You are on page 1of 7

Journal of Personality and Social Psychology Copyright 2000 by the American Psychological Association, Inc.

2000, Vol. 78, No. 5, 906-912 0022-3514/00/$5.00 DOI: 10.1037//0022-3514.78.5.906

Pro-Norm and Anti-Norm Deviance Within and Between Groups

Dominic Abrams Jos6 M. Marques


University of Kent at Canterbury University of Porto

Nicola Bown Michelle Henson


University of Leeds University of Kent at Canterbury

Participants evaluated other individuals who deviated in either an anti- or pro-normative direction relative
to normative members. In Study 1, in-group gender-normative members were rated more positively than
deviant members. The pro-norm deviant was viewed as more attractive than the anti-norm deviant. In
Study 2 anti-norm in-group deviants were evaluated more negatively than anti-norm out-group deviants
even though both held identical attitudes. In both studies, despite objective equivalence, pro-norm
deviance was perceived as less "atypical" than anti-norm deviance. Judgments and reactions to deviance
depend on group membership and the direction of deviance, not just its magnitude. Evaluations of
deviants are also related to perceivers' identification with their own group. These findings are consistent
with our model of subjective group dynamics.

Research on group deviance usually focuses on group members would be ambiguous for other home team supporters to know
whose behavior or characteristics reduce the group's distinctive- when to cheer. Conversely, some home team supporters might
ness, contradict its stereotype, or threaten its goals (Levine, 1989). vehemently deny that their team player committed a foul, even
According to self-categorization theory, in-group norms embody though the evidence was clear that the foul had been committed.
the behaviors or features that members should adopt and that Supporting the opposition and excessive in-group support may
maintain distinctiveness from the out-group (Turner, Hogg, takes, both be viewed as highly atypical, but they are likely to carry
Reicher, & Wetherell, 1987). In-group members who do not follow different connotations for the value of social identity (cf. Sherman,
such norms will reduce the clarity of intergroup boundaries and, Hamilton, & Lewis, 1999) for home team supporters. One form of
hence, should attract less positive evaluations than more normative deviance may invite a negative reaction, the other might attract
members. However, deviants who are pro-normative should be tacit approval.
judged as more prototypical than those who are anti-normative, In the present article, we investigate how individuals react when
because the former are more distant from the out-group. Therefore prescriptive norm deviation occurs in a direction that is inconsis-
the perceived prototypicality of pro-norm deviants should be tent with (anti) or consistent with (pro) the prevailing norm for the
heightened when intergroup comparisons are salient. target group. We define pro-norm deviants as group members who
Marques, Abrams, Paez, and Martinez-Taboada (1998) distin- deviate from their group's norm by holding an extreme position
guished between descriptive in-group norms and prescriptive that, nonetheless, supports the group's aims or ethos. We define
norms. Descriptive norm deviance simply reduces the clarity of anti-norm deviants as group members who reject their group's
group prototypes and thus of intergroup boundaries. For example, norms and/or who favor the normative attitudes or behavior of a
if half the spectators at a football game chose to wear the colors of contrasting group.
neither team it would be less easy to distinguish between rival Social identity theory (e.g., Tajfel, 1981), holds that people are
supporter groups. Prescriptive norm deviance has implications motivated to sustain a positive social identity. Marques et al.'s
both for distinctiveness and for the legitimacy of positive evalua- (1998) model of subjective group dynamics holds that one way
tions of the in-group because it threatens the validity of the group members do this is by maintaining the subjective validity
in-group consensus. For example, if some people who were wear- and legitimacy of in-group norms. We define subjective group
ing home team colors cheered when the opposition played well, it dynamics as a process by which people maximize and sustain
descriptive intergroup differentiation while simultaneously maxi-
mizing and sustaining the relative validity of prescriptive in-group
Dominic Abrams and Michelle Henson, Department of Psychology, norms through intragroup differentiation.
University of Kent at Canterbury, Canterbury, United Kingdom; Jose M. Relative validity is increased by evidence supporting the in-
Marques, Department of Psychology and Educational Sciences, University group norm relative to evidence that supports contrasting norms.
of Porto, Porto, Portugal; and Nicola Bown, Leeds University Business An interesting consequence of subjective group dynamics is that
School, University of Leeds, Leeds, united Kingdom.
people should like out-group members whose relative support for
Correspondence concerning this article should be addressed to Dominic
Abrams, Department of Psychology, University of Kent at Canterbury, in-group norms boosts relative validity, whereas they should dis-
Canterbury CT2 7NP, United Kingdom, or to Jos6 M. Marques, FPCE-UP, like in-group members whose relative rejection of in-group norms
R. do Camp Alegre, 1055, P-4150 Porto, Portugal. Electronic mail may be undermines relative validity. In fact, in-group--favoring deviants
sent to d.abrams@ukc.ac.uk or to marques@psi.up.pt. should be favored more than in-group-rejecting deviants, even

906
PRO-NORM AND ANTI-NORM DEVIANCE 907

when both deviate by the same amount and regardless of whether Measures
the deviant is an in-group or out-group member.
Participants rated each of the three targets using a 9-point semantic
differential scale to indicate how masculine-feminine, passive-dominant,
Hypotheses and Overview of Studies unintelligent-intelligent, and incompetent-competent they were. These
constituted the manipulation checks. Next, participants were asked to
We propose that reactions to deviants involve at least two indicate their evaluations of each of the three targets on four items using a
distinct processes. First, group members should perceive the extent 5-point scale (1 = not at all, 5 = very much). These items asked how much
of deviation at the intragroup level. Noticing that a group member participants liked the target, would like to work with the target, saw
has deviated is not sufficient to provoke negative evaluations and themselves as having a similar personality to the target, and had a lot in
judgments of the deviants (Marques et al., Expt. 3, 1998). Instead, common with the target. Next, participants rated the career prospects and
we believe that perceivers evaluate deviants with reference to the appropriate pay and conditions for each of the 3 targets. All items were
responded to using a 5-point scale (1 = not, 5 = very). They were asked
validity of salient in-group norms. Whereas self-categorization
how high a chance the target had for promotion, how successful their career
theory simply predicts that less prototypical members should be
would be, how far they will progress in their career, and what payment
evaluated less positively, we propose that such evaluations are band their salary should fall within (from 10,000 in bands of 2,000 up to
systematically related to the direction of such deviation. The 18,000+). They were also asked what percentage pay rise the target
studies reported in this article examine whether this prediction is should receive at their annual review (1%-5%), and, if promoted to
upheld in intragroup and intergroup contexts. manager, what their salary band should be (from 10,000 rising in bands of
In Study 1, we examined judgments of normative and deviant 3,000 to 22,000+). They were asked what type of contract the target
members of the gender in-group. In Study 2, we examined judg- should receive (3 months, 6 months, 1 year, 2 years, 3 years), how long the
ments of in-group and out-group normative and deviant members organization would want to keep them (1 = less than most, 5 = more than
who hold different attitudes toward asylum seekers in the United most), and how satisfied the company would be with their performance
(1 = not very satisfied, 5 = very satisfied). The next set of items asked how
Kingdom. We assume that deviants who favor the in-group nor-
similar the target was to other members of the department (1 = not very
mative direction help validate prescriptive group norms and that an
similar, 5 = very similar) and how much they had in common with the rest
underlying motive for evaluations of these members is to sustain of the department (1 = nothing, 5 = a lot).
social identity. This leads us to predict that both normative mem-
bers and pro-in-group deviants should be evaluated more posi-
Results
tively than deviants who oppose the in-group norm. Moreover, the
more individuals identify with their group, the more they should Data Reduction
favor in-group and out-group members that are closer to the
After conducting factor and reliability analyses we computed
in-group norm.
composite indexes by averaging across sets of measures relating to
each target (pro-norm, normative, and anti-norm). All composite
Study 1 indexes had acceptable internal reliability and distinctiveness from
one another. There were no significant main effects or interactions
Method involving participants' gender, and therefore this factor was not
included in the analyses reported below. The details are available
Participants
from the first author on request.
Sixty male and sixty female students (age 17-18 years) at a tertiary
college participated in the study in groups of 20-30 in each session. Manipulation Checks

An averaged masculinity and dominance index of stereotypic


Procedure extremity was coded so that a higher score represents greater
Participants were asked to take the role of a personnel officer in an consistency between the target's femininity-masculinity and that
insurance company (Alpha). Their task was to form impressions of the which is stereotypical for their group (masculinity for male targets,
employees in the Shipping Section (all of the participant's own gender). femininity for female targets). An analysis of variance (ANOVA)
They were told that the supervisor was about to leave and the 7 other revealed a significant effect of target, F(2, 238) = 1942.86, p <
members could be considered for promotion. Three members had applied, .001, accurately reflecting the profiles (see Table 1). Pro-norm
each of whom had worked for the company for approximately 5 years, had targets were rated as higher (M = 7.93) than normative targets
performed well, and was well qualified. The company ostensibly had (M = 5.03), t(119) = 42.09, p < .001, who were in turn higher
obtained information about all 7 employees in the course of their depart- than anti-norm targets (M = 2.16), t(l19) = 36.34, p < .001. W e
mental review.
did not want levels of perceived competence to be confounded
A booklet presented bar charts that used a 9-point continuum to depict
with gender extremity and normativeness. As intended, there were
the 7 employees' standing on 7 dimensions. All 7 employees had very
similar levels of intelligence, optimism, politeness, and competence (rated no significant differences in the competence ratings, F(2,
across 10 office tasks). Five (normative) members were very similar on the 238) = 1.19, or the intelligence rating, F(2, 238) = 0.55, of the
dimensions of dominance, masculinity, and emotionality, averaging 5.2 on three targets.
these dimensions. One deviant target was described as much higher (av-
eraging 8.0), and the other was much lower (averaging 2.0) on those Target Fit
dimensions. Thus, depending on the participant's own gender, one deviant
was pro-in-group normative and the other was anti-in-group normative in The extent to which targets were similar to, and had much in
comparison with the normative employees. common with, other members of the department were combined
908 ABRAMS, MARQUES, BOWN, AND HENSON

Table 1 The advancement prospects of the pro-norm and anti-norm devi-


Means and Standard Deviations of Evaluations of In-Group ants did not differ, t(l19) = 0.35.
Members as a Function of Deviance

Target Rewards

Pro-norm Normative Anti-norm This index averaged ratings of appropriate pay, percentage pay
rise, promotion pay, and appropriate length of contract. The effect
Index M SD M SD M SD of target, F(2, 238) = 14.18, p < .001, was significant. The
Stereotypic Extremity 7.93 0.71 5.03 0.27 2.16 0.87 normative target was rewarded more highly (M = 3.75) than the
Competence 5.60 1.07 5.67 0.82 5.52 1.15 pro-norm deviant (M = 3.42), t(l19) = 4.32, p < .001, and than
Target Fit 2.87 0.87 4.00 0.79 2.66 0.87 the anti-norm deviant (M = 3.40), t(l19) = 5.23, p < .001.
Similarity to Self 2.65 0.93 2.85 0.81 2.15 0.73 Rewards for pro-norm and anti-norm deviants did not differ sig-
Attractiveness 3.13 0.78 3.58 0.56 2.80 0.76
Advancement 3.42 0.81 3.82 0.70 3.38 0.82 nificantly, t(119) = 0.22.
Rewards 3.42 0.71 3.75 0.62 3.40 0.68

Note. The scale ranged from 1 (low) to 5 (high), except the scale for Discussion
Stereotypic Extremity and Competence ranged from 1 (low) to 9 (high).
The manipulation checks verified that both the pro-norm and
anti-norm deviants were seen as equally more gender extreme
relative to the normative employee and that all targets were per-
into a measure of fit. There was a significant effect of target, F(2, ceived as equally competent. Both deviants were downgraded on
238) = 93.97, p < .001. The normative member was judged as expectations of advancement and rewards relative to the normative
more fitting to their group (M = 4.00) than either the pro-norm target. This appears to be consistent with the idea that both types
(M = 2.87), t(l19) = 10.85, p < .001, or the anti-norm of deviants attracted similar (negative) judgments on these "ob-
(M = 2.66), t(l19) = 12.40, p < .001, deviants. However, the jective" measures. However, other aspects of the results lead to a
pro-norm deviant was also viewed as significantly more fitting different conclusion.
than the anti-norm deviant, t(119) = 2.03, p = .045. Both the anti-norm and pro-norm deviant were seen as fitting
less well to the group than was the normative employee. However,
Self-Target Similarity participants made less distinction between the fit of the normative
member and the pro-norm deviant than they did between the
This index included ratings of similarity of personality, amount normative member and the anti-norm deviant. This supports the
in common, and extent of shared views and opinions with the hypothesis that pro-norm deviants are viewed as more typical than
target. The effect of target was significant, F(2, 238) = 23.15, p < they really are. The evaluations of the targets amplify this pro-anti
.001. Self was viewed as more similar to the normative target distinction. Despite being equally deviant, the pro-norm deviant
(M = 2.85) than to the pro-norm deviant (M = 2.65), was perceived as more attractive than the anti-norm deviant. This
t(l19) = 1.82, p = .071, and than to the anti-norm deviant is consistent with the idea that the pro-norm deviant conflicts less
(M = 2.15), t(l19) = 7.45, p < .001, and as more similar to the with norms that define in-group social identity.
pro-norm than to the anti-norm deviant, t(119) = 4.46, p < .001. One explanation for the differential attractiveness of pro- and
anti-norm deviants in Study 1 might rest on our use of gender to
Target Attractiveness define the group. Even though we did not provide any explicit
information about targets' sexuality, participants may have in-
This index averaged across items that measured liking, willing-
ferred that a member of their gender who displayed opposite-
ness to work with, willingness to have as a friend, and perceived
gender traits might be a lesbian or a gay man, and this would
friendliness. The effect of target was significant, F(2,
convey additional stereotypical information. This might have ren-
238) = 39.55, p < .001. The normative member was rated as more
dered anti-norm deviants less attractive than gender-extreme per-
attractive (M = 3.58) than the pro-norm deviant (M = 3.13),
sons who deviated in a pro-norm direction (cf. Lindsey & Zakahi,
t(199) = 5.97, p < .001, and the anti-norm deviant (M = 2.80),
1996; Madon, 1997). We have no direct evidence for such an
t(199) = 9.79, p < .001, respectively. Moreover, the pro-norm
interpretation, but it remains consistent with our hypothesis that
deviant was seen as more attractive than the anti-norm deviant,
people are sensitive to in-group members who deviate toward
t(l19) = 3.10, p = .002.
out-group norms because such deviants challenge the subjective
validity of in-group norms.
Advancement
This index averaged ratings of the target's future success, pro- Study 2
motion prospects, progress, prospective length of tenure, and likely
organizational satisfaction with their work. The effect of target was Our second study used a different manipulation of objective
significant, F(2, 238) = 11.09, p < .001. The normative target was deviance to allow us to compare evaluations of deviants who hold
evaluated as having better advancement prospects (M = 3.82) than identical attitude positions but different positions relative to their
the pro-norm deviant (M = 3.42), t(l19) --- 3.83, p < .001, and group norm. In addition, we examined the effects of evaluating
than the anti-norm deviant (M = 3.38), t(l19) = 4.61, p < .001. in-group or out-group targets in the context of an explicit inter-
PRO-NORM AND ANTI-NORM DEVIANCE 909

group comparison in which the norms for both groups were also the status quo, namely that the current level of admissions for asylum
made explicit. seekers set by the present Government was about right." This reflected
We added three critical predictions to the hypotheses from actual pretested attitudes regarding the asylum issue. Participants were then
Study 1. First, we hypothesized that an out-group member who told that the survey would be continued and that they would be asked to
deviates toward the in-group (anti-norm) will be evaluated as less give their impressions of other people who had participated.
typical of the out-group than normative out-group members and
than out-group members who deviate in the out-group direction
Global Evaluations
(pro-norm). Second, the out-group anti-norm deviant will be
judged as more attractive than out-group members who are nor- participants completed a set of measures that asked (on a scale of 1 =
mative or are pro-norm deviant. Third, we predicted that an out- not at all, 7 = extremely) how favorable they felt toward BAICO, the
group anti-norm deviant will be evaluated more positively than an extent to which they considered psychologists and the BAICO as two very
in-group anti-norm deviant even when both have exactly the same similar groups, how much they would like to belong to BAICO, how
attitude. This is because, relative to other out-group targets, out- favorable they felt toward psychologists, and how much they agreed with
group anti-norm deviants provide validation o f in-group norms. In psychologists' and with BAICO's views about asylum.
turn, relative to other in-group targets, in-group anti-norm deviants
weaken the validity of the in-group norm. To test these ideas we
Judgments of Group Members
used a 2 (group presented: in-group vs. out-group) 3 (target
evaluated: anti-norm deviant, normative, pro-norm deviant) Participants viewed 6 group members' (labeled A to F) responses to
between-within design. In pretesting our materials, 40 psychology the 10 attitude items used in the pilot test, as well as each member's
students indicated their agreement with each of 10 attitude state- specific recommendations about the percentage change in numbers granted
ments about national policy (more open-more closed) on immi- asylum. In the in-group condition the group members were all psychology
grants seeking asylum. A composite measure ( a = .86) had a mean students; in the out-group condition they were all nonpsychologist BAICO
score of 3.65 (SD = 0.92) on a 7-point bipolar scale. The students members. Attitude responses were represented on 21-point bipolar scales,
made recommendations for British government policy on the per- which allowed us to manipulate deviance precisely. Four (normative)
centage increase or decrease in people granted asylum, using a members were depicted as holding normative beliefs for their group.
21-point scale in steps of,5%, ranging from - 5 0 % to +50%. The Members B and E were deviant. Both were equally divergent from the
normative members. Their attitudes fell outside of the range of the nor-
mean recommended change was + 5 % (SD = 3.02). Thus, students
mative members. Moreover, the anti-norm targets in the in-group and
had a neutral attitude overall.
out-group conditions actually shared the same position on the scale.
In the in-group condition, the 4 normative members' (psychology stu-
Method dents) positions corresponded to the (transposed) scale positions of psy-
chology students in the pilot study. Their attitude positions each aver-
Participants and Procedure aged 11, with a range of 10-12 across the attitude items. The
recommended percentage change in people granted asylum averaged 0%
Forty-one University of Kent psychology students (9 men, 32 women) across the normative members, with a range from - 5 % to +5%. Deviants'
completed a brief self-description questionnaire, designed to make identity attitudes were determined so that when transposed to a 7 point scale, they
as a psychology student salient. Then they completed a five-item measure fell outside the 95% confidence interval (and differed significantly) relative
of identification with psychology whose items were based on results from to attitudes held by psychology students. The pro-norm deviant averaged 8
previous research (Abrams, 1992; Marques et al., 1998). Next, there was a
(range 7-9) on the attitude items, and recommended a 20% increase in the
brief introduction to the topic of asylum seeking in Europe. A newspaper
numbers granted asylum. The anti-norm deviant averaged 14 (range 13-15)
article was described, containing a series of graphs and tables reporting the
and recommended a 20% reduction in the numbers granted asylum.
number of people granted asylum in different European countries. Britain
In the out-group condition, the 4 normative (BAICO) members each
was the middle of the range in terms of number of people granted asylum.
averaged 17 on the attitude scales, with a range of 16-18 across the items.
Participants then read a document entitled, "Survey Research on Asylum
The recommended percentage change in people granted asylum averaged
Seeking." This presented two opposing views regarding asylum regulations
-30% across the normative members, with a range from -25% to -35%.
in Britain. An introductory paragraph highlighted the importance and
The pro-norm deviant averaged 20 (range 19-21) on the attitude items, and
emotiveness of the issues involved, and provided (accurate) statistical
recommended a 40% decrease in the numbers granted asylum. The anti-
information that 194,000 asylum seekers enter Britain each year, of which
norm deviant averaged 14 (range 13-15) and recommended a 20% reduc-
21% (41,000) are granted asylum. The second paragraph gave a series of
tion in the numbers granted asylum. Note that the anti-norm positions in the
strong reasons why the numbers should be increased. The third paragraph
provided information from a bogus survey conducted among the British in-group and out-group condition are identical.
Association of Immigration and Customs Officers (BAICO). This stated Next, participants evaluated each of the 6 members using a series of
reasons the numbers granted asylum should be reduced (arguments were rating scales. The policy manipulation check asked participants to describe
matched for strength and number against those in the preceding paragraph). the policy orientation of the member (1 = open, 7 = closed). Remaining
At the end of this paragraph we included the sentence, items asked participants to rate (1 = not at all, 7 = extremely) how typical
the member was of their wider category, how favorable they felt toward the
In this survey the responses of the large majority of members reflected member, how similar each member was to themselves, how well the
the official policy of the BAICO, namely that the proportion of member would support the views of psychology students if they registered
asylum seekers who are ultimately granted permission to stay in for a psychology degree at Kent, and what share of 300 should be paid to
Britain should be reduced by at least 30%, to 27,000. each member if all 6 were to sit on a panel of appeal judges for asylum
applications. Finally, participants rated how similar the 6 people from the
The final paragraph mentioned a survey conducted among British psychol- survey were to one another (1 = not at all, 7 = extremely) and how much
ogy students and stated that the results indicated "widespread support for they had in common with one another (1 = very little, 7 = very much).
910 ABRAMS, MARQUES, BOWN, AND HENSON

Results t(22) = 7.23, p < .001, respectively. The anti-norm deviant was
also significantly different from the normative members in both
Global Evaluations conditions, t(17) = 7.71, p < .001, and t(21) = -6.09, p < .001,
In-group identification was similar in the in-group (M = 5.60, respectively. This confirms the effectiveness of the deviance
SD = 1.20) and out-group (M = 5.26, SD = 0.90) conditions, the manipulation.
two groups were rated as being generally dissimilar (M = 2.35, Typicality. Following a significant main effect of target, F(2,
SD = 1.10), and participants did not wish to belong to the out- 78) = 21.29, p < .001, paired t-tests revealed that normative
group (M = 2.41, SD = 1.13). When asked to rate the two groups members and pro-norm deviants were seen as equally typical of
separately, participants were more favorable toward the in-group their wider group, (M = 5.06 and 4.97, respectively), t(40) = 0.41,
(M = 5.45, SD = 0.93) than toward the out-group (M = 3.68, but that normative members were seen as significantly more typ-
SD = 1.12), F(1, 38) = 69.26,p < .001, and agreed more with the ical than anti-norm members (M = 3.43), t(40) = -7.15, p <
in-group (M = 4.44, SD = 1.23) than with the out-group .001. This pattern was repeated both within the in-group condition,
(M = 3.63, SD = 1.24), F(1, 39) = 9.70,p < .004. Thus the global t(40) = 0.57, t(40) = -3.84, p < .001, respectively, and the
information successfully conveyed an image of an out-group with out-group condition, t(40) = 0, t(40) = -6.15, p < .001,
which participants felt less in agreement and less favorable toward respectively.
relative to the in-group. Perceived Similarity of Members to Self. There was a signif-
Previous research suggested that in-group bias would be asso- icant main effect of group, F(1, 39) = 8.74,p < .005, qualified by
ciated with higher prior identification (Abrams, 1992, 1999; Jetten, a significant Group Target interaction, F(2, 78) = 14.72, p <
Spears, & Manstead, 1996; Lindeman, 1997; Marques et al., .001. In-group members were rated as more similar to self than
1998). Identification was significantly associated with general out-group members when they were pro-norm, F(1, 39) = 32.22,
favorability toward the in-group, r(40) = .49, p < .001, and with p < .001, or normative, F(1, 39) = 6.56, p < .02, but this pattern
the difference between in-group and out-group favorability, was reversed (nonsignificantly) if the targets were anti-norm, F(1,
r(40) = .42, p < .009, but not with favorability toward the 39) = 3.48, p < .08. Ratings of targets differed within the in-group
out-group, r(40) = - . 0 9 . condition, F(2, 78) = 4.11, p < .02. Pro-norm and normative
members were rated as equally similar to self, t(17) = 0.14, but
normative members were rated as more similar than anti-norm
Judgments o f Group Members
members to self, t(17) = -2.23, p < .04. Ratings also differed
We averaged the ratings for the 4 normative members and within the out-group condition, F(2, 78) = 13.83, p < .001.
analyzed the remaining variables using a Group Presented (in- Pro-norm members were rated as less similar than normative
group vs. out-group) Target (pro-norm deviant vs. averaged members to self, t(22) = -4.50, p < .001, but normative members
normative vs. anti-norm deviant) ANOVA. Group is a between- were rated as less similar than anti-norm members to self,
participants variable and target is a within-participants variable. t(22) = 2.34, p < .03.
Table 2 shows the means and standard deviations for these Favorability to Members. There was a significant Group
analyses. Target interaction, F(2, 78) = t9.20, p < .001; see Figure 1.
Policy Manipulation Check. There was a significant effect of Within levels of target, in-group and out-group members were
group, F(1, 38) = 22.22, p < .001, and a significant Group evaluated differently if they were pro-norm, F(1, 39) = 12.07, p <
Target interaction, F(2, 76) = 100.17, p < .001. The out-group .001, or anti-norm, F(1, 39) = 15.60, p < .001, but not if they were
targets were perceived to have more closed policies than the normative, F(1, 39) = 0.38. Moreover, evaluations of targets
in-group targets. The pro-norm deviant was significantly different differed within both the in-group condition, F(2, 78) = 7.29, p <
in the relevant direction from the normative targets in the in-group .001, and the out-group condition, F(2, 78) = 13.15, p < .001. In
and out-group target conditions, t(17) = -7.21, p < .001, and the in-group condition, pro-norm and normative members were

Table 2
Means and Standard Deviations of Judgments of Pro-Norm, Normative, and Anti-Norm Group
Members as a Function of Group Membership

In-group target Out-group target

Pro-norm Normative Anti-norm Pro-norm Normative Anti-norm

Measure M SD M SD M SD M SD M SD M SD
Policy Manipulation Check 2.50 0.92 4.12 0.72 5.72 0.89 6.65 1.27 5.18 0.76 3.59 1.44
Typicality 4.61 1.42 4.36 0.90 3.06 1.14 5.43 1.41 5.43 0.71 3.70 1.26
Perceived Similarity of Members to Self 4.39 1.46 4.33 1.07 3.22 1.77 2.04 1.19 3.25 1.52 4.17 1.50
Favorability to Members 4.61 1.65 4.13 0.93 3.17 1.25 2.91 1.47 3.92 1.11 4.65 1.15
Anticipated Agreement With In-Group 4.44 1.69 4.28 0.93 3.17 1.47 2.52 1.41 3.70 1.07 4.26 1.05
Recommended Payment for Members 54.06 28.98 49.75 12.07 35.18 19.37 35.65 16.54 51.30 6.39 63.04 22.70

Note. The Recommended Payment for Members measure involved distributing 300 among the group members. For the Policy Manipulation Check, the
scale endpoints were 1 (lenient) and 7 (closed). All other measures ranged from 1 (not at all) to 7 (very much~extremely).
PRO-NORM AND ANTI-NORM DEVIANCE 911

members of either group. However, identification had divergent


Pro-Norm
associations with judgments of targets who were pro- versus
, ~ m. Normative anti-norm. We encapsulated this in an index of the difference
4.8
between favorability toward pro- and anti-norm deviates. Differ-
. o . . o . Anti-Norm entiation was positively associated with identification for in-group
4.4 ~ * * * * *
targets, r(18) = .55, p < .05, but negatively associated with
4.0 ~ ****** identification for out-group targets, r(23) = - . 3 8 , p < .08. These
two correlations differ significantly (z = 2.98, p < .005). In other
3.6
.... **** words, there was a clear relationship between identification and
relative favorability toward in-group-validating deviants regard-
3.2
less of whether these are in-group or out-group members.
2.8
I 1 Discussion
In-group Out-group
In Study 2, participants felt more favorable toward, and in
Figure 1. Study 2: Favorability toward in-group and out-group members agreement with, the in-group than the out-group. Consistent with
as a function of deviance from their respective group norms. intergroup differentiation predicted by social identity theory (cf.
Abrams, 1992; Lindeman, 1997), this difference was larger among
participants who identified more strongly with the in-group.
rated equally favorably, t(17) = 1.44, but anti-norm members were
rated less favorably than normative members, t(17) = -2.71, p < Perceptions of Difference
.02. In the out-group condition, pro-norm members were rated less
favorably than normative members, t(22) = -5.79, p < .001, and Individual targets whose policy orientation deviated either to-
normative members were rated less favorably than anti-norm ward or away from their own group norm were perceived to
members, t(22) = 2.51, p < .02. deviate to the same degree in both the out-group and in-group
Anticipated Agreement with In-Group. There was a significant condition (similar to the gender-extremity results in Study 1).
Group Target interaction, F(2, 78) = 17.33, p < .001. Within However, anti-norm deviants were regarded as being more atypical
levels of target, agreement from in-group and out-group members than pro-norm deviants, consistent with (but more pronounced
was rated differently if they were pro-norm, F(1, 39) = 15.79, p < than) the pattern of Findings for "fit" in Study 1. It seems that
.001, or anti-norm, F(1, 39) = 7.74, p < .001, but not if they were judgments of fit and typicality reflect perceived prototypicality, as
normative, F(1, 39) = 3.37, p < .08. Within the in-group condi- conceptualized in self-categorization theory (Turner et al., 1987).
tion, targets were judged differently, F(2, 78) = 6.52, p < .002. According to self-categorization theory, the metacontrast of
Pro-norm and normative members were judged to be equally in within-group and between-group differences should mean that
agreement, t(17) = 0.43, but normative members were expected to both normative and pro-norm members are more prototypical than
agree more than anti-norm members, t(17) = -2.84, p < .02. In anti-norm members. The former are closer to the position that
the out-group condition, targets were also judged differently, F(2, encapsulates the difference between in-group and out-group in
78) = 13.60, p < .001. Pro-norm members were assumed to agree relation to perceived intragroup differences. The stronger finding
less than normative members, t(22) = -5.00, p < .001, but in Study 2 is attributable to the increased salience of intergroup
normative members were not seen as very different from anti-norm context, which would be expected to polarize the prototypical
members, t(22) = 1.83, p < .09. position for the in-group (Abrams & Hogg, 1990). The results for
Recommended Paymentfor Members. There was a significant judgments of self-target similarity in both studies support the idea
Group Target interaction, F(2, 76) = 12.18, p < .001. In-group that these judgments are associated with self-categorization. In
and out-group members were paid differently if they were pro- both studies participants regarded the in-group pro-norm deviant
norm, F(1, 38) = 6.47, p < .02, or anti-norm, F(1, 38) = 16.63, as more similar to self than the in-group anti-norm deviant.
p < .001, but not if they were normative, F(1, 38) = 0.28.
Payments differed within both the in-group condition, F(2, Evaluations
76) = 3.85, p < .03, and the out-group condition, F(2,
76) = 10.05, p < .001. In-group pro-norm and normative members In both studies, evaluations were affected by targets' closeness
were paid equally, t(16) = 0.50, but normative members were paid to the in-group prototype more than by group membership. The
more than anti-norm members, t(16) = -2.72, p < .02. Out-group in-group pro-norm deviant was rated more positively than the
pro-norm members were paid less than normative members, anti-norm deviant, and in Study 2, as positively as the normative
t(22) = -4.16, p < .001, and normative members were paid less target. In Study 2, out-group pro-norm deviants were evaluated
than anti-norm members, t(22) = 2.10, p < .05. more negatively than normative members, and anti-norm deviants
were evaluated more positively than normative members. This
pattern is consistent with a motivation to validate the in-group
Identification and Ratings of Members
norm. It is in line with our idea that anti-norm out-group deviants
We examined the correlation between in-group identification help to undermine the validity of out-group norms and strengthen
and favorability measures within each condition. Identification had in-group norms. This idea is supported by the fact that out-group
only a weak association with judgments regarding normative anti-norm deviants were evaluated and rewarded substantially
912 ABRAMS, MARQUES, BOWN, AND HENSON

more positively than in-group anti-norm deviants, despite the fact Abrams, D. (1999). Social identity, social cognition, and the self: The
that both expressed identical attitudes. As well as ruling out a flexibility and stability of self-categorization. In D. Abrams & M. A.
similarity-attraction interpretation, this discrepancy demonstrates Hogg (Eds.), Social identity and social cognition (pp. 197-229). Oxford,
that the meaning of deviance is derived not only from the intra- England: Blackwell.
group but also from the intergroup context. This is clearly in line Abrams, D., & Hogg, M. A. (1990). Social identification, self-
categorization and social influence. European Review of Social Psychol-
with subjective group dynamics.
ogy, 1, 195-228.
Jetten, J., Spears, R., & Manstead, A. S. R. (1996). Intergroup norms and
Identification and Global Judgments o f Groups intergroup discrimination: Distinctive self-categorization and social
In Study 2, consistent with our hypothesis that identification identity. Journal of Personality and Social Psychology, 71, 1222-1233.
motivates intragroup differentiation, and also in line with subjec- Levine, J. M. (1989). Reaction to opinion deviance in small groups. In P. B.
Paulus (Ed.), Psychology of group influence (pp. 187-231). Hillsdale,
tive group dynamics, higher identification was associated with
NJ: Erlbaum.
greater favorability toward deviants that endorsed a more in-group
Lindeman, M. (1997). Ingroup bias, self-enhancement and group identifi-
normative direction. This supports the findings in Marques et al.'s
cation. European Journal of Social Psychology, 27, 337-355.
(1998) previous research. Lindsey, A. E., & Zakahi, W. R. (1996). Women who tell and men who
ask: Perceptions of men and women departing from gender stereotypes
Conclusions during initial interaction. Sex Roles, 34, 767-786.
Madon, S. (1997). What do people believe about gay males? A study of
The present research shows how individuals subjectively con-
stereotype content and strength. Sex Roles, 37, 663-685.
struct deviance when they evaluate group members. In an intra- Marques, J. M., Abrams, D., Paez, D., & Martinez-Taboada, C. (1998). The
group context without explicit intergroup comparison, members role of categorization and in-group norms in judgments of groups and
are sensitive to anti-norm and pro-norm deviance. However, they their members. Journal of Personality and Social Psychology, 75, 976-
evaluate anti-norm deviance more negatively. In an intergroup 988.
context this tendency is more extreme. Pro-norm deviance among Sherman, S. J., Hamilton, D. L., & Lewis, A. L. (1999). Perceived entita-
in-group members is largely disregarded and such members are tivity and the social identity value of group memberships. In D. Abrams
treated as if they were typical. Anti-norm deviance from in-group & M. A. Hogg (Eds.), Social identity and social cognition (pp. 80-110).
members is regarded very negatively and is punished. Judgments Oxford, England: Blackwell.
of out-group members seem to be harsher for pro-norm deviants Tajfel, H. (1981). Human groups and social categories. Cambridge, En-
and more positive for anti-norm deviants. Subjective group dy- gland: Cambridge University Press.
namics seem to afford protection to in-group pro-norm deviants Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M.
and a welcoming hand to out-group anti-norm t:leviants. (1987). Rediscovering the social group: A self-categorization theory.
Oxford, England: Blackwell.
References
Abrams, D. (1992). Processes of social identification. In G. Breakwell Received July 2, 1999
(Ed.), Social identity and the self-concept (pp. 57-100). San Diego: Revision received November 30, 1999
Academic Press. Accepted November 30, 1999

You might also like