Sie sind auf Seite 1von 10



Outgroup Accountability in the Minimal

Group Paradigm: Implications for Aversive
Discrimination and Social Identity Theory

Michael Dobbs
University of Arizona
William D. Crano
Claremont Graduate University

The minimal group paradigm (MGP) is a popular method of nature of the relationship between the social group and
testing intergroup phenomena. Originally created to facilitate the individual’s role in it. Early work emphasized compe-
discovery of conditions necessary and sufficient to produce tition over scarce resources as a primary impetus for dis-
ingroup favoritism, early MGP results suggested that simply criminatory intergroup behavior (Campbell, 1965;
grouping people was sufficient to cause discrimination. More Sherif, 1966; Sherif, Harvey, White, Hood, & Sherif,
recent research has uncovered factors that cause MGP-based dis- 1961). More recent studies have been guided by SIT,
crimination to disappear. The present experiment examined which seeks to explain the cognitive and motivational
outgroup accountability as explicatory of these variations. It was processes that individuals engage in when acting as
found that requiring justification for allocations attenuated dis- group members (Tajfel & Turner, 1979, 1986).
crimination. Outgroup accountability-based attenuation was The MGP has been an important adjunct to progress
especially evident when the allocator was of majority (vs. minor- in the formation and evaluation of SIT. Developed by
ity) status, as expected on the basis of aversive racism consider- Tajfel (1970) and his associates (Tajfel, Billig, Bundy, &
ations. Allocators of minority status tended to discriminate more Flament, 1971), the MGP is used as a convenient para-
when made accountable to the outgroup. Implications of these digm for conducting intergroup research in the labora-
results for theories of social identity and aversive racism are tory. This article describes an MGP experiment that
discussed. examines causal factors underlying the common and
originally counterintuitive finding that merely categoriz-
ing people into distinct groups is a necessary and suffi-
I ntergroup relations has long been a central research cient condition for eliciting ingroup favoritism. We sug-
gest that the mere categorization effect might be
theme in social psychology. The minimal group para-
digm (MGP) was developed to facilitate such research. particularly easy to achieve in the MGP, which typically
Recently, however, the MGP itself has come under close does not include many common characteristics of real
methodological scrutiny. As a result, questions have been intergroup contexts. Specifically, the research is con-
raised regarding the interpretability and generality of cerned with the possibility that the absence of account-
results produced in the MGP. The present research was ability to outgroup members in the standard MGP might
undertaken to provide data relevant to an alternative increase the likelihood of discrimination. When group
interpretation of results obtained in the MGP. The
research is framed in the context of two competing theo- Authors’ Note: This research was supported in part by a grant from the
retical perspectives that relate to intergroup phenom- National Institute on Drug Abuse (Grant No. R01-DA12578-01). We
are grateful to Rachel Smith for her help in conducting the research.
ena: social identity theory (SIT) (Tajfel & Turner, 1979,
Requests for reprints may be addressed to either author at the Depart-
1986) and aversive racism theory (Dovidio, in press; ment of Psychology, Claremont Graduate University, 123 E. 8th Street,
Gaertner & Dovidio, 1986). Claremont, CA 91711; e-mail:
Intergroup research has progressed through a num- PSPB, Vol. 27 No. 3, March 2001 355-364
ber of phases, each with a unique perspective on the © 2001 by the Society for Personality and Social Psychology, Inc.


Downloaded from at The University of Iowa Libraries on March 12, 2015

members are accountable to the outgroup for their ments, Diehl (1988, 1989, 1990) examined a number of
actions, results contrary to the usual findings are potential alternative explanations for discrimination in
expected to ensue. the MGP, including derivations from belief congruence
theory (Rokeach, 1960), equity theory (Walster, Walster, &
Tajfel’s Minimal Group Paradigm Berscheid, 1978), and category differentiation theory
Tajfel’s (1978) initial research, which was (Doise, 1978). Results showed that SIT was the most par-
designed to discover a baseline intergroup condition simonious explanation in most cases; however, certain
of non- discrimination, was envisioned as the first of a elements of real group interaction not modeled in the
series of experiments that would begin by stripping away typical MGP substantially attenuated outgroup discrimi-
all characteristics normally associated with groups: nation. In particular, Diehl (1989) found that partici-
face-to-face interaction between members, conflicts of pants tended to reciprocate fairness to outgroup mem-
interest, previous hostility between groups, and so on. In bers whom they believed had been fair to them. This
subsequent experiments, these characteristics were to be suggests the possibility that the lack of interindividual
systematically layered on until discrimination was mutuality in the standard MGP might foster discrimina-
observed, to establish the necessary and sufficient condi- tion, perhaps because participants need not justify their
tions to produce ingroup favoritism (Oakes, Haslam, & allocation choices.
Turner, 1994). The results of the first minimal group Another line of research that has questioned the asso-
experiments were surprising and unexpected: Partici- ciation between social identity and positive distinctive-
pants favored members of their own ingroup at the ness as well as the mere categorization effect is found in
expense of the outgroup despite the lack of any apparent the work of Mummendey and her associates (Blanz,
reason for doing so. Group membership was completely Mummendey, & Otten, 1995; Buhl, 1999; Mummendey
anonymous; the groups were artificial, created on the et al., 1992; Otten, Mummendey, & Blanz, 1996). Among
basis of trivial criteria; and the participants had no prior other factors, this research has examined the effect of
history of intergroup conflict. Discrimination was not allocating negative incentives (rather than positive
expected until later in the research program when some rewards) in the MGP. In Tajfel’s initial delineation of
key element (or combination of elements) of realistic minimal group requirements, penalties (in addition to
groups was added to the variable mix. Based on the or in lieu of rewards) are specifically mentioned as
results of early minimal group studies, Tajfel and his acceptable indicators of discrimination (see Tajfel et al.,
associates concluded that group formation and inter- 1971, pp. 153-154). However, the majority of MGP stud-
group discrimination had occurred as a result of social ies since the early 1970s have employed only positive
categorization per se; they were not due to other factors (usually monetary) reward allocations. Research com-
such as interpersonal friendships within the groups. paring positive with negative allocations across different
Later studies that implemented the paradigm replicated settings, levels of group categorization, salience of cate-
the phenomenon (Allen & Wilder, 1975; Tajfel, 1974; gorization, and relative group size and status shows that
Turner, 1975). participants allocating positive rewards and evaluating
others on positive dimensions display the familiar
Social Identity Theory
ingroup favoritism found in past MGP studies. However,
Tajfel and Turner’s (1979, 1986) SIT was developed this effect disappears when negative allocations (such as
both to capitalize on and to make sense of the MGP the time an ingroup or outgroup member must spend in
results. The theory posits that individuals strive to an uncomfortably noisy room) are made under the stan-
achieve or maintain a positive social identity, defined as dard (baseline) conditions.
“that part of an individual’s self-concept which derives Discrimination on positive outcomes, and its absence
from his knowledge of his membership in a social group when negative allocations and evaluations are made, is
(or groups) together with the value and emotional signifi- termed the positive-negative asymmetry bias in inter-
cance attached to that membership” (Tajfel, 1978, p. 63). group discrimination. Obviously, positive-negative asym-
Often, this positive valence can be achieved through an metry is not meant to imply that discrimination
appropriate intergroup social comparison. As pre- expressed in terms of negatively valenced behavior does
sented, SIT is primarily a motivational theory that posits not exist but rather that conditions beyond simple cate-
that self-esteem drives individual behavior in intergroup gorization may be necessary to provoke it. Mummendey
settings. Such an interpretation fits with the results of the (1995) refers to this possibility as the “aggravation
early minimal group experiments. hypothesis” because the conditions that intensify dis-
Tests of SIT over the past 18 years have proved generally crimination in the positive case (i.e., salient categoriza-
supportive, although some of the theory’s central proposi- tion and inferiority and/or minority status of the
tions have received mixed results. In a series of experi- ingroup) are needed simply to call it forth in the nega-

Downloaded from at The University of Iowa Libraries on March 12, 2015

tive case. Thus, it appears that the mere categorization patterns of intergroup bias such as those observed in the
effect does not hold under all circumstances. As sug- typical MGP may become evident. Although we do not
gested by Diehl (1990), some of the characteristics of the speculate on its source, we may indeed possess a ten-
classic MGP, such as the consistent use of positive dency to discriminate against outgroups. However,
resource allocations (and, we might add, the absence of expression of ingroup favoritism may be masked when
outgroup accountability), might present a misleading circumstances require public allocations of resources or
picture of the ease with which discriminatory behavior is merely careful thought about normative expectations.
elicited. When these conditions are absent, ingroup/outgroup
In explaining the mechanisms at work in the positive- discrimination is more probable. This pattern of bias or
negative asymmetry effect, Mummendey (1995) specu- nonbias in allocation behavior follows logically from
lated that the different valences of allocations or evalua- aversive racism theory; for present purposes, we refer to
tion dimensions prime different cognitive styles or dif- this composite allocation/evaluation pattern as aversive
ferent styles of information processing. “Negative discrimination.
stimuli may lead to a more analytic, more careful, more
accurate way of processing the information in the social Accountability
situation provided in the experiments” (Mummendey, Consistent with both aversive discrimination and
1995, p. 667). This more careful cognitive style could Mummendey’s perspective, research on the effects of
lead to the decision to discriminate less, or not at all. accountability on individual judgment may shed light on
Careful cognitive deliberation may lead to the processes that operate in the MGP. Tetlock (1983,
superordinate recategorization of the allocation target. 1992) found that “accountability to an individual of
As Otten et al. (1996) speculate, this recategorization unknown views motivates people to consider arguments
process might personalize the allocation interaction, and evidence on both sides of [an] issue in order to pre-
thus lessening the likelihood of discrimination. pare themselves for a wide variety of possible critical
Aversive Racism reactions to their views” (p. 75). Earlier, Cvetkovitch
(1978) claimed that accountability led to “less intuitive”
The study of aversive racism provides an alternative and “more analytic” modes of thought (p. 155). In the
means of conceptualizing allocation asymmetries in the MGP, participants are physically isolated from one
MGP. Throughout the past decade or so, proponents of another, anonymous, and have no anticipation of inter-
aversive racism theory have reported research findings acting with and/or justifying their choices to members
on real-world discrimination that may have relevance for of the ingroup or outgroup. In real-life situations, this
understanding the results of the MGP laboratory (e.g., normally is not the case. Individuals do typically interact
Dovidio, Mann, & Gaertner, 1989; Gaertner & Dovidio, with one another and information about group mem-
1986; Murrell, Dietz-Uhler, Dovidio, Gaertner, & Drout, bership often is readily available, if not blatantly appar-
1994). According to the aversive racism perspective, ent (as with gender or race). In such situations, account-
many White people in the United States today readily ability is usually inferred to some degree and, as a result,
express a belief in egalitarianism, fairness, and equality. whereas one’s ingroup may be privately favored, the
In some situations, however, underlying prejudice behavioral manifestations of that favoritism may be sub-
against Black people is manifest. Such discrimination tle or less likely to emerge (Dovidio & Gaertner, 1996).
occurs as a result of ambivalence on the part of Whites: a Whereas accountability to an outgroup member may
conflicting desire to conform outwardly to socially cause people to think about the expression of their atti-
accepted norms of fairness and equality between races tudes and actions more carefully, and not necessarily
and a tendency to view Black people in a prejudicial suppress them, the more thoughtful consideration of
manner as a result of various cognitive, historical, and group bias may restrain obvious favoritism or
sociocultural factors. Aversive racists avoid overt displays discrimination.
of racial favoritism when the normative structure of a sit- In this way, accountability can also be viewed as simi-
uation is clear, but when the situation is normatively lar to the concept of identifiability (to the ingroup or the
ambiguous, racial discrimination may occur (Gaertner & outgroup), as considered in the Social Identity
Dovidio, 1986). Deindividuation (SIDE) Model (Spears & Lea, 1994).
Although aversive racism studies have focused on race Within SIDE, identifiability is related to an individual’s
as the basis of intergroup distinction, the logical frame- freedom to express the contents of a salient identity.
work of this perspective lends itself to a broader interpre- However, the issue of to whom the individual is account-
tation, as Crocker, Major, and Steele (1998) have able is an important moderator in this context. Whereas
observed. When a situation is normatively ambiguous or accountability to ingroup members may result in height-
anonymous, fairness pressures are absent and familiar ened conformity to ingroup norms and a more devout

Downloaded from at The University of Iowa Libraries on March 12, 2015

adherence to ingroup identity maintenance strategies the group as a minority, in competition with the larger
(Noel, Wann, & Branscombe, 1995), being accountable collective. Otten et al. (1996) suggested that low compe-
to outgroup members may suppress the expression of tence and minority status constitute threats to social
ingroup-favoring actions. In this study, we consider the identity. Accordingly, participants under such threats
latter. Reicher and Levine (1994) showed that individu- may be prone to discrimination against an outgroup.
als are more likely to endorse attitudes and behaviors of The aversive discrimination approach makes no
an outgroup when in the presence of, and identifiable strong prediction for competence or group status
by, a member of this group, especially when under the (main) effects. However, it fosters a prediction of an
threat of punishment for holding a counterattitudinal interaction of group status with outgroup accountability;
position. Thus, highly identifiable individuals “may be discrimination should be exacerbated among un-
more inhibited to engage in behavior that is proscribed accountable majority members. The approach does not
by or detrimental to the outgroup for fear of disapproval predict this same discriminatory behavior for anony-
or sanction” (Spears & Lea, 1994, p. 447). It follows, mous members of the minority.
then, that introducing outgroup accountability in the The final possibility to be considered involves the
MGP may attenuate or conceal ingroup favoritism, just interaction of competence and group size. The aversive
as introducing negative allocations did in discrimination model does not make a clear prediction
Mummendey’s research. However, giving people a justi- in this instance. SIT, however, holds that the minority
fiable reason to favor or defend their own ingroup may tendency to discriminate against the majority will be
cause discrimination to become manifest, consistent exacerbated when their competence is threatened. The
with Mummendey’s hypothesis. double threat to identity in the minority-incompetent
condition of the study should pose the most extreme
Central Variables and Hypotheses menace to social identity. Thus, nonaccountable partici-
pants classified as members of incompetent minority
This experiment compares the effects of a typical min-
groups should exhibit the greatest discriminatory
imal group treatment with one that adds a layer of
accountability, whereby participants anticipate that they
will be required to justify their allocation choices to
outgroup members. In addition to accountability, the METHOD

research makes use of two other treatment variables, Participants

minority/majority group status and task competence,
that have relevance to a comparison of SIT, the positive- Ninety-seven undergraduate communication stu-
negative asymmetry hypothesis, and aversive discrimina- dents at the University of Arizona served as participants
tion theory. The manipulated variables promote different and were given extra credit and a small monetary reward
predictions and facilitate comparison of the competing for their service. The study was conducted in groups that
models. ranged in size from 10 to 21 participants.
Predictions regarding the effects of accountability on
discrimination are not readily derived from either SIT or
aversive discrimination theory. However, as will be Participants were randomly assigned to conditions in
argued, aversive discrimination does suggest a possible a 2 (accountability vs. no accountability) × 2 (high com-
interaction of outgroup accountability with group status. petence vs. low competence) × 2 (majority vs. minority)
A main effect of accountability can be derived from factorial design. Effects of the independent variables on
Mummendey’s positive-negative asymmetry hypothesis monetary allocations, the dependent measure, were
if we assume that outgroup accountability motivates assessed via Tajfel matrices (see Allen & Wilder, 1975;
more careful thought and that this added consideration Billig & Tajfel, 1973; Bourhis, Sachdev, & Gagnon, 1994;
is a driving force behind the allocation asymmetries that Diehl, 1988; Otten et al., 1996; Tajfel et al., 1971). These
she and her colleagues have discovered (Buhl, 1999). matrices are widely used; their application in the present
As noted, SIT holds that group factors that affect research allows comparison of results with previous min-
self-evaluations will have an important bearing on dis- imal group studies.
crimination. Thus, an individual whose ingroup is
threatened in some way is expected to reassert a positive
social identity by devaluating or disadvantaging a com- After arriving, participants were seated individually at
peting outgroup. One way a group may be threatened is desks that were separated by shoulder-high dividers.
by showing that its members are incompetent vis-à-vis a They could not see the other participants. To create a
comparison group. Another way may be to characterize basis for dividing participants into groups, they com-

Downloaded from at The University of Iowa Libraries on March 12, 2015

pleted a trivial line estimation task. Although group One matrix type pits the FAV strategy (a combination
assignment was random, participants were led to of the MD and MIP strategies) against the MJP strategy.
believe the division was based on overestimation or under- The two remaining matrix types pit FAV against P and
estimation of the line lengths. All were given feedback MD against MIP + MJP.
concerning their individual results on the line estima- Participants always awarded points to two other anon-
tion task as well as their membership in either the ymous players, never directly to themselves. They were
overestimator or underestimator group. In addition, the told they would receive the amount two other (anony-
cover passage supplied participants with their group mous) participants awarded them. Placement of in- and
member identification letter (all participants were listed outgroup awardees was counterbalanced in the matri-
as member “N”). ces, and the columns were reversed for each of the vari-
Participants were led to believe that their ingroup was ous matrices, resulting in 3 (matrix types) × 2 (ingroup
of either high or low competence in line estimation. This or outgroup in top row position) × 2 (column reversal)
ability was defined as characteristic of overestimators or matrices. In all, participants completed 12 matrices.
underestimators. They were told that their group tended In the accountability condition, participants were
to be more (or less) accurate judges in perceptual tasks, told to wear their name tag sticker and that after they
such as line estimation. In addition, they were told that filled out their booklets, they would discuss their alloca-
their group represented a numerical majority (“80% of tion decisions with two members of the outgroup. This
the general population is also an over/underestimator”) manipulation formed the central accountability factor.
or minority (“20% of the general population . . . ”). In the standard (control) condition, no mention was
When participants finished reading the cover page, made of other group members or revealing group mem-
they were instructed to answer 10 items taken from bership during the experiment itself.
Heatherton and Polivy’s (1991) State Self-Esteem Scale When all participants finished the matrices, the book-
(SSES) and then to read, as the experimenter carefully lets were collected and participants were carefully
recited aloud, the instructions for completing the alloca- debriefed. In the interest of fairness, each participant
tion matrices that followed. As a manipulation check, received $1.
each participant was asked to write his or her group
name and group membership identification letter on RESULTS
the next page before continuing on to the matrices. In
the accountability condition, participants also wrote Manipulation Checks
their group name and participant identification letter, as
The manipulation check data reveal that all partici-
well as their real name, on a name tag sticker that was
pants, save one, correctly identified their group mem-
clipped to the booklet. These outgroup-accountable
bership. The errant individual’s data were excluded
participants were told they would later meet with mem-
from further analysis. A second analysis was conducted
bers of the opposite group.
to determine the effect of the manipulations on partici-
When completing the allocation matrices, partici-
pants’ self-esteem. Heatherton and Polivy’s (1991) SSES
pants apportioned points (supposedly worth 25 cents
purports to assess momentary instantiations of an indi-
each) between an anonymous member of their own
vidual’s overall well-being. The 10 items taken from the
ingroup and an anonymous member of the outgroup.
SSES were combined (α = .76) and analyzed in a 2
Each matrix pits particular arrangements of allocation
(outgroup accountability: present/ absent) × 2 (relative
strategies against each other, as follows:
ingroup size: minority/majority) × 2 (ingroup compe-
tence: high/low) ANOVA. The analysis revealed a main
• maximum ingroup profit (MIP), in which the most pos- effect for group size, F(1, 89) = 3.66, p < .06; majority par-
sible points are awarded to the ingroup member, regard- ticipants displayed higher self-esteem scores than those
less of outgroup points; assigned to the minority condition (X = 40.56 and 38.48,
• maximum joint profit (MJP), in which the most total respectively). The analysis also revealed a statistically sig-
points is awarded to both ingroup and outgroup mem- nificant Accountability × Competence interaction, F(1,
ber, regardless of which receives more; 89) = 5.80, p < .02; participants in the high-competent
• maximum difference in favor of ingroup (MD), in which
group who were accountable for their actions to
the difference between amount awarded to ingroup
member and amount awarded to outgroup member is
outgroup members displayed higher self-esteem than
the greatest, with the ingroup member receiving more; incompetent, but accountable, participants (X = 41.43 vs.
• favoritism (FAV), which is a combination of the MIP and 38.19, respectively), F(1, 89) = 4.49, p < .05. These
MD strategies; and results suggest that participants in the minority condi-
• parity (P), in which each person receives an equal tion, or who were both accountable and a member of an
amount of money. incompetent group, experienced identity threat. The

Downloaded from at The University of Iowa Libraries on March 12, 2015

experimental treatments generally functioned as TABLE 1: Mean Pull Scores for Each Strategy Across All Participants
planned. Strategy Orientation M (SD) t Value z Value

Allocation Analyses FAV on MJP 1.98 (4.20) 4.65*** 4.46***

MD on MIP + MJP 1.80 (3.85) 4.61*** 3.98***
Allocation decisions on the Tajfel matrices were ana- FAV on P 1.53 (4.06) 3.70*** 3.54*
lyzed using pull scores, which measure participants’ MJP on FAV 0.21 (3.81) 0.55 0.43
preferences of one allocation strategy over another (see MIP + MJP on MD 1.27 (5.14) 2.43* 2.34***
Bourhis et al., 1994). Six pull scores were calculated for P on FAV 5.77 (5.09) 11.18*** 7.39***
each participant: the pull of FAV on MJP, MD on MIP + NOTE: t tests represent difference from a hypothetical mean of zero; Z
MJP, FAV on P, MJP on FAV, MIP + MJP on MD, and P on scores obtained via Wilcoxon ranked pairs test compare the two rank
FAV. The first three of these strategies favor the ingroup components of each pull score. Significant values suggest that the strat-
egy was used at greater than chance levels. FAV = ingroup favoritism,
over the outgroup. The remaining strategies (MJP on MJP = maximum joint profit, MD = maximum difference in favor of
FAV, MIP + MJP on MD, and P on FAV) represent deci- ingroup, MIP = maximum ingroup profit, and P = parity.
sions that are more fair and economically rational. In *p < .05. ***p < .001.
testing pull score data, Bourhis et al. (1994) suggest both
within- and between-subjects analytic designs be used to shows that participants in the accountable majority con-
provide a complete data depiction. In the within-subjects ditions did not make use of any of the discriminatory
analysis, each pull score is examined via a t test, which strategies. In the nonaccountable conditions, each sub-
pits the mean pull against a hypothetical mean of zero. group employed at least one of the discriminatory strate-
This analysis indicates intensity of a given strategy in the gies. A more precise parametric appraisal of the effects
reward allocation process. Next, effects of the manipu- of the manipulated variables on variations in distribu-
lated between-groups variables are investigated for those tion behaviors is provided in the between-subjects analy-
strategies whose pull scores exceed zero. In the present ses that follow.
study, this analysis will require two 2 (accountable/not
accountable) × 2 (relative ingroup size: majority vs. Between-Subject Analyses:
minority) × 2 (group competence: high vs. low) Discriminatory Distribution Strategies
MANOVAs on the combined discriminatory and the Pull scores on the three discriminatory distribution
combined nondiscriminatory strategies. strategies (FAV on MJP, MD on MIP + MJP, and FAV on P)
Within-Subject Analyses were analyzed in a 2 (outgroup accountability) × 2 (rela-
tive ingroup size) × 2 (ingroup competence) MANOVA.
Each of the six allocation strategies was analyzed via The MANOVA revealed a significant main effect of
one-sample t tests. This test was used to determine the accountability, F(3, 87) = 2.71, p < .05. This result sug-
difference between the observed mean pull score on gests the general tendency of outgroup-accountable par-
each strategy and a hypothetical value of zero, which sug- ticipants to discriminate less across the three strategies
gests the absence of any appreciable pull. The results of than those not held accountable to the outgroup.
these analyses are presented in Table 1. As shown, signifi- Univariate ANOVAs on each of the three discriminatory
cant pull scores were found for five of the six allocation strategies disclosed a significant main effect of outgroup
strategies. Only the maximum joint profit on favoritism accountability for the FAV on MJP strategy, F(1, 89) =
(MJP on FAV) pull score, an egalitarian strategy, was not 4.49, p < .05. Participants who were made accountable to
significantly different from zero. These results suggest the outgroup for their allocations were significantly
that with one exception, participants employed both dis- less discriminatory than were nonaccountable partici-
criminatory and egalitarian strategies in the allocation of pants (M pull = 1.13 vs. 2.94 for accountable and
resources. nonaccountable participants, respectively; higher scores
A second within-subjects analysis also was calculated. indicate greater discriminatory pull). Neither the com-
In this (nonparametric) analysis, the two rank compo- petence nor the group status manipulation produced
nents of the pull scores were compared via Wilcoxon significant univariate main effects.
ranked pairs tests. The results of these nonparametric The MANOVA on the discriminatory strategies also
comparisons are presented in the final column of Table 1. disclosed two substantial first-order interactions. The
As shown, both the Wilcoxon and the one-sample in- first was the Multivariate Outgroup Accountability × Rel-
dependent t-test analyses yielded essentially identical ative Ingroup Size interaction, F(3, 87) = 2.18, p < .10.
results. Following previous analyses (e.g., Mullin & Univariate ANOVAs revealed significant Accountability ×
Hogg, 1998; Otten et al., 1996), Table 2 summarizes the Group Status interactions for both FAV on P and MD on
mean pull score differences (from zero) within each of MIP + MJP, F(1, 89) = 4.12 and 5.30, respectively, both ps <
the extended experimental treatment cells. The table .05. These interactions, presented in Figure 1, show that

Downloaded from at The University of Iowa Libraries on March 12, 2015

TABLE 2: Mean Pull Scores by Accountability, Relative Ingroup Size, and Relative Group Status

Minority Majority
Allocation Strategy Low Status High Status Low Status High Status

Not accountable
FAV on MJP 1.50 (3.52)* 3.15 (6.25) 3.41 (5.06)* 3.68 (4.33)*
MD on MIP + MJP –0.38 (2.51) 3.19 (5.02)* 4.00 (4.42)* 2.05 (2.23)*
FAV on P .92 (2.54) 1.15 (5.84) 2.05 (5.02) 2.45 (4.90)
MJP on FAV –0.73 (5.45) 0.54 (2.87) 1.77 (6.47) 0.04 (2.64)
MIP + MJP on MD 2.23 (6.98) 1.88 (5.68) 1.45 (6.30) 1.95 (6.01)
P on FAV 8.31 (3.95)*** 2.85 (5.56) 3.86 (7.07) 6.64 (4.84)**
FAV on MJP 0.32 (2.93) 2.32 (3.70) 0.46 (3.90) 1.42 (2.41)
MD on MIP + MJP 2.14 (4.10) 2.86 (3.26)* 0.82 (3.88) 0.42 (3.41)
FAV on P 3.64 (3.54)** 1.68 (4.00) 0.11 (3.52) 0.88 (2.13)
MJP on FAV –0.77 (4.20) 0.50 (1.25) 0.61 (2.63) –0.19 (3.05)
MIP + MJP on MD 0.86 (4.39) 1.68 (5.33) 0.39 (3.31) –0.12 (3.25)
P on FAV 5.91 (3.77)*** 6.68 (4.82)** 5.54 (4.21)*** 6.42 (5.24)**

NOTE: Standard deviations are in parentheses. Pull scores for each allocation strategy can range from –12 to +12. FAV = ingroup favoritism, MJP =
maximum joint profit, MD = maximum difference in favor of ingroup, MIP = maximum ingroup profit, and P = parity. Cell ns range from 11 to 14.
*p < .05. **p < .01. ***p < .001, two-tailed t tests.

majority participants were significantly less discrimina-

tory when held accountable to the outgroup (minority)
for both the FAV on P, t(95) = 1.50, p < .07, and the MD on
MIP + MJP strategies, t(95) = 2.24, p < .025. This pattern
was not found among minority allocators.
The Relative Group Size × Competence interaction
was the final substantial multivariate effect found in the
MANOVA on discriminatory strategies, F(3, 87) = 4.51,
p < .005. Univariate ANOVAs on the three strategies
revealed that this interaction was significant for the MD
on MIP + MJP strategy, F(1, 89) = 4.81, p < .05. Inspection
of this interaction, graphically depicted in Figure 2,
reveals a set of results that, although unanticipated,
may prove useful for theory development. The interac-
tion suggests that minority participants discriminated
more than majority when both were members of high-
competence groups, t(95) = 1.66, p < .05. However, in the
low-competence condition, minority participants
tended to discriminate less than the majority, t(95) =
1.44, p < .08.
The MANOVA on the egalitarian/economic strate-
gies (MJP on FAV, MIP + MJP on MD, and P on FAV)
revealed no significant multivariate main effects or


The initial within-subjects analyses demonstrated that

the participants made substantial use of both discrimina-
tory and egalitarian strategies. However, outgroup Figure 1 Outgroup Accountability × Relative Group Size interac-
accountability impeded discrimination, especially tion: Pull of FAV on P (top view) and MD on MIP + MJP
among members of the majority, a finding anticipated by (bottom view).
NOTE: Higher scores indicate greater discrimination. FAV = ingroup
the aversive discrimination model. Multivariate analysis favoritism, P = parity, MD = maximum difference in favor of ingroup,
of the three discrimination strategies confirmed the MIP = maximum ingroup profit, and MJP = maximum joint profit.

Downloaded from at The University of Iowa Libraries on March 12, 2015

approach holds that White aversive racists are mindful of

the appropriateness of their behavior in situations that
feature clear normative guidelines, but this mindful con-
sideration is attenuated in situations that are less well
defined. Finally, it is possible that outgroup-accountable
participants become repersonalized, moving from
group-level identification to the individual level. In a
mindful, normatively unambiguous, personalized con-
text, the complementary tendencies to identify with an
ingroup and discriminate against an outgroup would
diminish. Although these possibilities were not tested
directly, the results suggest provocative possibilities that,
at a minimum, encourage future research on the mecha-
nisms responsible for accountability effects.
Whereas both positive-negative asymmetry and
aversive racism produce predictions regarding the main
Figure 2 Ingroup Competence × Relative Group Size interaction:
Pull of MD on MIP + MJP. effect of accountability, SIT does not. However, SIT does
NOTE: Higher scores indicate greater discrimination. MD = maximum form the basis for predictions of the main effects of both
difference in favor of ingroup, MIP = maximum ingroup profit, and relative group size and group competence on alloca-
MJP = maximum joint profit.
tions. The theory fosters the prediction that an individ-
ual characterized as belonging to an incompetent
expectation that participants held accountable to the group, or thrust unexpectedly into minority status, may
outgroup for their decisions show less ingroup favorit- suffer a threat to social identity. Ellemers, Van Rijswijk,
ism than those not held accountable. Univariate analysis Roefs, and Simons (1997) point out that the degree of
revealed that this result held only for one of the three dis- perceived stability and legitimacy in an intergroup situa-
criminatory strategies (FAV on MJP), but among these tion featuring groups of unequal status may moderate
strategies, FAV on MJP is arguably the most self-serving. the perception of threat. As such, membership in a
At one end of the FAV on MJP matrix, one can obtain the low-status group (i.e., minority and/or low competence)
highest ingroup profit for one’s group along with the does not automatically constitute a social identity threat.
maximum positive between-groups differential (the FAV However, our self-esteem data show that participants in
strategy). This outcome contrasts with an outgroup- this study did perceive such a threat when in a minority
favoring allocation at the other end of the choice matrix. or an accountable, low-competence group. This should
The realistic choice for even mildly discriminatory par- prompt the individual to discriminate more than a per-
ticipants appears obvious in this case. The proper choice son of majority status and/or high competence. Con-
is not so evident in the other discriminatory matrices. trary to these expectations, the analysis revealed no sig-
For example, to generate a positive differential in the nificant multivariate or univariate main effects for either
MD on MIP + MJP matrix, the discriminating allocator is group competence or group status.
forced to take less, whereas FAV on P demands a choice The absence of group competence and relative group
of clear favoritism over equality rather than outgroup size main effects does not mean that the variables had no
favoritism. Thus, whereas a significant pull score for any impact on allocation behavior. They did, but their
of the discriminatory strategies indicates bias, the FAV impact was moderated, as shown in the significant inter-
on MJP strategy may represent the most obvious form of action effects obtained. The aversive discrimination per-
favoritism available. This is consistent with Buhl’s (1999) spective fosters the expectation that majority-status allo-
approach, which used only the FAV on MJP strategy cators will discriminate less when held accountable to
when agglomerating results of earlier selected MGP the outgroup for their actions. SIT makes a different pre-
studies in his meta-analytic summary. diction: Experiencing a threatened social identity, mem-
Three possibilities suggest themselves as responsible bers of a minority group should discriminate more in gen-
for the outgroup accountability effect. Consistent with eral, regardless of accountability. The positive-negative
Mummendey’s (1995) reasoning, those who are aware asymmetry hypothesis would seem to support a similar
that they will be held accountable for, and possibly asked (main effect) prediction. The results provide strong sup-
to defend, their allocation behavior to outgroup mem- port for the aversive discrimination prediction: Account-
bers may adopt a more careful thought-processing style. ability interacted with group status. Majority allocators
The accountability results are consistent, too, with discriminated more when not held accountable to the
expectations based on aversive racism theory. This outgroup for their actions, whereas minority members

Downloaded from at The University of Iowa Libraries on March 12, 2015

did not discriminate differentially on the basis of attenuation of discrimination. Although this hypothesis
outgroup accountability. In short, consistent with was not directly tested here, research suggests that the
aversive racism, in the absence of normative guidelines, accountability construct represents a more direct link-
majorities discriminated; when made accountable to the age to careful thought than allocation valence
outgroup for their actions, majority bias was attenuated. (Cvetkovich, 1978; Tetlock, 1992).
Similar accountability variations on discrimination were
not found in members of minority status. This pattern of Conclusions
findings fits well with expectations based on aversive rac- The results of this experiment suggest that the tradi-
ism; it is less successfully anticipated by the other theoret- tional conceptualization of a minimal group is at least
ical orientations considered to this point. partially misleading. Instead of establishing a
The lack of main effects of relative group size and ground-level intergroup situation, the MGP appears to
competence was contrary to expectations based on SIT. present respondents with a scenario in which discrimi-
The particular form of the interaction of these two vari- nation is fostered, so that adding certain layers of real
ables also was unanticipated. It was expected that partici- group characteristics (in this case, outgroup account-
pants in the minority/incompetent group treatment ability) to the paradigm can result in less, rather than
condition would experience the greatest identity threat more, discrimination. Thus, when Tajfel and his asso-
and, thus, the greatest need to discriminate against the ciates attempted to remove these intergroup charac-
outgroup (in this case, the majority). Such discrimina- teristics, they were, in fact, unwittingly adding certain
tion was not evident. In fact, the opposite was found. elements, such as positive reward allocations and
When the minority was competent, it tended to discrimi- n o n a c c o u n t a b i l i t y, w h i c h s e r v e t o p r o m o t e
nate more than the majority. When informed that they discrimination.
were part of an incompetent group, minority members Our findings suggest that it might be fruitful to take a
tended to discriminate less than their majority counter- closer look at the MGP. Arguably, a central goal of
parts in their allocation behavior. These findings lend lit- research in intergroup relations is to understand the
tle support for the standard SIT approach. Of course, it processes at work in negative social interactions, to
could be argued that the theory does not necessarily understand how and why we discriminate, so that harm-
require that minority or members of incompetent ful real-world discrimination might be reduced. By ana-
groups discriminate against the outgroup. Other mecha- lyzing behavior through the MGP, a technique funda-
nisms of self-enhancement may be employed. The rever- mentally characterized by trivial criteria and unrealistic
sal of the predicted results, however, is less comforting conditions, we may be limiting ourselves in the pursuit of
than a mere null finding and should motivate serious this important understanding. The baseline conditions
consideration. established in the early minimal group experiments may
Accountability Effects and the MGP have led to a baseline understanding that may be unreal-
istic. By systematically moving toward more realistic
In examining the results of this experiment, it is groups and examining various group characteristics
apparent that accountability played a powerful role in across studies, the knowledge we already have can be
participants’ allocation decisions. Consistent with our combined with new discoveries about the effects of the
expectations, the findings indicated that discrimination various components of group membership and the psy-
was attenuated when participants felt they would have to chological, behavioral, and communicative effects they
justify their allocation decisions to outgroup members. have both in isolation and in interaction with each other.
Such results support the hypothesis initially proposed by In this way, existing theories might be refined and
Mummendey and her associates that discrimination in expanded to better predict and explain intergroup
terms of a mere categorization effect may obtain only relations.
under certain conditions that incidentally happened to The direction of research being advocated is neither
be present in the classic minimal group studies. When meant to diminish SIT nor to discount the role of SIT
certain factors or layers (such as negative allocations or, and the MGP in furthering understanding and promot-
in the present case, accountability) are added to the min- ing research. Rather, it is suggested that the considerable
imal intergroup situation, discrimination disappears. energies that have been devoted to these issues be har-
Thus, this experiment may have uncovered a new kind of nessed and perhaps redirected so that the ultimate aim
inconsistency effect that, in keeping with Mummendey’s of the field may be achieved. If studied logically and sys-
(1995) nomenclature, can be thought of as an “account- tematically, the utility and possible integration of other
ability asymmetry effect.” The accountability findings empirically supported perspectives such as aversive dis-
are consistent with Mummendey’s speculation that crimination and the positive-negative asymmetry effect
more careful thought is at least partly responsible for the can lead to a better understanding of intergroup rela-

Downloaded from at The University of Iowa Libraries on March 12, 2015

tions. Tajfel (1978) once referred to the original mini- Mullin, B., & Hogg, M. A. (1998). Dimensions of subjective uncertainty
in social identification and minimal intergroup discrimination.
mal group studies as “crutches” on which to base further British Journal of Social Psychology, 37, 345-365.
thinking (p. 77). It seems as though research since his Mummendey, A. (1995). Positive distinctiveness and social discrimina-
death has largely forgotten his important advice. It is tion: An old couple living in divorce. European Journal of Social Psy-
chology, 25, 671-688.
hoped that the results of this experiment will serve as a Mummendey, A., Simon, B., Dietze, C., Grunert, M., Haeger, G.,
reminder of his wise observation. Kessler, S., Lettgen, S., & Schaferhoff, S. (1992). Categorization is
not enough: Intergroup discrimination in negative outcome allo-
cation. Journal of Experimental Social Psychology, 28, 125-144.
REFERENCES Murrell, A. J., Dietz-Uhler, B. L., Dovidio, J. F., Gaertner, S. L., &
Drout, C. (1994). Aversive racism and resistance to affirmative
Allen, V. L., & Wilder, D. A. (1975). Categorization, belief similarity, action: Perceptions of justice are not necessarily color blind. Basic
and group discrimination. Journal of Personality and Social Psychology, and Applied Social Psychology, 15, 71-86.
32, 971-977. Noel, J. G., Wann, D. L., & Branscombe, N. R. (1995). Peripheral
Billig, M. G., & Tajfel, H. (1973). Social categorization and similarity in ingroup membership status and public negativity toward
intergroup behaviour. European Journal of Social Psychology, 3, 27-52. outgroups. Journal of Personality and Social Psychology, 68, 127-137.
Blanz, M., Mummendey, A., & Otten, S. (1995). Perceptions of relative Oakes, P. J., Haslam, S. A., & Turner, J. C. (1994). Stereotyping and social
group size and relative group status on intergroup discrimination reality. Oxford, UK: Blackwell.
in negative evaluations. European Journal of Social Psychology, 25, Otten, S., Mummendey, A., & Blanz, M. (1996). Intergroup discrimina-
231-247. tion in positive and negative outcome allocations: Impact of stimu-
Bourhis, R. Y., Sachdev, I., & Gagnon, A. (1994). Intergroup research lus valence, relative group status, and relative group size. Personality
with the Tajfel matrices: Methodological notes. In M. P. Zanna & J. M. and Social Psychology Bulletin, 22, 568-581.
Olson (Eds.), The psychology of prejudice: The Ontario symposium Reicher, S., & Levine, M. (1994). On the consequences of
(Vol. 7). Hillsdale, NJ: Lawrence Erlbaum. deindividuation manipulations for the strategic communication of
Buhl, T. (1999). Positive-negative asymmetry in social discrimination: self: Identifiability and the presentation of social identity. European
Meta-analytical evidence. Group Processes and Intergroup Relations, 2, Journal of Social Psychology, 24, 511-524.
51-58. Rokeach, M. (1960). The open and closed mind. New York: Basic Books.
Campbell, D. T. (1965). Ethnocentric and other altruistic motives. In Sherif, M. (1966). Group conflict and co-operation: Their social psychology.
D. Levine (Ed.), Nebraska symposium on motivation (Vol. 13). Lincoln: London: Routledge Kegan Paul.
University of Nebraska Press. Sherif, M., Harvey, O. J., White, B. J., Hood, W. R., & Sherif, C. W.
Crocker, J., Major, B., & Steele, C. (1998). Social stigma. In D. T. (1961). Intergroup conflict and cooperation: The Robber’s Cave experi-
Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychol- ment. Norman: University of Oklahoma Book Exchange.
ogy (4th ed., Vol. 2, pp. 504-553). Boston: McGraw-Hill. Spears, R., & Lea, M. (1994). Panacea or panopticon? The hidden
Cvetkovitch, G. (1978). Cognitive accommodation, language, and power in computer-mediated communication. Communication
social responsibility. Social Psychology, 41, 149-155. Research, 21, 427-459.
Diehl, M. (1988). Social identity and minimal groups: The effects of Tajfel, H. (1970). Experiments in intergroup discrimination. Scientific
interpersonal and intergroup attitudinal similarity on intergroup American, 223, 96-102.
discrimination. British Journal of Social Psychology, 27, 289-300. Tajfel, H. (1974). Social identity and intergroup behavior. Social Science
Diehl, M. (1989). Justice and discrimination between minimal groups: Information, 13, 65-93.
The limits of equity. British Journal of Social Psychology, 28, 227-238. Tajfel, H. (1978). Social categorization, social identity, and social com-
Diehl, M. (1990). The minimal group paradigm: Theoretical explana- parison. In H. Tajfel (Ed.), Differentiation between social groups
tions and empirical findings. In W. Stroebe & M. Hewstone (Eds.), (pp. 61-76). New York: Academic Press.
European review of social psychology (Vol. 1). Chichester, UK: Wiley. Tajfel, H., Billig, M. G., Bundy, R. F., & Flament, C. (1971). Social cate-
Doise, W. (1978). Groups and individuals: Explanations in social psychol- gorization and intergroup behavior. European Journal of Social Psy-
ogy. Cambridge, UK: Cambridge University Press. chology, 1, 149-177.
Dovidio, J. F. (in press). Combating bias at its roots: Individual and Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup
intergroup approaches. In S. Oskamp (Ed.), Reducing prejudice and conflict. In W. G. Austin & S. Worchel (Eds.), The social psychology of
discrimination. Mahwah, NJ: Lawrence Erlbaum. intergroup relations (pp. 33-47). Monterey, CA: Brooks/Cole.
Dovidio, J. F., & Gaertner, S. L. (1996). Affirmative action, uninten- Tajfel, H., & Turner, J. C. (1986). The social identity theory of inter-
tional racial biases, and intergroup relations. Journal of Social Issues, group behavior. In S. Worchel & W. G. Austin (Eds.), Psychology of
52, 51-75. intergroup relations (pp. 7-24). Chicago: Nelson-Hall.
Dovidio, J. F., Mann, J. A., & Gaertner, S. L. (1989). Resistance to affir- Tetlock, P. E. (1983). Accountability and complexity of thought. Jour-
mative action: The implication of aversive racism. In F. A. nal of Personality and Social Psychology, 45, 74-83.
Blanchard & F. J. Crosby (Eds.), Affirmative action in perspective Tetlock, P. E. (1992). The impact of accountability on judgment and
(pp. 83-102). New York: Springer-Verlag. choice: Toward a social contingency model. In M. P. Zanna (Ed.),
Ellemers, N., Van Rijswijk, W., Roefs, M., & Simons, C. (1997). Bias in Advances in experimental social psychology (Vol. 25, pp. 331-376). San
intergroup perceptions: Balancing group identity with social real- Diego, CA: Academic Press.
ity. Personality and Social Psychology Bulletin, 23, 186-198. Turner, J. C. (1975). Social comparison and social identity: Some pros-
Gaertner, S. L., & Dovidio, J. F. (Eds.). (1986). The aversive form of rac- pects for intergroup behavior. European Journal of Social Psychology, 5,
ism. In Prejudice, discrimination, and racism (pp. 61-89). Orlando, FL: 5-34.
Academic Press. Walster, E., Walster, G. W., & Berscheid, E. (1978). Equity: Theory and
Heatherton, T. F., & Polivy, J. (1991). Development and validation of a research. Boston: Allyn and Bacon.
scale for measuring state self-esteem. Journal of Personality and Social
Psychology, 60, 895-910.
Received May 24, 1999
Revision accepted March 8, 2000

Downloaded from at The University of Iowa Libraries on March 12, 2015