Bulletin
http://psp.sagepub.com/
Published by:
http://www.sagepublications.com
On behalf of:
Additional services and information for Personality and Social Psychology Bulletin can be found at:
Email Alerts: http://psp.sagepub.com/cgi/alerts
Subscriptions: http://psp.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://psp.sagepub.com/content/35/3/309.refs.html
A high need for uniqueness undermines majority influence. Need for uniqueness (a) is a psychological state in
which individuals feel indistinguishable from others and
(b) motivates compensatory acts to reestablish a sense of
uniqueness. Three studies demonstrate that a strive for
uniqueness motivates individuals to resist majority
influence. In Study 1, the need for uniqueness was measured, and it was found that individuals high in need for
uniqueness yielded less to majority influence than those
low in need for uniqueness. In Study 2, participants
who received personality feedback undermining their
feeling of uniqueness agreed less with a majority (vs.
minority) position. Study 3 replicated this effect and
additionally demonstrated the motivational nature of
the assumed mechanism: An alternative means that
allowed participants to regain a feeling of uniqueness
canceled out the effect of high need for uniqueness on
majority influence.
Keywords:
esearch on social influence has gathered considerable evidence that majority sources are usually
highly influential. A majoritys high consensus determines what is seen as normal and correct. Asked to
articulate an opinion on a subject that they have never
thought about, individuals will conform to what they
perceive as the majority option, presumably following
the heuristic, If everybody thinks that way, there must
be something to it. Conversely, being too similar to others has been described as evoking aversive feelings and
motivating compensatory action. Under certain conditions, it may not subjectively appear gratifying to think
or say what most others do. We presently propose that
one such condition is met when recipients of influence
309
Downloaded from psp.sagepub.com at Universitaets- und on September 4, 2012
310
Oakes, Reicher, & Wetherell, 1987) or when the recipients are willing to show leniency toward the minority
group (Crano, 2001) or want to compensate the minority for unjustified wrongs they have suffered from in the
past (Moscovici & Perez, 2007).
Obviously, these models of minority influence unanimously share the assumption that minorities can overcome their inherently negative position by compensating
for low consensus. The idea of compensation is perhaps
most clearly stated in Latans social impact theory
(Latan & Wolf, 1981; Nowak, Szamrej, & Latan,
1990), according to which the minority can surmount
the disadvantage of being small in number by the immediacy (e.g., physical closeness) and strength (e.g., social
power) of their appeal. Accordingly, everything else
held constant, the majority always has the advantage of
representing higher consensus than the minority. Thus,
mechanisms proposed in these models to enable minorities to exert influence actually also lead to greater
majority influence. For example, a consistent behavioral
style not only enhances minority influence. In Aschs
(1956) experiments, as well, consistency had a strong
impact on the amount of influence, leading Moscovici
and colleagues to label conformity an effect of consistency (Moscovici, Lage, & Naffrechoux, 1969; p. 377).
Similarly, categorization of the influence agent as similar to oneself can increase both minority and majority
influence (David & Turner, 1996). In Baker and Pettys
(1994) research, both minority and majority influence
increased when sourceposition imbalance motivated
recipients to extensively process convincing arguments
put forward for the respective position. The review by
Wood and colleagues (1994) provides more of these
examples where a variable produced a main effect on
judgments, parallel for minority and majority sources,
but no interaction with consensus.
Thus, our approach is different from concurrent
models of minority influence in that we seek conditions
that operate differently under varying consensus conditions. Analogous to Moscovicis (1980, p. 210) notion
that what the minority does is bad, because there are
few who do it, we seek conditions under which the
majority option becomes less attractive, in that what
the majority does is bad because there are too many
who do it. We propose that one such condition is a
high NfU.
311
STUDY 1
Method
Overview. Participants received a text including the
results of an alleged opinion poll in which consensus
information was manipulated. The main dependent
variable was their attitude toward the issue referred to
in the opinion poll. Participants then received a questionnaire that included the four items of the SANU
(Lynn & Snyder, 2002). We expected that participants
with a high NfU would agree more with the low-consensus position, whereas participants low in NfU would
agree more with the high-consensus position. At the
conclusion of the experiment, participants answered an
open-ended suspicion probe. They were then thoroughly
debriefed and paid 2 for their participation. Materials
were presented in appropriate German translations of
the following descriptions.
Participants and design. Participants were 44
students (17 female, 27 male; mean age 22.2) of various
majors at the Chemnitz University of Technology,
Germany. They volunteered for a study on text comprehension. Students were randomly assigned to one
of the two influence conditions (minority vs. majority
influence). NfU was assessed by a questionnaire including the SANU at the end of the experimental session.
Manipulation of minority and majority influence.
Participants received a text on an exhibition on the
history of eating and drinking in the German railway.
Attached to the text, the sheet presented a diagram
showing the alleged results of an opinion poll on the
question of whether dining cars should be disestablished
in the German railway. The diagram displayed that,
depending on the condition, either 79% (majority) or
21% (minority) agreed. Operationalizations were chosen to prevent recipients from (a) identifying with the
influence group and (b) processing message arguments.
Thus, (a) no information was given as to what sample
or how many individuals had been polled, and (b) the
text neither referred to the diagram nor presented
312
Results
None of the students reported any suspicion about
the cover story. Thus, data analyses are based on the full
sample of N = 44 cases. Because this study, as well as
Studies 2 and 3, revealed no gender effects, this variable
will not receive further attention. Degrees of freedom
slightly vary as a result of occasionally missing data.
Manipulation check. Participants in the minority
condition estimated the percentage of respondents in
favor of the issue significantly lower, M = 20.60%, than
participants in the majority condition, M = 76.58%,
F(1, 42) = 1375.92, p < .001. Thus, the manipulation
of consensus was successful.
SANU. Because one item lowered the reliability of
the total scale, only three items were combined to form
a single index of need for uniqueness ( = .63). Scores
were independent of the consensus factor, F < 1.
Attitudes. The three attitude items were combined to
form a single index ( = .98). It was used as a dependent variable in a regression analysis with effect-coded
consensus condition (minority 1, majority +1), centered NfU, and an interaction term of consensus by NfU
as independent variables (Cohen, Cohen, West, &
Aiken, 2003, p. 375). This regression analysis produced
the expected interaction, = .29, p .05, but no
Minority
Majority
Attitude Index
5
4
3
2
1
- 1 SD
+ 1 SD
Discussion
This study provides initial support for the connection
between individuals NfU and their readiness to be
influenced by high consensus: Low-NfU participants
yielded to majority influence, whereas recipients with
high NfU scores did not. These effects emerged even
though the influence group was not specified in any way
that would likely raise feelings of identification. In fact,
the influence group was defined by nothing else than
consensus information displayed as the result of an
opinion poll. Furthermore, participants had not
received any information on the issue other than the
alleged opinion poll result. Thus, differential message
processing can be ruled out as the mechanism underlying the observed effect on attitudes.
However, there are also some weaknesses in the present study that need to be addressed. First of all, the
NfU measure yielded a reliability score that was not
fully satisfactory. Additionally, we see another liability in the fact that NfU was merely measured. It might
STUDY 2
Study 2 tested the hypothesis that manipulating NfU
by bogus feedback would affect social influence consistent with our reasoning. To experimentally vary NfU,
we developed a test-feedback method closely shaped
after Snyder and Fromkin (1980; Fromkin, 1970). A
pilot study was conducted to check whether the false
feedback had the intended effect on NfU.
Pilot Study
Participants and overview. Participants included 42
students (29 female, 13 male; mean age 22.7 years) of
various majors at the University of Bonn, Germany.
They volunteered for two allegedly unrelated studies.
The first study was introduced as an evaluation of a
recently developed personality questionnaire. Participants
first filled in an alleged personality questionnaire and
received bogus feedback to manipulate their NfU. This
was followed by a second short questionnaire study
containing the same questionnaire as used in Study 1,
including the four SANU items. Students were randomly
assigned to one of two feedback conditions and participated in groups of up to 4. At the conclusion, they were
thoroughly debriefed.
Manipulation of NfU. The participants filled in a
32-item questionnaire presented on a computer screen.
The questionnaire was constructed with regard to face
validity and supposedly measured four central personality traits. After completion, the computer displayed a value for each of these traits on a stanine
scale. The participants were asked by written instruction to copy these values from the screen on a paper
results sheet. For each dimension on the results sheet,
the margin between the values 4 and 6 was gray colored, indicating the values including 62% of the norm
group. High-NfU participants received a 5 on three of
the four dimensions and either a 4 or a 6 on the fourth.
Control participants received one score in this area (4)
and more diverse scores on the other three dimensions
(3, 7, and 8). In addition, a remark on the screen
informed participants about the alleged percentage of
identical answers in a reference population of 10,000
313
Main Study
Overview. To manipulate NfU, participants received
the same false feedback as in the pilot study. They were
then exposed to majority versus minority influence.
This resulted in a 2 (high NfU vs. control) 2 (high vs.
low consensus) factorial design. The main dependent
variable was the attitude toward the target issue. We
expected that under high NfU, participants would yield
significantly less to majority influence. At the conclusion, participants answered an open-ended suspicion
probe, were then thoroughly debriefed, and were paid
3 for their participation.
Participants and design. Participants were 62
students (35 female, 26 male; mean age 23.9 years) of
various majors at the University of Bonn, Germany.
They volunteered for two allegedly unrelated studies.
The first study was introduced as an evaluation of a
recently developed personality questionnaire, followed
by a second study on text comprehension. Students
were randomly assigned to one of the four conditions
and participated in groups of up to 8.
Manipulation of majorityminority influence. Participants received written instructions that the second study
explored mechanisms of text comprehension. Their task
was to read a text and form a general impression.
The headline of the text stated, A majority (minority)
of students of the University of Bonn argue for Curutao
Lake as a rewarding holiday destination. The text went
on to state that an informational event about Curutao
Lake had taken place at the University of Bonn. After
the presentation, the audience had been asked their
opinion about Curutao Lake. It turned out that a
majority of 83% (vs. a minority of 17%) had considered Curutao Lake a rewarding vacation spot. The text
continued presenting arguments that were brought forward by those in favor of the vacation spot. The issue
was chosen because it has been extensively pretested
and successfully used in other experiments on majority
influence (Erb et al., 1998; Erb, Bohner, Werth,
Hewstone, & Reinhard, 2006).
Dependent measures and specific hypotheses. A
booklet contained all dependent measures. The first
four questions on text comprehension (e.g., The text
was easy to comprehend) were filler items to maintain
the cover story. Then, three attitude items assessed the
overall attitude toward the vacation resort, for example,
I would recommend Curutao Lake as a destination of
travel. Seven further items tapped into participants
attitudes toward several aspects of the vacation spot.
Responses to each attitude item were collected on identical scales from 1 (do not agree at all) to 9 (fully agree).
An interaction was predicted for attitude measures such
that majority influence in the control condition would
disappear (or even reverse) in the high-NfU condition.
After that, participants were asked to estimate the
percentage of respondents who were in favor of Curutao
Lake (manipulation check). They then reported their age,
gender, major, and semester. Finally, two open-ended
questions tapped into suspicion regarding the true intent
of the study.
Results
We dropped two cases from analyses. A non-native
speaker reported severe comprehension problems; the
computer of another participant had a power shortage.
None of the other participants reported any suspicion.
The analyses are thus based on N = 60 cases. Dependent
measures were subjected to 2 2 ANOVA with the factors consensus (high vs. low) and NfU (high vs. control).
Manipulation checks. All participants had completed
the personality questionnaire, and their answers showed
no specific pattern that gave rise to suspicion. They all
had copied their results to the result sheets correctly and
314
Minority
Majority
B
AB
AB
A
4
0
Control
High NfU
Feedback Manipulation
Figure 2
therefore actively noticed the feedback. For the consensus measure, the ANOVA yielded a single main effect of
consensus, F(1, 56) = 674.54, p < .001, and no other
effects were significant, both p > .10. Participants estimated 20.60% consensus in the minority condition and
83.13% consensus in the majority condition. Thus, the
manipulation of consensus was also successful.
Attitude measures. All attitude items displayed a similar pattern of means. Nine of them were combined to
form a single attitude index ( = .81). The ANOVA on
this index yielded the predicted two-way interaction of
consensus by NfU, F(1, 56) = 6.47, p < .02, but no
main effects, both p > .21. Means (Figure 2) indicate
that in the control group, the majority was more influential than the minority, whereas in the high-NfU condition, this pattern reversed. Thus, the main hypothesis
of this experiment was confirmed.
Post hoc tests of simple effects revealed that the majority was less influential under high NfU than in the control
condition, t(56) = 2.69, p < .005. All other effects did
not reach the adjusted significance level for multiple comparisons. However, on a more descriptive level, participants in the control condition agreed more with the
majority than the minority, t(56) = 1.86, p < .04,
whereas the minority was marginally more influential
than the majority under high NfU, t(56) = 1.72, p < .09.
Discussion
Our main hypothesis was supported by the predicted
interaction of NfU by consensus on attitude judgments.
STUDY 3
We have argued that a high NfU motivated participants in Studies 1 and 2 to resist majority influence. A
major assumption common to all well-established
accounts of human motivation is that a perceived discrepancy between a desired and an existing state motivates activities to eliminate it (e.g., Higgins, 1987;
Kruglanski, 1996). Another important motivation principle is that a desired state can be achieved through different means (Heider, 1958). Different means can be
functionally equivalent and hence can substitute each
other. Thus, given that the pattern of attitudes under high
NfU in Study 2 resulted from a motivational force, the
negative effect of high NfU on majority influence should
be canceled out if participants were given the opportunity
to regain uniqueness through alternative means prior to
the influence attempt. Study 3 tested this hypothesis.
Method
Overview. Study 3 followed Study 2 in most respects.
However, one further condition was added that was
presumed to satisfy a high NfU prior to the influence
setting. In addition to the control group and the highNfU group (as in Study 2), a third group was created that
received the high-NfU manipulation but then was given
an alternative means to regain uniqueness. Furthermore,
the persuasive setting was changed to replicate the findings from Study 2 with a different topic and a different
influence group.
Participants and design. A total of 161 students (52
female and 109 male; mean age 23.1 years) of various
majors at the University of Bonn, Germany, volunteered in
exchange for 3 or partial course credit. The design was
a 3 (control vs. high NfU vs. high NfU plus regain uniqueness) 2 (majority vs. minority influence) between-subjects
factorial. Students were randomly assigned to one of the
six conditions and participated in groups of up to 8.
315
Results
All participants had completed the personality questionnaire, and their answers showed no specific pattern
that gave rise to suspicion. They all had copied their
results to the result sheets correctly.
Manipulation check. The manipulation check for the
consensus information revealed that 10 participants
had not indicated the consensus information correctly.
They were excluded from further analyses. Experimental
dropout was independent of condition, 2(5, N = 161) =
7.53, p > .18. The analyses are thus based on N = 151
cases. However, all results remained stable when the
10 cases were included in control analyses. None of the
remaining participants reported any suspicion.
As for the consensus estimates, in one cell, all participants reported the correct answer, producing a variance
of zero in this cell. Such heterogeneous variances are a
violation of the prerequisites to calculate a 2 3
ANOVA. We therefore conducted a one-way ANOVA
over all 6 conditions with a post hoc Dunnett-C test for
heterogenous variances. The ANOVA revealed a significant effect of condition on consensus, F = 19,174.67,
p < .001. The Dunnett-C tests showed that all three
experimental conditions that received the low-consensus
information gave significantly lower estimations of consensus than each of the three experimental conditions
with high-consensus information, all p < .001. None of
the conditions with identical consensus information differed from each other, all p > .20. Participants estimated
16.77% consensus in the minority condition and
83.04% consensus in the majority condition. Thus, the
manipulation of consensus was successful.
Attitude measures. Items were combined to form a single attitude index ( = .88). The ANOVA yielded the
predicted two-way interaction of consensus by NfU, F(1,
145) = 6.13, p < .004, but no main effects, both p >
.22. The majority was less influential in the high-NfU
condition than in either the control or in the highNfU/regain-uniqueness condition (Figure 3). Simple tests
showed that the majority was clearly more influential
than the minority in the control condition, t(145) = 3.16,
p < .005, whereas this reversed under high NfU, t(145) =
1.80, p < .04. Thus, the hypothesis that individuals
under high NfU yield less to majority influence was fully
supported, and again some weak evidence was found that
high NfU actually increased minority influence. For the
10
Attitude Toward Rhine Tunnel
316
Minority
Majority
8
B
6
B
AB
AB
A
0
Control
High Nfu
Regain
Uniqueness
Feedback Manipulation
Figure 3
regain-uniqueness condition, the results appeared somewhat less clear. Although participants agreed more with
the majority than did high-NfU participants, there was
only a slight tendency toward majority influence; means
did not significantly differ between the two consensus
conditions, t(145) = .78, p > .22. Presently it appears
that the regain-uniqueness procedure was not sufficiently
strong to totally resolve the NfU and thus may have left
participants in this condition with conflicting motives
that overall nullified the effect on attitudes. Nonetheless,
the hypothesis that an alternative means to regain
uniqueness would cancel out the effect of a high NfU on
majority influence was confirmed.
Discussion
The central hypothesis of this study was confirmed,
as we found the predicted interaction of NfU by consensus on attitude judgments. In the high-NfU condition, majority influence was undermined. The hypothesis
that high NfU motivates recipients to refuse agreement
with the majority received strong support. When highNfU participants were given an opportunity to regain
uniqueness by listing personality characteristics that
would describe them as unique, the blockade of majority influence disappeared. This is in line with the widely
acknowledged notion that a motivational state becomes
weaker or even disappears once the desired state has
been achieved through goal fulfilment or satisfaction
(e.g., Higgins, 1987; Kruglanski, 1996). The fact that
GENERAL DISCUSSION
This research demonstrated that majority influence
can be undermined when participants experience a high
NfU, either measured or experimentally manipulated.
The effect arose with different topics and different
delineations of the respective influence source, including
plain consensus delivered by an opinion poll result, a
socially relevant in-group of fellow students, and a
socially rather irrelevant out-group of nonspecified residents of an area different from where the targets of
influence would reside. Furthermore, operationalizations were chosen not to produce conditions of
enhanced message processing (e.g., Moscovici, 1980),
leniency toward the influence group (e.g., Crano, 2001),
the attribution of courage and commitment (e.g., Sigall
et al., 2006), and so forth, put forward by models to
explain minority influence. In fact, the influence sources
did not differ in any such variable but merely in the consensus on the respective position. Thus, the observed
block of majority influence under high NfU cannot
alternatively be explained by any concurrent model of
minority influence. Moreover, the proposed motivational character of the presented effect not only is in line
with the theoretical assumptions of uniqueness theory
(Snyder & Fromkin, 1980), research on deindividuation
(e.g., Zimbardo, 1969), and optimal distinctiveness
(Brewer, 1991; Hornsey & Jetten, 2004) but also was
empirically demonstrated. In Study 3, an alternative
means to regain uniqueness eliminated the effect of high
NfU on majority influence (adhering to the widely
accepted notion of mutual substitutability of different
means to fulfill the same motivational goal). Thus, feeling too similar to others appears to be a powerful motivational force that can make a majority position
subjectively unattractive, just for the sake of refusing the
most common position represented by high consensus.
Our finding that the strong majority advantage
reported throughout the literature was undermined by
high NfU may shed some new light on classical findings. In Aschs (1956) seminal study, for example, one
fourth of the participants never yielded to majority
influence. Postexperimental interviews provided some
narrative evidence that high NfU might have had an
impact on the willingness to conform. One completely
independent participant confessed, I was getting
almost a sadistic pleasure out of being different (Asch,
1956, p. 37). The participant went on to explain why he
displayed a nonconformist response style: Long years
317
318
REFERENCES
Asch, S. E. (1956). Studies of independence and conformity: I. A
minority of one against a unanimous majority. Psychological
Monographs, 70, 1-70.
Axsom, D., Yates, S., & Chaiken, S. (1987). Audience response as a
heuristic cue in persuasion. Journal of Personality and Social
Psychology, 53, 30-40.
Baker, S. M., & Petty, R. E. (1994). Majority and minority influence:
Source-position imbalance as a determinant of message scrutiny.
Journal of Personality and Social Psychology, 67, 5-19.
Bond, R., & Smith, P. B. (1996). Culture and conformity: A metaanalysis of studies using Aschs (1952b, 1956) line judgment task.
Psychological Bulletin, 119, 111-137.
319
320