%), non-work related teams
(13%), family members and relatives (7%), ideology groups (6%), social categories (5%), and
others (4%). The offenders were most often male (80% male, 16% female, 4% unidentified) and
slightly more balanced for the victims (57% male, 35% female, 8% unidentified)
7
.
Additionally, 72% of the participants were physically present at the time of event
whereas the rest were not (e.g., they had heard about the incidence on the news or from someone
else). In 88% of the incidents there was a presence of an audience of ingroup members 95% of
the time, outgroup members 90% of the time, and others 70% of the time. To participants’
knowledge, 55% of offenders suffered some kind of consequences (e.g., fired from the job, fine,
and jail time), whereas 21% of the victims received some kind of compensation (e.g., money,
7
To test potential impacts of gender of participants, of an offender, and of victim, a 2
(participant’s gender) X 2 (offender’s gender) X 2 (victim’s gender) ANOVA was conducted on
punishment and devil protection. There was a main effect of participant’s gender on punishment,
F(1, 83) = 6.64, p < .05., in which males were more punitive (M = 5.00, SD = 1.05) than female
(M = 4.42 , SD = 1.37). In the main hypothesis testing reported, analyses were repeated with
participant’s gender as a covariate. Substantive results did not change, and therefore reported
results omit gender in the model.
43
apology)
8
.
Descriptive statistics and intercorrelations among key dependent variables are reported in
Table 2. As expected, the shame and anger composite exhibit a moderately high correlation with
the guilt score. Similarly, as previously noted, there were high intercorrelations among self-threat,
general group-threat, and intergroup-threat items.
Table 2
Descriptive Statistics and Intercorrelations among Variables Examined in Experiment 2
Mean SD 1 2 3 4 5 6 7 8
1. Punishment 4.84 1.18 -
2. Protection 3.49 1.51 -.26** -
3. Shame and anger 6.33 1.64 .53** -.27** -
4. Guilt 5.55 1.80 .34** .08 .62** -
5. Positive affect 3.21 2.02 -.04 .33** -.17 .04 -
6. Self-threat 4.09 1.44 .13 .36** .22* .31** .28** -
7. Group-threat 4.67 1.35 .08 .23* .34** .40** .01 .68** -
8. Intergroup-threat 4.69 1.31 .16 .23* .28** .44** .06 .62** .79** -
9. Group identity 5.11 1.17 .23* -.05 -.03 .07 .03 -.07 .03 .09
Note. **p < .01; *p < .05.
Manipulation Checks
On the perceived intergroup competitiveness/cooperativeness measure administered at
the end of the survey, a 2 (intergroup goal structure: cooperative vs. competitive) X 2 (offender’s
group membership: ingroup vs. outgroup) ANOVA only showed a main effect of intergroup goal
structure, F(1, 98) = 7.29, p < .01. Participants perceived greater competition between the
ingroup and the outgroup in the competitive intergroup condition (M = 5.27, SD = 0.32) than in
8
Although potentially influential, there were no reliable main effects of the group type, physical
presence of participants at the incident, the presence of audience, and whether there was a
consequence to the offender as well as a compensation to the victim on punishment and
protection. Hence, none of these were used as a covariate in the main analyses.
44
the cooperative incident condition (M = 4.05, SD = 0.322). There was neither main effect nor an
interaction with offender’s group membership (Fs < 1).
As mentioned earlier, responses were discarded when the offender’s group membership
was not consistent with the description of the incident that was assigned by the experimental
condition. There were 7 cases wherein those who failed the manipulation check were nonetheless
retained in the data base because of a clear indication in their description of the incident that they
had properly recalled the offender’s group membership in relation to the event. Since there are
times that the offender’s membership might have had changed after the incident (e.g., fired from
the company because of the wrongdoing), it was reasonable to do so despite their failure to
identify the offender’s group membership. There were also 24 cases wherein participants
recalled an incident when an outgroup member acted offensively toward an ingroup member
instead of another outgroup member. These were excluded from the analysis. As a result, there
was a total of 103 participants (32 in the cooperative ingroup offender, 31 in the competitive
ingroup offender, 19 in the cooperative outgroup offender, and 21 in the competitive outgroup
offender conditions), remained in the analysis
9
.
9
The percentage of participants who were discarded did not differ by experimental condition, χ2
(3) = 1.93, p = .59. Nonetheless, despite use of procedures that randomly assign participants to
conditions, in the final sample of 185 participants there were 109 persons in the ingroup offender
condition whereas there were 76 in the outgroup offender condition. This presumably occurred
because participants who were assigned to the outgroup offender condition (i.e., an instruction to
recall an instance wherein an outgroup member committed an offensive act) were more likely
than those assigned to the ingroup offender condition to choose not to participate in the study.
Indeed, 914 Mturk users initially clicked to view the study (and thereby fired a pixel to recode
their entry to the Qualtrics survey system). There is no definitive way, however, to determine
whether or not they actually viewed the manipulation. One can only infer that there was attrition
bias with respect to the manipulation of the offender’s group membership both because of the
differential sample sizes reported above and because of the fact that one ordinarily has less
knowledge about outgroup than ingroup members, making it likely that it is more difficult to
think of an offensive act that was committed by an outgroup member. See Appendix C for the
instructions.
45
Offender’s Group Membership, Intergroup Goal Structure, and Level of Group
Identification as Predictors of Action Tendencies, Group-Based Emotions, and Threat
Perceptions
A series of moderated regression analyses was conducted to test the hypothesized 3-way
interaction on action tendencies (punishment and protection), group-based emotions (the shame
and anger composite and guilt scores), and threat perceptions (intergroup-threat, general group-
threat, and self-threat). In these analyses, group identification was centered and entered in the
first step of a hierarchical analysis along with the dichotomous variables of offender’s group
membership (ingroup vs. outgroup) and intergroup goal structure (cooperation vs. competition).
The second step included the cross-products of all possible combinations between these three
variables entered in the first step (three 2-way interaction terms). The final step included the 3-
way interaction term between the three variables. A cross-over interaction was predicted
wherein level of group identification positively predicts a punitive tendency toward an ingroup
offender under the cooperative intergroup condition whereas it decreases it in the competitive
intergroup condition. The reverse of this pattern was predicted for outgroup offender conditions.
Action tendencies: punishment and protection. The moderated regression analysis of
punishment indicated main effects of an offender’s group membership (greater punishment for
outgroup offenders than ingroup offenders), β = -.23, p < .05, and level of group identity, b = .21,
p < .05 that were qualified by the predicted 3-way interaction, β = 1.26, p < .001. When the
component two-way interactions were examined separately within the intergroup goal structure
conditions, there was a marginally reliable 2-way interaction for the cooperative intergroup
condition, β = .48, p = .09, and a reliable 2-way interaction for the competitive intergroup
condition, β = -.96, p < .001. Furthermore, the simple slopes representing the relationship
46
between the level of group identification and punishment were examined within each of the
2(offender) X 2(intergroup goal structure) conditions. These analyses provided support for a
black sheep effect under the cooperative condition and a devil protection effect under a
competitive condition. Under cooperation, the level of group identification had a positive
relationship with punishment of an ingroup offender and directionally opposite relationship
toward an outgroup offender (viz. a black sheep effect). On the other hand, under competition,
the reverse was true wherein the level of group identification had a negative relationship with
punishment of an ingroup offender and a positive relationship with that of an outgroup offender
(viz. a devil protection; see Table 3 for coefficients and Figure 2 for the patterns of effects).
Table 3
Simple Slopes Representing the Relationship between level of Group Identity and Action
Tendencies, Group-Based Emotions, and Perceived Threats in Cooperative and Competitive
Intergroup Goal Structures for Ingroup and Outgroup Offenders
Note. ** p < .01; * p < .05; ^ p < .10; Ingroup = Ingroup offender condition; Outgroup =
Outgroup offender condition; Interaction = a 2-way interaction between offender’s group
membership and level of group identification.
Cooperation Competition
Outcome variables Ingroup Outgroup Interaction Ingroup Outgroup Interaction
Action Tendencies
Punishment
0.36* -0.12 .48^ -0.32^ 0.77**
-.96**
Protection
-0.38* 0.18 -.37 0.44* -0.31
.51*
Group-based Emotions
Shame and Anger
0.23 -0.19 .39 -0.31^ 0.08
-.32
Guilt
0.09 0.26 -.15 -0.1 0.23
-.32
Positive Affect
0.14 0.31 -.13 -0.13 -0.26
-.09
Perceived Threats
Intergroup-threat
0.31^ -0.24 .64* 0.13 0.38
-.13
General Group-threat
0.4* -0.23 .63* 0.27 -0.06
.24
Self-Threat
0.03 0.05 .11 -0.03 0.04
.02
47
Figure 2. A 3-way interaction between offender’s group membership, intergroup goal structure,
and level of group identification (at 1 SDs above and below its mean) on punishment.
When the same moderated regression was conducted on protection, it too indicated a 3-
way interaction, β = -.83, p < .01. When the component two-way interactions were examined
separately within the offender’s group membership conditions, there was a reliable 2-way
interaction under competition, β = .51, p < .05, but not under cooperation, β = -.37, p = .21.
Furthermore, the simple slope analyses parallel to those of punishment provided complementary
support for the devil protection and black sheep effects. Under cooperation, although the 2-way
interaction was not reliable, the level of group identification had a negative relationship with
protection of an ingroup offender whereas it had a directionally positive relationship with that of
an outgroup offender (viz. a black sheep effect). On the other hand, under competition, the level
of group identification had a positive relationship with protection of an ingroup offender and a
directionally negative relationship with that of an outgroup offender (viz. a devil protection
effect; see Table 3 and Figure 3).
1.0
2.0
3.0
4.0
5.0
6.0
7.0
Low High
Punishment
Group Identification
Ingroup offender
Competitive
Outgroup offender
Competitive
Ingroup offender
Cooperative
Outgroup offender
Cooperative
48
Figure 3. A 3-way interaction between offender’s group membership, intergroup goal structure,
and level of group identification (at 1 SDs above and below its mean) on protection.
Group-based emotions: shame, anger, and guilt. When the same moderated regression
analysis was applied to the shame and anger composite, it yielded the predicted 3-way
interaction, β = .60, p < .05. When the same component two-way interactions were examined
separately within the intergroup goal structure conditions, there were only directional 2-way
interactions under cooperation, β = .39, p < .18, and under competition, β = -.32, p = .16. Within
the offender’s group membership conditions, however, there was a 2-way interaction for the
ingroup offender condition, β = .45, p < .05, but not for the outgroup offender condition, β = -.18,
p = .43. The simple slopes analyses indicated a pattern similar to that found for punishment and a
pattern opposite to that found for protection (see Table 3). When the same analyses were
repeated for guilt and positive affect, there were neither a 3-way interaction, other 2-way
interactions, nor main effects
10
.
10
Guilt was highly correlated with the shame and anger composite (r = .62) and therefore the
analysis on the shame and anger composite was repeated while controlling for guilt. The result
1.0
2.0
3.0
4.0
5.0
6.0
7.0
Low High
Protection
Group Identification
Ingroup offender
Competitive
Outgroup offende
Competitive
Ingroup offender
Cooperative
Outgroup offender
Cooperative
49
Threat perceptions: self-threat, general group-threat, and intergroup-threat. When
the same moderated regression analyses were conducted on perceptions of self-threat, general
group-threat, and intergroup-threat separately, a main effect of offender’s group membership was
found across the three types of threat, β = .31, .42, and .39, ps < .01 respectively, wherein these
threats were higher when the offender was an ingroup member than an outgroup member. Due to
ones respective levels of association or connection with the two groups, these main effects are
reasonable, given that one is much more likely to be held responsible for the actions of ones own
group (or an ingroup offender) than those of another group (or an outgroup offender).
The predicted 3-way interaction only emerged on the intergroup-threat, β = .56, p < .05.
When the component two-way interactions were examined separately within the intergroup goal
structure conditions, there was a 2-way interaction in the cooperative intergroup condition, β = -
.64, p < .05, but not in the competitive counterpart, β = -.16, p = .53. The simple slopes analyses
indicated a pattern similar to that found for punishment and a pattern reversed from that found
for protection (see Table 3).
On the other hand, general group-threat showed a 2-way interaction between offender’s
group membership and level of group identification, β = .45, p < .01, wherein, regardless of the
intergroup goal structure with the outgroup, level of group identification was positively related to
perceived general group-threat only when the offender was an ingroup member, β = .35, p < .01.
This effect is reasonable, given that an intergroup goal structure with a particular outgroup would
not be expected to be diagnostic of an ingroup culpability from the eyes of a third-party group.
That is, when a member of a group harms a member of another group, even when they are in
conflict, the offender group would be seen as immoral by a third party. Since there was no
showed that the 3-way interaction remained significant, beta = .45, p < .05, above and beyond
the main effect of guilt, beta = .62, p < .001.
50
conceptually relevant 3-way interaction effect for the self-threat or the general group-threat,
mediation analyses for these variables were omitted.
Intergroup-threat and the Shame and Anger Composite as Mediators of Black Sheep and
Devil Protection Effects
The preceding analyses showed similar patterns of support for the predicted 3-way
interaction on measures of action tendencies of punishment and protection, group-based
emotions of shame and anger, and the perception of intergroup-threat associated with the
offender. To explore the processes underlying black sheep versus devil protection effects,
separately within the cooperative and competitive intergroup conditions, perception of
intergroup-threat and the shame and anger composite were examined as serial mediators of the
influence of group identification on punishment and protection using Hayes’ (2013) SPSS Macro.
Under the cooperative intergroup condition, group identification was predicted to increase the
perception of intergroup-threat, and consequently induce feelings of shame and anger toward the
ingroup offender, which, in turn, will elicit punitive tendencies toward ingroup offenders. Under
the competitive condition, on the other hand, group identification was expected to augment
intergroup-threat, leading to a stronger suppression of group-based shame and anger, which is
concomitantly accompanied by an increased justification of an ingroup offender’s actions (i.e.,
devil protection).
Punishment of ingroup offender: black sheep effect. Under the cooperative intergroup
condition, a serial mediation was found in which intergroup-threat preceded group-based shame
and anger in mediating the relationship between group identification and punishment (Indirect
effect = .06, 95% bootstrap CIs [.01, .22], See Figure 4). The relationship between group
identification and punishment became insignificant when both intergroup-threat and the shame
51
and anger composite were controlled (b = .13, p = .25).
Figure 4. Intergroup-threat and group-based shame and anger as serial mediators of the black
sheep effect under cooperative intergroup condition. The model reports unstandardized
regression weights. ** p < .01; * p < .05.
Backwards mediation models were tested to examine the possibility of bidirectional
effects between intergroup-threat, the shame and anger composite, and punishment. When the
order of threat and emotion was reversed, mediation was no longer significant (Indirect effect
= .06, 95% bootstrap CIs [-.00, .26]). Furthermore, when the shame and anger composite was
examined as an outcome with intergroup-threat and punishment as mediators, mediation was
again not significant (Indirect effect = .16, 95% bootstrap CIs [-.01, .44]). This supports the
theoretically postulated temporal order of the mediators in the model.
Protection of ingroup offender: devil protection effect. When the same serial
mediation model was used to examine devil protection under the competitive intergroup
condition, only the shame and anger composite was found to mediate the relationship between
the level of group identification and devil protection (Indirect effect = .14, 95% bootstrap CIs
[.02, .41], See Figure 5). The relationship between group identification and protection became
insignificant when the shame and anger composite was controlled (b = .22, p = .17).
Group
Identification
Punishment
Intergroup
Threat
Shame
Anger
.44**
(.13)
.49*
.38** .31*
52
Figure 5. Intergroup-threat and group-based shame and anger as serial mediators of the devil
protection effect under competitive intergroup condition. The model reports unstandardized
regression weights. ** p < .01; * p < .05.
When a backwards mediation model was examined with the shame and anger composite
as an outcome instead of protection, mediation was also significant (Indirect effect = -.15, 95%
bootstrap CIs [-.46, -.01]). This suggests the presence of bidirectional effects between group-
based emotions of shame and anger and protection of an ingroup offender under the competitive
intergroup condition or that the shame and anger composite and protection are conceptually
isomorphic. The correlation between the protection of ingroup offender and the shame and anger
composite was only moderate under competition, r = -.40, p < .05, which suggests that the
former account is more likely than the latter.
When the same mediation model was examined under the cooperative intergroup
condition, a serial mediation was found in which intergroup-threats preceded group-based shame
and anger in mediating the relationship between group identification and protection of an
ingroup member (Indirect effect = -.07, 95% bootstrap CIs [-.33, -.01]). The relationship between
group identification and protection became insignificant when both intergroup-threat and the
shame and anger composite were controlled (b = -.06, p = .70), indicating that full mediation
Group
Identification
Devil
Protection
Intergroup
Threat
Shame
Anger
.37*
(.22)
.60*
-.40* .09
-.34*
53
occurred. Note that the same serial mediation path predicted punishment in the opposite direction,
providing a complimentary support for the occurrence of the black sheep effect and its reverse,
devil protection.
Parallel to the backward mediation results for punishment, a serial mediation was no
longer significant neither when the order of threat and emotion was reversed (Indirect effect
= .01, 95% bootstrap CIs [-.02, .11]) nor when the shame and anger composite was examined as
an outcome (Indirect effect = .03, 95% bootstrap CIs [-.03, .14]). Note that the correlation
between protection and punishment of an ingroup offender was only moderate under cooperation,
r = -.35, p < .05 (overall r across the experimental conditions = -.33, p = .06).
Experiment 2 conceptually replicated Experiment 1 by further identifying the moderating
effect of the strength of ones group identification on the respective emergence of black sheep
versus devil protection effects in cooperative and competitive intergroup goal structures
respectively. This evidence supports hypothesis 3. A cross-over 3-way interaction on punishment
showed that, under the cooperative intergroup goal structure, the stronger the participant’s group
identification, the greater the punishment of an ingroup offender and the less that of an outgroup
offender; by contrast under the competitive intergroup goal structure, the stronger their group
identification, the less the punishment of an ingroup offender and the greater that of an outgroup
offender. Again, not only did Experiment 2 conceptually replicate an outgroup leniency effect
under cooperation, but it also showed a reliable black sheep effect that was driven by the harsher
treatment recommended for an ingroup offender by stronger identifiers. Moreover, devil
protection, as measured by justification, minimization, and protection of an ingroup offender was
positively predicted by the level of group identification under the competitive condition and
negatively predicted by it under the cooperative condition; by contrast, the opposite trend was
54
found for an outgroup offender.
For the group-based shame and anger, similar effects were reliably found in the ingroup
offender conditions but not in outgroup offender conditions. This is reasonable given that group-
based shame is particularly concerned with group identity threat and therefore a targeting of
shame toward an outgroup member becomes less meaningful conceptually. Furthermore, the lack
of reliable 3-way interaction effects for the guilt and positive affect scores showed discriminant
validity for the effect of the experimental manipulations on group-based emotions.
In regards to intergroup-threat, similar effects were reliably found in the cooperative
intergroup condition but not in the competitive outgroup conditions. This is consistent with the
notion that the basis of the black sheep effect specifically rests on concern about the ingroup’s
reputation among outgroup members under cooperation but not under competition. Moreover,
the lack of reliable 3-way interaction effects for the self-threat and the general group-threat
provided discriminant validity for the kinds of threat involved in reactions to an ingroup vs.
outgroup offender.
More importantly, Experiment 2 clarified the process by which a black sheep effect
occurs under a cooperative intergroup goal structure. Under a cooperative relationship with an
outgroup, the higher the participants’ group identification, the more concerned they were with
the ingroup’s loss of reputation among outgroup members, which led to stronger feelings of
shame and anger toward an ingroup offender. In turn, these latter emotional reactions led ingroup
members to advocate stronger punishment of the offender. On the other hand, under a
competitive relationship with an outgroup, the effect of group identification on a devil protection
effect was mediated by group-based shame and anger such that the higher the participants’ group
identification, the less the group-based shame and anger felt toward an ingroup offender. This, in
55
turn, resulted in devil protection – leniency toward an ingroup offender. Note that a backward
mediation analysis further indicated that under competition, feelings of shame and anger are
suppressed as a result of protecting an ingroup offender.
Theoretical perspectives suggest that group identification does not moderate perceptions
of intergroup-threat when groups are in a competitive goal structure, an expectation confirmed
by the data. Hence, intergroup threat had no mediating role in accounting for the obtained devil
protection effect. These findings provide support for the view that when high identifiers are
faced by an ingroup offender and the intergroup goal structure is a competitive one, ingroup
members exhibit devil protection effects as a defense mechanism that promotes group solidarity.
Furthermore, this was contrasted with the parallel analyses conducted for the cooperative
condition wherein intergroup threat did mediate the suppression of the devil protection effect.
That is, under cooperation, the stronger the group identification, the stronger the intergroup
threat and the greater the shame and anger, which then resulted in a lower level of protection of
an ingroup member. This result compliments the mediation path observed for punishment of an
ingroup offender under cooperation.
Chapter 4: General Discussion
Two experiments manipulated the intergroup goal structure and an offender’s group
membership (ingroup vs. outgroup). Under a cooperative intergroup goal structure, analyses of
retributive reactions toward an ingroup offender who aggressed toward an outgroup member
provide evidence for the occurrence of a black sheep effect. Under a competitive intergroup goal
structure, its reverse (termed a devil protection effect) occurred. More specifically, in experiment
1, under a cooperative goal structure, participants felt greater shame and anger toward an ingroup
offender than an outgroup offender who had similarly inflicted harm upon outgroup members – a
56
black sheep effect. On the other hand, under a competitive goal structure, participants were more
likely to avoid feeling shame and anger, as well as guilt, toward the ingroup offender by
comparison with that toward the outgroup offender – a devil protection effect. In experiment 2,
the preceding effects were moderated by level of group identification, wherein the stronger the
participant’s identification with the ingroup, the larger the resulting magnitudes of a black sheep
effect (obtained under cooperation) and a devil protection effect (obtained under competition).
Furthermore, Experiment 2 provided evidence about the mechanisms that underlie the obtained
black sheep and devil protection effects. Under a cooperative intergroup goal structure, high
group identifiers more strongly felt that an ingroup offenders act threatened their ingroup’s moral
integrity in the eyes of the outgroup. In turn, these feelings of intergroup threat elicited feelings
of shame and anger, and led to harsher punishment of the ingroup offender who had aggressed
toward outgroup members. This threat of moral laxity was unassociated with the devil protection
effect -- an effect only obtained under the competitive intergroup goal structure, wherein the
more strongly group members identified with their ingroup, the more they suppressed shame and
anger and engaged instead in justification, minimization, and protection of an ingroup offender.
Perhaps, this suppression of shame and anger was also the result of devil protection as indicated
by the backward mediation analysis. These effects were observed in a variety of group types,
ranging from a college affiliation within an University (Experiment 1), work and non-work
related teams, to friendship groups (Experiment 2) in which participants were either confronted
with an incident of an offender on their campus (Experiment 1) or guided to recall an incident of
an offender from their past (Experiment 2).
The results are consistent with the previous research on outgroup leniency, which showed
that a black-sheep-like effect is driven by leniency toward a low status outgroup that poses little
57
threat to the positive distinctiveness of an ingroup (Braun & Gollwitzer, 2010; Shephard et al., in
press; van Prooijen & Lam 2007). Cooperative and low status outgroups may share the property
of reducing the need for positively differentiating ones ingroup from an outgroup or altering the
manner by which it is expressed. This allows them to display their egalitarian moral concern,
thereby enabling them to avoid an otherwise defensive reaction in the face of an offensive
outgroup member. Indeed, as discussed earlier, those studies which yielded outgroup leniency
had examined low status outgroups that are particularly non-threatening to the participants’
group (e.g., a weak minority group instead of an emerging low status group that is increasing in
number in the population or a minority group with a history of violence in the community).
Future studies need to examine the possibility that outgroup leniency will be attenuated when a
low status outgroup is threatening to change the status quo. Such an outcome conceptually
overlaps with the effect of competitive intergroup goals structure in the present study.
An important difference between the present study and studies of both outgroup leniency
and other black sheep studies concerned with retributive justice is that it made the outgroup the
target of the offender's harmdoing as a constant feature of the design. Past research has
confounded it in many cases. This feature allowed us to systematically examine the focal effects
of the intergroup goal structure and the offender’s group membership, as well as the specific
conditions in which the punishment of an ingroup offender influences the way in which outgroup
members perceive the ingroup as a whole.
In addition to promoting outgroup leniency, the current research suggests that a
cooperative intergroup goal structure increases punitive reactions toward an ingroup offender by
increasing concern about maintaining the ingroup’s moral integrity in the eyes of the outgroup.
The discrepancy between this finding and that of the previous research that shows a lack of such
58
an effect on ingroup offenders is perhaps due to the fact that the current research specifically
constrained the target of harm, making it consistently directed toward the outgroup in both
studies. As previously noted, prior research made the target of harm neither salient nor
consistent (e.g., Braun & Gollwitzer, 2010) and hence may have occluded the effects of a
concern about the ingroup’s moral integrity in the eyes of the outgroup. As Experiment 2
demonstrated, rather than a threat to one’s individual identity or a threat to a general group
identity, a type of threat directly associated with the ingroup’s moral integrity in the eyes of
others elicited retributive reactions toward an ingroup offender. That is, when they take the
perspective of an outgroup member whose fellow group members were offended by one of their
fellow ingroup members, people derogate and punish an ingroup offender, despite its potential
damage to one’s positive group distinctiveness. This finding is complementary to Cehajic and
Brown’s (2010) finding that good-quality intergroup contact predicted acknowledgment of in-
group responsibility by increasing perspective taking and decreasing ingroup justification
(perception that one’s ingroup equally suffered).
The role of outgroup perspective taking in inducing ingroup directed blame is studied in
the emotion literature. Research on collective guilt suggests that the beneficial effects of group-
based emotion in promoting positive intergroup relations may be limited to an outgroup-focused
form of guilt or to the distinct emotional responses of empathy or sympathy (e.g., Figueiredo et
al., 2010; Thomas, McGarty, & Mavor, 2009; Zebel et al., 2009). Despite the possibility that this
kind of empathetic guilt predicts reconciliation, however, it cannot be induced easily among
people who strongly identify with the offender’s ingroup. For example, Zebel et al. (2009) found
that instructing Dutch participants to imagine themselves as a slave when reading about the
Dutch slave trade or the treatment of Jews during WWII decreased collective guilt among
59
individuals with strong national identity whereas it increased that of weak identifiers. In the
current research, however, strong group identifiers, by comparison with weak identifiers, were
more concerned about a loss of the ingroup’s positive reputation among outgroup members, and
therefore were more likely to take blame for their fellow ingroup offender. This is perhaps based
on the possibility that they voluntarily engaged in perspective taking when in a cooperative goal
structure with their outgroup. Cooperation and positive intergroup contacts are known to
promote a more accurate (i.e., less ethnocentric) perspective taking of the views of outgroup
members than are competitive contacts or individualistic efforts (e.g., Johnson & Johnson, 1989,
2000). When forced to take the perspective of harmed outgroup members following an
imposition of harm by a fellow ingroup member, people are motivated to avoid or distance
themselves from the harm, or even worse, dehumanize outgroup members to justify their
ingroup’s wrongdoing (Castano & Giner-Sorolla, 2006). This is driven by our pervasive need for
a positive group identity (Tajfel & Turner, 1980). On the other hand, the current research
suggests that, when a cooperative goal structure is salient, it reduces one’s need for positive
distinctiveness. This allows people to focus on the loss to the outgroup and take the blame for the
harm by punishing an ingroup offender instead of distancing themselves from the harmdoer’s
group.
Until recently, a majority of research on the roles of group-based emotions in intergroup
relationships exclusively focused on guilt for its adaptive and prosocial function (e.g.,
Baumeister et al., 1994). Here, because of their theoretical links to moral identity threat and their
role in promoting actions that support reconciliation and change, the present research focused on
group-based shame and anger, while also examining guilt. Group-based shame and anger were
more sensitive than guilt to manipulations of the intergroup goal structure and an offender’s
60
group membership. The shame and anger composite was moderately correlated with guilt (rs
= .52 in Experiment 1 and .62 in Experiment 2) and in Experiment 1 the two measures showed
similar patterns of results. Furthermore, while controlling for guilt, shame and anger had a
mediational role in the obtained black sheep effect (Experiment 2). The current research is
consistent with studies of group-based shame and ingroup-directed anger (Iyer et al., 20007;
Leach et al., 2006; Schmader & Lickel, 2006; Thomas et al., 2009; see also Harth et al., 2008)
and suggests that a combination of the two will trigger retributive action toward an ingroup
wrongdoing. In turn, this leads to a remediation of the harm done to an intergroup relationship.
Limitations and Future Directions
A large number of respondents were discarded from Experiment 2. A total of 82
responses out of 185 were discarded because of participants’ failure to follow the instruction.
This may not be surprising given that the sample comes from Mechanical Turk users who were
paid very little to complete simple tasks online (hence large variance in intrinsic motivation in
completing a study). Perhaps, the task of recalling a specific event was too complicated or
cognitively taxing for Mechanical Turk users who were not intrinsically motivated. However,
this is a well-known cost of the use of Mechanical Turk and its data has been shown to be
reliable in replications across various laboratory studies (see Casler, Bickel, & Hackett, 2013)
and its validity is commonly not a threat (see Paolacci, Chandler, & Ipeirotis, 2010).
More concerning, perhaps, is that Mechanical Turk users did not provide their stories
(and so did not participate in the study at all) when they saw the instruction to recall an event
wherein an outgroup member offended another member of that group. Although there was no
definitive measure to tell whether or not they declined to participate in the study because of this
particular experimental condition, the fact that there were fewer responses collected for the
61
outgroup offender condition in comparison to the ingroup offender condition posed a threat of
attrition bias. Perhaps it is not intuitive for people to think about their own group and its
relationship to an outgroup in which an offence takes place. Nonetheless, Experiment 1
compensated for this limitation by examining a condition wherein participants were confronted
with an offender in a manner that precluded attrition bias.
Making the outgroup the target of the harm as a constant feature of the present research,
despite its clear contribution to a conceptually meaningful outcome, limited the scope of
phenomena to which the current findings apply. For instance, if the harm is targeted toward an
ingroup member, devil protection effect and outgroup leniency are unlikely regardless of which
type of intergroup goal structure connects groups. This expectation rests on the assumption that
group members, particularly high identifiers, are likely to be primarily concerned about the harm
done to the ingroup, rather than focusing on reputational implications of the harmdoer’s
association with their group.
Experiment 1 failed to detect either black sheep effect or devil protection effects on
measures of retributive attitudes toward an offender. As discussed earlier, participants provided
their opinion about an offender in anticipation of contact with an outgroup member in order to
discuss the matter. This may have activated their shared University identity with the outgroup
member and/ or simply induced agreeableness with the outgroup member in their expressed
opinion and attitudes about the offence. However, the hypothesized relationship did emerge on
measures of emotion, such as shame, anger, and guilt. Perhaps this selective confirmation of the
hypotheses occurred because participants did not expect to have to share or express emotional
reactions during the intergroup discussion that they were made to anticipate. Hence, responses on
these latter measures were not constrained by an imagined outgroup audience. Nonetheless,
62
when, in Experiment 2, past behaviors were recalled in response to an actual instance of having
observed an offender, both black sheep and devil protection effects were found. In the incidents
recalled, there was a mixture of ingroup members, outgroup members, and others in the audience
(except about 10 % of the incidents being privately observed). Publicity of a threat, provocation,
or wrongdoing is felt more intensely in the presence of an audience (Vasquez et al., 2013). Thus,
the effects of an audience should be further examined in order to (a) better understand the
circumstances under which black sheep and devil protection effects may emerge and (b) the
conciliatory functions that they serve in intergroup relationships.
63
References
Abrams, D., Marques, J., Bown, N., & Dougill, M. (2002). Anti-norm and pro-norm deviance in
the bank and on the campus: Two experiments on subjective group dynamics. Group
Processes & Intergroup Relations, 5(2), 163-182.
Abrams, D., Marques, J.M., Bown, N.J., & Henson, M. (2000). Pro-norm and anti-norm
deviance. Journal of Personality and Social Psychology, 78, 906–912.
Abrams, D., Rutland, A., & Cameron, L. (2003). The Development of Subjective Group
Dynamics: Children's Judgments of Normative and Deviant In ‐Group and Out ‐Group
Individuals. Child Development, 74(6), 1840-1856.
Abrams, D., Rutland, A., Cameron, L., & Ferrell, J. (2007). Older but wilier: In-group
accountability and the development of subjective group dynamics. Developmental
Psychology; Developmental Psychology, 43(1), 134.
Asbrock, F. (2010). Stereotypes of social groups in Germany in terms of warmth and
competence. Social Psychology, 41, 76–81.
Asbrock, F., Lemmer, G., Wagner, U., Becker, J., & Koller, J. (2009). Das Gefühl macht den
Unterschied—Emotionen gegenüber “Ausländern” in Ost- und Westdeutschland [The
feeling makes the difference—Emotions toward “foreigners” in East and West Germany]. In
W. Heitmeyer (Ed.), Deutsche Zustände Folge 7 (pp. 152–167). Frankfurt: Suhrkamp.
Bègue, L. (2001). Social judgment of abortion: A black-sheep effect in a Catholic sheepfold. The
Journal of social psychology, 141(5), 640-649.
Branscombe, N. R., & Miron, A. M. (2004). Interpreting the ingroup's negative actions toward
another group: Emotional reactions to appraised harm.
Braun, J., & Gollwitzer, M. (2012). Leniency for out ‐group offenders. European Journal of
Social Psychology, 42(7), 883-892.
64
Brewer, M. B. (1999). The psychology of prejudice: Ingroup love and outgroup hate? Journal of
Social Issues, 55(3), 429-444.
Bernard, J. L. (1979). Interaction between the race of the defendant and that of the jurors in
determining verdicts. Law and Psychological Review, 5, 103-111.
Brown, R., & Cehajic, S. (2008). Dealing with the past and facing the future: Mediators of the
effects of collective guilt and shame in bosnia and herzegovina. European Journal of Social
Psychology, 38(4), 669-684.
Buhrmester, M., Kwang, T., & Gosling, S. D. (2011). Amazon's mechanical turk a new source of
inexpensive, yet high-quality, data? Perspectives on Psychological Science, 6(1), 3-5.
Casler, K., Bickel, L., & Hackett, E. (2013). Separate but equal? A comparison of participants
and data gathered via Amazon’s MTurk, social media, and face-to-face behavioral testing.
Computers in Human Behavior, 29(6), 2156-2160.
Castano, E., & Giner-Sorolla, R. (2006). Not quite human: Infrahumanization in response to
collective responsibility for intergroup killing. Journal of Personality and Social
Psychology, 90(5), 804.
Čehajić, S., & Brown, R. (2010). Silencing the past effects of intergroup contact on
acknowledgment of in-group responsibility. Social Psychological and Personality Science,
1(2), 190-196.
Chan, M. K., Louis, W. R., & Hornsey, M. J. (2009). The effects of exclusion and reintegration
on the evaluation of deviant opinion holders. Personality and Social Psychology Bulletin,
35(12), 1619-1631.
Coull, A., Yzerbyt, V. Y., Castano, E., Paladino, M. P., & Leemans, V. (2001). Protecting the
65
ingroup: Motivated allocation of cognitive resources in the presence of threatening ingroup
members. Group processes & intergroup relations, 4(4), 327-339.
Darley, J. M., & Pittman, T. S. (2003). The psychology of compensatory and retributive justice.
Personality and Social Psychology Review, 7(4), 324-336.
Dovidio, J. F., Gaertner, S. L., Isen, A. M., & Lowrance, R. (1995). Group representations and
intergroup bias: Positive affect, similarity, and group size. Personality and Social
Psychology Bulletin, 21, 856-865.
Dovidio, J. F., Gaertner, S. L., & Validzic, A. (1998). Intergroup bias: Status, differentiation, and
a common in-group identity. Journal of Personality and Social Psychology, 75, 109-120.
Eidelman, S., & Biernat, M. (2003). Derogating black sheep: Individual or group protection?.
Journal of Experimental Social Psychology, 39(6), 602-609.
Eidelman, S., Silvia, P. J., & Biernat, M. (2006). Responding to deviance: Target exclusion and
differential devaluation. Personality and Social Psychology Bulletin, 32(9), 1153-1164.
Feather, N. T., & Oberdan, D. (2000). Reactions to penalties for an offence in relation to ethnic
identity, responsibility, and authoritarianism. Australian Journal of Psychology, 52(1), 9-16.
Feather, N. T., & Souter, J. (2002). Reactions to mandatory sentences in relation to the ethnic
identity and criminal history of the offender. Law and Human Behavior, 26(4), 417-438.
Festinger, L. (1950). Informal social communication. Psychological Review, 57(5), 271.
Figueiredo, A., Doosje, B., Valentim, J.P., & Zebel, S. (2010). Dealing with past colonial
conflicts: How perceived characteristics of the victimized outgroup can influence the
experience of group-based guilt. International Journal of Conflict and Violence, 4(1), 89-
105.
Foley, L. A., & Chamblin, M. H. (1982). The effect of race and personality on mock jurors'
66
decisions. Journal of Psychology, 112, 47-51.
Gaertner, S. L., & Dovidio, J. F. (2000). Reducing intergroup bias: The Common Ingroup
Identity Model. Philadelphia, PA: Psychology Press.
Gaertner, S. L., Dovidio, J. F., Banker, B. S., Houlette, M., Johnson, K. M., & McGlynn, E. A.
(2000). Reducing intergroup conflict: From superordinate goals to decategorization,
recategorization, and mutual differentiation. Group Dynamics: Theory, Research, and
Practice, 4, 98-114.
Gaetner, S. L., Dovidio, J. F., Nier, J. A., Ward, C. M., & Banker, B. (1999). Across cultural
divides: The value of a superordinate identity. In D. A. Prentice & D. T. Miller (Eds.),
Cultural divides: Understanding and overcoming group conflict (pp. 173-212). New York:
Russell Sage Foundation.
Giner-Sorolla, R., Castano, E., Espinosa, P., & Brown, R. (2008). Shame expressions reduce the
recipient’s insult from outgroup reparations. Journal of Experimental Social Psychology,
44(3), 519-526.
Gollwitzer, M., & Keller, L. (2010). What you did only matters if you are one of us: Offenders’
group membership moderates the effect of criminal history on punishment severity. Social
psychology, 41(1), 20.
Gordon, R. A. (1993). The Effect of Strong Versus Weak Evidence on the Assessment of Race
Stereotypic and Race Nonstereotypic Crimes1. Journal of Applied Social Psychology, 23(9),
734-749.
Gordon, R. A., Bindrim, T. A., McNicholas, M. L., & Walden, T. L. (1988). Perceptions of blue-
collar and white-collar crime: The effect of defendant race on simulated juror decisions. The
Journal of Social Psychology, 128(2), 191-197.
67
Goto, N., & Karasawa, M. (2011). Identification with a wrongful subgroup and the feeling of
collective guilt. Asian Journal of Social Psychology, 14(4), 225-235.
Gray, D. B., & Ashmore, R. D. (1976). Biasing influence on defendants' characteristics on
simulated sentencing. Psychological Reports, 38, 727-738.
Harth, N. S., Kessler, T., & Leach, C. W. (2008). Advantaged group's emotional reactions to
intergroup inequality: The dynamics of pride, guilt, and sympathy. Personality and Social
Psychology Bulletin, 34(1), 115-129.
Hutchison, P., & Abrams, D. (2003). Ingroup identification moderates stereotype change in
reaction to ingroup deviance. European Journal of Social Psychology, 33(4), 497-506.
Hutchison, P., Abrams, D., Gutierrez, R., & Viki, G. T. (2008). Getting rid of the bad ones: The
relationship between group identification, deviant derogation, and identity maintenance.
Journal of Experimental Social Psychology, 44(3), 874-881.
Iyer, A., Schmader, T., & Lickel, B. (2007). Why Individuals Protest the Perceived
Transgressions of Their Country The Role of Anger, Shame, and Guilt. Personality and
Social Psychology Bulletin, 33(4), 572-587.
Johnson, D. W., & Johnson, R. (1989). Cooperation and Competition: Theory and research.
Edina, MN: Interaction Book Co.
Johnson, D. W., & Johnson, R. (2000). The three Cs of reducing prejudice and discrimination. In
S. Oskamp (Ed.), Reducing prejudice and discrimination (pp.139–268). Mahwah, NJ:
Erlbaum.
Johns, M., Schmader, T., & Lickel, B. (2005). Ashamed to be an American? The role of
identification in predicting vicarious shame for anti-Arab prejudice after 9–11. Self and
Identity, 4(4), 331-348.
68
Jussim, L., Coleman, L. M., & Lerch, L. (1987). The nature of stereotypes: A comparison and
integration of three theories. Journal of Personality and Social Psychology, 52(3), 536.
Khan, S., & Lambert, A. J. (1998). Ingroup favoritism versus black sheep effects in observations
of informal conversations. Basic and applied social psychology, 20(4), 263-269.
Kemmelmeier, M. (2005). The Effects of Race and Social Dominance Orientation in Simulated
Juror Decision Making1. Journal of Applied Social Psychology, 35(5), 1030-1045.
Kerr, N., Hymes, R., Anderson, A., & Weathers, J. (1995). Defendant-juror similarity and mock
juror judgments. Law and Human Behavior, 19, 545-567.
Klein, K., & Creech, B. (1982). Race, rape, and bias: Distortion of prior odds and meaning
changes. Basic and Applied Social Psychology, 3, 21-33.
Leach, C. W., Iyer, A., & Pedersen, A. (2006). Anger and guilt about ingroup advantage explain
the willingness for political action. Personality and Social Psychology Bulletin, 32(9), 1232-
1245.
Leidner, B., & Castano, E. (2011). Morality shifting in the context of intergroup violence.
European Journal of Social Psychology, 42(1), 82-91.
Lewis, A. C., & Sherman, S. J. (2003). Hiring you makes me look bad: Social-identity based
reversals of the ingroup favoritism effect. Organizational Behavior and Human Decision
Processes, 90(2), 262-276.
Lickel, B., Miller, N., Stenstrom, D. M., Denson, T. F., & Schmader, T. (2006). Vicarious
retribution: The role of collective blame in intergroup aggression. Personality and Social
Psychology Review : An Official Journal of the Society for Personality and Social
Psychology, Inc, 10(4), 372-390.
69
Lickel, B., Schmader, T., Curtis, M., Scarnier, M., & Ames, D. R. (2005). Vicarious shame and
guilt. Group Processes & Intergroup Relations, 8(2), 145-157.
Lickel, B., Steele, R. R., & Schmader, T. (2011). Group ‐based shame and guilt: Emerging
directions in research. Social and Personality Psychology Compass, 5(3), 153-163.
Marques, J., Abrams, D., Paez, D., & Martinez-Taboada, C. (1998). The role of categorization
and in-group norms in judgments of groups and their members. Journal of Personality and
Social Psychology, 75(4), 976.
Marques, J.M., Abrams, D., & Serôdio, R.G. (2001). Being better by being right: Subjective
group dynamics and derogation of in-group deviants when generic norms are undermined.
Journal of Personality and Social Psychology, 81, 436–447.
Marques, J. M., & Yzerbyt, V. Y. (1988). The black sheep effect: Judgmental extremity towards
ingroup members in inter ‐and intra ‐group situations. European Journal of Social
Psychology, 18(3), 287-292.
Marques, J.M., Yzerbyt, V.Y., & Leyens, J.-P. (1988). The “black sheep effect”: Extremity of
judgment toward ingroup members as a function of group identification. European Journal
of Social Psychology, 18, 1–16.
Matheson, K., Cole, B., & Majka, K. (2003). Dissidence from within: Examining the effects of
intergroup context on group members’ reactions to attitudinal opposition. Journal of
Experimental Social Psychology, 39(2), 161-169.
Miller, D. T., & Vidmar, N. (1981). The social psychology of punishment reactions. The justice
motive in social behavior, 145, 172.
Miron, A. M., Branscombe, N. R., & Biernat, M. (2010). Motivated shifting of justice standards.
Personality and Social Psychology Bulletin, 36(6), 768-779.
70
Morton, T. A., Postmes, T., & Jetten, J. (2007). Playing the game: strategic considerations and
responses to normative and deviant group members. European Journal of Social Psychology,
37, 599-616.
Paolacci, G., Chandler, J., & Ipeirotis, P. G. (2010). Running experiments on amazon mechanical
turk. Judgment and Decision making, 5(5), 411-419.
Pfeifer, J. E., & Ogloff, J. R. E (1991). Ambiguity and guilt determinations: A modern racism
perspective. Journal of Applied Social Psychology, 21, 1713-1725.
Pfeifer, J. E., & Ogloff, J. R. (2003). Mock juror ratings of guilt in Canada: Modern racism and
ethnic heritage. Social Behavior and Personality: an international journal, 31(3), 301-312.
Pinto, I. R., Marques, J. M., Levine, J. M., & Abrams, D. (2010). Membership status and
subjective group dynamics: Who triggers the black sheep effect?. Journal of personality and
social psychology, 99(1), 107.
Schmader, T., & Lickel, B. (2006). The approach and avoidance function of guilt and shame
emotions: Comparing reactions to self-caused and other-caused wrongdoing. Motivation and
Emotion, 30(1), 42-55.
Sargent, M. J., & Bradfield, A. L. (2004). Race and Information Processing in Criminal Trials:
Does the Defendant’s Race Affect How the Facts Are Evaluated?. Personality and Social
Psychology Bulletin, 30(8), 995-1008.
Sommers, S. R., & Ellsworth, P. C. (2000). Race in the courtroom: Perceptions of guilt and
dispositional attributions. Personality and Social Psychology Bulletin, 26, 1367–1379.
Schruijer, S. G., & Lemmers, L. (1996). Explanations and evaluations by Turks and Dutchmen
of norm violating ingroup and outgroup behaviour. Journal of community & applied social
psychology, 6(2), 101-107.
71
Shepherd, L., Spears, R., & Manstead, A.S.R. (2013). ‘This will bring shame on our nation’: The
role of anticipated group-based emotions on collective action. Journal of Experimental
Social Psychology, 49, 42-57.
Shepherd, L., Spears, R., & Manstead, A.S.R. (in press). When does anticipating group-based
shame lead to lower ingroup favoritism? The role of status and status stability. Journal of
Experimental Social Psychology.
Sweeney, L. T., & Haney, C. (1992). The influence of race on sentencing: A meta ‐analytic
review of experimental studies. Behavioral Sciences & the Law, 10(2), 179-195.
Tangney, J. P., Wagner, P. E., Hill-Barlow, D., Marschall, D. E., & Gramzow, R. (1996).
Relation of shame and guilt to constructive versus destructive responses to anger across the
lifespan. Journal of Personality and Social Psychology, 70(4), 797.
Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. The Social
Psychology of Intergroup Relations, 33(47), 74.
Taylor, T. S., & Hosch, H. M. (2004). An examination of jury verdicts for evidence of a
similarity-leniency effect, an out-group punitiveness effect or a black sheep effect. Law and
Human Behavior, 28(5), 587-598.
Thomas, E. F., McGarty, C., & Mavor, K. I. (2009). Transforming “apathy into movement”: The
role of prosocial emotions in motivating action for social change. Personality and Social
Psychology Review, 13(4), 310-333.
Tylr, T. R., Boeckmann, R. J., Smith, H. J., & Huo, Y. J. (1997). Social justice in a diverse
society. Westview Press.
72
van Prooijen, J. W. (2006). Retributive reactions to suspected offenders: The importance of
social categorizations and guilt probability. Personality and Social Psychology Bulletin,
32(6), 715-726.
van Prooijen, J. W., & Lam, J. (2007). Retributive justice and social categorizations: The
perceived fairness of punishment depends on intergroup status. European Journal of Social
Psychology, 37(6), 1244-1255.
Vasquez, E.A., Pedersen, W.C., Bushman, B.J., Kelley, N.J., Demeestere, P., & Miller, N.
(2013). Lashing out after stewing over public insults: The effects of public provocation,
provocation intensity, and rumination on triggered displaced aggression. Aggressive
Behavior, 13, 13-29.
Vidmar, N. (2002). Retribution and revenge. Handbook of justice research in law, 31-63.
Zebel, S., Doosje, B., & Spears, R. (2009a). How perspective-taking helps and hinders group-
based guilt as a function of group identification. Group Processes & Intergroup Relations,
12(1), 61-78.
73
APPENDIX A
Texts of the Newspaper Articles for Experiment 1: Manipulation of Intergroup Goal
Structure (Dornsife Participants’ Version)
Cooperative Intergroup Goal Structure
Marshall students and Dornsife students interact in a number of different ways. Most
importantly, they are working towards creating a university environment that will benefit both
Marshall students and Dornsife students.
Recent economic difficulties, however, have hindered their efforts in obtaining new
facilities and more computer labs. Dornsife Student Council has proposed a plan to obtain money
that will then be used for creation of new facilities and computer labs. They suggested that
willing Dornsife and Marshall students should volunteer a little of their time by working in
various positions offered throughout the University. The proceeds of their work should go
towards a fund that will later be used for the creation of new facilities and computer labs.
Marshal students agree with this plan and believe that this is a great idea.
On other matters as well, opinions of Dornsife and Marshall students are similar. For
example, Marshall students believe that the Office of Academic Affairs should increase the
funding and allocate more scholarships to Dornsife and Marshall students. Marshall Business
Student Government proposed that this could be done at the expense of other student funding
available on campus. Specifically, the Office of Academic Affairs could cut the funds allocated
to students’ initiated events that are not directly relevant to their studies in order to obtain
necessary funds. Marshall students argue that such a demand is reasonable considering the fact
that better facilities and more computer labs are necessary to meet the current needs of Dornsife
and Marshall. Dornsife Student Council agrees with such proposals and they do not seem to be
troubled by the apparent pay-cut on various student organized activities. Dornsife students argue
74
that students’ run organizations have received many benefits throughout the past and prioritizing
educational demands are necessary for the success of both Dornsife and Marshal schools.
Competitive Intergroup Goal Structure
Marshall students and Dornsife students interact in a number of different ways. Most
importantly, they are working towards creating a university environment that will benefit both
Marshall students and Dornsife students.
Recent economic difficulties, however, have hindered their efforts in obtaining new
facilities and more computer labs. Dornsife Student Council has proposed a plan to obtain money
that will then be used for creation of new facilities and computer labs. They suggested that
willing Dornsife and Marshall students should volunteer a little of their time by working in
various positions offered throughout the University. The proceeds of their work should go
towards a fund that will later be used for the creation of new facilities and computer labs.
Business students however, disagree with this plan. Representatives of Marshall Business
Student Government said that they believe that the students should keep the money that they earn
instead of devoting it to the fund.
Opinions of Dornsife and Marshall students differ in other ways as well. For example,
Marshall students believe that the Office of Academic Affairs should increase the funding and
allocate more scholarships to Marshall students. Marshall Business Student Government
proposed that this could be done at the expense of Dornsife students. Specifically, the Office of
Academic Affairs could cut the funds allocated to Dornsife students in order to obtain necessary
funds. Marshall students argue that such a demand is reasonable considering the fact that
Marshall students after graduation contribute more to the economic well-being of society
compared to Dornsife students. Dornsife Student Council finds their beliefs unfounded and
75
motivated by greed. Dornsife students argue that their fields of study benefit the society to the
same extent if not more than Marshall students. Specifically, Dornsife students point to their
contributions to the advancement of knowledge and science.
76
APPENDIX B
An Opinion Sheet under the Guise of Facilitating the Face-to-Face Discussion
OPINION SHEET
*You will be asked to show this sheet during the discussion. Please write legibly.
What do you think about the proposal by Dornsife Student Council?
What do you think about the proposal by Marshall Business Student Government?
What do you think about the student secretary who was caught embezzling?
Others:
77
APPENDIX C
Instructions for Recalling Incidents of Ingroup and Outgroup Offenders in a Cooperative
or Competitive Goal Structure
Ingroup Offender and Competitive Intergroup Goal Structure Condition
78
Ingroup Offender and Cooperative Intergroup Goal Structure Condition
79
Outgroup Offender and Competitive Intergroup Goal Structure
80
Outgroup Offender and Cooperative Intergroup Goal Structure
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Asset Metadata
Creator
Onuki, Mayuko
(author)
Core Title
Black sheep or devil protection?: the effects of cooperative and competitive intergroup goal structures on group members reactions to an offensive act when committed by an ingroup or an outgroup ...
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Psychology
Publication Date
09/12/2014
Defense Date
07/28/2014
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
Anger,black sheep effect,goal structures,group emotions,intergroup relationships,OAI-PMH Harvest,offenders,punishment,Shame
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Miller, Norman (
committee chair
), John, Richard S. (
committee member
), Read, Stephen (
committee member
), Wiltermuth, Scott (
committee member
), Wood, Wendy (
committee member
)
Creator Email
mayukonuki@gmail.com,monuki@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-476939
Unique identifier
UC11287126
Identifier
etd-OnukiMayuk-2933.pdf (filename),usctheses-c3-476939 (legacy record id)
Legacy Identifier
etd-OnukiMayuk-2933.pdf
Dmrecord
476939
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Onuki, Mayuko
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
black sheep effect
goal structures
group emotions
intergroup relationships
offenders