Word count 5007 - Universität Bielefeld

advertisement
Persuasion by Majorities and Minorities
Running Head: CONVERSION THEORY VERSUS OBJECTIVE CONSENSUS
Persuasion by Majorities and Minorities:
Delineating the Scope of Conversion Theory and Objective Consensus Explanations
Gerd Bohner
Amanda Dykema-Engblade
Universität Bielefeld
Loyola University Chicago
R. Scott Tindale
Helen Meisenhelder
Loyola University Chicago
Loyola University Chicago
Word count 5,007
Author Note
Gerd Bohner, Abteilung für Psychologie, Universität Bielefeld, Germany; Amanda
Dykema-Engblade, R. Scott Tindale, and Helen Meisenhelder, Department of Psychology,
Loyola University Chicago. This research was partially funded by the National Science
Foundation (#SES-0136332). We would like to thank Hans-Peter Erb for his thoughtful
comments on a previous draft. Portions of this research were presented at Small Group
Meetings of the European Association of Experimental Social Psychology on "Minority
Influence Processes", Oxford, UK, September 22-26, 2003, and on "The Psychology of
Minorities", London, June 27-29, 2003. Correspondence should be addressed to Gerd Bohner,
Universität Bielefeld, Abteilung für Psychologie, Postfach 100131, 33659 Bielefeld, Germany
(e-mail: gerd.bohner@uni-bielefeld.de), or to Amanda Dykema-Engblade, Department of
Psychology, Loyola University Chicago, 6525 N. Sheridan Rd., Chicago, IL 60626 (e-mail:
adykema@luc.edu).
1
Persuasion by Majorities and Minorities
2
Abstract
Conversion theory and the objective consensus approach were contrasted in explaining effects
of minority and majority sources on message processing and attitude judgments. Participants
read a persuasive message containing weak, ambiguous, or strong arguments that was attributed
to either a majority or a minority source. To prime either social comparison or accuracy
concerns, the source was introduced as being either similar to participants or knowledgeable.
The dependent variables were cognitive responses to the message and post-message attitudes.
Under social comparison priming, the results supported conversion theory: Participants
systematically processed minority messages, as indicated by effects of argument strength on
thought valence and attitudes, whereas they generally accepted the majority message regardless
of argument strength. Under accuracy priming, results supported the consensus approach: For
ambiguous arguments, participants showed biased processing based on source consensus,
whereas they showed contrast effects for weak (strong) arguments attributed to a majority
(minority) source.
Persuasion by Majorities and Minorities
3
Persuasion by Majorities and Minorities:
Consensus as a Cue to Message Validity versus Normative Functions of Consensus
Social consensus is an important determinant of human judgments. While this statement
may be noncontroversial, theorists disagree about how consensus affects attitude judgments and
information processing, and about the direction of its effects (e.g., Mackie, 1987; Moscovici,
1976, 1980; for discussion see Erb, Bohner, Rank, & Einwiller, 2002).
Theories of Minority and Majority Influence
According to Moscovici's conversion theory (1980), conflict between the majority and
minority groups is the key to explaining influence processes. Majority support for a given
position induces normative conflict, whereas minority support instigates informational conflict.
Recipients who are exposed to majority influence are thought to engage in social comparison.
Their main concern is thought to be the social conflict between their own position and the
majority's position, and message content is assumed to be of little importance in determining
outcomes. Recipients will show conformity with the majority independent of the majority's
arguments. Quite differently, recipients who are under minority influence are thought to solve
the informational conflict they experience through an active validation process in which issuerelevant information processing prevails. Hence, the minority message will be elaborated
extensively, and this extensive processing may eventually lead to conversion, i.e. attitude
change in the direction of the minority's position (see also Maass & Clark, 1984; Nemeth,
1986). Importantly, however, validation attempts will not necessarily lead to conversion.
Although this aspect was not emphasized by Moscovici (1980), research guided by process
theories of persuasion clearly demonstrated that the outcome of effortful message validation
depends on the cogency of the available evidence: Whereas strong evidence will lead to
acceptance of the minority's position, weak evidence will lead to its rejection (e.g., Bohner,
Persuasion by Majorities and Minorities
4
Moskowitz, & Chaiken, 1995; Martin & Hewstone, 2003; Ziegler, Diehl, Zigon, & Fett, 2004).
In contrast to Moscovici's ideas, there is a long tradition of approaches that emphasize
the informational value of consensus information, even in the absence of social conflict. Within
attribution theory, Kelley (1967) suggested that high consensus for a given position makes
people “know that they know” and – all else being equal – promotes entity attributions.
Accordingly, recipients under majority (vs. minority) influence should infer high message
validity and be concerned with the issue under consideration rather than with the influence
source (see Bohner, Erb, Reinhard, & Frank, 1996; Bohner, Frank, & Erb, 1998). In a similar
vein, Mackie's (1987) objective consensus approach asserts that the majority position is usually
perceived as reflecting "objective reality". Mackie predicted and found that recipients pay close
attention to majority messages and process their content more extensively than that of minority
messages. Finally, Erb, Bohner, Schmälzle, and Rank (1998) suggested that individuals use
mere consensus regarding an attitudinal position to form an initial judgment about this
position's validity. In terms of process analyses, Erb and his colleagues' research most clearly
demonstrated that recipients of majority (vs. minority) messages indeed formed positive (vs.
negative) initial judgments based on perceived consensus. These initial judgments then biased
recipients’ subsequent message processing: Mixed evidence led to more positive cognitive
responses and attitudes when it came from a majority than when it came from a minority.
The consensus-related theories thus share the idea that a high degree of consensus in
favor of a position creates an expectation in recipients that this position should be valid, which
in turn causes recipients to focus on the issue under consideration. In this regard they are
diametrically opposed to conversion theory, which suggests that majority influence (i.e., high
consensus) instigates recipients to focus on social comparison, which goes along with a lack of
issue-related considerations. The fact that supporting evidence has been reported for both
Persuasion by Majorities and Minorities
5
conversion theory (for reviews, see Moscovici, 1980; Wood, Lundgren, Ouellette, Busceme, &
Blackstone, 1994) and the consensus approaches (Erb et al., 1998; Erb & Bohner, 2001;
Mackie, 1987), suggests that each approach may be valid under certain conditions.
The aim of our current research is to explain more precisely what these conditions might
be and to demonstrate that both types of effect may be obtained in a single experimental design.
Researchers in each school of thought have tended to test their model under specific conditions
that may be particularly conducive to producing the model's predicted result pattern, but
without explicating these conditions. Evidence for conversion by minorities and compliance to
majorities was usually demonstrated under conditions where the influence group's position was
either clearly discrepant from recipients' initial attitude on involving issues (see Moscovici,
1976, 1980) or even conflicting with "physical reality", such as when the influence group
claimed to perceive blue slides as green (e.g., Moscovici, Lage, & Naffrechoux, 1969).
Research in the consensus tradition usually focused on novel issues (e.g., Mackie, 1987), where
recipients could not have any clear expectations about the amount of consensus on the issue;
also, to demonstrate that social conflict is not a necessary condition for consensus to affect
judgments, researchers have studied influence groups that were of little social relevance to
participants (Erb et al., 1998, 2002).
One might thus be tempted to conclude that a useful way of delimiting the scope of each
theory would be by applying each to particular topics and / or types of group. However, we
believe that such a distinction by content areas would be less satisfactory than a distinction
based on psychological constructs and processes (see Kruglanski & Thompson, 1999, for a
similar argument regarding modes of persuasion). One psychological construct that seems to be
useful in distinguishing between the conversion theory and objective consensus approaches is
perceived message discrepancy, as was shown by Erb and his colleagues (2002). Holding topic
Persuasion by Majorities and Minorities
6
and influence group constant, these researchers independently varied level of consensus and
message discrepancy. They predicted that minority sources would instigate greater message
scrutiny than majority sources for highly discrepant messages, whereas the opposite pattern
would obtain for modrately discrepant messages. Two independent studies supported these
predictions, showing that message discrepancy indeed moderates the effects of consensus on
message processing.
In the present paper we extend this research by studying a different moderator of
consensus effects, namely the framing of the influence setting as one in which accuracy versus
social comparison concerns prevail. Furthermore, we use an expanded experimental paradigm
by studying both unambiguous and ambiguous messages. Before describing the method in
detail, we will elaborate on each of these aspects and present our specific hypotheses.
The Present Research
We generally predict that message processing will depend on whether recipients are
mainly concerned with potential social conflict, or mainly with determining message validity on
the basis of consensus information. In natural situations, the relative strength of these different
concerns may depend on various factors, of which the amount of message discrepancy is just
one. To test their impact in a more generic way than had previously been done, we used a
priming procedure to activate either social comparison concerns or accuracy concerns, holding
constant factors such as the topic, the discrepancy of the message, and the influence group.
If perceived consensus is used as a cue to message validity whenever accuracy concerns
prevail, then consensus information should produce similar effects as other validity cues, for
example information about source expertise (e.g., Bohner et al., 1995; Chen & Chaiken, 1999).
We therefore modeled our design after recent research in which the effects of expertise cues on
message processing had been investigated (Bohner, Ruder, & Erb, 2002). Drawing on the
Persuasion by Majorities and Minorities
7
heuristic-systematic model's predictions about the interplay of source information and message
arguments (see Bohner et al., 1995), Bohner and his colleagues proposed that information about
low vs. high source expertise would create corresponding expectancies about message validity.
The effect that these expectancies would have on message processing was predicted to depend
on features of the message. For ambiguous messages, biased processing consistent with
expectancies was predicted, with resulting attitudes being thus assimilated to the valence of the
source information. For unambiguously strong or weak messages, however, expectancies were
predicted to function as a standard of comparison, resulting in contrast effects if message
quality was discrepant from the expectancies induced by the expertise cue (i.e. strong
arguments coming from a nonexpert or weak arguments coming from an expert). Bohner et al.'s
results supported these hypotheses: Message recipients reported less favorable thoughts and
attitudes after reading weak arguments presented by a high (vs. low) expertise source, but
reported more favorable thoughts after reading strong arguments presented by a low (vs. high)
expertise source. Conversely, recipients' thoughts and attitudes were more favorable when a
high expertise source presented ambiguous arguments than when a low expertise source did so.
Following up on the above considerations and findings, the study we conducted featured
an extended version of the experimental paradigm by Bohner et al. (2002). Participants in
whom either social comparison concerns or accuracy concerns had been primed received
information about either high or low source consensus and then read a message containing
strong, ambiguous, or weak arguments. Later they reported their thoughts about the message
and their attitudes toward the message position.
We hypothesized that recipients in whom social comparison concerns were primed
would show a processing pattern as predicted by conversion theory. The hallmark of this
pattern is systematic processing of minority messages with agreement being a direct function of
Persuasion by Majorities and Minorities
8
argument quality, but general agreement with majority messages independent of their quality.
In the case of minority messages, we should therefore observe that cognitive responses and
post-message attitudes are most positive if the minority's arguments are strong, less positive if
those arguments are ambiguous, and the least positive if those arguments are weak. In the case
of majority messages, however, strong, ambiguous, and weak arguments should cause hardly
any differences in the valence of cognitive responses and all lead to relatively high agreement.
This specific interaction pattern of argument quality and source status is depicted in Figure 1
(Panel b). The numbers in Figure 1 represent the weights for a priori contrasts to be used in data
analysis.
By contrast, we hypothesized that recipients in whom accuracy concerns are primed use
consensus information as a cue to message validity ("high consensus implies correctness") in
much the same way as recipients in the studies by Bohner et al. (2002) had used expertise
information. Their initial consensus-based expectancies should interact with their subsequent
processing of message arguments. Specifically, in the case of ambiguous arguments we should
observe an assimilative processing bias as a function of consensus-based expectancies, with a
majority source leading to more positive cognitive responses and attitudes than a minority
source. In the case of unambiguous strong or weak arguments, however, we should observe
contrast effects if consensus-based expectancies and message quality are discrepant: In these
conditions (i.e. weak arguments coming from a majority or strong arguments coming from a
minority), thoughts and attitude judgments should be opposite to the cue-based expectancies,
with minorities leading to more positive cognitive responses and attitudes than majorities (see
Bohner et al., 1995). This specific interaction pattern of argument quality and source status is
depicted in Figure 1 (Panel a). The numbers again represent the weights for a priori contrasts to
be used in data analysis.
Persuasion by Majorities and Minorities
9
Priming Accuracy Concerns versus Social Comparison Concerns: A Note on Methodology
Our aim of creating concerns with accuracy versus social comparison in our participants
met with some difficulty, because the manipulation had to be more subtle than those used in
previous research. Simply instructing participants to pay close attention to content or
intensifying their feeling of accountability (cf. Tetlock, 1985) to raise accuracy concerns, for
examples, might easily result in systematic processing of message content independent of
minority or majority source status. Similarly, presenting an identity-relevant in-group as
message source and directly instructing participants to think about their social relationship to
that group to raise social comparison concerns might have eliminated any effect of message
content. We therefore resorted to the use of an influence group that (a) could be presented as
knowledgeable about the persuasion topic, thereby raising accuracy concerns in a relatively
subtle manner, and (b) did not constitute a clear-cut in-group but could be presented as similar
to participants, thereby raising social comparison concerns in a subtle manner. We are aware
that this manipulation of knowledgeability versus similarity may be open to alternative
interpretations, but they seemed sufficiently subtle to enable an initial investigation of our
hypotheses without overshooting the mark.
Method
Adapting and extending the experimental paradigm used by Bohner et al. (2002), we
exposed participants to strong, weak, or ambiguous arguments from either a majority or a
minority source. In addition, we primed participants to perceive the source either mainly in
terms of their knowledge regarding the domain of interest, thus making accuracy concerns
salient, or mainly in terms of their similarity to the participants, thus making social comparison
concerns salient. The main dependent variables assessed were participants' cognitive responses
to the message and their post-message attitudes. The expected result patterns for each priming
Persuasion by Majorities and Minorities
10
condition, as outlined above, are represented in Figure 1; the numbers in the figure represent
contrast weights that will be applied in the analyses.
Participants and Procedure
Two hundred and forty two students at Loyola University Chicago participated for
course credit in a study allegedly dealing with "text comprehension". Participants were
randomly assigned to the conditions of a 2 (priming: accuracy, social comparison) x 2 (source
status: minority, majority) x 3 (argument quality: strong, weak, ambiguous) betweenparticipants factorial design. The number of participants per condition ranged from18 to 21.
They were asked to read a message about the construction of a traffic tunnel in the city of
Rotterdam, and learned that the message source was a group of Rotterdam university students.
Rotterdam students were described either as knowledgeable about the issue or as similar to
Loyola students; participants also learned that the arguments they would read were endorsed by
either a minority or a majority of Rotterdam students. Then participants were asked several
questions related to their expectations of the source. Next, participants had one minute to read a
strong, weak, or ambiguous message favoring the construction project. Then they were asked to
list all of their thoughts that had come to mind while reading the message. Finally, participants
reported their attitudes toward building the tunnel. At the completion of the study, participants
were debriefed and thanked for their participation.
Independent Variables
Source status and salience manipulation. Prior to reading the message in favor of
building the tunnel, participants received information about the source. In the accuracy priming
conditions, the following description was given: "The students at the University of Rotterdam
are both active in and informed regarding local politics and environmental issues in Rotterdam.
They regularly attend urban planning and development meetings to voice their opinions and
Persuasion by Majorities and Minorities
11
stay abreast of current issues. They have researched both the environmental and economic
consequences of building the tunnel." In the social comparison priming conditions, the
following description was given: "These students, much like you, are members of an
introductory psychology class. They live, study, and work in a highly populated city that is
culturally diverse and rich in both business and cultural opportunities. Their interests and
activities are quite similar to students here at Loyola." Half of the participants were given the
majority source version of the description, which indicated that 85% of the Rotterdam students
were in favor of the proposal, while the other half were given the minority source version,
which indicated that this was true of 15% of the Rotterdam students.
Argument quality. The message contained four arguments. Depending on condition, all
four arguments were strong, all four arguments were weak, or two arguments were strong and
two were weak (ambiguous). For example, in the strong argument condition, participants read
that "the tunnel will bring great advantages for residents because the traffic will be reduced by
about 80%" and that "the volume of construction orders will increase by about 45%". In the
weak argument condition, participants read that "the tunnel will bring great advantages for
residents because the traffic will be reduced by about 4%" and "the volume of construction
orders will increase by about 0.6%". In the ambiguous argument condition, arguments taken
from the strong and weak versions (two each) were presented in alternating order. The message
versions had been extensively pilot tested and used in previous research (see Bohner et al.,
2002; Erb et al, 1998).
Dependent Variables
Manipulation checks: Expected message validity and perceived similarity of source.
After receiving information about the source but before reading the message, participants
responded to one filler question ("How well informed are you personally about the topic 'tunnel
Persuasion by Majorities and Minorities
12
construction in Rotterdam'?"). Then they answered three questions designed to assess their
expectancies regarding message validity (“How valid will the students’ arguments be?”; “How
convincing will the students’ arguments be?”; “How expert do you think the students are with
respect to the topic 'tunnel construction in Rotterdam'?”) and one question designed to assess
the perceived similarity of the source to themselves (“How similar are the college students from
University of Rotterdam to Loyola college students, like you?”). Each question was
accompanied by a response scale ranging from 1, "not at all" to 9, "very". It should be noted
that these manipulation checks, although not directly assessing concerns with accuracy or social
comparison, tap into processes assumed to underlie the effects predicted by the objective
consensus and conversion models, respectively. According to the consensus model, recipients
form expectancies of message validity that guide their processing, and according to conversion
theory, recipients are concerned with the similarity of self and influence group.
Thought listing. After reading the message, participants were given three minutes to list
any thoughts they had had while reading the text. To do so, they received a sheet with ten boxes
and were asked to use one box per thought. It was pointed out that they could use as many or as
few boxes as they thought appropriate. After listing their thoughts, participants were asked to
read over their thoughts and place a "+" sign in the margin if the thought was positive or in
favor of the text, the tunnel project, or the students. Likewise, they were asked to place a "–"
sign in the margin if the thought was negative or against the text, the tunnel project, or the
students.
Attitudes. After completing the thought listing task, participants were asked to indicate
their attitude toward the construction of the traffic tunnel in Rotterdam on four items: “The
tunnel in Rotterdam should be built”; “Building the tunnel would bring great advantages for the
economy”; “The tunnel would bring great advantages for the residents”; and “Harmful
Persuasion by Majorities and Minorities
13
consequences of exhaust fumes could be reduced considerably by the tunnel”. All responses
were made on nine point rating scales ranging from 1, “do not agree at all”, to 9, “completely
agree”.
Results
Manipulation Checks
Expected message validity. An index was formed by averaging participants' responses to
the three items on expected message validity (Cronbach's  = .78); this index was subjected to a
2 (priming: accuracy, social comparison) by 2 (source status: minority, majority) by 3
(argument quality: strong, weak, ambiguous) analysis of variance (ANOVA; MSE = 1.84).
Participants for whom accuracy was primed (M = 5.73) reported higher expectations regarding
message validity than did participants for whom social comparison was primed (M = 5.23), F
(1, 229) = 8.10, p = .005. Also, participants in the majority condition (M = 5.67) reported higher
expectations regarding message validity than did participants in the minority condition (M =
5.29), F (1, 229) = 4.53, p = .034. Although the priming by source status interaction was not
significant, F (1, 229) = 1.16, the simple effect of source status was significant only when
accuracy had been primed (Mmajority = 6.00 vs. Mminority = 5.45), F (1, 229) = 4.94, p = .027, but
not when social comparison had been primed (Mmajority =5.32 vs Mminority = 5.13), F < 1. Thus,
the priming manipulation worked as intended: Accuracy priming led to higher expectations of
message validity, in particular if the source was a majority.
Perceived similarity. An ANOVA on the single item assessing perceived similarity
(MSE = 3.59) yielded only the intended main effect of priming: Participants for whom social
comparison was primed (M = 6.11) rated the source as being more similar to themselves than
did participants for whom accuracy was primed (M = 5.20), F(1, 229) = 13.97, p < .001.
Overall, then, the salience manipulation produced the appropriate pre-message expectations.
Persuasion by Majorities and Minorities
14
Attitudes and Cognitive Processing
The four attitude items were averaged to form a composite attitude score (Cronbach’s 
= .82). A valenced index of cognitive responses was formed by subtracting, for each participant,
the number of negative thoughts generated from the number of positive thoughts generated.
Both variables were subjected to a series of contrast analyses associated with our hypotheses.
First, for each dependent variable we computed overall contrast effects for all twelve
conditions, using the contrast coefficients of Figure 1. In a second step, we analyzed the data
from each priming condition separately, again using the contrast coefficients representing the
hypothesized pattern (i.e. Figure 1, Panel a for the accuracy conditions and Figure 1, Panel b for
the social comparison conditions). Finally, we applied the alternative contrast coefficients (i.e.,
Figure 1, Panel a for the social comparison conditions and Figure 1, Panel b for the accuracy
conditions) in order to explore how well each model could explain the data in conditions
"outside of its scope". For each contrast analysis, we also performed tests of residual betweencondition effects to assess not only whether contrasts were significant, but also whether they
provided a good fit to the observed data (see Abelson & Prentice, 1997).
Thought valence. The pattern of means for thought valence closely matched our
predictions (see Figure 2). The overall contrast was highly significant, F (1, 228) = 13.37, p <
.001 (MSE = 17.26). It also fit the data very well, as indicated by a nonsignificant test of
residual between-condition effects, F (10, 228) = 0.41, ns.
The results of separate contrast analyses for each priming condition are presented in
Table 1. When accuracy was primed, the z-shaped contrast effect predicted by the consensus
model was significant as predicted and represented the observed data pattern very well (see
Table 1 and Figure 2, Panel a), whereas the conversion theory contrast was not significant, also
as predicted. Thus, minorities produced more positive thoughts than did majorities for both
Persuasion by Majorities and Minorities
15
strong and weak arguments (a contrast effect), but produced less positive thoughts than
majorities for the ambiguous arguments (an assimilation effect). Conversely, when social
comparison was primed, participants seemed to process the minority messages more carefully
than the majority messages (see Table 1 and Figure 2, Panel b), the results resembling more the
fan-shaped pattern predicted by conversion theory. Here, the consensus model contrast
explained hardly any variance, whereas the conversion theory contrast was stronger as
predicted, although, considered on its own, it failed to reach significance.
An unexpected feature of the results under social comparison priming is the seeming
reversal of strong versus weak message means (M = 1.22 vs. 2.25) for the majority source.
However, this difference was nonsignificant, t(228) = -0.76, p = .447, for the pairwise
comparison. (Statistical power to detect a medium sized effect was .97; Faul & Erdfelder,
1992).
Attitudes. The pattern of means for participants' attitudes is presented in Figure 3. An
inspection of Figure 3 suggests that the data pattern under social comparison priming closely
matched our predictions: Participants' attitudes were differentially affected by the quality of
minority arguments, whereas the majority messages were all similarly persuasive regardless of
their quality (see Figure 3, Panel b). Under accuracy priming the match appears to be less good.
Although the attitude data conformed to the expected pattern for the strong and ambiguous
arguments, majorities presenting weak arguments unexpectedly tended to lead to more, rather
than less, positive attitudes than minorities (see Figure 3, Panel a), although this difference is
much less pronounced than under social comparison priming. We will get back to this point in
the Discussion section. This slight deviation from predictions notwithstanding, the overall
contrast was both highly significant, F (1, 228) = 30.77, p < .001 (MSE = 2.20), and provided a
good fit to the data, as indicated by a nonsignificant test of residual between-condition effects,
Persuasion by Majorities and Minorities
16
F (10, 228) = 0.97, ns.
The results of separate contrast analyses for each of the priming conditions (see Table 1)
were as follows: When accuracy was primed, the consensus model contrast was significant as
predicted and explained a larger proportion of variance than the conversion theory contrast,
although the latter was significant as well. Residual effects were significant in each case,
indicating that neither contrast fully captured the differences between conditions. When social
comparison was primed, the conversion theory contrast was significant as predicted and
provided a good fit to the data, its residual variance test being nonsignificant. The consensus
model contrast, as expected, was not significant and did not fit the observed data pattern, as
indicated by a highly significant residual effect.
Discussion
The results of this study were generally in line with our predictions. Most importantly,
the processing consequences of minority and majority source information differed depending
on the context in which messages were embedded. When social comparison concerns had been
made salient by pointing out similarities between the source and the recipients, the results were
in line with the predictions of conversion theory (Moscovici, 1980) in combination with the
insights of process models of persuasion (e.g., Bohner et al., 1995): Recipients thoroughly
scrutinized minority arguments, and both their thoughts and their attitude judgments reflected
the quality of those arguments; however, recipients accepted majority arguments independent of
their quality.
It is important to note that this pattern emerged with a novel issue of low personal
relevance, in the absence of any social or perceptual conflict, and in a setting where any direct
interaction between recipients and the influence group was obviously impossible. In other
words, none of the conditions that were often described as necessary for conversion and
Persuasion by Majorities and Minorities
17
compliance effects to occur was present. Our results thus suggest that high salience of social
comparison concerns may be sufficient for such effects to occur. This both clarifies and
expands the range of conditions for which conversion theory makes valid predictions.
When accuracy concerns had been made salient by referring to the expertise of the
influence group, a different result pattern emerged. Consensus information functioned in much
the same way as did expertise information in previous work (Bohner et al., 2002). Recipients
formed expectancies about message validity based on the amount of consensus present; then
they processed the message arguments in light of these expectations, which resulted in
assimilation effects for the ambiguous message, and in contrast effects for the unambiguous
message when argument quality was discrepant from expectations. However, the contrast
pattern emerged more clearly for thought valence than for attitudes. Specifically, weak
arguments presented by a majority still caused moderately positive atttitude judgments, even
though, as predicted, the cognitive responses to the message were the most negative in this
condition. It seems possible then, that majorities have the power to cause favorable judgments
despite the presence of other, more negative information (see, e.g., Darke, Chaiken, Bohner,
Einwiller, Erb, & Hazlewood, 1998). The somewhat mixed data at the group level may indicate
that different subgroups of participants were differentially responsive to the priming, such that
our idealized predicted patterns failed to be fully supported. Another possibility is that both
consensus-based expectancies and social comparison processes were driving the results in the
accuracy priming conditions for all participants, with different weights of each applied to the
different dependent variables. In other words, a majority's normative influence may come into
play even when accuracy concerns prevail, preventing people from making negative attitude
judgments although their issue-related thoughts are rather negative. Further research is certainly
necessary to follow up on these conjectures.
Persuasion by Majorities and Minorities
18
An interesting lead for potential follow-up investigations comes from a comparison of
the present results with the precursor study by Bohner et al. (2002). Using high vs. low
expertise cues, these authors also found the most clear-cut z-shaped pattern in their thought
valence results, whereas the pattern of post-message attitudes was less congenial to predictions.
The divergence of the attitude data, however, took an interestingly different shape in the two
studies. Specifically, our present data suggest that a majority presenting weak arguments failed
to produce more negative attitudes than a minority presenting weak arguments; the predicted
contrast effect seemed to be absent. An expert presenting weak arguments, however, did
produce a contrast effect, being much less influential than a non-expert presenting the same
weak arguments (see Bohner et al., 2002, Figure 5). Conversely, Bohner et al. found that a nonexpert with strong arguments failed to evoke positive contrast, being no more influential than
an expert with strong arguments, whereas in the present study the predicted contrast pattern was
observed for a minority presenting strong arguments. Thus, it may be the case that high
expertise coupled with weak arguments generally violates expectancies to a greater degree than
does low expertise coupled with strong arguments. The opposite may generally be true for
consensus, where high consensus coupled with weak arguments may violate expectancies less
than low consensus coupled with strong arguments. This possible asymmetry in persuasion cues
should be addressed in future studies.
At a more general level, our findings demonstrate once more that persuasion is a
complex sequential process in which inferences drawn from early information may alter the
way in which subsequent information is perceived and elaborated (see Bohner et al., 1995;
Chen & Chaiken, 1999; Kruglanski & Thompson, 1999; Petty & Wegener, 1998). One
powerful type of "early information" in this respect are source cues; whereas previous studies
have repeatedly demonstrated that different cues may evoke different inferences, our present
Persuasion by Majorities and Minorities
findings extend this line of research by showing that even identical source cues may have
dramatically different effects depending on the social context in which they are evaluated.
19
Persuasion by Majorities and Minorities
20
References
Abelson, R. P., & Prentice, D. A. (1997). Contrast tests of interaction hypotheses.
Psychological Methods, 2, 315-328.
Bohner, G., Erb, H.-P., Reinhard, M.-A., & Frank, E. (1996). Distinctiveness across topics in
minority and majority influence: An attributional analysis and preliminary data. British
Journal of Social Psychology, 35, 27-46.
Bohner, G., Frank, E., & Erb, H.-P. (1998). Heuristic processing of distinctiveness information
in minority and majority influence. European Journal of Social Psychology, 28, 855860.
Bohner, G., Moskowitz, G. B., & Chaiken, S. (1995). The interplay of heuristic and systematic
processing of social information. European Review of Social Psychology, 6, 33-68.
Bohner, G., Ruder, M., & Erb, H.-P. (2002). When expertise backfires: Contrast and
assimilation effects in persuasion. British Journal of Social Psychology, 41, 495-519.
Chen, S., & Chaiken, S. (1999). The heuristic-systematic model in its broader context. In S.
Chaiken & Y. Trope (Eds.), Dual-process theories in social psychology (pp. 73-96).
New York: Guilford.
Darke, P. R., Chaiken, S., Bohner, G., Einwiller, S., Erb, H.-P., & Hazlewood, J. D. (1998).
Accuracy motivation, consensus information, and the law of large numbers: Effects on
attitude judgment in the absence of argumentation. Personality and Social Psychology
Bulletin, 24, 1205-1215.
Erb, H.-P., & Bohner, G. (2001). Mere consensus in minority and majority influence. In N. K.
De Vries & C. K. W. De Dreu (Eds.), Group consensus and minority influence:
Implications for innovation (pp. 40-59). Oxford, UK: Blackwell.
Erb, H.-P., Bohner, G., Rank, S., & Einwiller, S. (2002). Processing minority and majority
Persuasion by Majorities and Minorities
21
communications: The role of conflict with prior attitudes. Personality and Social
Psychology Bulletin, 28, 1172-1182.
Erb, H.-P., Bohner, G., Schmälzle, K., & Rank, S. (1998). Beyond conflict and discrepancy:
Cognitive bias in minority and majority influence. Personality and Social Psychology
Bulletin, 24, 620-633.
Faul, F., & Erdfelder, E. (1992). GPOWER: A priori, post-hoc, and compromise power
analyses for MS-DOS [Computer program]. Bonn: University of Bonn, Department of
Psychology. [On-line: http://www.psycho.uni-duesseldorf.de/aap/projects/gpower/]
Kelley, H. H. (1967). Attribution theory in social psychology. In D. Levine (Ed.), Nebraska
Symposium on Motivation (Vol. 15, pp. 192-238). Lincoln: University of Nebraska
Press.
Kruglanski, A. W., & Thompson, E. P. (1999). Persuasion by a single route: A view from the
unimodel. Psychological Inquiry, 10, 83-109.
Maass, A., & Clark, R. D. III. (1984). Hidden impact of minorities: Fifteen years of minority
influence research. Psychological Bulletin, 95, 428-450.
Martin, R., & Hewstone, M. (2003). Majority versus minority influence: When, not whether,
source status instigates heuristic or systematic processing. European Journal of Social
Psychology, 33, 313-330.
Moscovici, S. (1976). Social influence and social change. London: Academic Press.
Moscovici, S. (1980). Toward a theory of conversion behavior. Advances in Experimental
Social Psychology, 13, 209-239.
Moscovici, S., Lage, E., & Naffrechoux, M. (1969). Influence of a consistent minority on the
responses of a majority in a color perception task. Sociometry, 32, 365-380.
Nemeth, C. J. (1986). Differential contributions of majority and minority influence.
Persuasion by Majorities and Minorities
22
Psychological Review, 93, 23-32.
Tetlock, P. E. (1985). Accountability: The neglected social context of judgment and choice.
Research in Organizational Behavior, 7, 297-332.
Wood, W., Lundgren, S., Ouellette, J. A., Busceme, S., & Blackstone, T. (1994). Minority
influence: A meta-analytic review of social influence processes. Psychological Bulletin,
115, 323-345.
Ziegler, R., Diehl, M., Zigon, R., & Fett, T. (2004). Source consistency, distinctiveness, and
consensus: The three dimensions of the Kelley ANOVA model in persuasion.
Personality and Social Psychology Bulletin, 30, 352-364.
Persuasion by Majorities and Minorities
23
Table 1
Results of Contrast Analyses by Priming Condition.
Accuracy Priming
Contrast
Fa
Social Comparison Priming
Residual Effects
Fb
Contrast
Fa
Residual Effects
Fb
0.49
0.01
1.75+
2.22
1.53
2.01
1.55
Consensus Model Contrast
8.47**
3.19***
2.49
3.79***
Conversion Theory Contrast
6.22*
3.42***
Thought Valence
Consensus Model Contrast
Conversion Theory Contrast
12.56***
Attitudes
27.63***
1.28
Note: Statistics for those contrasts that were predicted to be significant and to fit the data well
are shown in boldface.
a
Degrees of freedom are 1 and 228 for thought valence, and 1 and 230 for attitudes.
b
Degrees of freedom are 10 and 228 for thought valence, and 10 and 230 for attitudes.
*** p < .001; ** p < .01; * p < .05; + p < .10.
Persuasion by Majorities and Minorities
Figure Captions
Figure 1. Predicted results for accuracy priming, based on the consensus approach (Panel a),
and for social comparison priming, based on conversion theory (Panel b).
Figure 2. Pattern of means of thought valence in accuracy priming (Panel a) and social
comparison priming conditions (Panel b).
Figure 3. Pattern of means of attitudes in accuracy priming (Panel a) and social comparison
priming conditions (Panel b).
24
Persuasion by Majorities and Minorities
(a) Pattern Predicted for Accuracy Priming
4
3
+3
Response
2
+1
+1
1
Strong
Ambiguous
Weak
0
-1
-1
-1
-2
-3
-3
-4
Minority
Majority
Source
(b) Pattern Predicted for Social Comparison Priming
2
Response
1
+1
+1
+1
+1
0
-1
-1
-2
-3
-3
-4
Minority
Majority
Source
Strong
Ambiguous
Weak
25
Persuasion by Majorities and Minorities
26
Thought Valence
(a) Thought Valence Results in Accuracy Priming Conditions
3.5
3
2.5
2
1.5
1
0.5
0
-0.5
-1
3.10
2.19
2.16
0.70
0.21
Strong
Ambiguous
Weak
-0.86
Minority
Majority
Source
(b) Thought Valence Results in Social Comparison Priming Conditions
4
Thought Valence
3
2
2.25
2.25
2.05
1.33
1.22
1
0
0.05
-1
Minority
Majority
Source
Strong
Ambiguous
Weak
Persuasion by Majorities and Minorities
27
(a) Attitude Results in Accuracy Priming Conditions
8
Attitude
7
6
6.79
6.46
6.47
6.06
5.43
5.77
Strong
Ambiguous
Weak
5
4
Minority
Majority
Source
(b) Attitude Results in Social Comparison Priming Conditions
8
6.93
6.61
6.33
Attitude
7
6.48
6
5.71
5
4.74
4
Minority
Majority
Source
Strong
Ambiguous
Weak
Download