Social Constraint and Self-Doubt: Mechanisms of Social Network
inﬂuence on Resistance to Persuasion
Lindsey Clark Levitan
Shepherd University
Those around us have a profound inﬂuence on our political attitudes and attitude strength, such that people
whose social networks include a variety of perspectives have weaker, less deeply entrenched attitudes than
those who are surrounded by like-minded others. In particular, those embedded in attitudinally heterogeneous
networks are more open to changing their views. The nature and mechanisms of this network inﬂuence on
openness to attitude change remain unclear. A survey experiment examines two mechanisms proposed by prior
literature: (1) social doubt triggered by network members’ dissent and (2) social constraint to maintain similar
attitudes. It also provides some data on the more commonly assumed mechanism, (3) information exchange.
Results strongly support social constraint and are mixed on social doubt. This contrasts with the theoretical
emphasis of much previous interdisciplinary social network research, which has focused primarily on
information exchange, to the detriment of other mechanisms. Findings also indicate that like-minded social
network members solidify attitudes at least as much as dissent erodes them, suggesting that prior emphasis on
the inﬂuence of heterogeneous rather than attitudinally congruent networks is overstated. Implications for
political movements are discussed.
KEY WORDS: social networks, social inﬂuence, attitude strength, persuasion
Although a well-informed public making considered judgments is central to the democratic ideal,
in reality the public often falls short of this ideal (see Taber, 2003). For example, a large minority of
the public continued to judge President Obama to be foreign-born well after his election, despite all
available evidence (NYT/CBS poll, 2011; PPP poll, 2012). It is tempting to blame this on lack of
political sophistication, lack of access to information, or general disinterest, but these explanations
ring hollow in cases like this, where the evidence is so widely publicized and easy to comprehend.
The current research examines how the strength and endurance of our attitudes is inﬂuenced by those
around us. Speciﬁcally, a social environment in which one is surrounded by like-minded others may
suppress people’s willingness to change their attitudes, perhaps even in the face of valid and convincing
evidence. Concerns about maintaining positive relationships may make those embedded in attitudinally
congruent networks reluctant to contemplate incongruent information and amend their views
compared to those in more attitudinally diverse social networks.
The role of network attitudinal heterogeneity in opening our views to change has been repeatedly
established (e.g., Levitan & Visser, 2009; Ryan, 2013; Visser & Mirabile, 2004), but the mechanisms
of this inﬂuence remain relatively obscure, despite the important normative implications. If this
change arises from thoughtful consideration of information or arguments provided by others, this has
0162-895X VC 2017 International Society of Political Psychology
Published by Wiley Periodicals, Inc., 350 Main Street, Malden, MA 02148, USA, 9600 Garsington Road, Oxford, OX4 2DQ,
and PO Box 378 Carlton South, 3053 Victoria, Australia
Political Psychology, Vol. , No. , 201
doi: 10.1111/pops.12458
bs_bs_banner
957
39 4 8
positive implications for democracy (see Fishkin, 1991). If, on the other hand, this effect stems from
more purely social pressures, the outlook may be somewhat less rosy.
Although little research directly addresses this issue, several plausible mechanisms have
been proposed, one or more of which may be at work: (1) information exchange with social network
members, (2) a metacognitive process in which merely knowing that others disagree causes
self-doubt, and (3) a purely social process of constraint by which alternate viewpoints are suppressed
to maintain social harmony. Research evidence is ambiguous, however. Previous
research, particularly in political science, has often assumed an information exchange mechanism
and modeled how information spreads though networks (e.g., Huckfeldt, Beck, Dalton, &
Levine, 1995). Yet it is a mistake to assume that this is the only way network members can inﬂuence
us, as a wealth of psychological research attests. The remaining two proposed mechanisms
both have empirical support, but the same ﬁndings can be interpreted as supporting multiple
mechanisms and so provide deﬁnitive evidence for none. The current research will clarify the
mechanisms of social network inﬂuence by comparing them within a single study. Experimental
manipulations will be leveraged to focus on the two non-information-exchange mechanisms to
establish their viability in explaining network inﬂuences.
Tightly entangled with this question of mechanism lurks a concern so basic that it has generally
been implicitly answered in researchers’ writing, rather than explicitly examined in research:
Which types of networks are inﬂuential in opening or closing our minds to alternate perspectives?
Do attitudinally heterogeneous networks encourage weaker, more malleable attitudes, or
do congruent networks strengthen them, leaving people less open to alternate views than they
would be if left to their own devices? Although research has established that both heterogeneous
and congruent groups of strangers have effects in an in-person persuasion context (e.g., Klar,
2014), there is sparse research directly addressing which of the opposing ends of the heterogeneity
spectrum drive naturally occurring network effects, particularly on openness to outside
perspectives.
The current research proposes that being surrounded by like-minded others suppresses people’s
willingness to change their opinions. In particular, it proposes that concerns about maintaining positive
relationships make those surrounded by like-minded others more reluctant to reconsider their
views than those in more attitudinally diverse social networks, thereby suppressing attitude change
that might otherwise freely occur.
Social Network Inﬂuence
Political attitudes are not maintained in isolation. The inﬂuence of those around us has long been
studied, particularly with respect to close social network members with whom we interact regularly
and discuss important matters (Berelson, Lazarsfeld, & McPhee, 1954; Huckfeldt, 2014; Huckfeldt &
Sprague, 1987). Our social networks socialize us politically (Settle, Bon, & Levitt, 2011), inﬂuencing
our attitudes (Huckfeldt, Mendez, & Osborn, 2004), votes (e.g., Ryan, 2010), and behavior (e.g.,
McClurg, 2006; Mutz, 2002a), and we carry this socialization with us as we consider political
situations.
Of particular importance are network effects that strengthen or weaken attitudes. “Strong
attitudes” are deﬁned as attitudes that are resistant to persuasion, persistent over time, and which inﬂuence
thought and behavior (Krosnick & Petty, 1995), whereas weak attitudes are more ﬂexible and
less impactful. A wealth of evidence demonstrates that individuals’ attitudes are more malleable and
less impactful when some network members hold differing attitudes (attitudinally heterogeneous networks)
than when all network members agree with them (attitudinally congruent networks). Those in
congruent networks are more likely than those in heterogeneous networks to show an impact of their
attitudes on behaviors including vote choice (Ryan, 2010) and political participation (Mutz, 2002a),
more likely to have stable attitudes (Huckfeldt & Sprague, 2000; Levitan & Visser, 2009), and more
2 Levitan958
likely to be resistant to persuasion (Visser & Mirabile, 2004). Overall, those embedded in more likeminded
networks have stronger attitudes that are less open to change than those in more attitudinally
heterogeneous networks.
One might assume that this relation is spurious or reverse-causal, with those who are most
invested in their attitudes choosing to associate with similar others. Ample research, however, supports
the causal role of network composition. Social networks are generally formed based on social
and demographic characteristics (Sinclair, 2012) and sheer proximity (Festinger, Schachter, & Back,
1950) rather than political agreement. Experimental studies demonstrate that randomly assigned networks
inﬂuence attitude change (e.g., Visser & Mirabile, 2004). Additionally, in-depth longitudinal
and quasi-experimental studies starting at or before a networks’ inception show that individuals’ attitudes
are notably inﬂuenced by those of network members (Lazer, Rubineau, Chetkovich, Katz, &
Neblo, 2010; Levitan & Visser, 2009; see also Bello & Rolfe, 2014). In contrast, the hypothesis that
people select similar others to associate with has been notoriously difﬁcult to conﬁrm outside of limited
laboratory circumstances (see a special 1992 issue of Communications Monographs 59(2) for
thorough discussion; Festinger et al., 1950; Levinger, 1972).
Mechanisms of Network Inﬂuence
Given the at least partially causal inﬂuence of network members on openness to attitude
change, the question then becomes how network members weaken or strengthen attitudes. Several
mechanisms have been suggested, but these mechanisms have often been taken for granted
without thorough testing. Moreover, evidence consistent with each is generally also consistent
with another mechanism, meaning that neither is deﬁnitively supported. As an additional issue,
most research has assumed that speciﬁcally heterogeneous networks inﬂuence openness to attitude
change. The current research seeks to examine these mechanisms of network inﬂuence, as
well as the role of attitudinally heterogeneous versus congruent social networks in encouraging
and suppressing attitude change.
Information exchange with network members is the predominant mechanism assumed in research
on political networks either implicitly or explicitly. As explained by the progenitors of much recent
networks research, “Political behavior may be understood in terms of individuals tied together by, and
located within, networks, groups, and other social formations that largely determine their opportunities
for exchange of meaningful political information” (Huckfeldt & Sprague, 1987, p. 1197; see also
Burt, 1987). In this line of reasoning, network members provide us with information and arguments
supporting their own views. This is expected to reinforce the attitudes of those embedded in congruent
networks, and to generate ambivalence and ultimately persuasion in people embedded in heterogeneous
networks.
Supporting research has shown that individuals whose networks are more heterogeneous are better
informed about opposing positions (Huckfeldt et al., 2004; Mutz, 2002b). Similarly, laboratory
experiments show that preferences can be inﬂuenced by exchanging information with group members
(e.g., Ahn, Huckfeldt, & Ryan, 2010; Klar, 2014). Still, attitude change is not a function of arguments
alone, but of our cognitive responses to those arguments (e.g., accepting, counterarguing; Greenwald,
1968; Petty & Cacioppo, 1986). Moreover, information exchange alone cannot account for the ﬁnding
that individuals in heterogeneous networks are more persuaded by the same arguments than are those
in more congruent networks (Levitan & Visser, 2009; Visser & Mirabile, 2004).
The social doubt mechanism relies more clearly on advances in persuasion research, particularly
dual-process models (Chen & Chaiken, 1999; Petty & Cacioppo, 1986). This mechanism suggests
that networks inﬂuence how deeply individuals consider information, even information they are
exposed to outside of the network context. Speciﬁcally, it proposes that heterogeneous networks instigate
greater systematic processing of available information. “Diversity of opinions among important
others may be taken as a signal that something is amiss, and that one’s attitude needs to be reMechanisms
of Network Inﬂuence 3959
evaluated. . . [motivating] people to scrutinize new attitude relevant information in an effort to identify
and adopt the most valid attitude” (Levitan & Visser, 2008, p. 642). In this case, network members’
attitudes, rather than their arguments, inﬂuence individuals by instilling self-doubt, which motivates
deeper examination of issue-relevant information. People judge the validity of their opinions by comparison
with similar others through social comparison processes (Festinger, 1954). Therefore, merely
knowing that network members hold diverse attitudes, even without knowing why, should instill
socially motivated doubt and encourage attitude reassessment and (where warranted) attitude change.
After all, how could these well-esteemed yet disagreeing network members be so mistaken? Alternately,
if we all agree, how could we all be wrong (e.g., social proof; Cialdini, 1995)?
Supporting evidence for this social doubt perspective shows that the relation between network
attitudinal composition and attitude change is partially mediated by certainty, and ambivalence (Visser
& Mirabile, 2004), such that individuals embedded in heterogeneous networks doubt their attitude
more than individuals embedded in congruent networks. Nor do individuals respond to networks
regardless of their arguments, as might be expected with pure social conformity. Instead, those
embedded in like-minded networks change their attitudes little, even in the face of cogent arguments,
whereas those in more heterogeneous networks scrutinize information and change their attitudes only
for strong arguments (Levitan & Visser, 2008). Moreover, those in heterogeneous social networks are
actually more likely to seek out new information (Levitan & Wronski, 2014) to alleviate their uncertainty.
This indicates that heterogeneous networks instigate deeper, more central, and systematic processing
of relevant information, as would be expected if knowing that network members hold differing
attitudes generates doubt.
Perhaps most interestingly, a social constraint mechanism is also plausible, but it has been understudied
in the network literature. This mechanism proposes that individuals are motivated to maintain
attitudes that are in step with the network (see Visser & Mirabile, 2004), even when network members
are not present. Individuals are chronically motivated to get along with others, a motivation which
continues to have inﬂuence when network members are not present because network members may
later hold them accountable for attitude change that happened in their absence. This idea is rooted in
classic research on social inﬂuence and conformity (Deutsch & Gerard, 1955; Kelley, 1952; Kelman,
1958; Kelman & Hamilton, 1989; Ulbig & Funk, 1999). Importantly, however, this social inﬂuence
need not be conformity as classically studied. Although iconic social inﬂuence research focuses on
bringing deviant members into the fold (e.g., Asch, 1955; Schachter, 1951), social constraint also
affects those who already agree because attitudes can serve a “social adjustment” function (Smith,
1956; see also Katz, 1960). In this case, social constraint can keep people who already agree from
straying, for example, by preventing full consideration of information that might encourage less congenial
attitudes. To date, this potential mechanism has not yet been tested with respect to network
inﬂuence, especially in terms of network composition suppressing change in those who initially agree.
Research consistent with social constraint demonstrates that humans have a deep-seated “need to
belong” (Baumeister & Leary, 1995). When unmet, this need leads to greater sensitivity to social
stimuli (Gardner, Pickett, & Brewer, 2000) in order to foster stronger relationships. Individuals therefore
seek to emulate members of reference groups (Kelley, 1952) in order to maintain positive social
relationships (“identiﬁcation,” Kelman & Hamilton, 1989) and avoid penalties ranging from mild derogation
to ostracism for falling out of step with group members’ views (e.g., compliance, Kelman &
Hamilton, 1989; Schachter, 1951). In short, people worry that others will tease, be angry, or simply
think them stupid. At minimum, individuals tend to remain silent about minority views they hold
(Noelle-Neuman, 1974). At the extreme, they may express views held by group members even in the
face of overwhelming objective evidence (Asch, 1955, although see Ross, Bierbrauer, & Hoffman,
1976). Ultimately, individuals internalize these views (Kelman & Hamilton, 1989), although classic
research often focuses on publicly conformity rather than private acceptance.
4 Levitan960
Such possibilities have received little attention in accounts of network inﬂuence. Still, recent ﬁndings
provide some support in that individuals embedded in attitudinally congruent networks are less
tolerant of alternate viewpoints (Mutz, 2002b, although see Bloom & Bagno-Moldavsky, 2015).
Those whose personalities are especially conﬂict-avoidant also show greater effects of network composition
on political behavior (Mutz, 2002a). Additionally, individuals exhibit lasting attitude change
after exposure to strangers known to hold disagreeing views, even without discussion of the reasons
for those views (Carlson & Settle, 2016; Levitan & Verhulst, 2016). In short, several lines of research
suggest that individuals may come to their views because they want to agree with others, but ultimately,
this social constraint mechanism has not been tested with respect to networks’ inﬂuence on
attitudes and openness to attitude change.
Ambiguous evidence. Unfortunately, the evidence for any one of these processes of network
inﬂuence is far from overwhelming, particularly in the light of the other two. Evidence that individuals
in heterogeneous networks know more about opposition views (supporting information exchange)
may be due to greater information seeking and scrutiny (as with social doubt) or differences in willingness
to examine counterattitudinal information at all (as with social constraint). Correspondingly,
evidence for the social doubt perspective, such as differences in certainty, may be due to differences
in the knowledge received during network interactions (consistent with information exchange) or due
to interpersonal ambivalence (Priester & Petty, 2001) (consistent with social constraint). As for social
constraint, it is a promising direction, but little research has explicitly examined this mechanism for
network inﬂuence. What ﬁndings there are, though suggestive, may be more relevant to responses to
an immediate group in a laboratory than to broader network composition altering an individual’s
mindset and attitude strength in a way that persists when network members are absent.
The Impact of Attitudinally Congruent Social Networks
An important step in disentangling these mechanisms is to reconsider our conceptualization of
network composition to bring the role of attitudinally congruent networks to the forefront. Network
inﬂuence tends to be framed in terms of heterogeneous networks providing information, signaling
potential problems with attitudes, and applying social pressures. Even the commonly used terms
“network heterogeneity” (e.g., Huckfeldt & Sprague, 1987) and “cross-cutting exposure” (Mutz,
2006) emphasize the attitudinally diverse side of the spectrum, rather than the attitudinally congruent
end. This may seem to be a semantic concern, but language constrains thought (Whorf, 1956), and
this terminology lends itself to discussion of models in which heterogeneous networks instigate
change (e.g., by bringing new information), rather than models in which congruent networks inhibit
change (e.g., by blocking consideration of information). Nonetheless, attitudinally congruent network
members may be driving inﬂuences behind network effects.
An effect of any of these mechanisms could theoretically result from support of others who
agree as much as subversion from those who do not, but this is especially true of social constraint.
Individuals’ network compositions tend to range from complete agreement to a mix of
views (e.g., Huckfeldt et al., 2004). These heterogeneous networks, which include some disagreement,
leave room for multiple perspectives because it is less clear what attitudes might be
“unacceptable.” It is difﬁcult to rock a boat already laden with a variety of attitudes, and so one
need not fear holding deviant attitudes. A congruent network, on the other hand, sends a clear
signal about which attitude is appropriate and is less tolerant of alternate views (Laumann, 1973;
Mutz, 2002b). Congruent networks may therefore be the motivating end of the scale, suppressing
openness to alternate perspectives. Thus, network effects may be characterized by congruent network
suppression of natural change, tethering an individual to a particular attitude, whereas heterogeneous
networks may not so much encourage change as fail to prevent attitudes from
drifting or changing in response to outside inﬂuences.
Mechanisms of Network Inﬂuence 5961
Overall, it is not clear whether heterogeneous or congruent networks or both are key in understanding
network-induced attitude change and maintenance, nor does research unambiguously support
any of the proposed mechanisms. It may be that only one of these mechanisms underlies the effect of
network attitudinal composition upon openness to attitude change, or that each independently contributes
to the effect.
The Current Study
The current research seeks to disentangle these mechanisms of social network inﬂuence by use of
a less causally ambiguous experimental methodology. This survey experiment particularly emphasizes
the previously underexamined social constraint mechanism, and the associated possibility that being
embedded in an attitudinally congruent network suppresses change, rather than (or perhaps in addition
to) heterogeneous networks encouraging change.
Multiple issues complicate research into mechanisms of any effect, but best practice is to manipulate
the proposed mediator of the effect rather than to measure the mediator (for detailed discussion,
see Bullock, Green, & Ha, 2010). As such, an experimental manipulation designed to disrupt a proposed
mechanism provides a strong test of the mechanism. If the relation between the independent
variable and the dependent variable is also disrupted, this provides evidence that the proposed mechanism
is indeed at work.
Experimental manipulations were therefore used to disrupt two potential mechanisms, while the
third was assessed correlationally. As described above, the proposed social constraint mechanism
operates through individuals’ need to belong and fear of social sanctions. Individuals in congruent networks
have greater fear of subsequent social sanctions for changing their views, and so avoid attitude
change. This mechanism was disrupted by an experimental intervention reassuring participants about
the strength of their relationships. If social constraint is at work, this should relieve it by minimizing
concern about maintaining relationships when considering attitudes. If participants believe they are
well-liked, and they will maintain positive relationships long into the future, then they do not need to
worry about keeping their attitudes in step with those of network members to maintain those relationships.
This should reduce or eliminate the relation between network composition and attitude strength.
Social doubt was disrupted by reassuring participants about their relative knowledge. Thus, the
dissent of network members can no longer be taken as an indicator that participants may be relatively
uninformed and need to reevaluate their views. If the social doubt mechanism is at work, this should
reduce or eliminate the relation between network composition and openness to attitude change. Participants’
knowledge was also assessed, providing evidence relevant to the information exchange
mechanism.
Additionally, the question of whether heterogeneous networks encourage change or congruent
networks suppress it was assessed by examining the speciﬁc pattern of results when network
inﬂuence was disrupted. If severing the link between networks and attitude change disproportionately
reduces attitude change of those whose networks are heterogeneous, this supports the idea
that heterogeneous networks enable or encourage change. If severing this link disproportionately
increases attitude change within congruent networks, this indicates that congruent networks suppress
attitude change.
Method
In order to uncover the processes behind network inﬂuences on resistance to persuasion, a
national probability sample participated in a survey experiment. Prescreened participants ﬁrst reported
their attitudes and knowledge about an issue (capital punishment) and described network members’
attitudes. Participants then experienced a randomly assigned condition (social afﬁrmation, knowledge
afﬁrmation, or control). Resistance to persuasion was then assessed. Manipulation checks and control
6 Levitan962
variables were also included. An outline of the procedure is depicted in Figure 1. Following participation,
respondents were thanked and debriefed.
Participants and Prescreening
Four hundred and ﬁfty-three participants were recruited from a nationally representative panel
initially recruited through a combination of random-digit-dialing and address-based sampling by
Knowledge Networks.1
Potential participants ﬁrst completed a screening item asking the extent to which they favored or
opposed capital punishment on a fully labeled 7-point bipolar scale. Participants who did not indicate
pro-capital punishment attitudes were excluded from the study, so that the subsequent persuasive message
would be counterattitudinal. (Participants cannot be persuaded to change their attitudes if they
Figure 1. Study procedure. [Color ﬁgure can be viewed at wileyonlinelibrary.com]
1
With funding from Time-sharing Experiments for the Social Sciences (TESS), see www.knowledgenetworks.com.
Mechanisms of Network Inﬂuence 7963
already agree.) As expected based on previous studies, 67% of those screened qualiﬁed for the study
(see Table S1 in the online supporting information for sample characteristics).2
Initial Attitude Measure
After the single-item prescreening, qualiﬁed participants reported their capital punishment attitudes
on an additional three 7-point semantic differential items, with anchors of good/bad, foolish/
wise, and harmful/beneﬁcial. All four items were averaged to form a measure of initial attitudes
(Cronbach’s alpha 5 .86). Participants who did not indicate pro-capital punishment attitudes on this
full four-item measure despite a positive prescreening response were excluded. This measure allows
for a reliable assessment of resistance to persuasion.
Factual Knowledge
Next, participants completed a four-item knowledge battery about capital punishment. Items
follow recommendations for assessing domain-speciﬁc knowledge (Iyengar, 1986) and were pretested
to include both easy and difﬁcult questions. Additionally, some questions whose answer
undermined pro-death-penalty arguments were included (see Appendix S3 in the online supporting
information).
Social Network Composition
Next, social network attitudinal composition was assessed by asking participants to list up to ﬁve
individuals with whom they interact regularly and discuss important matters (used in the General
Social Survey; see Burt, 1984). Eighty percent of participants listed ﬁve network members. They
were then asked to report the capital punishment attitudes of each network member on the screening
item described above.3
Systematic research indicates that individuals are reasonably accurate when
reporting their perceptions across a variety of issues and candidates, particularly for close network
members, (e.g., Goel, Mason, & Watts, 2010; Huckfeldt & Sprague, 1987; Levitan & Visser, 2009;
Mutz, 2006), although it is beyond the scope of this article to assess accuracy of network perceptions.
This article will remain agnostic about whether it is network members’ actual attitudes or individuals’
perceptions of those attitudes that is responsible for the previously established inﬂuence of network
composition. It is likely that participants’ perceptions of their network members are key:
Presumably one cannot be inﬂuenced by views that others keep well-hidden. Nonetheless, the current
study does not offer a test of this possibility, but it instead focuses on the mechanisms of the established
effect.
In order to calculate degree of network heterogeneity, the absolute difference between the
participant’s attitude and those of each network member was calculated, and those differences
were averaged across network members (see Levitan & Visser, 2009). The measure was scaled
to range from 0 (no disagreement with network members) to 1, (maximum possible disagreement).
As with prior studies (e.g., Huckfeldt et al., 2004; Huckfeldt & Sprague, 1987; Levitan &
Wronski, 2014; Visser & Mirable, 2004), network composition was skewed toward attitudinal
2
Several participants were excluded for providing invalid responses, speciﬁcally nonpositive views on the subsequent
full capital punishment measure (22 participants), leaving the IV blank (41), completing it improperly (e.g., listing
“Facebook” as a network member, four participants), leaving the DV blank (1), or requesting that their responses be
ignored (1). With the exception of the ﬁnal two, these responses occurred prior to the manipulation and therefore do
not compromise random assignment.
3
Some participants answered “don’t know” with reference to network member attitudes. They were then prompted to
guess. The analyses presented here do not include guesses, but substantively similar results were obtained when
guesses were included, although results were weaker, as expected with more error-prone measures. Thirty-four participants
reported guessing on all network members. The remaining participants guessed on 18% (< 1 per network) on
average. Additionally, seven individual “network members” were deleted for noncompliance (e.g., listing a group as
one network member). All network measures occurred prior to the manipulations.
8 Levitan964
congruence, with the vast majority falling in networks ranging from complete agreement to an
even mix of views.4
Experimental Manipulations
Participants then experienced one of three randomly assigned conditions: social afﬁrmation,
knowledge afﬁrmation, or control.
Social afﬁrmation manipulation. If social constraint is indeed at work, then people are maintaining
network-congruent attitudes to service belongingness needs. In this case, reassuring participants of
their present and future connectedness will satisfy belongingness needs and reduce fear of rejection
and other social consequences of holding aberrant views. This will disrupt the effects of social constraint
and make individuals more open to alternate attitudes. Participants in the social afﬁrmation
condition (see Appendix S4 in the online supporting information) ﬁrst described a time when they felt
particularly close to network members (adapted from Maner DeWall, Baumeister, & Schaller, 2007).
Previous research using this social acceptance manipulation found signiﬁcantly less desire to
“connect” with others in this condition, which is indicative of satisﬁed belongingness needs (Maner
et al., 2007).5
Participants were then told that, due to requests from prior participants, they would be
given personality feedback purportedly based on their responses. The feedback explained that they
would, in general, have rewarding social relationships, close friends, and a happy marriage (adapted
from Twenge, Baumeister, Tice, & Stucke, 2001, studies 1 and 3, with an added reference to close
friends). This manipulation has been previously used to establish a sense of belonging in contrast with
social rejection and generates results similar to studies which employ acceptance or rejection by live
group members (Twenge et al., 2001). Based on prior research, the social afﬁrmation manipulation
therefore is expected to reassure participants about their social acceptance in their network and establish
a feeling of belongingness and reduce concerns about social rejection.
Knowledge afﬁrmation manipulation. If the social doubt mechanism is in effect, such that those
in attitudinally heterogeneous networks doubt their own judgments more, an intervention telling them
that their level of knowledge on the issue is relatively high should counteract this doubt. Participants
in the knowledge afﬁrmation condition therefore received positive feedback about the earlier knowledge
battery, also purportedly requested by prior participants. Speciﬁcally, participants were told that
they “know more about the death penalty than 91% of Americans” (see Appendix S4 in the online
supporting information). If one knows so much about capital punishment, the dissent of network
members no longer implies that one is misinformed, disrupting social doubt.
Control condition. Control condition participants moved directly to the persuasive message.
Resistance to Persuasion
Next, participants read a counterattitudinal persuasive message regarding capital punishment
taken from Visser and Mirabile (2004, study 3), who adapted it from arguments disseminated by
organizations such as the ACLU (see Levitan & Visser, 2008, Appendix A.1. for full text). Each paragraph
was presented on a separate screen, to facilitate reading. Attitudes were then reassessed on the
same four-item measure as initial attitudes, described above (Cronbach’s alpha 5 .94). Resistance to
persuasion was calculated by subtracting postmessage attitudes from initial attitudes, such that positive
numbers indicate change in the direction of the persuasive message. The result was then scaled to
range from 21 to 1.
4
For further information on the distribution and correlates of network composition, see Appendices S1 and S2 in the
online supporting information.
5
Individuals who claimed to never have been close to network members or left the question blank were excluded from
analyses relevant to this manipulation (26), since they effectively skipped the manipulation. Results were similar,
although predictably weaker, when these participants were included. This occurred prior to receiving any feedback
from a manipulation.
Mechanisms of Network Inﬂuence 9965
Control Variables
To account for the possibility that the generally positive nature of the manipulations might inﬂuence
persuasion (see Schwarz et al., 1991a), participants described their mood on a 7-point, fully
labeled, bipolar scale with anchors of very negative (sad, angry, upset) and very positive (happy,
proud, enthusiastic). To conﬁrm that the knowledge afﬁrmation effectively inﬂuenced perceived
knowledge, participants were asked how knowledgeable they consider themselves to be about capital
punishment on a 5-point scale ranging from not at all to extremely. Demographic information, including
age, had previously been assessed by Knowledge Networks.
Results
In order to replicate and extend prior results, network composition and control variables were
used to predict attitude change in an OLS regression. Age and its quadratic effects were included to
account for life-cycle differences in attitude strength (Visser & Krosnick, 1998). Network composition
marginally predicted persuasion, such that as individuals’ networks became more attitudinally heterogeneous,
their attitudes changed more (Table 1, Model 1). According to a priori analyses of controlcondition
participants, as network heterogeneity increased, attitude change in response to the persuasive
message signiﬁcantly increased as well (Table 2, Column 1), replicating prior ﬁndings. The relatively
weaker effect averaged across conditions suggests that the experimental manipulations may,
indeed, reduce this network inﬂuence. To examine this possibility, we now turn to an examination of
the impact of the experimental conditions.
Social Constraint
In order to examine the mechanisms of network inﬂuence more closely, a similar regression analysis
was conducted, this time including as predictors dummy variables for each experimental condition,
and the two-way interaction of network composition with each condition dummy (Table 1,
Model 3). Consistent with the social constraint perspective, the interaction of social afﬁrmation and
Table 1. Impact of Network Heterogeneity Upon Attitude Change and Moderation by Social Afﬁrmation and Knowledge
Afﬁrmation
Model 1:
Network Heterogeneity
and Controls
Model 2:
Experimental
Conditions Included
Model 3:
Interactions
Included
(Constant) 0.18 (.04)** 0.19 (.04)** 0.19 (.04)**
Network heterogeneity 0.10 (.05)†
0.10 (.05)†
0.18 (.08)*
Knowledge afﬁrmation 20.02 (.02) 20.02 (.02)
Social afﬁrmation 0.01 (.02) 0.01 (.02)
Knowledge afﬁrmation 3Network heterogeneity 20.07 (.11)
Social afﬁrmation 3Network heterogeneity 20.28 (.14)*
Male 5 1 20.03 (.01)* 20.03 (.01)* 20.03 (.01) †
Party identiﬁcation (Democrat high) 0.05 (.02)** 0.05 (.02)** 0.05 (.02)**
Age 20.07 (.10) 20.08 (.10) 20.08 (.10)
Age squared 0.15 (.11) 0.16 (.11) 0.15 (.11)
Education 20.05 (.02)* 20.05 (.02)* 20.05 (.02)*
Income 20.03 (.03) 20.03 (.03) 20.03 (.03)
Metropolitan area 5 1 0.04 (.02)* 0.04 (.02)* 0.04 (.02)*
Mood 20.12 (.03)** 20.13 (.03)** 20.12 (.03)**
R2
.12 .13 .14
Note. Standard errors are indicated in parentheses. Gender and living in a metropolitan area and experimental conditions
were dummy coded as indicated above. Network composition was mean-centered before calculating the interaction terms
(see Aiken & West, 1991). All other variables were scaled from 0 to 1. †p  .1, *p  .05, **p  .01.
10 Levitan966
network composition signiﬁcantly predicted attitude change, suggesting that the social afﬁrmation
manipulation disrupted the relation between network composition and openness to attitude change.
A visual examination of Figure 2 shows that participants embedded in attitudinally congruent networks
exhibited substantially more attitude change on average when in the social afﬁrmation condition
than in the other two conditions. It appears that these participants have been freed from social
concerns constraining their attitudes to agreement with network members. Indeed, those embedded in
attitudinally congruent networks exhibited more than 50% more attitude change than participants in
other conditions. In contrast, those embedded in more heterogeneous networks exhibited similar
amounts of attitude change across conditions. Those in the social afﬁrmation condition exhibited
roughly the same amount of change across network compositions, suggesting that social afﬁrmation
eliminated network effects.
A priori analyses of the social afﬁrmation condition conﬁrm that the social afﬁrmation manipulation
eliminated the signiﬁcant effect of network composition on persuasion observed in the control
Figure 2. Impact of network heterogeneity across conditions. Plotted points are the minimum, one standard deviation
below and above the mean level of network heterogeneity, the ﬁrst and third quartiles, and the mean. Note that heterogeneity
was mean-centered. Asterisks indicate level of signiﬁcance of the marginal effect of the social afﬁrmation condition
over the control, as indicated by simple effects tests (Aiken & West, 1991). † p  .1, *p  .05. [Color ﬁgure can be
viewed at wileyonlinelibrary.com]
Table 2. Impact of Network Heterogeneity Upon Attitude Change in Each Experimental Condition
Control
Condition
Social Afﬁrmation
Condition
Knowledge Afﬁrmation
Condition
(Constant) .28 (.07)** .19 (.09)* .04 (.06)
Network heterogeneity .19 (.08)* 2.10 (.13) .11(.07)
Male 5 1 .00 (.02) 2.06 (.03)* 2.02 (.02)
Party identiﬁcation (Democrat high) .06 (.03)* .07 (.04) .04 (.03)
Age 2.36 (.17)* 2.05 (.21) .24 (.16)
Age squared .43 (.19)* .15 (.24) 2.15 (.17)
Education 2.06 (.04) †
2.08 (.05) 2.01 (.04)
Income 2.03 (.05) 2.06 (.07) 2.03 (.05)
Metropolitan area 5 1 .02 (.03) .04 (.04) .03 (.03)
Mood 2.22 (.05)** 2.04 (.07) 2.09 (.05) †
R2
.21 .14 .12
Note. Gender and living in a metropolitan area were dummy coded as indicated above. Other variables were scaled 0–1.
†
p  .1, *p  .05, **p  .01.
Mechanisms of Network Inﬂuence 11967
condition (Table 2, Column 2), providing strong support for the social constraint mechanism. Simple
effects tests following Aiken and West (1991) conﬁrm that participants with congruent networks in
the social afﬁrmation condition did indeed experience signiﬁcantly more attitude change than those in
either of the other two conditions.
In other words, participants who were reassured about the strength of their relationships exhibited
striking attitude change compared to others with similar networks receiving the same political information.
Once social concerns were relieved by afﬁrming participants’ social relationships, participants
embedded in congruent networks showed higher levels of attitude change previously only observed in
individuals embedded in more heterogeneous networks. This supports the hypothesis that relationship
concerns constrain individuals in congruent networks to remain attitudinally in step with the network
and therefore resistant to attitude change.
This ﬁnding is also noteworthy given that prior research has frequently focused on the role of heterogeneous
networks in instigating attitude change rather than the role of like-minded others in preventing
change by making individuals less willing to consider other perspectives. This is, however,
consistent with prior literature showing that individuals embedded in heterogeneous networks are
more willing to seek out and more deeply consider issue-relevant information (Levitan & Visser,
2008; Levitan & Wronski, 2014).
Social Doubt
In contrast, evidence for the self-doubt mechanism was decidedly mixed. In the full model, the
interaction of network composition and knowledge afﬁrmation did not approach signiﬁcance (p 5 .57;
Table 1, Model 3) suggesting that the knowledge manipulation did not modify network inﬂuence.
This result is somewhat softened by a priori analyses, in which knowledge afﬁrmation did disrupt the
effect of network composition on persuasion, such that it was no longer signiﬁcant (network effect:
p 5 .16; Table 2, Column 3). Thus, the social doubt mechanism may have some merit. In Figure 2, it
appears that those in more heterogeneous networks exhibited less attitude change in this condition,
consistent with the idea that being connected to network members with differing opinions ordinarily
generates doubt about one’s knowledge and increases willingness to consider other perspectives. Still,
without a signiﬁcant interaction between knowledge afﬁrmation and network composition, the mechanism
cannot be considered clearly supported.
The lack of strong social doubt effects might have been due to the difﬁculty of some knowledge
questions. To assess the possibility that these weak results were due to a failure of the manipulation, a
manipulation check analysis was conducted. If the knowledge afﬁrmation did indeed bolster participants’
conﬁdence in their own level of knowledge, then those in the knowledge afﬁrmation condition
should have reported feeling more knowledgeable than those in other conditions. As anticipated, the
dummy variable for the knowledge afﬁrmation condition signiﬁcantly predicted subsequent selfreported
level of knowledge in a regression analysis controlling for network heterogeneity, the social
afﬁrmation condition, and all previous control variables, b 5 0.22, SE 5 0.06, p  .001. Thus, the
knowledge afﬁrmation was successful in increasing participants’ perception of their own knowledge,
but this did not translate into a signiﬁcant weakening of network composition’s inﬂuence.
Still, it is possible that while the participants believed the feedback that they knew more than
others, they were still not conﬁdent that they knew enough because the difﬁculty of the questions
asked (M 5 49% questions correct) highlighted their lack of knowledge (see Schwarz et al., 1991b).
This possibility is exempliﬁed by one participant’s report: “I’m very surprised that I ranked in the
91st percentile . . . it’s scary to think how many people could be mis-informed about [capital punishment].”
This clearly indicates that the participant believed the feedback, but felt that this amount of
knowledge was distressingly insufﬁcient.
These knowledge afﬁrmation results also provide a useful comparison to rule out alternate
explanations for social afﬁrmation effects, such as mood boosting. In both experimental conditions,
12 Levitan968
participants received positive information about themselves, which could plausibly be self-afﬁrming
or mood boosting. Therefore, if positive feedback in general was responsible for the above social constraint
ﬁndings, one would expect a similar pattern of results in the knowledge afﬁrmation condition.
Instead, the knowledge afﬁrmation manipulation actually leans toward reducing open-mindedness (as
per social doubt), rather than increasing it as the social afﬁrmation manipulation did (see Figure 2).
Additional analyses including predictors from prior models show that neither condition had an appreciable
inﬂuence on participants’ moods (ps >.5). Furthermore, experimental effects remain despite
controlling for mood (Tables 1 and 2). Thus, these results strongly suggest that speciﬁcally social
feedback and not merely positive feedback, is responsible for the observed social constraint effects.
Information Exchange
As previously stated, this design does not provide an experimental test of the informationexchange
mechanism. It does, however, allow us to examine some relevant information, speciﬁcally
whether initial knowledge mediates the relation between network composition and attitude change.
An OLS regression used network heterogeneity and control variables to predict initial knowledge.
This did not yield supporting evidence. Participants who were embedded in heterogeneous networks
correctly answered no more capital punishment knowledge questions than others, b 5 20.002,
SE 5 0.07, p 5 .98. Logistic regression analyses were also conducted predicting correct responses to
individual knowledge items using network composition and control variables. Network composition
marginally predicted one knowledge item in the wrong direction (whether any women had been executed
in the last 25 years), b 5 21.65, SE 5 0.89, p 5 .06. Network composition did not signiﬁcantly
predict correct answers to any other question (ps >.15, with coefﬁcients in both directions). Network
composition notably failed to predict items testing knowledge which might be construed as counterattitudinal,
in contrast with prior research. Thus, the Baron and Kenny (1986) mediation of network
effects by knowledge fails. Sobel tests conducted with the control condition alone and the full sample
also failed, ps > .15. Although these results do not disprove the information exchange perspective,
they do suggest a need for further study regarding how pervasive it is.
Discussion
People think about politics differently when close others hold the same attitudes. These results
illuminate the complex role of social networks, not just as a source of information, but as a source of
goals and motivations that alter how we interact with the world. These results are particularly important
in that careful, deliberate consideration of political attitudes is crucial to the democratic ideal
(e.g., Fishkin, 1991). These ﬁndings do not challenge prior ﬁndings that individuals embedded in
more heterogeneous social networks hold more ﬂexible views (Huckfeldt & Sprague, 2000), or that
bringing diverse groups of ordinary citizens together can yield normatively positive outcomes (e.g.,
List, Fishkin, Luskin, & McClean, 2005). Instead, they explain why this may be the case, with noteworthy
implications for how to stimulate greater thought and shake uniﬁed groups’ presumption of
rectitude. In particular, these results highlight a uniquely social role of networks in guiding attitudes
by demonstrating that belongingness needs motivate individuals to maintain attitudes that are similar
to those of their network members.
Particularly important in these results is the ﬁnding that, among those whose social afﬁliation
goals have been met, attitudinal ﬂexibility is the rule, rather than attitude stability. These results ﬁrmly
demonstrate the role of attitudinally congruent networks in socially motivating suppression of attitude
change. Prior research has generally framed the impact of network composition in terms of network
heterogeneity, rather than network attitudinal congruence, implying that heterogeneous networks were
generating inﬂuence (by bringing deviating individuals into the fold and providing novel information).
The current study, however, reveals that attitudinally congruent networks suppress attitude change at
Mechanisms of Network Inﬂuence 13969
least as much as, if not more than, heterogeneous networks encourage it. Given the same strong arguments,
those surrounded by like-minded others will still be less persuaded because of social rejection
and belongingness concerns. When those concerns are alleviated, those in congruent networks are just
as open to change as their heterogeneous network counterparts.
This effect of social constraint may, in part, explain the oft-observed attitudinal similarity among
close network members, in that attitudes that are consonant with those of network members have a
quasi-gravitational pull. Once people have similar attitudes, social motivations dictate that they are
much less likely to change them, but those in more attitudinally heterogeneous networks are not so
constrained and may continue to hold unstable views unless, over time, their attitudes drift closer to
those of network members, and thence spiral into agreement.
Limitations
It should be noted that these results do not completely rule out other mechanisms of social network
inﬂuence, although they are unique in supporting the social constraint mechanism with ﬁndings
that cannot be attributed to other proposed mechanisms. It is possible that both heterogeneous and
congruent networks impose countervailing inﬂuences through different mechanisms: congruent networks
applying social inﬂuences to maintain congenial views, while heterogeneous networks bring
new information and generate doubt and ambivalence. One should not generalize from null results,
and it may well be that additional mechanisms simply remained undetected here. Evidence for the
social doubt mechanism in particular was mixed. Similarly, we have uncovered evidence that the attitudinally
congruent end of the spectrum is inﬂuential, but we must not repeat prior mistakes and simply
assume that the other end of the scale of network composition is an inactive baseline. Such
possibilities await further research.
The information exchange results especially require further consideration. The current study did
not experimentally test the information exchange account, but these survey results differed notably
from previous results (Huckfeldt et al., 2004; Mutz, 2002b). This may be due to differences in methodology.
Prior studies counted the number of offered “rationales” for views, potentially capturing
thoughts and hearsay rather than the factual knowledge, as tested here. Additionally, the current study
used closed-ended questions, so participants in heterogeneous networks may have known more, but
not in these speciﬁc areas. As such, the inﬂuence of information exchange may have gone undetected.
Alternately, it may only occur on a subset of salient topics. It may be useful for future research to
examine what kind of information is exchanged and when.
With respect to social doubt, mixed ﬁndings may have been a function of a weak manipulation that
did not truly convince participants that they knew enough, as discussed above, although the manipulation
check limits such concerns. Regardless of what other mechanisms may also be at work, these results
strongly support a social constraint account and cannot be accounted for by the other mechanisms.
Of course, full understanding and support of the mechanisms of any effect are not the work of a single
study, but of a broad body of research conducted using multiple paradigms (Bullock et al., 2010). As
with any single study, the possibility of alternate explanations cannot be completely ruled out, only vigilantly
guarded against. Nonetheless, this study has followed such recommendations as can be practically
followed in a single component of a program of research and should be given what credence such a
practice allows. Importantly, this research has manipulated proposed mechanisms directly and independently,
rather than only manipulating the initial causal factor (networks) and then measuring proposed
mediators. Additionally, the current research has compared proposed mechanisms within the same dataset
and taken pains to ensure that manipulations are inﬂuencing only the mechanism in question. These
steps minimize both theoretical and statistical concerns about correctly identifying mechanisms.6
6
See Bullock et al. (2010) for a detailed analysis of hurdles in identifying mechanisms.
14 Levitan970
A more speciﬁc concern that often plagues the area of social networks is direction of causality. It
is tempting to suppose that individuals who differ in their level of open-mindedness or attitude
strength have differing likelihoods of selecting into heterogeneous networks. This concern has limited
relevance here on several counts. First, it cannot explain the current result that the relationship
between open-minded attitude change and social network composition differs as a function of randomly
assigned experimental condition. There is no plausible explanation for how a social manipulation
(irrelevant to capital punishment) might inﬂuence attitude change without some causal role of the
network itself, particularly when the impact of the manipulation is conditional, depending on an individual’s
network (assessed prior to the manipulation). More broadly, as noted in the introduction,
much prior research supports the causal role of networks in inﬂuencing attitudes and openness to attitude
change, using experiments, quasi-experiments, and longitudinal methods (e.g., Lazer et al., 2010;
Levitan & Visser, 2009; Visser & Mirabile, 2004), yet those same studies and others have failed to
ﬁnd evidence supporting the role of political attitudes in network construction (see also a special issue
of Communication Monographs, 1992). Even if individuals intend to form bonds with politically similar
others, the lack of constraint among attitudes across political issues (Converse, 1964) would render
this nearly impossible because bringing a network into congruence in one area, such as on economic
policy, would not inherently generate agreement in other areas, like foreign policy, social issues, or
here capital punishment. Add to this the host of network members who are effectively chosen for us
(coworkers, family, spouse’s friends, neighbors, etc.; see also Festinger et al., 1950; Sinclair, 2012),
and it is perhaps not surprising that so many studies have failed to ﬁnd evidence that individuals
choose politically congruent network members. Taken together, selective network construction certainly
becomes an implausible explanation for how an experimental manipulation could negate the
relation between network composition and openness to persuasion in the current study.
Some readers may wonder whether the mere act of listing social network member views might
be enough to activate or “prime” social pressures, thus making our self-report measure itself an inﬂuence
on openness to attitude change. This serious concern would undermine the broader body of
research that relies on self-report assessments of network composition. Fortunately, previous research
has experimentally examined this exact possibility. Visser and Mirabile (2004) manipulated the order
in which participants responded to a counterattitudinal message and reported on their social network,
such that some participants’ attitude change was assessed before reporting on their networks, and
others reported on their networks ﬁrst. Neither the impact of order of measures nor its interaction with
network heterogeneity approached signiﬁcance in two separate studies. Therefore, the act of reporting
network composition did not signiﬁcantly alter network inﬂuence. Openness to attitude change is
increased as network heterogeneity increases, regardless of whether participants have explicitly been
asked about their network, and reporting on one’s network does not signiﬁcantly change the size of
the relation.
These limitations notwithstanding, the experimental test here supports the idea that social network
composition inﬂuences resistance to persuasion in part through individuals’ motivations to maintain
positive social relationships. Further, these results remind us that network inﬂuence is not an
inevitable result of being exposed to close others with differing views. Network inﬂuence has boundary
conditions and can be nulliﬁed by circumstances which alter an individual’s construal of the situation
(see also Bloom & Levitan, 2011). This research also highlights the importance of going beyond
information and cognitive processing when attempting to understand attitudes. To fully understand
attitudes and persuasion, we must take into account the social context, including social goals, motivations,
sanctions, and rewards.
Broader Implications
The current study has implications not only for the basic literatures on public opinion, persuasion,
and social networks, but also for a variety of ﬁelds and subﬁelds with an interest in improving group
Mechanisms of Network Inﬂuence 15971
decision-making. These results echo the classic groupthink theory (Janis, 1982) in that even groups of
otherwise intelligent experts can make foolhardy and conformist decisions, in part due to their desire
to maintain group members’ positive regard. If attitudinally congruent networks are actively or even
incidentally suppressing dissent, then information alone will not be sufﬁcient to dislodge misinformed
views and encourage good decisions. Instead, a more psychological approach is required, such as
making people feel more secure about the steadfast positive regard of their network (as in the current
study) or perhaps emphasizing support outside their network for attitudes with which immediate network
members might not agree.
These results also suggest a different spin on motivated reasoning with respect to attitudes
(Kunda, 1990; Redlawsk, 2002; Taber & Lodge, 2006): We may be attached to our attitudes not for
the sake of consistency alone, but also for the sake of social adjustment. Therefore, the goal of
approaching and maintaining attitudes that are congenial to the network may also underlie motivated
reasoning. This has special relevance for the ﬂourishing literature on political networks, which has
generally emphasized accuracy goals. These results highlight that uniquely social goals can motivate
individuals to “go along to get along” and avoid attitudes that might rock the boat (see also DeWall,
Visser, & Levitan, 2006). When their social afﬁliation goals are sated, however, individuals feel less
social constraint and are free to consider less congenial attitudes, as seen here.
It would be taking these results too far, however, to infer that networks of similar others are inherently
damaging to the process of maintaining correct attitudes. Much like students conferring after an
exam, if we all agree, we’re more likely to be right. Thus, our network members may hold us to a correct
position, saving us time and energy (cognitive misers that we are), and making political action
less complicated and more frequent (Mutz, 2002a). Of course, the crowd will not always be correct,
and therein lies the dilemma.
Conclusions
Factors underlying the maintenance of attitudes, rather than attitude change, should receive
renewed attention as research continues. We turn to others to help us form attitudes and make decisions,
but so too do we respond to them once those attitudes have formed. Those around us play a role
in solidifying our attitudes and judgments just as they contribute to eroding and changing them. This
role of networks in suppressing our willingness to consider alternate viewpoints may ultimately be
responsible for many perplexing phenomena, including various situations in which individuals continue
to hold views contrary to all available evidence. “Birthers’” continued belief that President
Obama was not born in the United States despite objective disproof is likely an extreme example of
the more common social process examined above, whereby individuals are reluctant to consider information
that might bring their attitudes out of step with those around them. Attitudes are formed and
maintained in a complex social environment, one whose effects we are only beginning to understand.
ACKNOWLEDGMENTS
Thanks are due to colleagues at the Stony Brook University Political Psychology Workshop,
who made helpful comments during the conceptualization of this research. Data was collected
through an award from Time-sharing Experiments for the Social Sciences, NSF Grant 0818839,
Jeremy Freese and James Druckman, Principal Investigators. This research was also supported by a
professional development stipend grant from Shepherd University. Correspondence concerning this
article should be addressed to Lindsey Levitan, Department of Psychology, Shepherd University,
Shepherdstown, WV, 25443. E-mail: llevitan@shepherd.edu
16 Levitan972
REFERENCES
Ahn, T., Huckfeldt, R., & Ryan, J. B. (2010). Communication, inﬂuence, and informational asymmetries among voters.
Political Psychology, 31(5), 763–787.
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Thousand Oaks, CA:
Sage.
Asch, S. E. (1955). Opinions and social pressure. Scientiﬁc American, 193(5), 31–35.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research.
Journal of Personality and Social Psychology, 51(6), 1173–1182.
Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal attachments as a fundamental
human motivation. Psychological Bulletin, 117(3), 497–529.
Bello, J., & Rolfe, M. (2014). Is inﬂuence mightier than selection? Forging agreement in political discussion networks
during a campaign. Social Networks, 36, 134–146.
Berelson, B. R., Lazarsfeld, P. F., & McPhee, W. N. (1954). Voting: A study of opinion formation in a presidential campaign.
Chicago, IL: University of Chicago Press.
Bloom, P. B. N., & Bagno-Moldavsky, O. (2015). The conditional effect of network diversity and values on tolerance.
Political Behavior, 37(3), 623–651.
Bloom, P. B. N., & Levitan, L. C. (2011). We’re closer than I thought: Social network heterogeneity, morality, and
political persuasion. Political Psychology, 32(4), 643–665.
Bullock, J. G., Green, D. P., & Ha, S. E. (2010). Yes, but what’s the mechanism? (Don’t expect an easy answer).
Journal of Personality and Social Psychology, 98(4), 550–558.
Burt, R. S. (1984). Network items & the General Social Survey. Social Networks, 6(4), 293–339.
Burt, R. S. (1987). Social contagion and innovation: Cohesion versus structural equivalence. American Journal of
Sociology, 92(6), 1287–1335.
Carlson, T. N., & Settle, J. E. (2016). Political chameleons: An exploration of conformity in political discussions.
Political Behavior, 38(4), 817–859.
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, NY: Guilford Press.
Cialdini, R. B. (1995). Principles and techniques of social inﬂuence. In A. Tesser (Ed.), Advanced social psychology
(pp. 257–282). New York, NY: McGraw-Hill.
Converse, P. E. (1964). The nature of belief systems in mass politics. In Ideology and discontent (pp. 206–261).
New York, NY: Free Press.
Deutsch, M., & Gerard, H. B. (1955). A study of normative and informational social inﬂuences upon individual
judgment. Journal of Abnormal & Social Psychology, 51(3), 629–636.
DeWall, C. N., Visser, P. S., & Levitan, L. C. (2006). Openness to attitude change as a function of temporal perspective.
Personality and Social Psychology Bulletin, 32(8), 1010–1023.
Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7(2), 117–140.
Festinger, L., Schachter, S., & Back, K. (1950). Social pressures in informal groups; a study of human factors in housing.
Oxford, United Kingdom: Harper.
Fishkin, J. S. (1991). Democracy and deliberation. New Haven, CT: Yale University Press.
Gardner, W. L., Pickett, C. L., & Brewer, M. B. (2000). Social exclusion and selective memory. Personality and Social
Psychology Bulletin, 26(4), 486–496.
Goel, S., Mason, W., & Watts, D. J. (2010). Real and perceived attitude agreement in social networks. Journal of Personality
and Social Psychology, 99(4), 611–621.
Greenwald, A. G. (1968). Cognitive learning, cognitive response to persuasion, and attitude change. In A. G. Greenwald,
T. C. Brock & T. M. Ostrom (Eds.), Psychological foundations of attitudes (pp. 147–170). New York, NY: Academic
Press.
Huckfeldt, R. (2014). Networks, contexts, and the combinatorial dynamics of democratic politics. Advances in Political
Psychology, 35(S1), 43–68
Huckfeldt, R., Beck, P. A., Dalton, R. J., & Levine, J. (1995). Political environments, cohesive social groups, and the
communication of public opinion. American Journal of Political Science, 39(4), 1025–1054.
Huckfeldt, R., Mendez, J. M., & Osborn, T. (2004). Disagreement, ambivalence, & engagement: The political consequences
of heterogeneous networks. Political Psychology, 25(1), 65–95.
Huckfeldt, R. & J. Sprague (1987). Networks in context: The social ﬂow of political information. American Political Science
Review, 81(4), 1197–1216.
Mechanisms of Network Inﬂuence 17973
Huckfeldt, R., & Sprague, J. (2000). Political consequences of inconsistency: The accessibility and stability of abortion
attitudes. Political Psychology, 21(1), 57–79.
Iyengar, S. (1986). Whither political information. NES Pilot Study Report, No. nes002253.
Janis, I. L. (1982). Groupthink: Psychological studies of policy decisions and ﬁascoes. Boston, MA: Houghton Mifﬂin.
Katz, D. (1960). The functional approach to the study of attitudes. Public Opinion Quarterly, 24(2), 163–204.
Kelley, H. H. (1952). Two functions of reference groups. In G. E. Swanson, T. M. Newcomb & E. L. Hartley (Eds.),
Handbook of social psychology (2nd ed., pp. 410–414). New York, NY: Holt, Rinehart & Winston.
Kelman, H. C. (1958). Compliance, identiﬁcation, and internalization three processes of attitude change. Journal of Conﬂict
Resolution, 2(1), 51–60.
Kelman, H. C., & Hamilton, L. V. (1989) Dynamics of authority. In Crimes of obedience: Toward a social psychology
of authority and responsibility (pp. 103–135). New Haven, CT: Yale University Press.
Klar, S. (2014). Partisanship in a social setting. American Journal of Political Science, 58(3), 687–704.
Krosnick, J. A., & Petty, R. E. (1995). Attitude strength: An overview. In R. E. Petty & J. A. Krosnick (Eds.), Attitude
strength: Antecedents and consequences (pp. 1–24). Mahwah, NJ: Lawrence Erlbaum.
Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108(3), 480–498.
Laumann, E. O. (1973). Bonds of pluralism: The form and substance of urban social networks. New York, NY: Wiley-
Interscience.
Lazer, D., Rubineau, B., Chetkovich, C., Katz, N., & Neblo, M. (2010). The coevolution of networks and political attitudes.
Political Communication, 27(3), 248–274.
Levinger, G. (1972). Little sand box and big quarry: Comment on Byrne’s paradigmatic spade for research on interpersonal
attraction. Representative Research in Social Psychology, 3, 3–19.
Levitan, L. C., & Verhulst, B. (2016). Conformity in groups: The effects of others’ views on expressed attitudes and attitude
change. Political Behavior, 38(2), 277–315.
Levitan, L. C., & Visser, P. S. (2008). The impact of the social context on resistance to persuasion: Effortful versus
effortless responses to counter-attitudinal information. Journal of Experimental Social Psychology, 44(3), 640–649.
Levitan, L. C., & Visser, P. S. (2009). Social network composition and attitude strength: Exploring the dynamics within
newly formed social networks. Journal of Experimental Social Psychology, 45(5), 1057–1067.
Levitan, L. C., & Wronski, J. (2014). Social context and information seeking: Examining the effects of network attitudinal
composition on engagement with political information. political behavior, 36(4), 793–816.
List, C., Luskin, R. C., Fishkin, J. S., & McLean, I. (2013). Deliberation, single-peakedness, and the possibility of meaningful
democracy: Evidence from deliberative polls. Journal of Politics, 75(01), 80–95.
Maner, J. K., DeWall, C. N., Baumeister, R. F., & Schaller, M. (2007). Does social exclusion motivate interpersonal
reconnection? Resolving the “porcupine problem.” Journal of Personality and Social Psychology, 92(1), 42–55.
McClurg, S. D. (2006). The electoral relevance of political talk. American Journal of Political Science, 737–754.
Mutz, D. C. (2002a). The consequences of cross-cutting networks for political participation. American Journal of Political
Science, 46(4), 838–855.
Mutz, D. C. (2002b). Cross-cutting social networks: Testing democratic theory in practice. American Political Science
Review, 96(1), 111–126.
Mutz, D. C. (2006). Hearing the other side: Deliberative versus participatory democracy. Cambridge, MA: Cambridge
University Press.
Noelle-Neumann, E. (1974). The spiral of silence: A theory of public opinion. Journal of Communication, 24(2), 43–51.
NYT/CBS Poll. (2011, April 21). The President, Congress and deﬁcit battles. Retrieved from http://www.cbsnews.com/
stories/2011/04/21/politics/main20056282.shtml?tag5contentMain;contentBody
Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. In L. Berkowitz (Ed.), Advances
in experimental social psychology. San Diego, CA: Academic Press.
Priester, J. R., & Petty, R. E. (2001). Extending the bases of subjective attitudinal ambivalence: Interpersonal and intrapersonal
antecedents of evaluative tension. Journal of Personality and Social Psychology, 80(1), 19–34.
Public Policy Polling (PPP). (2012, March 18). Romney up big in Illinois. Retrieved from http://www.publicpolicypol-
ling.com/pdf/2011/PPP_Release_IL_318.pdf
Redlawsk, D. (2002). Hot cognition or cool consideration? Testing the effects of motivated reasoning on political decision
making. Journal of Politics, 64(04), 1021–1044.
Ross, L., Bierbrauer, G., & Hoffman, S. (1976). The role of attribution processes in conformity and dissent: Revisiting
the Asch situation. American Psychologist, 31(2), 148–157.
Ryan, J. B. (2010). The effects of network expertise and biases on vote choice. Political Communication, 27(1), 44–58.
18 Levitan974
Ryan, J. B. (2013). An experimental study of persuasive social communication. Political Communication, 30(1), 100–116.
Schachter, S. (1951). Deviation, rejection, and communication. Journal of Abnormal & Social Psychology, 46(2),
190–207.
Schwarz, N., Bless, H., & Bohner, G. (1991a). Mood and persuasion: Affective states inﬂuence the processing of persuasive
communications. Advances in Experimental Social Psychology, 24, 161–199.
Schwarz, N., Bless, H., Strack, F., Klumpp, G., Rittenauer-Schatka, H., & Simons, A. (1991b). Easy of retrieval as information:
Another look at the availability heuristic. Journal of Personality and Social Psychology, 61(2), 195–202.
Settle, J. E., Bond, R., & Levitt, J. (2011). The social origins of adult political behavior. American Politics Research,
39(2), 239–263.
Sinclair, B. (2012). The social citizen. Chicago, IL: University of Chicago Press.
Smith, M. B., Bruner, J. S., & White, R. W. (1956). Opinions and personality. New York, NY: Wiley.
Taber, C. S. (2003). Information processing and public opinion. In D. O. Sears, L. Huddy, & R. Jervis (Eds.), Oxford
Handbook of Political Psychology (pp. 433–476). New York, NY: Oxford University Press.
Taber, C., & Lodge, M. (2006). Motivated skepticism in the evaluation of political beliefs. American Journal of Political
Science, 50(3), 755–769.
Twenge, J. M., Baumeister, R. F., Tice, D. M., & Stucke, T. S. (2001). If you can’t join them, beat them: Effects of
social exclusion on aggressive behavior. Journal of Personality and Social Psychology, 81(6), 1058–1069.
Ulbig, S. G., & Funk, C. L. (1999). Conﬂict avoidance and political participation. Political Behavior, 21(3), 265–282.
Visser, P. S., & Krosnick, J. A. (1998). Development of attitude strength over the life cycle: Surge and decline. Journal
of Personality and Social Psychology, 75(6), 1389–1410.
Visser, P. S., & Mirabile, R. R. (2004). Attitudes in the social context: The impact of social network composition on
individual-level attitude strength. Journal of Personality and Social Psychology, 87(6), 779–795.
Whorf, B. L. (1956). In Language, thought, and reality: Selected writings of Benjamin Lee Whorf. J. B. Carroll (Ed.).
Cambridge, MA: MIT Press.
Supporting Information
Additional supporting information may be found in the online version of this article at the publisher’s
website:
Appendix S1. Sample Characteristics
Table S1. Sample Characteristics
Appendix S2. Correlates of Network Composition
Table S2. Correlates of Network Composition in Current Study
Appendix S3. Measure of Knowledge About Capital Punishment
Appendix S4. Manipulations
Mechanisms of Network Inﬂuence 19975