IS OUTGROUP PREJUDICE FUNDAMENTAL?
EXPLORING INTERGROUP BIAS IN THE MINIMAL GROUP PARADIGM
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the Degree of Doctor of
Philosophy in the Graduate School of The Ohio State University
By
Michael John McCaslin, M.A.
Graduate Program in Psychology
The Ohio State University
2010
Dissertation Committee:
Richard E. Petty, Advisor
Russell H. Fazio
William A. Cunningham
Copyright
Michael John McCaslin
2010
ABSTRACT
Previous research using the minimal group paradigm (MGP) has shown that group categorization per se is sufficient to elicit a preference for one’s ingroup over the outgroup (Tajfel et al., 1971). Social psychologists have generally believed this relative bias to be the result of ingroup favoritism rather than outgroup derogation, but the lack of an appropriate control group in prior studies has made it difficult to definitely determine the direction of bias. Making this methodological adjustment in the current research revealed evidence indicating that explicit (Experiments 1 and 2) and implicit (Experiment
3) intergroup bias are a function of both ingroup favoritism and outgroup derogation. In addition, results from Experiments 1 and 2 showed that less favorable outgroup attitudes were most likely to be reported by those low in outgroup identification, a variable that has received relatively little attention in the literature. Together, these findings suggest that outgroup derogation is an important determinant of intergroup attitudes, particularly for certain individuals.
ii
ACKNOWLEDGMENTS
I would first like to extend my deepest gratitude to those who served as faculty in the Social Psychology Area during my time at Ohio State. Thank you for creating an intellectually supportive environment that encouraged the pursuit of excellence. In particular, I would like to thank the members of my committee, Russ Fazio and Wil
Cunningham, for their time and effort. And a special thank you is given to Tim Brock, whose efforts helped ensure that I could come to Ohio State in the first place. In addition, I would like thank the members of the Petty lab and the Group for Attitudes and
Persuasion for their insightful suggestions and comments throughout the various stages of development of this research. I greatly appreciate all your helpful feedback.
I would also like to thank my advisor, Rich Petty, for his steadfast guidance and generous support over the last several years. Rich, your commitment and dedication to your students is unparalleled, and I feel extremely fortunate to have been mentored by someone as willing to share his time, resources and intellectual knowledge as you have been. I could not ask for a better advisor and colleague.
Finally, I would like to thank my wife and family, whose love and support is matched only by their patience and understanding. Jamie, Mom, Teresa, Sarah, and Phil
– thank you for everything. iii
VITA
May 13, 1980...... Born – Omaha, Nebraska
2003...... B.A. Psychology, Saint Louis University
2003-2004...... University Fellow, The Ohio State University
2004-2005...... NIMH Predoctoral Trainee, The Ohio State University
2005...... M.A. Psychology, The Ohio State University
2005-2009...... Graduate Teaching Associate, Department of Psychology, The Ohio State University
2009-2010...... Graduate Research Associate, Nationwide Center for Advanced Customer Insights
PUBLICATIONS
McCaslin, M. J., & Petty, R. E. (2007). Persuasion. In R. Baumeister & K. Vohs (Eds.), Encyclopedia of social psychology (Vol. 2, pp. 665-669). Thousand Oaks, CA: Sage Publications.
Briñol, P., Petty, R. E., & McCaslin, M. (2009). Changing attitudes on implicit versus explicit measures: What is the difference? In R. E. Petty, R. H. Fazio, & P. Briñol (Eds.). Attitudes: Insights from the new implicit measures (pp. 285-326). New York, NY: Psychology Press.
McCaslin, M. J., & Petty, R. E. (2009). Attitude change. In D. Matsumoto (Ed.), Cambridge dictionary of psychology, Cambridge University Press.
iv
Petty, R. E., Briñol, P., Loersch, C., & McCaslin, M. J. (2009). The need for cognition. In M. R. Leary & R. H. Hoyle (Eds.), Handbook of individual differences in social behavior. New York: Guilford Press.
FIELDS OF STUDY
Major Field: Psychology
Specialization: Social Psychology
Minor Fields: Quantitative Psychology, Consumer Psychology
v
TABLE OF CONTENTS
Abstract...... ii
Acknowledgements...... iii
Vita...... iv
List of Tables...... viii
List of Figures...... ix
Chapters:
1. Introduction...... 1 Minimal Group Studies...... 3 Limitations of Prior Research...... 5 Control Group Studies...... 9 Current Research...... 13
2. Experiment 1...... 18
Method...... 19 Participants and Design...... 19 Procedure...... 19 Independent Variables...... 20 Dependent Variables...... 21 Results...... 22 Attitudes...... 22 Identification...... 22 Moderation Analyses...... 23 Discussion...... 24
vi
3. Experiment 2...... 29
Method...... 30 Participants and Design...... 30 Procedure...... 30 Independent Variables...... 30 Dependent Variables...... 31
Results...... 32 Attitudes Toward Group Categories...... 32 Attitudes Toward Group Members...... 32 Identification...... 33 Moderation Analyses...... 34 Discussion...... 36
4. Experiment 3...... 40
Method...... 40 Participants and Design...... 40 Procedure...... 41 Independent Variables...... 41 Dependent Variable...... 41
Results...... 42 Discussion...... 43
5. General Discussion...... 45 Future Directions...... 49
Notes...... 53
References...... 55
Appendices:
A. Alternative Moderation Analyses for Experiments 1 and 2...... 65
B. Illustrations...... 69
C. Sample Materials from Experiment 1...... 84
D. Sample Materials from Experiment 2...... 91
E. Sample Materials from Experiment 3...... 101
vii
LIST OF TABLES
Table 1. Multiple-choice allocation matrix with MD + MIP and MJP as opposing orientations...... 70
viii
LIST OF FIGURES
Figure 1. Mean explicit attitude ratings toward group categories in Experiment 1...... 71
Figure 2. Mean explicit attitude ratings toward group categories in Experiment 2...... 72
Figure 3. Mean explicit attitude ratings toward group exemplars in Experiment 2...... 73
Figure 4. Mean category outgroup derogation as a function of ingroup and outgroup identification in Experiment 2...... 74
Figure 5. Mean exemplar outgroup derogation as a function of ingroup and outgroup identification in Experiment 2...... 75
Figure 6. Implicit attitudes toward group categories as indicated by mean response latencies on the Brief IAT in Experiment 3...... 76
Figure 7. Mean attitudes toward the ingroup vs. control (ingroup favoritism) as a function of ingroup identification in Experiment 1...... 77
Figure 8. Mean attitudes toward the outgroup vs. control (outgroup derogation) as a function of outgroup identification in Experiment 1...... 78
Figure 9. Mean attitudes toward the ingroup vs. control (ingroup favoritism) as a function of outgroup identification in Experiment 2...... 79
Figure 10. Mean attitudes toward the ingroup vs. control (ingroup favoritism) as a function of ingroup identification in Experiment 2...... 80
Figure 11. Mean attitudes toward the outgroup vs. control (outgroup derogation) as a function of outgroup identification in Experiment 1...... 81
Figure 12. Mean attitudes toward ingroup members vs. control group members (ingroup favoritism) as a function of ingroup identification in Experiment 2...... 82
Figure 13. Mean attitudes toward ingroup members vs. control group members (ingroup favoritism) as a function of outgroup identification in Experiment 2...... 83
ix
CHAPTER 1: INTRODUCTION
Psychologists have long understood that forming relationships with other people
satisfies a number of basic human needs, and research has shown that people from
different cultures have a strong drive to cultivate and maintain associations with others
(Gardner, Pickett, & Brewer, 2000; Manstead, 1997). One way individuals connect with
others and derive a sense of belonging is by affiliating with a group. However, although
group membership can confer a number of advantages to those within a group, it might
also lead to prejudice against those outside the group. Indeed, membership in a group
often appears to be a catalyst for the kind of negative attitudes that lead to wars,
genocide, hate crimes, political and economic oppression (e.g., Sidanius, 1993), and other
hostile discriminatory behaviors (Sherif, Harvey, White, Hood, & Sherif, 1961).
Given the devastating social conflict that seems to come at the hands of group-
based prejudice, a great deal of work in psychology and beyond has focused on
investigating the underlying processes and consequences of social categorization (e.g.,
Tajfel, 1969). Over the years, different perspectives have emerged regarding the role that
mere categorization plays in fueling prejudice and discrimination. On the one hand,
some have argued that affiliation with an ingroup (a group to which one belongs) by definition engenders antagonism and aggression toward an outgroup (a group to which
1
one does not belong). For example, in his discussion of the nature of ethnocentrism,
Sumner (1906) concluded that the positive sentiments typically attributed to one’s ingroup are paralleled by feelings of negativity and hostility toward outgroups. That is, prejudice against outsiders is an unavoidable aspect of identifying with an ingroup. This perspective fits with the evolutionary principle of adaptive conservatism, which holds that people are predisposed toward distrusting anything unfamiliar or different
(Henderson, 1985; Mineka, 1992). As a result, it may be that anyone outside the familiar bounds of the ingroup is viewed with dislike and suspicion.
In contrast, social psychologists have generally rejected the position that ingroup love and outgroup hate are reciprocally related (Hewstone, Rubin, & Willis, 2002). For instance, drawing on the postulates of Allport (1954), Brewer (1999) proposed that, at a fundamental level, ingroup favoritism and outgroup derogation are orthogonal constructs that arise under distinct circumstances. Since humans are social beings who require the cooperation of others for long-term survival, Brewer (1999) argued that people are characterized by an obligatory interdependence. In order to survive, individuals must rely on others for information, assistance, and shared resources – and be willing to share each in return. This kind of cooperative system requires that people trust one another, but it is potentially dangerous to trust indiscriminately. Forming ingroups, however, is one way of creating an environment where the risks of non-reciprocation are minimized. By limiting the extension of trust only to mutually acknowledged ingroup members, people are more likely to cooperate with those who will return the favor. In this view, trust is not extended to outgroup members – but neither is any form of fundamental hostility.
2
Rather, it is argued that additional factors, such as a sense of moral superiority (Sidanius,
1993) or perceived intergroup threat (for a review of various potential sources of
perceived threat, see Riek, Mania, & Gaertner, 2006), must be present to elicit negative
attitudes toward the outgroup.
Due to the importance and primacy of the ingroup (Allport, 1954), the process of
group formation is believed to involve a differentiation of the ingroup from the outgroup,
rather than the outgroup from the ingroup (Brewer, 1979). So, if one were to imagine a
baseline situation where an individual sees himself/herself as distinct from an
undifferentiated group of others, creating ingroup and outgroup boundaries would lead to
a realignment of perceptions such that ingroup members would be drawn psychologically
closer to the self (less differentiated from the self than before), but the distance between
the self and outgroup would remain constant. Based on this logic, when groups are
formed, ingroup members are seen as more positive, but outgroup members are viewed
no more negatively than they had been before group labels were introduced (Brewer,
1979; 1999).
Minimal Group Studies
In an effort to understand group bias in its most fundamental form, much research
has explored the link between categorization and prejudice using an experimental tool
known as the Minimal Group Paradigm (MGP; Tajfel, Billig, Bundy, & Flament, 1971;
Tajfel & Billig, 1974). Because the dynamic between most real-world ingroups and outgroups is confounded with a number of factors (e.g., a prior history of conflict or ongoing competition for resources), it is often difficult to determine if existing intergroup
3
prejudice or discrimination is based on categorization per se or some other variable(s).
The MGP was developed as an attempt to examine prejudice and discrimination in a
“pure” setting free of the elements that frequently accompany real groups.
In the most commonly used form of this procedure, participants are brought into
the lab and assigned to one of two groups, ostensibly on the basis of some arbitrary
characteristic. In some studies, for example, participants are led to believe that their
placement in a group is based on a tendency to over- or underestimate the number of dots presented in a visual perception task, or a preference for paintings by the artist Klee to those of Kandinsky (e.g., Tajfel et al., 1971). In actuality, group membership is determined by random assignment. To ensure that the groups created in the lab remain minimal, the researchers also isolate each participant so that he/she has little or no interaction with members of either the ingroup or outgroup. Thus, participants have no pretext other than their group label to form impressions of ingroup and outgroup members.
After being categorized into one of two minimal groups, participants are most commonly asked to rate ingroup and outgroup members on various evaluative dimensions (e.g., Brewer & Silver, 1978) or divide specific resources (e.g., points, money) between ingroup and outgroup members (e.g., Tajfel et al., 1971; Tajfel & Billig,
1974). Across a substantial number of studies using the MGP (Brown, 1995; Tajfel &
Turner, 1986), participants have tended to rate ingroup members more favorably than outgroup members and/or give more resources to ingroup targets than to outgroup targets.
The dramatic implication of these results is that prejudice and discrimination appear to be
4
elicited from group categorization alone (c.f., Diehl, 1990; Gaertner & Insko, 2000;
Rabbie & Horwitz, 1969). By simply being placed into groups, even ones that have no apparent meaningful basis, people seem willing to view or treat certain people differently from others.
Limitations of Prior Research
The findings from the original MGP studies provided the foundation for the development of Tajfel and Turner’s (1986) Social Identity Theory (SIT), which states that people perceive their group as better than the outgroup in an effort to feel good about themselves. In addition, this work helped spur a long-standing interest in the consequences of categorization and the cognitive processes involved in stereotyping and prejudice (Fiske, 1998; Park & Judd, 2005). However, despite the contributions of the initial MGP studies and those that followed, the specific nature of the bias observed under minimal group conditions remains ambiguous. In particular, although a relative preference for the ingroup over the outgroup has been well established, it is unclear whether this bias is driven by ingroup favoritism (Allport, 1954; Brewer, 1979; 1999), outgroup derogation (Sumner, 1906), or some combination of both. This is because the experimental designs used in existing MGP studies make it difficult to definitively determine the direction of this effect. In a typical MGP experiment, participants are asked to rate or distribute resources between an ingroup and outgroup target, but no appropriate control group is included to decompose the relative bias in attitudes or behavior that emerges.
5
Instead researchers have primarily used different measurement approaches to try to infer the nature of the bias. For example, one of the most common dependent measures used in MGP studies is the allocation matrix (Tajfel, et al., 1971). Here, participants are presented with a two-row matrix of values representing resources (e.g., money or points) to be divided between two other individuals in the experiment.
Participants indicate how much they wish to allocate to each target by circling one column of the matrix, and the values are arranged so that choosing a particular column over others represents a specific distribution strategy. Through repeated allocations, general orientations in the way participants award resources to ingroup and outgroup members can be derived. Four orientations are possible (Bourhis, Sachdev, & Gagnon,
1994): 1) fairness/parity – equally distribute resources to ingroup and outgroup members,
2) maximum joint profit (MJP) – maximize resources given to both ingroup and outgroup members, 3) maximum ingroup profit (MIP) – award highest absolute number of resources to ingroup members regardless of what is given to outgroup members, and 4) maximum differentiation (MD) – maximize difference in resources in favor of the ingroup over the outgroup, but at a cost to maximum ingroup profit.
In general, participants tend to choose allocation strategies that favor ingroup members over outgroup members (e.g., Tajfel et al., 1971). However, the matrices are designed in a manner that confounds certain orientations (Bornstein, Crum, Wittenbraker,
Harring, Insko, 1983; Brewer, 1979). Specifically, participants are always asked to choose betweens columns that reflect one orientation and columns that confound two other orientations. For example, the dominant allocation strategy that emerges is one
6
based on choices that confound the MIP and MD orientations (as in Table 1). This makes it difficult to determine whether participants are motivated to favor the ingroup, discriminate against the outgroup, or possibly both.1 Despite this ambiguity, however, the allocation strategy that confounds MIP and MD has confusingly been defined in the literature at times as ingroup favoritism (e.g., Bourhis et al, 1994).
Some researchers have also tested to see if categorization can lead to outgroup derogation by having participants rate group targets on negative dimensions (e.g.,
Mummendey, Otten, Berger, & Kessler, 2000; Otten & Wentura, 2001) or distribute negative outcomes to group members (e.g., Amiot & Bourhis, 2003; Mummendey, et al.,
1992). The thinking here is that if participants feel hostile toward the outgroup, they should rate the outgroup more negatively than the ingroup and allocate more negative outcomes (e.g., punishments) to the outgroup than the ingroup. When such measures are used, however, the bias typically observed in MGP studies disappears or is much less pronounced (Buhl, 1999; Mummendey & Otten, 1998). This apparent reluctance to directly disparage or harm the outgroup has contributed to the belief that categorization per se is insufficient to arouse outgroup prejudice (e.g., Brewer, 1999). It is important to note, however, that even if participants’ assessment or treatment of the outgroup is not negative in an absolute sense, it is still possible for them to hold a bias against the outgroup. That is, the formation of social categories could result not only in a boost in positivity toward the ingroup, but also a decrease in favorability toward the outgroup.
This decrease in outgroup attitudes would indicate a clear bias against the outgroup, even if evaluations of the outgroup were neutral or positive in absolute terms.2
7
In addition to providing evidence of intergroup bias in the MGP on conscious,
explicit measures of attitudes and behavior, some research has also examined attitudes
using implicit measures designed to capture automatic evaluative reactions to groups (see
Petty, Fazio, & Briñol, 2009, for a review). Notably, research using these measures has
shown that a relative preference for the ingroup over the outgroup also exists at a
relatively automatic level (e.g., Ashburn-Nardo et al., 2001; Castelli, et al., 2004, Van
Bavel, Packer, & Cunningham, 2008). Moreover, as was the case with explicit measures,
a number of studies suggest that the bias observed on implicit measures is driven entirely
by ingroup favoritism. For example, Otten and Wentura, (1999) found that participants
made fewer errors in categorizing positive words when they had been subliminally
primed with their ingroup category label than when primed with the outgroup label.
Furthermore, the percentage of errors committed when participants were primed with
neutral words was significantly greater than when the ingroup was primed but was no
different than when the outgroup was primed. These results were interpreted as evidence
of 1) an implicit preference for the ingroup over the outgroup that 2) is driven solely by
ingroup favoritism. That is, implicit positivity is extended to the ingroup, whereas the
outgroup is associated with neutrality. However, although the outgroup might be
implicitly evaluated as neutral in an absolute sense, it is still possible that the inclusion of
a control group could reveal evidence of bias against the outgroup. That is, priming
participants with a control group category label in the Otten and Wentura (1999) procedure could have yielded error rates that were significantly smaller than the outgroup prime error rate (and perhaps even the same as the ingroup prime error rate). This pattern
8
of results would indicate that although the outgroup may be considered objectively
neutral (i.e., similar to a neutral prime), the difference between implicitly measured
ingroup and outgroup attitudes is based in part (or perhaps entirely) on outgroup
prejudice.
Otten and Moskowitz (2000) found evidence of implicit ingroup favoritism using
a paradigm measuring spontaneous trait inference (Uleman, Newman, & Moskowitz,
1996). The authors showed that reading sentences describing behaviors performed by a
minimal ingroup member facilitated the inference of positive, but not negative, traits implied by the sentences. In contrast, reading sentences about behaviors performed by a minimal outgroup member did not facilitate the inference of implied traits, regardless of the valence of the behavior. This pattern of results was interpreted as evidence of implicit positivity toward the ingroup and a lack of implicit negativity toward the outgroup. However, as noted earlier, a bias against the outgroup could still exist, even if the outgroup was not viewed negatively in absolute terms. For instance, reading about positive behaviors performed by a non-affiliated person could facilitate the inference of
positive traits, which would indicate prejudice against the outgroup. Indeed, if the degree
of facilitation of positive traits was the same for control and ingroup targets, then one
would conclude that the implicit intergroup bias was driven solely by outgroup prejudice.
Control Group Studies
In sum, without a control group for comparison, it is difficult to fully discern the direction of the explicit and implicit intergroup bias observed in the minimal group situation. But there a few cases where a control group has been implemented in the
9
MGP. For example, Van Bavel and Cunningham (2009, Experiment 2) assigned
Caucasian participants to minimal groups and measured both their implicit and explicit
evaluations of mixed race ingroup, outgroup, and unaffiliated targets. A preference for
the ingroup over the outgroup emerged on both types of measures. Control group
attitudes were statistically no different from outgroup attitudes, which indicated that this
relative difference was driven by ingroup favoritism alone. However, the focus of this
work was on understanding how currently relevant social categories might influence
automatic race bias, and the salience of race in this study may have produced effects that
obscured what would normally occur in a context in which minimal groups were the only
basis of differentiation. That is, in the absence of an automatic race bias, participants
may have evaluated the group targets differently. For instance, implicit attitudes toward
both the outgroup and control group were characterized by a preference for White vs.
Black targets. In fact, white outgroup members seemed to be rated the most favorable of
all White targets (whereas attitudes toward Black outgroup members appeared close to
neutrality, and unaffiliated Black targets seemed to be evaluated the least favorably of all
Black targets (whereas attitudes toward unaffiliated white targets appeared close to neutrality). If the influence of the race bias could be stripped away, attitudes toward the outgroup might be less positive than attitudes toward the control group, which would indicate outgroup prejudice. Similarly, the lack of a bias against the outgroup on explicit attitude measures could be due to a reluctance to derogate White outgroup members. In either case, it is not clear if the pattern of results observed by Van Bavel et al. (2009)
10
would generalize to a situation in which minimal group membership was not crossed with race identity.
In a more standard minimal group experiment, Valdesolo and DeSteno (2007) had participants rate hypocritical behaviors committed by either an ingroup member, outgroup member, or an unaffiliated individual. Hypocritical acts by outgroup members were perceived as more unfair than those committed by ingroup members, but equally unfair as those performed by unaffiliated targets. These findings again suggest the influence of ingroup favoritism rather than outgroup derogation, but it is possible that the fairness measure was not sensitive to participants’ feelings of prejudice toward the outgroup. Attitude items tapping liking may be more likely to capture the absence of positive affect toward the outgroup (Pettigrew & Meertens, 1995). Also, even if the observed intergroup bias is characterized by ingroup favoritism overall, there may be moderators that could be used to identify any individuals who are showing a bias against the outgroup.
For instance, a number of factors have been shown to amplify or attenuate intergroup bias (see Hewstone et al., 2002, for a review). Previous studies have found, for example, that heightened feelings of attachment with the ingroup tend to be associated with more positive ingroup attitudes and stronger intergroup bias, even under minimal group conditions (e.g., Gagnon & Bourhis, 1996; Mummendey et al., 2000; Perreault &
Bourhis, 1999). These findings are anticipated by SIT (Oakes, Haslam, & Turner, 1994) and have been interpreted as evidence that ingroup identification is related to ingroup favoritism. However, given the ambiguity of the direction of intergroup bias in the MGP,
11
it is unclear if ingroup identification also predicts outgroup prejudice in minimal group contexts, as it can when factors like perceived threat are present (e.g., Branscombe &
Wann, 1994).
Furthermore, although ingroup identification has received considerable attention in the literature, social psychologists have rarely tried to measure participants’ degree of identification with the outgroup. And in those few cases where outgroup identification has been assessed in the MGP, no attempt has been made to directly link this variable with participants’ attitudes. Instead, measures of outgroup identification have been used only as manipulation checks of the minimal group induction (Mummendey, et al., 2000;
Otten & Moskowitz, 2000). However, there is evidence to suggest that outgroup identification could be an important predictor of intergroup bias. In one study, Zhong,
Phillips, Leonardelli, and Galinsky (2008, Study 2) found that participants in a MGP who were induced to focus on their non-affiliation with the outgroup (negational identity), rather than their affiliation with the ingroup (affirmational identity), reported more negative attitudes toward the outgroup than the ingroup on negative trait ratings items.
Findings from follow-up studies suggested that when a negational identity is emphasized, people tend to contrast the self away from the outgroup, which leads to outgroup derogation. That is, focusing on a negational identity may elicit feelings of detachment from the outgroup that, in turn, foster prejudice against the outgroup. However, forcing participants to focus on a particularly type of identity (negational vs. affirmational) may have led to effects based on demand. Consequently, it would be useful to see if outgroup prejudice can be obtained when identity is not made salient. In addition, it would be
12
useful to measure outgroup identification to see if there are some people who naturally
hold a negational mindset. If so, perhaps those individuals are more are more likely to
exhibit outgroup prejudice. Finally, given that ingroup identification can predict
outgroup derogation in at least some situations (such as when threat is present, e.g.,
Branscombe & Wann, 1994), it would also be interesting to see if outgroup identification
is related in any way to ingroup favoritism.
Current Research
Although prior MGP studies have been able to demonstrate that categorization can lead to a relative difference in the way individuals view and treat ingroup and outgroup members on explicit and implicit measures, it is unclear whether this bias is the result of ingroup favoritism, outgroup derogation, or a joint impact of both. The methods of measuring intergroup prejudice and discrimination employed in previous studies seem unable to definitely determine the direction of this effect. A critical factor missing from
most prior MGP studies is the inclusion of an appropriate control group to compare to the
ingroup and outgroup targets. And as noted, previous studies that did include a control
group also contained other features that rendered the results ambiguous. Having
participants evaluate a group of unaffiliated individuals would provide a baseline for
determining the impact of ingroup and outgroup labels and allow for a concrete test of
whether preference for ingroup over outgroup members represents ingroup favoritism,
outgroup derogation, or both.
If participants evaluated the control group no differently than the outgroup, then
the relative ingroup-outgroup bias would be interpreted as ingroup favoritism. In
13
contrast, if attitudes toward the control group were equivalent to evaluations of the
ingroup, then one would conclude that the bias was based on outgroup derogation.
Finally, if the control group was evaluated as less positive than the ingroup but more
positive than the outgroup, then it would be apparent that that bias was bi-directionally driven by both ingroup favoritism and outgroup derogation. The use of a control group in this manner has been very helpful in gaining a clearer understanding of other phenomena, such as the impact of similarity/dissimilarity on attraction/repulsion effects (e.g.,
Rosenbaum, 1986; Singh & Tan, 1992) and whether being physically attractive or unattractive has a greater influence on how one is perceived and treated by others (Griffin
& Langlois, 2006).
The main purpose of the current research, therefore, was to replicate the minimal group situation and include an appropriate control group to test whether the observed intergroup bias is due to ingroup favoritism, outgroup derogation, or a combination of both. In this work, ingroup favoritism was conceptualized as greater positivity toward the ingroup relative to the control group, and outgroup derogation was defined as decreased positivity toward the outgroup compared to the control group.3 The minimal
group information provided to participants was the only group information available and
was not combined with any other details about the group. In Experiment 1, participants
were assigned to hypothetical minimal groups using a procedure developed by Pinter and
Greenwald (2004). Following the group assignment, participants were asked to report
their attitudes toward the ingroup, outgroup, and a group of individuals who were not
affiliated with either group (control). In Experiment 2, participants were assigned to
14
ostensibly real minimal groups and again instructed to provide their evaluations of the three target groups.
In both of these experiments, we also tested to see if participants’ degree of identification with the ingroup and outgroup might impact intergroup bias in the MGP.
Prior research suggests that greater ingroup identification leads to more positive ingroup attitudes and larger intergroup bias (e.g., Gagnon & Bourhis, 1996; Mummendey et al.,
2000; Perreault & Bourhis, 1999), and we hypothesized that higher levels of ingroup identification would be associated with increased ingroup favoritism in Experiments 1 and 2. We also tested to see if those higher in ingroup identification were more likely to express outgroup derogation. In addition, although previous research has not directly explored the relationship between outgroup identification and group attitudes in the
MGP, research by Zhong et al. (2008) suggests that making negational identities (i.e., group non-memberships) salient can lead to disidentification with the outgroup, and subsequently, outgroup derogation. But it is not clear if some people spontaneously disidentify with the outgroup and whether such individuals are more likely to show a bias against the outgroup in a minimal group context. Therefore, we measured participants’ level of outgroup identification to see if those who felt less identified with the outgroup would exhibit greater outgroup derogation. Finally, because so little is known about outgroup identification in the context of minimal groups, we also examined whether this variable was distinct from ingroup identification and if it was related in any way to ingroup favoritism.
15
In addition to showing that people prefer their ingroup to the outgroup on explicit measures of prejudice and discrimination, previous research has also shown that being placed into minimal groups can elicit intergroup bias on implicit measures (e.g.,
Ashburn-Nardo, Voils, & Monteith, 2001; Castelli, Zogmaister, Smith, & Arcuri, 2004;
Otten & Moskowitz, 2000; Otten & Wentura, 1999). The results of these studies suggest that preference for the ingroup over the outgroup occurs at a relatively automatic level.
Therefore, to provide a full test of the direction of intergroup bias in the MGP, in
Experiment 3 we measured implicit attitudes toward the ingroup, outgroup, and a control group to determine whether implicit intergroup bias is based on ingroup favoritism, outgroup derogation, or both and to see if that implicit bias would correspond to any explicit bias observed in the first two experiments. For example, perhaps people will express prejudice when they consciously think about the outgroup, but only favor the ingroup on an automatic level. Or perhaps people will correct for outgroup bias and appear non-prejudiced in their self-reported attitudes (Wegener & Petty, 1997; Wilson &
Brekke, 1994) but still exhibit prejudice in their implicit evaluations, where the ability to correct is limited. Or maybe both explicit and implicit measures will reveal the same type of bias. In any case, by including a control group in the MGP and measuring both explicit (Experiments 1 and 2) and implicit attitudes (Experiment 3), we hoped to gain a richer and more complete picture of minimal group bias than what is currently offered in the literature.
16
Taken together, the aim of these studies was to address the critical limitations of prior MGP studies and provide a clearer understanding of the nature and antecedents of prejudice arising from mere categorization.
17
CHAPTER 2: EXPERIMENT 1
In order to determine the direction of intergroup bias elicited by social
categorization, participants were assigned to minimal groups and asked to report their
attitudes toward not only the ingroup and outgroup, as is typically the case in MGP
studies (e.g., Brewer & Silver, 1978), but a non-affiliated control group as well.
Participants were randomly assigned to one of two hypothetical groups using a procedure from Pinter and Greenwald (2004), which has been shown to be an effective deception-
free minimal group induction that elicits the classic relative preference for the ingroup
over the outgroup (Pinter & Greenwald, 2010). After imagining themselves as a member
of either the Green group or the Blue group, participants reported their attitudes toward their ingroup, the outgroup, and a collection of individuals who were described as not belonging to any group. One of the challenges in developing an appropriate control to compare to the ingroup and outgroup involves defining a target that is not merely a third minimal group. In fact, intergroup bias seems to disappear in the context of a three- minimal group structure, presumably because an “us vs. them” distinction is less salient
(Hartstone & Augoustinos, 1995). Therefore, in Experiment 1 we attempted to introduce a control target that closely resembled the pre-categorization, unaffiliated individuals described by Brewer (1979). We then compared ratings toward this baseline with ratings
18
toward the ingroup and outgroup to determine if the relative preference in favor of the ingroup is based on ingroup favoritism, outgroup derogation, or a joint influence of both.
In addition, we sought to identify potential moderators of ingroup favoritism and outgroup derogation by measuring participants’ feelings of identification with the ingroup and outgroup. Prior work suggests that ingroup identification is positively related to ingroup attitudes and a stronger intergroup bias (e.g., Gagnon & Bourhis, 1996;
Mummendey et al., 2000; Perreault & Bourhis, 1999). Degree of identification with the outgroup is a construct that has not been directly linked with group attitudes in the MGP but plausibly is an antecedent to outgroup prejudice, given prior research on negational identities (Zhong et al., 2008). Thus, we hypothesized that higher levels of ingroup identification would be associated with greater ingroup favoritism (Oakes et al. 1994) and that lower levels of outgroup of identification would be associated with increased outgroup derogation.
Methods
Participants and Design
A sample of 72 introductory psychology students at Ohio State University completed the experiment in partial fulfillment of a course requirement. Participants were randomly assigned to a 2 (Group label: Green vs. Blue) X 3 (Target: Green vs. Blue vs. No Group) mixed design.
Procedure
All materials were presented with MediaLab software (Jarvis, 2006). Participants were led to the laboratory and randomly assigned to one of several personal computers
19
visually isolated by partitions. Then, screen instructions guided participants through the minimal group manipulation, where participants were assigned to either the Green or
Blue group and given information about a no group control. Next, participants reported their attitudes toward each of these three targets. Following this, all participants indicated their level of identification with the Green and Blue groups. Finally, the participants were debriefed and released.
Independent Variables
Group label. To manipulate group membership, participants were assigned to minimal groups using an adapted version of a procedure developed by Pinter and
Greenwald (2004). First, participants were asked to imagine that they were members of either the Green Team or the Blue Team (determined by random assignment), who were competing in a campus scavenger hunt. After imagining themselves as members of a hypothetical group, participants were shown the names of five fellow group members and given 45 seconds to memorize them. To strengthen group associations, participants then completed a self-group association task, where they were shown, one at a time in random order, the five names of their ingroup, fives names of the outgroup, as well as words related to self (i.e., I, me, myself) and words related to other (i.e., other, them, they).
Across two blocks of 48 trials, participants were told to categorize each word that appeared using one of two response keys on the keyboard. One key (‘F’) corresponded to
“self” words and names of ingroup members, while the other key (‘J’) corresponded to
“other” words and the names of outgroup members. The response keys were reversed in
20
the second block of trials. Participants were instructed to respond quickly while trying to
avoid mistakes. Response errors required correction for the program to continue.
Target group. Following the self-group association task, participants were presented with a list of names described through screen instructions as “students who are not on either the Green or Blue Team. They represent individuals who are not competing in the scavenger hunt.” This information was intended to create the perception of a hypothetical, non-affiliated control target. After receiving these statements, participants reported their attitudes toward the Green group, the Blue group, and those not in any group.
Dependent Variables
Attitudes. To assess their evaluations of each group, participants indicated on a 1 to 7 scale the extent they agreed with the following items (adapted from Pinter &
Greenwald, 2004), presented in random order: “I like the ____ group” and “I think the
____ group is good.” Participants completed each pair of items for the Green group (α =
.55), the Blue group (α = .86), and the control target (α = .63), which was represented as
“those not in a group.” The two items were then averaged together to create separate
attitude indices for each target group. Participants’ assessments of the Green group and
Blue group were then recoded (based on their group assignment) as evaluations of their
ingroup and outgroup.
Identification. To assess their level of identification with the Green and Blue
groups, participants indicated on a 1 to 7 scale the extent they agreed with the following
items (adapted from Pinter & Greenwald, 2004), presented in random order: “I feel
21
attached to the ____ group” and “I identify with ____ group.” Participants completed each pair of items for both the Green group (α = .69) and the Blue group (α = .63). These items were then averaged to create separate indices of identification for each group.
Participants’ degree of identification with the Green and Blue groups were then recoded
(based on their group assignment) to reflect their level of identification with their ingroup and outgroup.
Results
Attitudes. Participants’ evaluation of the ingroup, outgroup and control group were submitted to a within-subjects repeated measures ANOVA, which revealed a significant effect of target group, F(2, 142) = 21.13, p < .001 (see Figure1). Separate repeated measures ANOVAS were then conducted to compare each pair of groups. In line with prior MGP studies, participants rated the ingroup (M = 4.76, SD = 1.29) more positively than the outgroup (M = 3.60, SD = 1.07), F(1, 71) = 28.13, p < .001. In addition, participants evaluated the ingroup more favorably than the control group, (M =
3.87, SD = 1.08) F(1, 71) = 22.86, p < .001, partial η2 = .24, which indicates that the relative difference in attitudes toward the ingroup and outgroup is based on a preference for the ingroup. However, participants also rated the outgroup as marginally less favorable than the control group, F(1, 71) = 3.17, p = .08, partial η2 = .04, which suggests that the ingroup-outgroup discrepancy is partly driven by outgroup derogation, as well.
Identification. Participants’ assessments of their level of identification with the ingroup and outgroup (which were uncorrelated, r = -.03) were first submitted to a within-subjects repeated measures ANOVA. As expected, participants felt more
22
identified with their ingroup (M = 4.53, SD = 1.47) than the outgroup (M = 3.01, SD =
1.16), F(1, 71) = 45.88, p < .001 (e.g., Mummendey et al., 2000; Pinter & Greenwald,
2004). Next, we examined the relationship between the degree of group identification and group attitudes. Replicating prior research (e.g., Mummendey et al., 2000), ingroup identification was positively correlated with ingroup attitudes (r = .61, p < .001) but was not associated with outgroup attitudes (r = -.19). In addition, outgroup identification was correlated with outgroup attitudes (r = .49, p < .001) but not ingroup attitudes (r = -.15).
Moderation analyses. Next, we investigated the potential moderating role of ingroup and outgroup identification on each component of the observed relative intergroup bias (i.e., ingroup favoritism, outgroup derogation). First, attitudes toward the control target were subtracted from attitudes toward the ingroup to create an index of ingroup favoritism where larger values represent greater ingroup positivity. We then conducted a regression analysis with ingroup identification, outgroup identification, and their interaction term predicting ingroup favoritism. In line with SIT (e.g., Oakes et al.,
1994), a significant main effect of ingroup identification emerged, ß = .61, t(68) = 5.37, p
< .001, where those who felt more identified with their ingroup expressed greater ingroup favoritism. Neither outgroup identification nor the ingroup X outgroup identification interaction significantly predicted ingroup favoritism, ps ≥ .07.
Next, we created an index of outgroup derogation by subtracting outgroup attitudes from control attitudes (larger values equal greater bias against the outgroup).
We then conducted a regression analysis with ingroup identification, outgroup identification, and their interaction term predicting outgroup derogation, which revealed a
23
significant effect of outgroup identification, ß = -.30, t(68) = -2.63, p = .01. Thus, those
who were less identified with the outgroup expressed greater outgroup derogation.
Neither ingroup identification nor the interaction of ingroup and outgroup identification
had a significant impact on outgroup derogation, ps ≥ .50. (See Appendix A for
alternative moderation analyses).
Discussion
Previous studies exploring the effects of social categorization indicate that the creation of group boundaries elicits a fundamental bias in favor of one’s ingroup (over an outgroup), even if the basis for categorization is arbitrary (e.g., Tajfel et al., 1971).
However, although a relative preference for the ingroup has been well-documented, prior research has been unable to definitively determine whether the bias is fueled by ingroup favoritism, outgroup derogation, or both. Limitations in the existing MGP work simply do not permit a clear test of this question. Therefore, in Experiment 1, we attempted to clarify the direction of the bias by including a control group in the minimal group situation. By comparing participants’ attitudes toward a baseline group of unaffiliated individuals to their evaluations of the ingroup and outgroup, we were able to decompose the intergroup bias into its component parts.
Specifically, Experiment 1 found that the control group was rated as less positive than the ingroup, but there was a tendency for unaffiliated individuals to be rated more positively than those in the outgroup. The latter finding suggests that the difference between ingroup and outgroup attitudes was driven in part by ingroup favoritism and outgroup derogation. The prevailing view among social psychologists has been that
24
categorization elicits ingroup positivity but not necessarily outgroup derogation (Brewer,
1979; 1999; Hewstone et al., 2002). In support of this position, our findings indicate that the observed intergroup bias is primarily based on ingroup favoritism (partial η2 = .24).
However, the data from Experiment 1 suggest that outgroup derogation (partial η2 = .04) also seems to contribute to the relative difference in ingroup and outgroup attitudes. If replicable, this finding would indicate that outgroup derogation, like ingroup favoritism, could be a basic psychological phenomenon (Sumner, 1906).
In addition to providing a clearer picture of what drives intergroup bias in the
MGP, we were also able to identify important predictors of ingroup favoritism and outgroup derogation. Replicating prior research, we showed that increased identification was associated with more positive ingroup attitudes and greater ingroup favoritism.
More importantly, we showed that decreased identification with the outgroup was related to less favorable outgroup attitudes and greater outgroup derogation. That is, those who disidentifed with the outgroup were the most likely to show a bias against the outgroup.
Taken together, these findings suggest that intergroup bias is likely to be greatest for those who are simultaneously highly identified with their ingroup and highly disidentified with the outgroup.
It is important to note that the consequences of outgroup identification (which appears to be distinct from ingroup identification, r = -.03) on intergroup attitudes and behavior are often not considered by researchers (e.g., Mummendey et al., 2000; Otten &
Moskowitz, 2000), but the results of Experiment 1 suggest that this variable is a potentially important predictor of outgroup derogation. In particular, this variable may be
25
useful in identifying a bias against the outgroup where none seems to be. As noted
earlier, most MGP studies lack an appropriate control target for comparison. But in the
few cases when a control target is present, the observed intergroup bias often appears to
be driven solely by ingroup favoritism (e.g., Valdesolo & DeSteno, 2007). However,
measuring outgroup identification could help researchers identify those individuals who
do feel prejudiced against the outgroup, despite the arbitrary nature of the group
boundaries.
In sum, the results of Experiment 1 suggest that researchers could detect biases
against the outgroup in a minimal group situation if they were to include a control target
to decompose the relative intergroup bias and measure factors to help identify those most
likely to exhibit outgroup derogation. However, before considering these results to be definitive, several cautions should be noted. First, the minimal groups used in this study were hypothetical. Participants only imagined that they were members of the Green or
Blue group; they were not actually assigned to any group. As a result, it is possible that
their attitudes were based on how they thought they might feel toward each of the target
groups, but not necessarily how they would actually feel if they were made part of a real
group. Prior research has shown that the bias displayed by participants in the MGP can
be influenced by salient norms and scripts (e.g., Gaertner & Insko, 2001; Hertel & Kerr,
2001). Thus, participants might have imagined they would hold a bias against the
outgroup (an effect that was only marginal), but if they had actually been assigned to
minimal groups, perhaps only a norm to favor the ingroup would be activated. To
26
address this issue, participants in Experiment 2 were assigned to non-imaginary minimal groups.
Second, the groups participants were asked to hypothetically join were described as two different “teams” competing against each other in a campus scavenger hunt.
Previous studies have shown that certain factors, such as competition for resources, can lead to outgroup prejudice (e.g., Sherif et al., 1961; Worchel, Andreoli, & Folger, 1977).
Therefore, the marginal bias against the outgroup revealed by the implementation of a control target in Experiment 1 could be the result of the competition element described in our minimal group procedure. In the absence of perceived competition, perhaps attitudes toward the outgroup would have been no different than attitudes toward the control target, suggesting only the presence of ingroup favoritism. To explore this possibility, in
Experiment 2 we removed all traces of competition from our MGP. This change allowed us to see if a significant outgroup derogation effect emerged in a more pure, competition- free intergroup setting.
Third, in Experiment 1, we introduced a control group into a minimal group situation that was described as individuals who were not members of any group. The goal was to provide a baseline target that represented unaffiliated individuals (Brewer,
1979), rather than simply an additional minimal group. As noted earlier, prior work has shown that when more than two minimal groups are manipulated, intergroup bias disappears (Hartstone & Augoustinos, 1995). The fact that we were able to replicate the classic minimal group effect (i.e., relative bias in favor of the ingroup) suggests that the control group used in Experiment 1 was not perceived by participants as a third minimal
27
group. Nevertheless, we wanted to further demonstrate that the control target of
unaffiliated individuals serves as an effective baseline target to compare to the ingroup
and outgroup.
In Experiment 2, we first asked participants to rate each of the three group
categories (i.e., ingroup, outgroup, and control) and then report their attitudes toward
individual members of each group. Mixed in among the (randomly presented) names of
the target group members were the names of brand new individuals participants had not
seen before and who had not been linked with any category label. Ratings of these new
names provided a manipulation check of the effectiveness of our control group. Thus, if
our control group represents a true baseline, attitudes toward the unaffiliated individuals
were predicted to be no different than attitudes toward the brand new individuals. It was
expected that implementing these new procedures would allow us to replicate and extend
the findings of Experiment 1 and rule out potential alternative explanations. In addition, we included measures of ingroup and outgroup identity to replicate the moderating influence of these variables.
28
CHAPTER 3: EXPERIMENT 2
Participants in Experiment 2 followed a procedure that was very similar to the one
used in Experiment 1, but with several critical changes. First, we dropped the self-other
categorization task and gave participants the opportunity to memorize the names of all
three target groups. This was done to ensure that participants’ had equal exposure to each
of the groups. Second, participants were assigned to actual (rather than hypothetical)
minimal groups to see if the outgroup bias observed in Experiment 1 would still emerge
if participants believed they were a member of a real group. Third, the minimal group
procedure was altered to eliminate the possibility that perceived intergroup competition
could account for any findings indicating the presence of outgroup derogation. Finally,
in addition to rating each group category, participants reported their attitudes toward
individual members of each group and toward a set of brand new individuals. This
change allowed us to test the effectiveness of our control target and see if the findings of
Experiment 1 could be extended to the exemplar level. Specifically, we hypothesized
that the inclusion of a control target in our MGP would reveal the presence of both
ingroup favoritism and outgroup derogation on both category- and exemplar-level attitudes, and that the ingroup and outgroup attitudes would be influenced by ingroup and outgroup identification, respectively.
29
Method
Participants and Design
A sample of 121 introductory psychology students at Ohio State University completed the experiment in partial fulfillment of a course requirement. Participants were randomly assigned to a 2 (Group label: Green vs. Blue) X 3 (Target: Green vs. Blue vs. No Group) mixed design.
Procedure
All materials were presented with MediaLab software (Jarvis, 2006). After being led to the laboratory, participants completed a procedure similar to that used in
Experiment 1. First, participants were assigned to either the Green or Blue group and given information about a no group control. Next, participants reported their attitudes toward the three target groups, as well as the individual members of each group.
Participants then indicated their level of identification with the Green and Blue groups before being debriefed and released.
Independent Variables
Group label. Upon entering the laboratory, participants were informed that they would be randomly assigned to one of two groups: the Green group or the Blue group.
To determine which group they would join, participants were instructed to select one of two boxes on the computer screen. Each box was said to represent one of the groups, but they were unmarked so that participants did not know which box belonged to which group. After making their pick, participants were told they were now a member of the
Green or Blue group. (Membership was determined by random assignment, regardless of
30
which box was selected). Next, participants were presented with the names of five
ingroup members and given 45 seconds to memorize them (for the ostensible purpose of
recalling them later in the experiment). This memorization task was repeated for the
names of five outgroup members and five individuals described as not being affiliated
with any group (just as in Experiment 1).
Target. As in Experiment 1, participants reported their evaluations of three
different targets: the Green group, the Blue group, and those not in any group.
Dependent Variables
Attitudes toward group categories. To assess their evaluations of each group at
the category level, participants completed the same attitude measures used in Experiment
1. Responses to the Green group (α = .95), Blue group (α = .95), and control group (α =
.71) items were highly consistent, and thus, averaged together to create separate attitude indices for each group. Overall attitudes toward the Green group and Blue group were then recoded as evaluations of the ingroup and outgroup, as was done in Experiment 1.
Attitudes toward group members. Participants also reported their attitudes toward each individual member of the Green group, the Blue group, and those not in a group
(presented in random order) on a seven-point scale: (-3) dislike to (+3) like. To examine whether “those not in a group” was an effective control, participants also rated five new individuals whose names had not been presented earlier. Ratings of the members of the
Green group (α = .89), Blue group (α = .90), control group (α = .83), and New group (α =
.87) were highly consistent within each category, and thus, averaged to create separate
31
attitude indices for each target group. Finally, the Green and Blue group indices were
recoded to reflect participants’ evaluations of their ingroup and outgroup members.
Identification. To assess their level of identification with the Green and Blue
groups, participants completed the same pair of identification items used in Experiment 1.
Responses to the Green (α = .91) and Blue group (α = .88) items were highly consistent, and thus, averaged to create separate indices of identification for each group. These identification scores were then recoded to reflect identification with their ingroup and outgroup, as was done in Experiment 1.
Results
Attitudes toward group categories. Participants’ evaluation of the ingroup, outgroup and control group categories were first submitted to a within-subjects repeated measures ANOVA, which revealed a significant effect of target, F(2, 240) = 84.18, p <
.001 (see Figure 2). Follow-up repeated measures ANOVAS were then conducted to compare each pair of targets. As in Experiment 1, participants rated the ingroup (M =
5.64, SD = 1.24) more positively than both the outgroup (M = 3.79, SD = 1.41), F(1, 120)
= 101.68, p < .001, and the control group, (M = 4.16, SD = 1.18), F(1, 120) = 96.82, p <
.001, partial η2 = .45. In addition, participants reported less favorable attitudes toward the
outgroup than the control group, F(1, 120) = 11.10, p = .001, partial η2 = .09. Together,
these findings replicate the results of Experiment 1 indicating that the classic minimal
group effect is jointly driven by ingroup favoritism and outgroup derogation.
Attitudes toward group members. First, to determine whether the control target
used in this study provided an effective baseline, we submitted participants’ attitudes
32
toward the unaffiliated persons and the new individuals to a within-subjects repeated
measures ANOVA. No significant difference between the two targets was found F < 1,
which suggests that our control target does in fact represent a baseline target, rather than
simply a third minimal group. And given that the control and new name indices were not
statistically different, we collapsed them into a single control target to compare to the
ingroup and outgroup.
Entering participants’ attitudes toward members of the ingroup, outgroup, and
combined control group into a within-subjects ANOVA revealed a significant effect of
target, F(2, 240) = 42.96, p < .001 (see Figure 3). Separate repeated measures were then
performed to examine differences between each pair of targets. Replicating the classic
minimal group effect, participants rated ingroup members (M = 1.01, SD = 1.19) more favorably than outgroup members (M = -.15, SD = 1.22), F(1, 120) = 46.81, p < .001. In support of our view that this discrepancy is driven by both a preference for the ingroup and a prejudice against the outgroup, control individuals (M = .14, SD = .91) were evaluated more less positively than ingroup members, F(1, 120) = 48.17, p < .001, partial
η2 = .29, and more positively than outgroup members, F(1, 120) = 13.32, p < .001, partial
η2 = .10.
Identification. A within-subjects repeated measures ANOVA was conducted to
compare participants’ feelings identification with the ingroup and outgroup (which were
weakly correlated, r = -.29, p = .001). As in Experiment 1, participants felt more
identified with their ingroup (M = 5.34, SD = 1.36) than the outgroup (M = 3.10, SD =
1.38), F(1, 120) = 125.03, p < .001. In addition, ingroup identification was strongly
33
correlated with ingroup category (r = .70, p < .001) and ingroup exemplar attitudes (r =
.43, p < .001) and weakly correlated with outgroup category attitudes (r = .20, p = .03).
Outgroup identification was positively associated with outgroup category (r = .58, p <
.001) and outgroup exemplar attitudes (r = .41, p < .001) and weakly negatively
correlated with ingroup category (r = -.23, p = .01) and ingroup exemplar (r = -.28, p =
.002) attitudes.
Moderation analyses. We first examined whether ingroup and outgroup
identification would impact ingroup favoritism and outgroup derogation at the category
level. As in Experiment 1, control category attitudes were subtracted from ingroup
category attitudes to create an index of ingroup favoritism at the category level. A
regression analysis was then performed with ingroup identification, outgroup
identification, and their interaction term as predictors of ingroup favoritism. This
analysis revealed a significant effect of both ingroup identification, ß = .46, t(117) = 5.75, p < .001 and outgroup identification, ß = -.23, t(117) = -2.57, p = .01, such that higher levels of ingroup identification and lower levels of outgroup identification were associated with greater ingroup favoritism. The interaction term was not significant, p =
.82.
Next, an index of outgroup derogation at the category level was created by
subtracting outgroup category attitudes from control category attitudes, as in the previous
study. Entering outgroup derogation into a regression analysis with ingroup and
outgroup identification and their interaction term as predictors yielded a marginal effect
of ingroup identification, ß = .17, t(117) = 1.92, p = .06, and a significant effect of
34
outgroup identification, ß = -.26, t(117) = -2.67, p = .009. These findings indicate that
increased ingroup identification and decreased outgroup identification were both
associated with greater outgroup derogation. A significant ingroup X outgroup
identification interaction also emerged, ß = -.23, t(117) = -2.50, p = .01, which was
interpreted using procedures recommended by Aiken and West (1991). Decomposing the
interaction one standard deviation above and below the mean of the ingroup identification
index revealed a significant effect of outgroup identification on outgroup derogation for
those high in ingroup identification, ß = -.49, t(3117) = -5.01, p < .001, but not for those
low in ingroup identification, ß = -.03, t(117) = -.17, p = .86.
We then examined the influence of ingroup and outgroup identification on
ingroup favoritism and outgroup derogation at the exemplar level. First, control
exemplar attitudes were subtracted from ingroup exemplar attitudes to create an index of
ingroup favoritism. Then a regression analysis with ingroup identification, outgroup
identification, and their interaction term as predictors of ingroup favoritism was
conducted, which revealed significant main effects of both ingroup, ß = .21, t(117) =
2.41, p = .02, and outgroup identification, ß = -.27, t(117) = -2.77, p = .006, but no
significant interaction, p = .22. That is, as with the category level attitudes, stronger
identification with the ingroup and weaker identification with the outgroup predicted
enhanced ingroup favoritism. Next, we created an index of exemplar-level outgroup derogation by subtracting outgroup exemplar attitudes from control exemplar attitudes.
Then a regression analysis was performed with ingroup and outgroup identification and their interaction term predicting outgroup derogation. This analysis revealed a significant
35
effect of ingroup identification, ß = .26, t(117) = 2.99, p = .003, a marginal effect of outgroup identification, ß = -.17, t(117) = -1.70, p = .09, and a significant ingroup X outgroup identification interaction, ß = -.23, t(117) = -2.42, p < .02, which was decomposed with the same simple slopes analysis described previously (Aiken & West,
1991). The two main effects replicated the findings for the category level measure, as did the interaction. That is, as with the category level attitudes, the results of these analyses indicated that the expression outgroup derogation was influenced by outgroup identification, but only when ingroup identification was high, ß = -.39, t(117) = -3.98, p <
.001, and not when it was low, ß = .06, t(117) = .36, p = .72. (See Appendix A for alternative moderation analyses).
Discussion
As in Experiment 1, participants reported more favorable attitudes toward the ingroup than the outgroup, and the inclusion of a control group for comparison revealed that this difference was based on a bi-directional influence of ingroup favoritism and outgroup derogation. This pattern of results emerged not only for category-level attitudes, but also for evaluations of ingroup and outgroup exemplars. As in Experiment
1, the difference between ingroup and outgroup attitudes was primarily driven by ingroup favoritism, but there was also an effect for outgroup derogation, though it was smaller in magnitude. This pattern of results is consistent with prior work arguing that a preference for the ingroup is primary (Brewer, 1979; 1999), but it also suggests that minimal group categorization per se can elicit bias against the outgroup.
36
In addition, participants’ feelings of identification with the ingroup and outgroup influenced their expression of ingroup favoritism and outgroup derogation. As in
Experiment 1, greater ingroup favoritism (at both the category and exemplar level) was observed for those who reported higher levels of ingroup identification. But in contrast to the previous study, ingroup favoritism (at both the category and exemplar level) was also stronger among those with lower levels of outgroup identification. Replicating the results of Experiment 1, participants who felt less identified with the outgroup also exhibited greater outgroup derogation in their assessments of both the group categories and group exemplars (although the latter effect was marginal), but this was only the case when ingroup identification was high. This finding indicates that those who disidentify with the outgroup are the most likely to show a bias against the outgroup and that this is particularly true for those who feel strongly attached to their ingroup. Taken together, these results highlight the usefulness of measuring ingroup and outgroup identification in providing a more complete understanding of ingroup favoritism and outgroup derogation.
Importantly, we found evidence of outgroup derogation with actual (vs. hypothetical) minimal groups and in the absence of intergroup competition. This indicates that the outgroup bias we observed is likely not based on participants’ lay theories of how they might feel in an intergroup situation or a sense of threat aroused by perceived competition (e.g., Sherif et al., 1961; Worchel et al., 1977). Moreover, participants rated new individuals they had never seen before and control group exemplars in a similar fashion, which provides evidence that our control group is being perceived as a non-affiliated target, rather than simply a third minimal group.
37
Together, the findings from the first two experiment provide more compelling
support for the position that mere categorization is sufficient to elicit bias against the
outgroup. However, a number of studies have shown that a relative preference for the
ingroup over the outgroup can be observed at a relatively automatic level (e.g., Ashburn-
Nardo et al., 2001; Castelli, et al., 2004, Van Bavel, Packer, & Cunningham, 2008), and
recent work investigating implicit minimal group bias suggests that this effect may be
driven entirely by ingroup favoritism (e.g., Otten & Moskowitz, 2000; Otten & Wentura,
1999). The discrepancy between the results of Experiments 1 and 2, which show
evidence of outgroup derogation at an explicit level, and studies suggesting only the
influence of ingroup favoritism at an implicit level could be explained by several
different possibilities. First, perhaps it is possible to construct reasons to be biased
against the outgroup when thinking about the group targets at a conscious level, but on a
relatively automatic level, only favor the ingroup. Or maybe the outgroup derogation
documented in our first two experiments was due to demand effects that would be less
likely to influence responses on an implicit attitude measure. Indeed, prior research has
shown that the participants in MGP studies can be influenced by experimenter demand
(Berkowitz, 1994). Perhaps the inclusion of the control group in our minimal group
induction led people to infer that we were anticipating them to exhibit outgroup
derogation, and they responded accordingly in their self-reported attitudes.
A third possibility is that a bias against the outgroup exists at both the explicit and implicit level and that limitations in prior research have made it difficult to observe outgroup derogation on implicit measures. As described earlier, MGP studies using
38
implicit measures contain many of the same weaknesses as those employing explicit
measures. For example, many studies lack a control target to disambiguate the observed
relative preference for the ingroup over the outgroup (e.g., Ashburn-Nardo et al., 2001;
Castelli, et al., 2004). In other cases, researchers have concluded that the difference
between implicit ingroup and outgroup attitudes is based only on ingroup favoritism
because participants’ evaluations of the outgroup were objectively neutral (e.g., Otten &
Moskowitz, 2000; Otten & Wentura, 1999). That is, because ingroup attitudes were
positive and outgroup attitudes were not negative, the implicit intergroup bias that
emerges in the MGP has been assumed to be driven solely by liking for the ingroup.
However, as we explained before, the inclusion of a control group could reveal the
presence of bias against the outgroup, even if the outgroup is not evaluated negatively in
absolute terms. If implicit attitudes toward the outgroup were less favorable than
attitudes toward the control group, that would suggest that the relative intergroup bias is
based at least in part on outgroup derogation.
Therefore, to provide a full test of the direction of the minimal group bias effect and see if a bias against the outgroup also emerges at a relatively automatic level, in
Experiment 3 we assigned participants to minimal groups and measured their evaluations of the ingroup, outgroup, and control group using an implicit attitude measure. Evidence of outgroup derogation on an implicit measure will also help rule out the possibility that the results of the previous experiments were due to experimenter demand.
39
CHAPTER 4: EXPERIMENT 3
In Experiment 3, participants were assigned to minimal groups with the same procedure used in Experiment 2. But this time, we measured participants’ evaluations toward each of the three target groups with an implicit attitude measure (i.e., Brief IAT,
Sriram & Greenwald, 2009). Capturing participants’ (relatively) automatic evaluations of a control group in addition to their evaluations of the ingroup and outgroup was expected provide a more clear test of the nature of implicit intergroup bias in a minimal group situation. Given the pattern of results observed on explicit attitude measures in
Experiments 1 and 2, we predicted that the results of Experiment 3 would show that the implicit bias in favor of the ingroup would be also be driven by both ingroup favoritism and outgroup derogation. In addition, because the implicit attitude measure requires speeded responses that make conscious deliberation more difficult, the evaluations participants provided in this experiment were less likely to be influenced by demand.4
Methods
Participants and Design
A sample of 106 introductory psychology students at Ohio State University completed the experiment in partial fulfillment of a course requirement. Participants
40
were randomly assigned to a 2 (Group label: Green vs. Blue) X 3 (Target: Green vs. Blue vs. No Group) mixed design.
Procedure
All materials were presented with MediaLab and DirectRT software (Jarvis,
2006), and participants followed a procedure similar to those used in the first two experiments. As before, participants were randomly assigned to either the Green or Blue group and given information about a no group control. But in this experiment participants’ evaluations toward the three target groups were assessed with an implicit rather than explicit attitude measure. Following these procedures, participants were debriefed and released.
Independent Variables
Group label. Participants were randomly assigned to either the Green or Blue group using the same minimal group procedure employed in Experiment 2.
Target. As in the previous two experiments, participants were presented with three different targets: the Green group, the Blue group, and those not in any group.
Because Experiment 2 showed that attitudes toward unfamiliar names were the same as those not assigned to any group, this control condition was dropped.
Dependent Variable
Implicit attitudes. After being assigned to minimal groups, participants completed an adapted version of the Brief IAT (Sriram & Greenwald, 2009). Across four 32-trial blocks, participants were presented with different word stimuli and directed to press one response key (labeled on screen) whenever the stimulus matched one of two target
41
categories and to press a second key (not labeled on screen) for “anything else.” The
participants’ response latencies on each trial were recorded. Following a practice block,
participants completed three blocks designed to assess their implicit evaluations of the
ingroup, outgroup, and control group, respectively. In this part of the task, participants
were presented with positive and negative words and the names of members from each
target group. Participants were instructed to press one key (‘F’) whenever the stimulus
was a “Good” word or a member of a focal target group (for example, Green). They
were told to press another key (‘G’) for anything that did not match one of these
categories (in this example, negative words and the names of Blue group and control
group members). For each of the three critical blocks, a different target group was
categorized on the same response key as the good words, in a counterbalanced order.
Participants were instructed to respond quickly during the task while trying to avoid
mistakes. Errors required correction before the program could proceed to the next trial.
Prior to analysis, response times were averaged across all trials within each critical block
and recoded (based on participants’ group membership) as mean latencies for an ingroup,
outgroup, and control group block. Faster response times reflect more favorable
evaluations of the focal target group.
Results
Participants’ mean response latencies (in milliseconds) for each target block were first submitted to a within-subjects repeated measures ANOVA, which revealed a significant effect of target, F(2, 210) = 9.66, p < .001 (see Figure 4). Additional repeated measures ANOVAs were then performed to compare each pair of blocks. Participants
42
responded more quickly when ingroup members and good words were categorized on the same response key (M = 749.38, SD = 202.94) than when outgroup members and good words were paired together (M = 824.31, SD = 231.42), F(1, 105) = 21.51, p < .001, which suggests that participants evaluated the ingroup more favorably than the outgroup.
In addition, average response times when the control group was the focal target (M =
787.50, SD = 247.31) were significantly slower than when the ingroup was paired with good, F(1, 105) = 4.38, p = .04, partial η2 = .04, and significantly faster than when the outgroup was paired with good, F(1, 105) = 4.86, p = .03, partial η2 = .04.
IAT D scores were also calculated for each pair of blocks using procedures recommended by Greenwald, Nosek, and Banaji (2003). These analyses indicated that participants favored the ingroup over the outgroup (D = .20) and control group (D = .10) but preferred the control group to the outgroup (D = .14). All D scores were significantly different from zero, ps < .05. This pattern of relationships corresponds to the analyses conducted with the mean response latencies in suggesting that the difference in ingroup and outgroup attitudes is based on both ingroup favoritism and outgroup derogation.
Discussion
Prior research investigating implicit intergroup bias in the MGP yielded evidence of a relative preference for the ingroup over the outgroup (e.g., Ashburn-Nardo et al.,
2001; Castelli, et al., 2004) that, at least in some cases, seemed to be driven solely by ingroup favoritism (e.g., Otten & Moskowitz, 2000; Otten & Wentura, 1999). However, the lack of an appropriate control group in these studies made it difficult to definitively rule out the influence of implicit outgroup derogation in this phenomenon. Therefore, in
43
Experiment 3, we assigned participants to minimal groups and measured their attitudes toward the ingroup, outgroup, and control group with an implicit attitude measure. In support of previous literature, the results of this experiment indicate that people implicitly favor their ingroup. However, evidence suggesting that they are implicitly prejudiced against the outgroup also emerged. This pattern of results replicates our findings from the explicit attitude measures used in Experiments 1 and 2 and suggests (given the nature of the implicit attitude measure we used) that those prior results were likely not due to demand.
44
CHAPTER 5: GENERAL DISCUSSION
From ethnic cleansing in Nazi Germany to contemporary terrorist attacks, many of the most tragic instances of social conflict across human history plausibly can be linked in some way to group identity and bias. In recognition of this relationship, one of the most vibrant and enduring research traditions in social psychology has focused on investigating the fundamental processes and consequences of social categorization (Fiske,
1998; Park & Judd, 2005). A critical finding from this domain is that categorization per se is sufficient to elicit intergroup prejudice and discrimination. That is, group membership, even when it is based on some arbitrary pretext, can lead people to favor those in their ingroup over those in the outgroup (e.g., Tajfel et al., 1971). Importantly, social psychologists generally believe that this bias is fueled by ingroup favoritism and not outgroup derogation, unless other hostility instigating factors are present, such as competition for resources (Brewer, 1979;1999).
However, prior studies that have attempted to examine this phenomenon within the confines of a “pure,” minimal group setting have relied on methodological designs, procedures and dependent measures that make it difficult to tell if the relative bias is indeed solely the result of ingroup favoritism or if some degree of outgroup derogation also plays a role. In particular, most prior research has not employed an appropriate
45
control group to help discern the direction of the minimal group effect, or if a control
group was used, other factors in the experimental design rendered the results ambiguous.
In addition, many researchers have assumed that no outgroup derogation existed because
outgroup attitudes were not objectively negative. The purpose of the present research
was to shed light on the nature of intergroup bias by introducing a control target into the
MGP – one that could reveal prejudice against the outgroup even if outgroup attitudes were not negative in an absolute sense. Therefore, we implemented a baseline target
(which we showed in Experiment 2 was not simply a third minimal group) that allowed us to make the type of comparisons necessary to decompose the relative intergroup bias that emerges in the MGP. With this methodological tool, we were able to show across three experiments and on both explicit (Experiments 1 and 2) and implicit (Experiment 3) attitude measures that the classic MGP effect is a function of both ingroup favoritism and outgroup derogation.
Of the two types of bias, ingroup favoritism seems to carry more weight in driving the effect, which fits with current theories regarding the underlying motivations and consequences of social categorization (e.g., Brewer, 1979; 1999; see Hewstone et al.,
2002, for a review). But the fact that outgroup derogation also contributes to the bias suggests that the seeds of group-based conflict can take shape as soon as groups are formed – and without the influence of other factors known to incite outgroup prejudice.
This finding is not often anticipated by researchers, but neither is it necessarily contradictory with existing theory. For example, social identity theory (Tajfel & Turner,
1986) states that individuals are inherently motivated to elevate and protect the status of
46
the ingroup. Devaluing the outgroup (at least a little) would seem to be one effective way
of pursuing these goals. Also, the fact that outgroup derogation was evident on both
explicit and implicit measures suggests that our results are likely not due to experimenter
demand and points to the fundamental nature of outgroup bias. That is, mere categorization is sufficient to elicit prejudice against the outgroup at a relatively
automatic level. This finding is critical given that some of the best evidence that
intergroup bias in the minimal group situation is based on ingroup favoritism alone
comes from MGP studies that employ implicit measures (e.g., Otten & Moskowitz, 2000;
Otten & Wentura, 1999). By including a more appropriate control target in Experiment
3, however, we were able to show for the first time that implicit outgroup derogation can
be elicited in a minimal group context.
In addition to clarifying the direction of intergroup bias in the MGP, we also
investigated potential key predictors of ingroup favoritism and outgroup derogation.
Existing literature suggests that greater identification with the ingroup leads to more
positive ingroup attitudes and a larger intergroup bias across a variety of domains (e.g.,
Fleming & Petty, 2000; Gagnon & Bourhis, 1996; Mummendey et al., 2000; Perreault &
Bourhis, 1999). However, because prior research has not included an effective control
target in the MGP, it has not been possible to directly examine the relationship between
ingroup identification and ingroup favoritism. In contrast, the results of Experiment 1
and 2 showed that higher levels of ingroup identification predicted stronger ingroup
favoritism.
47
Furthermore, we were also able to identify a new variable that was useful in determining who is most likely to exhibit a bias against the outgroup. Although ingroup identification has received considerable attention in the literature (Hewstone et al., 2002), few studies measure participants’ degree of identification with the outgroup, and no study has attempted to link this variable with intergroup attitudes in the MGP. Rather, in rare cases when this variable has been captured, it has been used as a manipulation check of the effectiveness of the minimal group induction (e.g., Mummendey et al., 2000; Otten &
Moskowitz, 2000). But previous research exploring the impact of taking a negational vs. affirmational group identity suggests that outgroup identification is a plausible antecedent of outgroup bias. Specifically, Zhong et al. (2008) found that inducing participants to focus on a group membership they did not hold (negational identity) led them to derogate that outgroup, presumably because they were inclined to psychologically distance themselves from the outgroup. However, it is possible that the outgroup bias observed in their studies was due to demand from their identity manipulation, and we were interested in determining if some people spontaneously disidentify with the outgroup and whether lower levels of outgroup identification would predict outgroup derogation. Measuring outgroup identification and including a control target in our studies allowed us to test this possibility, and results from Experiments 1 and 2 indicated that those who felt less identified with the outgroup exhibited greater outgroup derogation. Identifying such individuals could be very useful to researchers, particularly in cases where, overall, participants seem to be motivated only by ingroup favoritism (e.g., Halevy, Bornstein, &
Sagiv, 2008; Valdesolo & De Steno, 2007). Finally, in Experiment 2, outgroup
48
identification also predicted ingroup favoritism, and there was only an impact of outgroup identification on outgroup derogation when ingroup identification was high, which points to the importance of measuring both ingroup and outgroup identification in gaining an understanding of the factors that contribute to ingroup favoritism and outgroup derogation.
Future Directions
By having participants evaluate a baseline control group and measuring their level of identification with the outgroup, we have been able provide a richer understanding of the bias that occurs under minimal group conditions. However, a number of important questions remain. For example, no link between explicit outgroup identification and implicit outgroup derogation was found in the current research, but future work should investigate whether implicit outgroup identification predicts implicit outgroup derogation. Nevertheless, our work suggests that those who naturally disidentify with the outgroup are more likely to derogate the outgroup in a minimal group setting (at the explicit level at least), and it will be useful for future research to investigate factors that may potentially promote or inhibit outgroup identification. For instance, certain personality variables, like social dominance orientation (Sidanius & Pratto, 1999) and right-wing authoritarianism (Altemeyer, 1988), that seem to have little impact on ingroup identification (e.g., Reynolds et al., 2007; c.f. Perrault & Bourhis, 1999) may be strongly associated with outgroup disidentification. On the other hand, individuals high in egalitarianism (Schwartz, 2004) might feel more attached to any outgroup and, as a result, be less likely to express outgroup derogation.
49
It will also be important to explore how the effects observed in the current
research relate to different downstream consequences and outcomes. For example,
behavioral measures were not included in our studies, but presumably, the bi-directional
attitudinal bias we captured would predict corresponding discriminatory behaviors. For
instance, individuals might be more/less willing to use or accept some product or object
that is associated with the ingroup, outgroup, or a control group (White & Dahl, 2006).
Also, those who are less identified with the outgroup in a minimal group situation might
chronically disidentify with real outgroups as well. That is, those who identify less with
a minimal outgroup might pervasively feel detached from any outgroup. If this is the
case, then those who report feeling less identified with the outgroup in a MGP may be
particularly likely to exhibit prejudice toward actual outgroups outside the laboratory.
In addition, because group differentiation occurs at a perceptual level as well as an evaluative and behavioral level (Doise, 1976), it would be interesting to see how a control group might shed light on different perceptual processes. For example, Van
Bavel et al. (2008) found greater neural processing of ingroup faces than outgroup faces,
following a minimal group induction, which indicated that the ingroup was perceptually
more relevant than the outgroup. Adding a control group to this study, however, could
provide additional understanding of these effects. For instance, if control faces received
the same degree of processing as outgroup faces, that would suggest the relative ingroup-
outgroup processing bias is based on the enhanced relevance of the ingroup label. On the
other hand, if control faces are processed to the same degree as ingroup faces, that would
indicate that it is the irrelevance of the outgroup label that accounts for the ingroup-
50
outgroup difference. Or perhaps control faces would receive the least amount of
processing of the three target groups, which would suggest that any type of affiliation is
more relevant than no affiliation, though perhaps for different reasons (i.e., because you
like the ingroup and fear the outgroup).
It would also be worthwhile to identify factors that could potentially exacerbate,
attenuate, or lead to a reversal of the bi-directional bias observed in the current research.
For instance, there could be cases where people wish to leave their ingroup and become a
member of a particular outgroup. In this type of situation, ingroup favoritism and
outgroup derogation could be replaced by outgroup favoritism and ingroup derogation.
For example, an individual could have a strong desire to become a member of a different
community because it is perceived as better in some way (e.g., more affluent, safe,
diverse, etc.) than the one to which he/she currently belongs. As a result, this person
might report more favorable attitudes toward the outgroup and less favorable attitudes
toward the ingroup (relative to a control group). Presumably, such effects would be
moderated by ingroup and outgroup identification. That is, those who feel less identified
with their ingroup should be most likely exhibit ingroup derogation, and those who feel
most identified with the outgroup should show the greatest degree of outgroup favoritism.
Of course, the control group provides diagnostic value only to the extent that
groups are favored or derogated relative to the control, and there may be situations in
which all groups are valued more or less than those who are unaffiliated. There could be
cases, for example, where individuals prefer both ingroup and outgroup members
significantly more than unaffiliated individuals. For instance, people might rate fans of
51
their favorite sports team and fans of a rival team more positively than those who are not
fans at all. In this situation, rooting for any team is perceived as preferable to rooting for
no team.5
In contrast, there could be cases where people prefer unaffiliated individuals to
both the ingroup and the outgroup. For example, believing that non-affiliation is a reflection of an admired trait like independence could lead people to evaluate the unaffiliated control group more favorably than either the ingroup or the outgroup. In this situation, being a member of no group is perceived as preferable to being a member of any group. It is important to note, however, that the scenarios described here seem most likely to arise with real world groups that are more than minimal in nature. Within the context of the minimal group situation, an unaffiliated control group should provide an appropriate comparison point for identifying group favoritism and derogation.
Ultimately, it is hoped that this line of research contributes to the development of effective strategies to reduce prejudice and discrimination. Based on our findings, it appears that mere group categorization is sufficient to elicit a bias against those in the outgroup on both a conscious and relatively automatic level. Additional factors known to arouse intergroup conflict, (e.g., perceived threat) do not appear to be necessary for this bias to occur, particularly for those who naturally disidentify with the outgroup. Indeed, such variables would likely serve to aggravate rather than instigate outgroup derogation.
Given the apparent fundamental nature of outgroup bias, tools to reduce prejudice seem even more needed than before.
52
NOTES
1 Alternative allocation measures have been proposed (e.g., Bornstein et al., 1983;
Brewer & Silver, 1978; Locksley, Ortiz, & Hepburn, 1980; Ng, 1981), but these are often criticized for having the same limitations as the Tajfel matrices (Bornstein et al., 1983) or for being susceptible to the influence of other factors, such as social norms (Gaertner &
Insko, 2001).
2 Neutral or even slightly positive attitudes toward group targets in the MGP would not necessarily be surprising, given the possibility of a positivity offset toward novel stimuli
(Cacioppo, Gardner, & Berntson, 1997).
3 As noted earlier, prejudice against the outgroup need not take the form of an attitude that is negative in an absolute sense. A significant difference between evaluations of the control group and outgroup, regardless of the absolute value of those ratings, is sufficient to indicate a bias against the outgroup. This is not to suggest that where outgroup attitudes objectively fall is inconsequential. Indeed, very different outcomes might result from outgroup attitudes that are negative in absolute terms than those that are neutral or
53
positive. The key issue is that the absence of outgroup negativity does not necessarily
indicate an absence of outgroup bias.
4 In addition, ingroup and outgroup identification were assessed in Experiment 3 with the
same measures used in the previous studies, but these variables had no impact on
participants’ implicit evaluations of the groups.
5 One could argue that this is due the formation of new group boundaries distinguishing sports fans (ingroup) from non-fans (outgroup).
54
REFERENCES
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting
interactions. Thousand Oaks, CA: Sage.
Allport, G. W. (1954). The nature of prejudice. Reading, MA: Addison-Wesley.
Altemeyer, B. (1988). Enemies of freedom: Understanding right-wing authoritarianism.
San Francisco: Jossey-Bass.
Amiot, C.E., & Bourhis, R.Y. (2003). Discrimination and the positive-negative
asymmetry effect: Ideological and normative processes. Personality and Social
Psychology Bulletin, 29, 597–608.
Ashburn-Nardo, L., Voils, C. I., Monteith, M. J. (2001). Implicit associations as the seeds
of intergroup bias: How easily do they take root? Journal of Personality and
Social Psychology, 81, 789-799.
Berkowitz, N. H. (1994). Evidence that subject’s expectancies confound intergroup bias
in Tajfel’s minimal group paradigm. Personality and Social Psychology Bulletin,
20, 184-195.
Bornstein G, Crum L, Wittenbraker J, Harring K, Insko CA, Thibaut J. (1983). On the
measurement of social orientations in the minimal group paradigm. European
Journal of Social Psychology, 13, 321-350.
55
Bourhis, R. Y., Sachdev, I., & Gagnon, A. (1994). Intergroup research with the Tajfel
matrices: Methodological notes. In M. P. Zanna, & J. M. Olson (Eds.), The social
psychology of prejudice: The Ontario symposium (Vol. 7, pp. 209–232). Mahwah,
NJ: Erlbaum.
Brewer, M. B. (1979). In-group bias in the minimal intergroup situation: A cognitive-
motivational analysis. Psychological Bulletin, 86, 307-324.
Brewer, M. B. (1999). The psychology of prejudice: Ingroup love or outgroup hate?
Journal of Social Issues, 55, 429-444.
Brown, R. J. (1995). Group processes: Dynamics within and between groups (2nd ed.).
Oxford, UK: Blackwell.
Branscombe, N. R., & Wann, D. L. (1994). Collective self-esteem consequences of
outgroup derogation when a valued social identity is on trial. European Journal of
Social Psychology, 24, 651-657.
Buhl, T. (1999). Positive-negative asymmetry in social discrimination: Meta-analytical
evidence. Group Processes and Intergroup Relations, 2, 51-58.
Cacioppo, J. T., Gardner, W. L., & Berntson, G. G. (1997). Beyond bipolar
conceptualizations and measures: The case of attitudes and evaluative space.
Personality and Social Psychology Review, 1, 3-25.
Castelli, L., Zogmaister, C., Smith, E. R., & Arcuri, L. (2004). On the automatic
evaluation of social exemplars. Journal of Personality and Social Psychology, 86,
373-387.
56
Diehl, M. (1990). The minimal group paradigm: Theoretical explanations and empirical
findings. In W. Stroebe & M. Hewstone (Eds.), European review of social
psychology (Vol. 1). Chichester, UK: Wiley.
Doise, W. (1976). L'articulation psychosociologique et les relations entre groupes.
Bruxelles: De Boeck.
Fazio, R.H., Jackson, J.R., Dunton, B.C., &Williams, C.J. (1995). Variability in
automatic activation as an unobtrusive measure of racial attitudes: A bona fide
pipeline? Journal of Personality and Social Psychology, 69,1013–1027.
Fiske, S. T. (1998). Stereotyping, prejudice, and discrimination. In D. T. Gilbert, S. T.
Fiske, & G. Lindzey (Eds.), Handbook of social psychology (4th ed., Vol. 2, pp.
357-414). Boston, MA: Mcgraw-Hill.
Fleming, M. A., & Petty, R. E. (2000). Identity and persuasion: An elaboration likelihood
approach. In D. J. Terry & M. A. Hogg (Eds.), Attitudes, behavior, and social
context: The role of norms and group membership (pp. 171-199). Mahwah, NJ:
Lawrence Erlbaum.
Gagnon, A., & Bourhis, R. Y. (1996). Discrimination in the minimal group paradigm:
Social identity or self interest? Personality and Social Psychology Bulletin, 22,
1289–1301.
Gardner, W. L., Pickett, C. L., & Brewer, M. B. (2000). Social exclusion and selective
memory: How the need to belong influences memory for social events.
Personality and Social Psychology Bulletin, 26, 486-496.
57
Gaertner, L., & Insko, C. A. (2000). Intergroup discrimination in the minimal group
paradigm: Categorization, reciprocation, or fear? Journal of Personality and
Social Psychology, 79, 77–94.
Gaertner, L., & Insko, C. A. (2001). On the measurement of social orientations in the
minimal group paradigm: Norms as moderators of the expression of in-group bias.
European Journal of Social Psychology, 31, 143–154.
Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding and Using the
Implicit Association Test: I. An Improved Scoring Algorithm. Journal of
Personality and Social Psychology, 85, 197-216.
Griffin, A. M., & Langlois, J. H. (2006). Stereotype directionality and attractiveness
stereotyping: Is beauty good or is ugly bad? Social Cognition, 24, 187-206.
Halevy, N., Bornstein, G., & Sagiv, L. (2008). “In-group love” and “out-group hate” as
motives for individual participation in intergroup conflict: A new game paradigm.
Psychological Science, 19, 405-411
Hartstone, M., & Augoustinos, M. (1995). The minimal group paradigm: Categorization
into two versus three groups. European Journal of Social Psychology, 25, 179-
193.
Henderson, R. W. (1985). Fearful memories: The motivational significance of forgetting.
In F. R. Brush & J. B. Overmier (Eds.), Affect, conditioning, and cognition:
Essays on the determinants of behavior (pp. 43-53). Hillsdale, NJ: Erlbaum.
58
Hertel G, Kerr NL. (2001). Priming ingroup favoritism: The impact of normative scripts
in the minimal group paradigm. Journal of Experimental Social Psychology, 37,
316-324.
Hewstone, M., Rubin, M., & Willis, H. (2002). Intergroup bias. In M. R. Rosenzweig &
L. W. Porter (Eds.), Annual review of psychology (Vol. 53, pp. 575-604). Palo
Alto, CA: Annual Reviews.
Jarvis, W. B. G. (2006). MediaLab and DirectRT (Version 2006) [Computer software].
New York, NY: Empirisoft.
Locksley A, Ortiz V, Hepburn C. (1980). Social categorization and discriminatory
behavior: Extinguishing the minimal intergroup discrimination effect. Journal of
Personality and Social Psychology, 39, 773-783.
Manstead, A. S. R. (1997). Situations, belongingness, attitudes, and culture: Four lessons
learned from social psychology. In G. McGarty & H. S. Haslam (Eds.), The
message of social psychology: Perspectives on mind and society (pp. 238-251).
Oxford, England: Blackwell.
Mineka, S. (1992). Evolutionary memories, emotional processing, and the emotional
disorders. In D. Medin (Ed.), The psychology of learning and motivation (Vol. 28,
pp. 161-206). San Diego, CA: Academic.
Mummendey, A., & Otten, S. (1998). Positive-negative asymmetry in social
discrimination. In M. Hewstone & W. Stroebe (Eds.), European review of social
psychology (Vol. 9, pp. 107–143). London: Wiley.
59
Mummendey, A., Otten, S., Berger, U., & Kessler, T. (2000). Positive-negative
asymmetry in social discrimination: Valence of evaluation and salience of
categorization. Personality & Social Psychology Bulletin, 26, 1258-1270.
Mummendey, A., Simon, B., Dietze, C., Grünert, M., Haeger, G., Kessler, S., Lettgen, S.,
& Schäferhoff, S. (1992). Categorization is not enough: Intergroup discrimination
in negative outcome allocation. Journal of Experimental Social Psychology, 28,
125-144.
Oakes, P. J., Haslam, A., & Turner, J. C. (1994). Stereotyping and social reality. Oxford,
UK: Blackwell.
Otten, S., & Moskowitz, G. B. (2000). Evidence for implicit evaluative in-group bias:
Affect-biased spontaneous trait inference in a minimal group paradigm. Journal
of Experimental Social Psychology, 36, 77–89.
Otten, S., & Wentura, D. (1999). About the impact of automaticity in the minimal group
paradigm: Evidence from affective priming tasks. European Journal of Social
Psychology, 29, 1049–1071.
Otten, S., & Wentura, D. (2001). Self-anchoring and in-group favoritism: An individual
profiles analysis. Journal of Experimental Social Psychology, 37, 525-532.
Park, B., & Judd, C. M. (2005). Rethinking the link between categorization and prejudice
within the social cognition perspective. Personality and Social Psychology
Review, 2, 108-130.
Perreault, S., & Bourhis, R. Y. (1999). Ethnocentrism, social identification, and
discrimination. Personality and Social Psychology Bulletin, 25, 92–103.
60
Petty, R. E., Fazio, R. H., & Briñol, P (Eds.) (2009). Attitudes: Insights from the new
implicit measures. New York, NY: Psychology Press.
Pinter, B. & Greenwald, A. G. (2004). Understanding implicit partisanship: Enigmatic
(but genuine) group identification and attraction. Group Processes and
Interpersonal Relations, 7, 283-296.
Pinter, B., & Greenwald, A. G. (2010). A comparison of minimal group induction
procedures. Manuscript submitted for publication.
Pettigrew, T. F., & Meertens, R. W. (1995). Subtle and blatant prejudice in western
Europe. European Journal of Social Psychology, 25, 57-75.
Rabbie, J. M., & Horwitz, M. (1969). Arousal of ingroup-outgroup bias by a chance win
or loss. Journal of Personality and Social Psychology, 13, 269-277.
Reynolds, K. J., Turner, J. C., Haslam, S. A., Ryan, M. K., Bizumic, B., & Subasic, E.
(2007). Does personality explain in-group identification and discrimination?
Evidence from the minimal group paradigm. British Journal of Social
Psychology, 46, 517-539.
Riek, B. M., Mania, E. W., & Gaertner, S. L. (2006). Intergroup threat and outgroup
attitudes: A meta-analytic review. Personality and Social Psychology Review, 10,
336-353.
Rosenbaum, M. E. (1986). The repulsion hypothesis: On the nondevelopment of
relationships. Journal of Personality and Social Psychology, 51, 1156-166.
61
Schwartz, S. H. (2004). Mapping and interpreting cultural differences around the world.
In H. Vinken, J. Soeters, & P. Ester (Eds.), Comparing cultures, dimensions of
culture in a comparative perspective (pp. 43– 73). Leiden, The Netherlands: Brill.
Sherif, M., Harvey, O. J., White, B., J., Hood, W. R., & Sherif, C. W. Intergroup conflict
and cooperation: The Robbers Cave experiment. Norman, OK: University Book
Exchange, 1961.
Sidanius, J. (1993). The psychology of group conflict and dynamics of oppression: A
social dominance perspective. In S. Iyengar & W. McGuire (Eds.), Explorations
in political psychology (pp. 183-219). Durham, NC: Duke University Press.
Sidanius, J., & Pratto, F. (1999). Social dominance: An intergroup theory of social
hierarchy and oppression. Cambridge, UK and New York, USA: Cambridge
University Press.
Singh, R., & Tan, L. S. (1992). Attitudes and attraction: A test of the similarity-attraction
and dissimilarity-repulsion hypotheses. British Journal of Social Psychology, 31,
227-238.
Sriram, N., & Greenwald, A. G. (2009). The Brief Implicit Association Test.
Experimental Psychology, 56, 283–294.
Sumner, W. G. (1906). Folkways. New York: Ginn.
Tajfel, H. (1969). Cognitive aspects of prejudice. Journal of Social Issues, 25, 79-98.
Tajfel, H., & Billig, M. (1974). Familiarity and categorization in intergroup behavior.
Journal of Experimental Social Psychology, 10, 159-170.
62
Tajfel, H. Billig, M.G., Bundy, R. P., & Flament, C. (1971). Social categorization and
intergroup behavior. European Journal of Social Psychology, 1, 149-178.
Tajfel, H., & Turner, J. C. (1986). The social identity theory of intergroup behavior. In S.
Worchel & W. G. Austin (Eds.), Psychology of intergroup relations (pp. 7-24).
Chicago: Nelson-Hall.
Uleman, J. S., Newman, L. S., & Moskowitz, G. B. (1996). People as flexible
interpreters: Evidence and issues from spontaneous trait inference. In M. Zanna
(Ed.), Advances in experimental social psychology (Vol. 28, pp. 211-280). San
Diego, CA: Academic Press
Valdesolo, P., & DeSteno, D. (2007). Moral hypocrisy: Social groups and the flexibility
of virtue. Psychological Science,18, 689-690.
Van Bavel, J. J., Packer, D. J, & Cunningham, W. A. (2008). The Neural Substrates of
In-Group Bias: A Functional Magnetic Resonance Imaging Investigation.
Psychological Science, 11, 1131-1139.
Wegener, D. T., & Petty, R. E. (1997). The flexible correction model: The role of naïve
theories of bias in bias correction. In M. P. Zanna (Ed.), Advances in
experimental social psychology (Vol. 29, pp. 141-208). San Diego: Academic
Press.
White, K., Dahl, D. W. (2006). To be or not be? The influence of dissociative reference
groups on consumer preferences. Journal of Consumer Psychology, 16, 404-414.
63
Wilson, T. D., & Brekke, N. (1994). Mental contamination and mental correction:
Unwanted influences on judgments and evaluations. Psychological Bulletin, 116,
117-142.
Worchel, S., Andreoli, V. A., & Folger, R. (1977). Intergroup competition and intergroup
attraction: The effect of previous interaction and outcome of combined effort.
Journal of Experimental Social Psychology, 13, 131-140.
Zhong, C., Phillips, K. W., Leonardelli, G. J., & Galinsky, A. D. (2008). Negational
categorization and intergroup behavior. Personality and Social Psychology
Bulletin, 34, 793-806.
64
APPENDIX A
ALTERNATIVE MODERATION ANALYSES FOR EXPERIMENTS 1 AND 2
Experiment 1
To investigate the potential moderating role of ingroup and outgroup identification on ingroup favoritism, a regression analysis was conducted in which group attitudes were predicted by ingroup identification, outgroup identification, and group target (coded 0 for control and 1 for ingroup). This analysis revealed a significant effect of group, ß = .36, t(136) = 5.12, p < .001, which was qualified by an ingroup identification X group target interaction, ß = .54, t(136) = 4.75, p < .001 (see Figure 7).
Decomposing this interaction using procedures recommended by Aiken and West (1991) revealed a significant difference between ingroup and control group attitudes when ingroup identification was high, ß = .68, t(140) = 6.89, p < .001, but not when it was low,
ß = .03, t(68) = .25, p = .80. That is, participants reported more favorable attitudes toward the ingroup relative to the control group (i.e., greater ingroup favoritism) when they felt more (vs. less) identified with the ingroup. No other significant effects emerged.
To determine whether ingroup and outgroup identification had any moderating influence on outgroup derogation, a regression analysis was performed in which group attitudes were predicted by ingroup identification, outgroup identification, and group
65
target (coded 0 for control and 1 for outgroup). Only a significant outgroup identification
X group target interaction emerged, ß = .25, t(136) = 2.23, p = .03, which showed that
there was a significant difference between outgroup and control group attitudes among
those who felt less identified with the outgroup, ß = -.36, t(140) = -2.73, p = .007, but not for those who felt more identified with the outgroup, ß = .12, t(140) = .88, p = .38 (see
Figure 8). Specifically, participants reported less favorable attitudes toward the outgroup compared to the control group (i.e., greater outgroup derogation) when they felt less (vs. more) identified with the outgroup.
Experiment 2
First, we examined whether identification with the ingroup and outgroup moderated the pattern of ingroup favoritism observed in participants’ category evaluations. To do so, category attitudes were entered into a regression equation with ingroup identification, outgroup identification, and group target (coded 0 for control and
1 for ingroup) as predictors. This analysis revealed a significant main effect of outgroup identification, ß = .22, t(234) = 2.81, p = .005, and group target, ß = .52, t(234) = 10.61, p
< .001. These effects were qualified by a significant outgroup identification X group
target interaction, ß = -.19, t(234) = -2.42, p = .02 (see Figure 9), which indicated that those who felt less identified with the outgroup reported more favorable attitudes toward the ingroup relative to the control group (i.e., greater ingroup favoritism), ß = .74, t(238)
= 9.77, p < .001, than those who felt more identified with the outgroup, ß = .31, t(238) =
4.06, p < .001. In addition, as in Experiment 1, a significant interaction between ingroup
66
identification and group target emerged, ß = .38, t(234) = 5.44, p < .001, which revealed greater ingroup favoritism among those who identified more with the ingroup, ß = .83, t(238) = 12.33, p < .001, than among those who identified less, ß = .22, t(238) = 3.20, p =
.002 (see Figure 10). In sum, these results suggest that both ingroup and outgroup identification contributed to the expression of ingroup favoritism.
Second, to investigate the impact of ingroup and outgroup identification on outgroup derogation at the category level, a regression analysis was conducted with ingroup identification, outgroup identification, and group target (coded 0 for control and
1 for outgroup) predicting category attitudes. This analysis revealed a significant effect of outgroup identification, ß = .24, t(234) = 2.54, p = .01, that was qualified by a marginally significant outgroup identification X group target interaction, ß = .17, t(234) =
1.79, p = .08 (see Figure 11). Decomposing this interaction showed that the outgroup was less favorably evaluated than the control group by those who felt less identified with the outgroup, ß = -.33, t(238) = -4.07, p < .001, whereas those who felt more identified with the outgroup did not express outgroup derogation, ß = .05, t(238) = .59, p = .56.
We then tested to see if the degree of identification with the ingroup and outgroup was related to ingroup favoritism at the exemplar level. Exemplar attitudes were submitted to a regression analysis with ingroup identification, outgroup identification, and group target (coded 0 for control and 1 for ingroup) as predictors. Significant effects emerged for ingroup identification, ß = .16, t(234) = 1.94, p = .05, outgroup identification, ß = .22, t(234) = 2.37, p = .02, and group target, ß = .36, t(234) = 6.25, p <
.001. More critically, both ingroup identification, ß = .18, t(234) = 2.19, p = .03, and
67
outgroup identification, ß = -.23, t(234) = -2.52, p = .01, interacted with group target. As with the category attitudes, greater ingroup favoritism was observed when participants were high in ingroup identification, ß = .55, t(238) = 6.97, p < .001, rather than low, ß =
.21, t(238) = 2.61, p = .01 (see Figure 12), and when they were low in outgroup
identification, ß = .61, t(238) = 7.48, p < .001, rather than high, ß = .15, t(238) = 1.82, p =
.07 (Figure 13). These results correspond to what was found on category attitudes in this
study and suggest that both ingroup and outgroup identification moderated the extent to
which participants favored the ingroup (relative to the control group).
Next, to determine if ingroup and outgroup identification had any impact on
outgroup derogation at the exemplar level, we performed a regression analysis with
ingroup identification, outgroup identification, and group target (coded 0 for control and
1 for outgroup) predicting exemplar attitudes. This analysis revealed a significant main
effect of outgroup identification, ß = .23, t(234) = 2.29, p = .02, but no significant
outgroup identification X group target interaction was found, ß = .10, t(234) = .96, p =
.34. Thus, in contrast to the results of the category evaluations in Experiment 1 and 2,
outgroup identification did not appear to influence the degree of outgroup derogation
expressed in participants’ attitudes toward the group exemplars.
Taken together, the results of Experiment 1 and 2 suggest that researchers trying
to identify to those who are most likely to favor the ingroup and derogate the outgroup
would be aided by measuring both ingroup identification and outgroup identification.
68
APPENDIX B
ILLUSTRATIONS
69
MD MIP MJP In 19 18 17 16 15 14 13 12 11 10 9 8 7
Out 1 3 5 7 9 11 13 15 17 19 21 23 25
Table 1. Multiple-choice allocation matrix with MD + MIP and MJP as opposing orientations
70
Figure 1. Mean explicit attitude ratings toward group categories in Experiment 1
71
Figure 2. Mean explicit attitude ratings toward group categories in Experiment 2
72
Figure 3. Mean explicit attitude ratings toward group members in Experiment 2
73
Figure 4. Mean category outgroup derogation as a function of ingroup and outgroup identification in Experiment 2
74
Figure 5. Mean exemplar outgroup derogation as a function of ingroup and outgroup identification in Experiment 2
75
Figure 6. Implicit attitudes toward group categories as indicated by mean response latencies on the Brief IAT in Experiment 3
76
Figure 7. Mean attitudes toward the ingroup vs. control (ingroup favoritism) as a function of ingroup identification in Experiment 1
77
Figure 8. Mean attitudes toward the outgroup vs. control (outgroup derogation) as a function of outgroup identification in Experiment 1
78
Figure 9. Mean attitudes toward the ingroup vs. control (ingroup favoritism) as a function of outgroup identification in Experiment 2
79
Figure 10. Mean attitudes toward the ingroup vs. control (ingroup favoritism) as a function of ingroup identification in Experiment 2
80
Figure 11. Mean attitudes toward the outgroup vs. control (outgroup derogation) as a function of outgroup identification in Experiment 1
81
Figure 12. Mean attitudes toward ingroup members vs. control group members (ingroup favoritism) as a function of ingroup identification in Experiment 2
82
Figure 13. Mean attitudes toward ingroup members vs. control group members (ingroup favoritism) as a function of outgroup identification in Experiment 2
83
APPENDIX C
SAMPLE MATERIALS FROM EXPERIMENT 1
84
Minimal Group Induction (Screen 1)
85
Minimal Group Induction (Screen 2)
86
Self-Group Association Task Instructions
87
Team Name Targets for Self-Group Association Task
88
Sample Trial from Self-Group Association Task
89
Sample Trial from Self-Group Association Task
90
APPENDIX D
SAMPLE MATERIALS FROM EXPERIMENT 2
91
Minimal Group Induction (Screen 1)
92
Minimal Group Induction (Screen 2)
93
Minimal Group Induction (Screen 3)
94
Minimal Group Induction (Screen 4)
95
Minimal Group Induction (Screen 5)
96
Minimal Group Induction (Screen 6)
97
Minimal Group Induction (Screen 7)
98
Minimal Group Induction (Screen 8)
99
Minimal Group Induction (Screen 9)
100
APPENDIX E
SAMPLE MATERIALS FROM EXPERIMENT 3
101
Brief IAT Sample Task Instructions
102
Positive and Negative Word Targets for Brief IAT
103
Sample Trial from Brief IAT
104
Sample Trial from Brief IAT
105