Conformity in Groups: The Effects of Others’ Views on Expressed Attitudes and Attitude Change

Levitan, Lindsey C.; Verhulst, Brad

doi:10.1007/s11109-015-9312-x

Conformity in Groups: The Effects of Others’ Views on Expressed Attitudes and Attitude Change

Original Paper
Published: 15 August 2015

Volume 38, pages 277–315, (2016)
Cite this article

Download PDF

Access provided by Autonomous University of Puebla

Political Behavior Aims and scope Submit manuscript

Conformity in Groups: The Effects of Others’ Views on Expressed Attitudes and Attitude Change

Download PDF

Lindsey C. Levitan¹ &
Brad Verhulst²

12k Accesses
44 Citations
203 Altmetric
25 Mentions
Explore all metrics

Abstract

Two experiments demonstrate the powerful influence of others’ views on individual attitudes and attitude expression. Those around us can influence our views through persuasion and information exchange, but the current research hypothesizes that exposure to alternate views even without discussion or exchange of persuasive arguments can also alter what attitudes are expressed, and even generate long term shifts in attitudes. In an initial study, naïve participants were asked their attitudes on a range of standard survey items privately, publicly in a group with trained confederates, and again privately following the group setting. Findings indicate significant attitudinal conformity, which was most pronounced when participants were faced with a unanimous (versus non-unanimous) group. The group experience continued to influence participants’ views when they were again asked their views in private. A second experiment varied whether participants heard views from live confederates or via computer, demonstrating that these effects could not be attributed only to issue-relevant information provided by or inferred from group members, and that attitude change persisted long after participants had left the laboratory. In summary, when people are asked their attitudes publicly, they adjust their responses to conform to those around them, and this attitude change persists privately, even weeks later. Accordingly, such purely social processes of attitude change may be every bit as important as more traditional cognitive informational processes in understanding where people’s political attitudes come from, and how they may be changed.

The potential of deliberative reasoning: patterns of attitude change and consistency in cross-cutting and like-minded deliberation

Article 10 July 2018

Inaccurate group meta-perceptions drive negative out-group attributions in competitive contexts

Article 11 November 2019

Respondents with more extreme views show moderation of opinions in multi-year surveys in the USA and the Netherlands

Article Open access 02 December 2023

Use our pre-submission checklist

Avoid common mistakes on your manuscript.

Those who surround us have a profound influence on how we understand the world in general, and politics in particular. Ideally, this influence happens through thoughtful exchanges of views and careful consideration, leading people to hold carefully constructed attitudes, make logical and informed decisions, and generally behave consistently with their underlying preferences (Fishkin 1991; Putnam 2000). Unfortunately, this is not always the case (Converse 1964). How, otherwise, can we account for the existence of political attitudes and movements that persist despite objective disproof of their beliefs? For example, “Birthers” insist that President Obama was not born in the US, despite various birth records, reports of those who were there, and even testimony of the Republican governor of the state in which he was born (Henig 2008; Malcolm 2008; PPP poll 2012). It therefore behooves us to look beyond pure logic and information, to more thoroughly consider the social roots of attitudes and judgments.

Research has taken two major directions in understanding why views are not a function of information alone. The first, motivated skepticism (Kunda 1990; Redlawsk 2002; Taber and Lodge 2006; see also Redlawsk and Lau 2013), establishes our tendency to arrive at desired (but not necessarily accurate) conclusions through phenomena like disconfirmation bias (Taber and Lodge 2006). Though ubiquitous, motivated skepticism has limits, and can generally not be stretched to explain disregard for overwhelming information (Kunda 1990; Redlawsk, et al. 2010). Additional, complementary explanations are required. A second direction of research therefore examines features of our social environment that can alter attitudes (Huddy 2004; Smith et al. 1956; Huckfeldt et al. 2004a) and can cause people to prioritize fitting in over correctness (see Baumeister and Leary 1995). Such social research provides some precedents for expressing erroneous judgments in the face of complete objective disproof, albeit with politically irrelevant stimuli (e.g. Asch 1951).

While those with whom we have social ties undoubtedly serve as sources of useful information and arguments, they also generate social pressure, and act as reference points for determining what is appropriate and correct. The result is often conformity, or attitude change towards the group as a result of social information rather than political information.^{Footnote 1} In order to fully understand public opinion and the attitudes citizens hold, we must look beyond overt persuasion and information exchange to examine social goals and influences that arise from simply knowing what others’ attitudes are, without any justification of those attitudes.

Several particularly deleterious effects make it essential to incorporate such social influence processes into our understanding of political attitudes. First, public conformity of one individual may spiral to others and lead to pluralistic ignorance in which the views of the majority are misperceived (Noelle-Neumann 1974; Prentice and Miller 1993; Todorov and Mandisodza 2004). As the appearance of unanimity grows, a corresponding intolerance for dissenting opinions develops (Mutz 2002b; Mutz and Mondak 2006) increasing social pressure for attitudinal homogeneity. Those who perceive (however inaccurately) that their views are supported by those around them will subsequently hold those attitudes more strongly (Visser and Mirabile 2004; Levitan and Visser 2009) and become more likely act on those attitudes through voting and other forms of political participation (Mutz 2002a, but see Huckfeldt et al. 2004b). Most dauntingly, public conformity may lead to private acceptance of expressed views, without due thought. The potential negative implications of this attitude change based upon blind conformity need little explanation. The democratic ideal is that people base their attitudes and votes on careful consideration of available information (e.g. Fishkin 1991), not simply on what attitudes are most popular.

The current studies demonstrate that attitude change through social pressure can arise simply from knowing what other people’s attitudes are, especially proximate others. Hearing justifications of these attitudes is not a necessary precondition for attitude change. We adopt the term “argument” to indicate such information about an issue other than the view of others, in keeping with the persuasion literature (see Shavitt and Brock, 1994, especially chapter 6). The current studies examine the extent to which brief interactions, in which few or no persuasive arguments change hands, can elicit public conformity and even generate enduring attitude change. We also examine features of the situation that can heighten this effortless social influence.

Social Influence

Our attitudes and behavior can be powerfully and often unwittingly shaped by the people around us (Hardin and Higgins 1996). Researchers have studied this premise from many perspectives involving different levels of analysis, mechanisms, and domains. Researchers have examined impersonal influence of mass media (e.g. Mutz 1998) and elites, as well as persuasion more generally (Zaller 1992), or have taken approaches focused more on social dynamics and relationships between individuals.

Social dynamics can unduly influence judgment and decision-making at the level of small group interactions because concurrence-seeking and concerns about group cohesion help generate negative phenomena like groupthink (Janis 1982; Tetlock et al. 1992). Similarly, small group dynamics can generate extreme views through group polarization (Myers and Lamm 1976). Experiments with small groups have been helpful not only in clarifying small group dynamics, but also in isolating specific variables to test their causal impact, demonstrating that characteristics such as expertise (Petty and Cacioppo 1986), gender (Eagly 1983), and status (Eagly 1983) can cause some individuals to be more influential than others.

At a more abstract level, research has examined the role of social identity in attitudes, judgments, and behaviors. Initial research demonstrated that people sometimes change their judgments (e.g. about distribution of resources, perceptions of others) merely due to being randomly assigned to group X or Y, both small groups of strangers (Tajfel and Turner 1986). These beginnings developed into a broad literature on the impact of various socially ascribed identities (race, gender, religion, etc.) on how we view the political landscape (see Huddy 2004).

The current research, like the literature on small groups, focuses on the influence of those around us, specifically people with whom we interact directly. Importantly, it also emphasizes the role of social situations, rather than exchange of arguments, in generating attitude change.

Personal Social Contact

Of late, social network research has made important strides in understanding how direct contact with close others and political discussants can influence attitudes and behavior (Huckfeldt et al. 2004a; Mutz 2006; Sinclair 2012; see Klofstad et al. 2013 on conceptualization and measurement). Our networks socialize us politically (Settle et al., 2011). We borrow information and expertise from them, which in turn influence our views (Ryan 2011) and our behavior (McClurg 2006). Disagreement with those around us can destabilize our views, generating attitude change (Huckfeldt and Sprague 2000; Levitan and Visser 2009), and reducing our likelihood of acting on those attitudes (Mutz 2002a, but see Huckfeldt et al. 2004b).

Various mechanisms are involved in this influence of close others. Structural equivalence theory emphasizes the role of information exchange with similar others, especially discussion which leads to a mutual understanding of costs and benefits (Burt 1987, pp. 1290–1291). Indeed, individuals embedded in attitudinally heterogeneous networks are more aware of opposing arguments (Huckfeldt et al. 2004b; Mutz 2002b). This argument-based influence has often been the focus of research (e.g. Ahn et al. 2010; Huckfeldt and Sprague 1987), but influence is not limited to argument-driven persuasion. Differences of opinion in one’s network also impact how we think about our views, influencing attitude accessibility (Huckfeldt and Sprague 2000), information seeking (Levitan and Wronski 2014), and how deeply we consider opposing information (Levitan and Visser 2008). This is partially because others serve as reference points for determining what attitudes and behaviors are correct and socially appropriate (Sherif 1936; Festinger 1954; Kelley 1952).

While those with whom we have social ties undoubtedly serve as sources of concrete arguments, social comparison theory proposes that others’ views are themselves used as information (Festinger 1954). This “informational influence” occurs when people use the views of others as heuristic cues or reference points to determine if their own views are likely to be correct (Deutsch and Gerard 1955). Despite the term “informational influence,” no arguments need to be exchanged – mere exposure to someone else’s opinion is enough to assess the accuracy of one’s own opinion, like students gauging whether their answer to a math question is correct by asking what others got. People intuitively believe that if more people share their view, then that view is more likely to be correct (Cialdini 1995). Conversely, learning that others hold discrepant attitudes, even without learning why, erodes confidence in that attitude (Visser and Mirabile 2004).

Informational influence is not the only means of influence without argumentation, however. Although people are motivated to be accurate and correct, they also have other motivations, including social motives (see for example Katz 1960; Taber and Lodge 2006). Humans are deeply social creatures, with an intense need to belong, which motivates people to affiliate with others, and to seek their positive regard (Baumeister and Leary 1995). People feel uncomfortable and ambivalent when their views are inconsistent with those of people they like (Priester and Petty 2001), and they are highly motivated to avoid this discomfort often through attitude change (Heider 1946). Such motivation to fit in is so powerful that mere postal mailings emphasizing group norms can influence people to vote (Gerber et al. 2008; Panagopoulos 2010; see also Cialdini et al. 1991). Indeed, some attitudes exist primarily to serve a “social adjustment” function (Smith et al. 1956; see also “value-expressive” function: Katz 1960), although any attitude might be affected by group norms.

“Normative influence” (Deutsch and Gerard 1955; Kelley 1952) follows along these lines, occurring when people we interact with serve as standards delineating what is appropriate and acceptable to the group, regardless of what is factually correct. The attitudes of others imply group norms, signaling what sorts of attitudes are proper and expected. People conform to these group norms to avoid social sanctions (Schachter 1951; Ulbig and Funk 1999), to remain in good standing with current social partners, or to join desirable new groups (Kelley 1952).

Consistent with the expectation that people will alter their views and behavior to fit in, one of the most remarkable findings of conformity research is that people sometimes express clearly incorrect judgments if those around them unanimously agree (Asch 1951; 1956; Deutsch and Gerard 1955). In Asch’s classic studies, participants were easily able to determine the relative lengths of printed lines when alone, but gave the wrong answer a stunning 37 % of the time in the presence of other “participants” who had just unanimously voiced a wrong answer (Asch 1956). A less touted, but also remarkable aspect of this research is that strangers had this influence. This suggests that, although research naturally focuses on close others, perhaps strangers and those with whom we have weak ties (who are more likely to disagree with us, Mutz and Martin 2001) have an underappreciated impact on our views.

While information exchange clearly affects attitudes, subtle features of the social context also have an impact. The underlying mechanisms of both normative and informational social influence rely on the knowledge of what others’ views are, without the necessity of any concrete arguments being exchanged. As such, the attitudes of others provide a social yardstick by which to evaluate both the correctness and appropriateness of one’s own preferences.

Conformity with Political Attitudes?

Political attitudes are quite different than the stimuli used in much of the conformity literature, allowing some to suggest that political attitudes may not be susceptible to such striking levels conformity as in classic conformity studies. This makes it especially important to empirically establish people’s susceptibility to political conformity.

Much of our knowledge of conformity is based on unimportant perceptual tasks (e.g. judging the lengths of lines), whereas political attitudes are quite important and self-relevant (Katz 1960). As attitude importance increases, people are more likely to act in accordance with that attitude, and are therefore less susceptible to attitude change (Boninger et al. 1995), and so people may be more willing to risk social conflict by posing a dissenting opinion. People may also be more willing to resist conforming politically because of the relative subjectivity of political issues. To disagree politically may be uncomfortable, but to disagree on objective matters risks “appearing incompetent, foolish, or even mad,” because one can be proven wrong (Ross et al. 1976, p. 149). The subjectivity of political views therefore reduces the social cost of non-conformity (see Ross et al. 1976).

Thus, there are valid reasons to doubt that the conformity so compellingly established in other domains will generalize to politics. Still, when there is no objective evidence, people are particularly reliant on social information to gauge the correctness of their attitudes and judgments (Festinger 1954; Kelley 1952; Deutsch and Gerard 1955), making it necessary and enlightening to test whether the purely social effects that psychologists tout are truly present with regard to political views.

Lasting Attitude Change?

Even temporary conformity in the face of disagreement has dramatic implications for the democratic process, but the possibility that it may presage lasting attitude change makes it critically important. Social information, even in brief interactions, can exert an influence on people’s perceptions even months later (Sherif 1936). Consequently, the attitudes that people possess may not only be a function of their initial attitude, but also of past social encounters.

Previous research on conformity and normative influence has often neglected the possibility of lasting changes (after all, one does not expect persistent inability to judge length of lines after a group interaction), but with political attitudes this perseverance is a real possibility. Beyond any direct impact of social influence, the mere act of expressing an attitude (or not) can solidify it (or weaken it). Publicly stated attitudes are more resistant to subsequent persuasion than views that remain unexpressed (Roese and Olson 1994). More critically, the simple act of expressing a counter-attitudinal view can shift one’s private attitude to be more similar to the expressed view, generating persistent attitude change (Bem 1972; Festinger 1957; Cooper 2007). Thus, we expect that conformity may produce some lasting private attitude change, as well as public conformity.

Factors Influencing Political Conformity and Subsequent Attitude Change

In addition to hypothesizing about social influence, we anticipate that several factors will affect people’s susceptibility to this influence. The range of views voiced by others is one such variable. The narrower the range of expressed attitudes, the more clearly the group norm is established. This makes deviation from that norm more evident, and normative influence more powerful. The more people agree amongst themselves, the less tolerant they are of disagreement (Laumann 1973; Mutz and Mondak 2006), and the more likely they are to apply pressure to deviants and ultimately ostracize them (Schachter 1951). When a group is unanimous, people therefore feel increased pressure to conform (Asch 1951). Even one other dissenter can make it easier to resist conforming (Nemeth and Chiles 1988). Reciprocally, when group members express a range of views, it becomes unclear whether or not a group norm exists and what that norm might be, even if the discordant group has the same ideological central tendency as the unanimous group. Dissention from the norm becomes less noticeable, more difficult to punish, and more acceptable. Hence, unanimity of others is expected to increase conformity.

The degree of difference between an individual’s views and those of group members is also of interest. When an individual’s views grow more distant from the group’s mean view, the distinction between the group norm and the individual’s view becomes more apparent. Clear deviants are the subject of greater social pressure, whereas minor deviations from the group’s attitudinal norm may go unnoticed by group members, thereby failing to elicit attempts at social influence (Schachter 1951). If individuals’ attitudes are close enough to the group to seem similar to those of others, this diffuses the social influence, reducing its impact (see Latane 1981). Additionally, assimilation effects may cause individuals who hold similar views to those around them to see little or no difference between their views and those of others (Hovland et al. 1949; see also Mussweiler 2003), and therefore feel no pressure to change to views with which they already “agree”. Ultimately, the more an individual is exposed to disagreement by close others (or perhaps even strangers) the more likely they are to change those views (Huckfeldt and Sprague 2000; Visser and Mirabile 2004). Therefore, the distance of the ego from the group is also expected to relate to conformity.

These same contextual variables are also expected to influence the degree of continued attitude change. Attitudes expressed in the group might be mere public conformity for the sake of getting along, but social motives for public agreement often influence private attitudes expressed outside of the group context, as well (see Wood 2000; Kelman 1958, 1989). Additionally, expressing views to others is known to encourage persistent attitude change through both self-perception and cognitive dissonance. Either individuals take the fact that they have expressed a view to indicate that they hold that view (Bem 1972), or they are uncomfortable with having expressed views at odds with their own, and so find ways to change their views (Festinger 1957, Cooper 2007). In either case, we anticipate that the likelihood of lasting attitude change should be greatest among those who initially conformed. As such, those factors that influence initial conformity should also influence likelihood of persistent change. We therefore expect that likelihood of persistent attitude change should increase when groups exhibit unity and when the ego is farther from the group mean both because these factors should affect normative and informational influence, and also because these factors first influenced expression of views within the group.

The Current Studies

The current studies (Stony Brook University IRB #82695) assess the degree to which people conform to the attitudes expressed by others without the necessity of exchanging persuasive arguments, and whether lasting attitude change results from this process. The above discussion points to two primary and several supplementary hypotheses.

Primary H1

Individuals in the presence of others will express attitudes more in line with the attitudes of those others than their initial attitudes had been.

Primary H2

Rather than reverting to their initial private attitudes after leaving the group context, individuals will exhibit persistent attitude change.^{Footnote 2}

Supplementary H3a

Likelihood of immediate public conformity will be greater when others are unanimous.

Supplementary H3b

Likelihood of persistent attitude change will be greater when others are unanimous.

Supplementary H4a

Likelihood of immediate public conformity will be greater when the individual’s attitude is more different from the group.

Supplementary H4b

Likelihood of persistent attitude change will be greater when the individual’s attitude is more different from the group.

Supplementary H5a

Likelihood of immediate public conformity will be due, at least in part, to normative influence.^{Footnote 3}

Supplementary H5b

Likelihood of persistent attitude change will be due, at least in part, to normative influence (see Footnote 3).

Two laboratory experiments were conducted during 2008 and 2009: a preliminary experiment and a second, expanded experiment. Study 1 examines the social pressures within a group and subsequent attitude change, as well as social and personality factors that enhance change. Study 2 demonstrates that the social pressures identified in study 1 were the result of not only informational influence, but also normative influence. Findings indicate that the physical presence of group members enhances the level of conformity considerably, indicating that participants changed their attitudes not just because the attitudes of others are useful information, but also because individuals feel especially motivated to hold attitudes that are similar to those with whom they interact. The second study also examines the longevity of these social influence effects over days and weeks, rather than minutes.^{Footnote 4}

Study 1 Introduction and Method

In order to understand the role of merely knowing others’ attitudes in encouraging conformity and attitude change, we examined the reactions of participants to small groups of others (confederates) with varying attitudes. We experimentally manipulated both the unanimity of confederates and the average attitude of the confederates. The latter manipulation allowed us to vary the attitudinal distance between the participant and confederates without compromising the internal validity of the experiment. Both of these experimental manipulations were conducted in a within-subjects design to allow efficient causal inferences. The within-subjects design allows greater power and efficiency in our design.^{Footnote 5}

After providing consent, participants reported their attitudes about various issues privately, via computer. They then heard the attitudes of a small group of confederates, and were asked to state their own views verbally in the presence of the confederates in order to establish whether merely knowing the views of strangers could alter their views. Participants then reported their attitudes again privately, via computer, to establish whether persistent attitude change was occurring, rather than simply public agreement. Participants were then thanked and debriefed.

Participants

We recruited 54 undergraduate participants from a large public university, 8 of whom were excluded from all analyses for indicating suspicion when probed during debriefing.^{Footnote 6} Of the remaining participants, 24 were male, and 23 identified as Democrats, 6 as Republicans, and 11 as independents.

Procedure and Cover Story

Participants were recruited in exchange for extra credit, under the auspices of a study about group-based political discussions, although no group discussion actually occurred. Participants were told that they would complete a short questionnaire, then engage in a group activity, and finally complete another questionnaire (Fig. 1). The consent form framed the initial, group, and post-group study phases as separate tasks.

Upon arriving at the lab, participants completed an attitude questionnaire (Online Resource, Appendix A) and filler questions anonymously on a computer. They were then taken to a separate room where three confederates were being seated for the anticipated group discussion. The same student confederates were used for every participant (2 male, 1 female, 1 minority). Once the naïve participant was seated, the experimenter told the group that in order to choose the best discussion topic he would ask them some “standard political attitudes questions” and requested that they respond verbally, one at a time, with the answer option only (i.e. “somewhat agree”), so that he could record their responses. These were the attitude questions previously asked, without the filler questions. The naïve participant by “chance” became the fourth participant in the group, such that each confederate answered a question in turn before them, ensuring that naïve participants always knew others’ responses prior to responding themselves. A round of responding to a particular item took a fraction of a minute. After the final verbal question, participants were told that in their specific case a discussion would not be useful (supposedly because their group’s attitudes were redundant with previous groups) and they were then taken back to their original room without any discussion of issues. The entire group portion of the study, including answering all 30 questions, took less than 15 min. All participants then completed a post-test questionnaire that consisted of the same survey questions embedded in a series of questionnaires about personality traits, demographics, and filler items to obscure the repeated questions.

Contextual Factors Influencing Conformity

Both unanimity of confederates views and the average attitude of those confederates were randomly assigned in within-subjects experimental manipulations. The manipulation of group unanimity kept the group average on an item constant, while altering the degree of pressure toward uniformity. On the unanimous condition items, confederates gave identical responses on a particular attitude item, whereas in the non-unanimous condition items, each confederate answered slightly differently from the other two (e.g. positions 1, 2, and 3 versus 2, 2, and 2). The confederates’ average response was also experimentally manipulated to vary along the continuum from agreeing to disagreeing with the item. To avoid issue-specific confounding, both unanimity and the ideological direction of the average group response across issues were counterbalanced (e.g. which items were unanimous varied across participants by random assignment to a condition schedule). Confederates gave only the response option of their answer (e.g. “somewhat favor”), and were trained to react neutrally both when responding and when others were speaking (passive expressions and no comments).

To gauge the dissimilarity of the participant from the group, the absolute attitudinal distance between a participant’s pre-discussion private response and the average response of the confederates was assessed for each item. This attitudinal distance varied as a function of the experimental manipulation of the average group attitude and participants’ initial attitudes. Therefore, although the distance between the individuals’ initial attitudes and the group’s mean response was not itself experimentally assigned, it varied probabilistically as a function of experimental treatment.

Individual Differences

We also assessed party identification and political sophistication to gain additional insights. Those who have greater political sophistication are better able to resist persuasion through counter-arguing (Zaller 1992), and so sophistication was assessed by summing the number of correct responses to 10 factual questions (Online Resource, Appendix D, Cronbach_α = 0.79; M = .53, SD = .27). Party identification was measured with the standard branched item from the ANES, in which Republicans are at the high end of the scale.

Conformity and Attitude Change Measures

Changes in participants’ response to the same attitude items across the phases of the study were used to assess conformity and persistent attitude change. Following Kelman (1958), we deemed a participant to have conformed if his or her verbal response in the group was more similar to the group’s average response than his or her initial private response. Similarly, we deemed participants to have experienced true attitude change if they continued to express views more congruent to those of the group even when the group (and the attendant social pressure) was no longer present.

Study 1 Results and Discussion

We present our analyses in several stages (conducted using Stata, throughout the paper). First, we provide some preliminary analyses to demonstrate that social influence and conformity are in fact occurring, and to what extent. We then examine which factors especially encourage conformity. Finally, we examine whether these social influence factors induce persistent attitude change.

Overall Attitude Change

To develop a broad picture of the extent of conformity and persistent attitude change, our initial analyses examined the overall magnitude of conformity in the group and private attitude change thereafter, regardless of experimental treatment. Specifically, we investigated the degree to which participants altered their views to be more similar to the average attitude of the group by correlating their initial attitudinal distance from the group mean with the amount of change toward the group. The resulting positive correlation indicates that when participants were assigned to a group that was more conservative than them on an issue, participants became more conservative, and when their group was more liberal, they became more liberal on that issue (top panel of Fig. 2). This strong positive relationship demonstrates that the more participants’ initial views differed from those of their group, the more they changed the attitudes they expressed in the group (r = 0.507, p < .001), consistent with Hypotheses 1 and 4a. For example, when the group was more conservative, perhaps opposing something the participant was neutral about (X = −2), the participant could be expected to become more conservative, typically covering about ½ the distance between their initial position and that of the group (Y = −1), and when the group was more liberal, the participant likewise became more liberal.^{Footnote 7}

Interestingly, attitude change persisted even after the group dispersed, consistent with Hypotheses 2 and 4b (Fig. 2, bottom panel). The more participants’ attitudes stood out from the group, the more they changed, even after participants were no longer in the group setting (r = 0.238, p < .001). Thus, the impact of social influence persisted even when the source of that influence was removed. The magnitude of the correlation for the post-discussion responses is noteworthy, though predictably not as large as the correlation for responses while in a group. The drop in the correlation when the participants were no longer with the group suggests some initial compliance that was purely intended to minimize social conflict. Nonetheless, people changed their attitudes overall to be more like those of the group, demonstrating social influence even in the absence of discussion or exchange of arguments.

While these data deserve more detailed examination (below), these preliminary results have important implications. The perseverance of the significant correlation between initial difference of opinion and subsequent attitude change even after participants were no longer with others demonstrates that participants changed their private views to be more like those of the group. They did not simply verbally “go along to get along” with the group and immediately revert back to their private attitudes after the group dissolved, as might have been expected based on compliance alone (Asch 1951, Kelman 1958, 1961). When people are in a group they conform politically, and this conformity translates into actual attitude change that persists beyond the group setting.

Level of Initial Conformity

Having examined the magnitude of conformity, we next examine some descriptive information about the frequency of overall initial conformity irrespective of condition, followed by more in-depth analyses of the factors that make such initial conformity more likely. On some issues, the participant’s initial attitude matched the average group attitude, as a consequence of random assignment of group views (as expected by chance ≈1/7th of the time, n = 183). In this case not changing one’s attitude is the expected response for conformists and for those who are unaffected by the group. Due to this ambiguity, these responses were excluded from all subsequent analyses. Note that this occurred randomly across conditions because of the within-subjects manipulation of group attitude, and that particular observations, but not participants, were removed. The excluded observations did not significantly differ based on any of the condition or individual difference variables. On the remaining 1197 responses, participants conformed most of the time (57 %) and polarized away from the group only 7 % of the time (Fig. 3). Indeed, participants conformed disproportionately often (χ² = 26.17, p ≤ .001), more often than polarizing and maintaining the same attitude combined. Importantly, every participant conformed on at least 6 of the 30 responses (20 %). This level of conformity is notably higher than the prior literature on non-political judgments where approximately ¼ of the participants are able to completely resist conformity pressures on all trials (Bond and Smith 1996).

Some might argue that some of these results (particularly the correlations) might be vulnerable to a “regression to the mean” interpretation, in that the effect might be a result of those who are most different from their group (i.e. most extreme) changing toward both the mean and their group. The counterbalancing of group issue positions make this account unlikely by making regression towards the mean and conformity toward the group statistically distinguishable. Specifically, some neutral or liberal participants who become more conservative are deemed to be conforming because their group mean was conservative on an item, but others are considered not conforming on that same item because their group mean was liberal (and vice versa for conservatives), cancelling out any overall effect. Thus, significant effects indicate change not to “the” mean (as in regression to the mean), but toward the different means of participants’ groups. Additionally, regression to the mean cannot account for significant differences between experimental conditions. Regression to the mean dictates a general drift of initially extreme scores toward the average, rather than two equivalent experimental groups showing differing results. Significant effects of condition therefore rule out this concern, and so we turn to analyzing factors that encourage conformity.

Conditions Encouraging Conformity

To examine the antecedents of conformity, we assessed whether or not participants’ attitudes expressed in the group were closer to the group average response than their initial attitude on each item. We chose this measure so as not to give undo weight to those who begin with the most extreme attitudes, and therefore have the most room to change. If a participant changed their attitude to align better with the group, the observation scored 1, and if not a 0.

Several statistical strategies were used to examine which factors encourage conformity, each yielding similar results to those reported here. These included different ways of coding the dependent variable and differing analyses. Attesting to the robustness of these results, parameter estimates changed minimally (generally within rounding error) and the substantive interpretation of the results was highly consistent across every method used.^{Footnote 8} ^, ^{Footnote 9} For the sake of clarity to the broadest audience, the simplest and most directly relevant analytic method is presented here. Thus, we report logistic regression analyses with nonconformity as the baseline condition (see Table 1, Col 1). Each response for each participant was treated as a separate observation. Responses were then clustered by the participant in the analysis to account for their non-independence. These analyses were conducted first on attitude change expressed in the group, then on attitudes expressed after the group phase to examine persistent attitude change.

Table 1 Logistic regression models explaining participant attitude change in Study 1

Full size table

Social Factors and Influence

In accordance with Hypotheses 3a and 4a, the contextual variables of unanimity and attitudinal distance from the group strongly influenced the likelihood that participants would conform on any given response (Table 1, Col 1). The probability of conforming significantly increases when the group is unanimous (predicted probability increases by 9 percentage points, holding all other variables at their means),^{Footnote 10} supporting the idea that unanimity sends a clear message about what is correct and appropriate, generating increased conformity. Additionally, the significant coefficient for attitudinal distance indicates that the more the group members’ average attitude differed from a participant’s initial attitude, the more likely that participant was to conform (the most discrepant being an estimated 39 percentage points more likely than the least discrepant). This is consistent with the idea that people who differ more feel more pressure to bring their views in line with those of others.

Individual Differences

Importantly, not every individual was equally likely to conform. The politically sophisticated were significantly less likely to conform (a predicted 26 percentage points less likely than the least sophisticated), suggesting that sophisticates are less susceptible to social influence (consistent with Zaller 1992). Importantly, although sophisticates were generally less likely to conform, sophistication did not completely inoculate participants against the social pressure created in the group setting. Predicted probabilities indicate that those with the highest levels of political sophistication could still be expected to conform on 45 % of trials, with all other variables held at their means (compared to the observed 57 % conformity across all respondents). There were no significant interactions between sophistication or party ID and unanimity or distance.^{Footnote 11}

Persistent Attitude Change

The presence of others had an immense impact on participants’ attitudes, even after those others were no longer present (Figs. 2, 3). Conformity was likely to persist as real attitude change, and even some of those who did not initially conform later showed some signs of social influence in an apparent “sleeper effect”.^{Footnote 12} To get a sense of the general frequency of persistent change, we again examine some basic statistics.

Prior Conformity

The most intriguing effect that emerges from these data is the immense effect of participants’ behavior while they were in the group upon their attitudes after they left the group (Fig. 3). Consistent with Hypothesis 2, participants who initially conformed on an issue were more likely than those who did not to retain attitudes more similar to those of group members’ after the group was no longer present (χ² = 266.17, p < .001). Specifically, 58 % of initial conformers’ attitudes remained more compatible with group members’. By comparison, if the participant initially resisted the pressure to conform, only 11 % exhibited attitude change later.

Although this analysis is somewhat coarse, the sheer magnitude of these findings attests to the impact of actual behavior on attitudes. To examine persistent socially-induced attitude change in more detail, we estimated another logistic regression, this time examining which factors influenced the degree to which conformity persisted as private attitude change rather than mere public lip-service (Table 1, Col 2). In this case, the dependent variable was persistent attitude change, with those observations on which a participant conformed in the group and continued to exhibit a changed attitude being coded as 1, while other observations were coded as 0.

Social Factors and Influence

Unanimity and disparity of views both had continuing significant effects on persistent attitude change (Table 1, Col 2) consistent with hypotheses 3b and 4b. On attitude items where the group was unanimous, respondents were predicted to be 6 percentage points more likely to exhibit continued attitude change. Further, the participants who were most different from their group were an estimated 32 percentage points more likely to exhibit persistent change after the group session had ended.

Overall, the attitudes expressed in the presence of strangers had impressive implications not just for conformity, but also for the attitudes participants later professed. As with initial conformity, unanimity and disparity of views also influence persistent attitude change in turn. Critically, this demonstrates the occurrence of persistent attitude change without any persuasive arguments being exchanged.

Study 2 Introduction

Study 1 results advance our understanding of social influence by demonstrating the power of brief, potentially offhand expressions of attitude to generate public conformity and private attitude change without any persuasive arguments. Still, important outstanding questions remain, most essentially the question of, “Why?” Although this attitude change is not the result of overt persuasion (since arguments were not exchanged), it remains unclear whether attitude change was the result of normative influence (change to fit in with the group, rather than because of new information inferred from others views). Additionally, other questions remain, particularly about the susceptibility of strong attitudes (vs weak or non-attitudes) to this effect, and the longevity of these effects.

The most prominent concern about study 1 is whether the changes in views might be the result of solely informational influence: taking the views of others as evidence that an attitude is correct (e.g. “Can these 3 people all be wrong?”), and not normative influence per se. Put another way, participants may have had primarily weak or non-attitudes, such that even the limited evidence of other people’s views was sufficient to sway them. If this were the case, attitude change might have been a result of the informational role of others’ views in indicating which view is correct. In the case of normative influence, attitude change would be due to the social information others’ attitudes provided about what others felt to be socially appropriate and expected. While informational influence is interesting in its own right for generating attitude change without supporting arguments, normative influence is still more striking in signifying the role of purely social situations in attitude formation and change. A changed triggered by normative influence is a change motivated neither by correctness nor even pro-attitudinal bias, but by fitting in.

This informational or non-attitude explanation is consistent with the observed buffering effect of political sophistication against conformity. This alternate explanation seems unlikely, however, given the overall high level of conformity and that analyses indicate that even political sophisticates (who could be expected to have stronger priors and more information) can still be expected to conform at relatively high rates. Additionally, sophisticates were not significantly less responsive to unanimity and attitudinal distance, and were not significantly less likely to experience persistent attitude change. Indeed, sophisticates showed some signs of being more responsive to the unanimity manipulation, which is at odds with a non-attitude explanation (footnote 11 and Online Resource, Appendix E). Study 2 allows us to more deeply examine this issue primarily by comparing the influence of exposure to the attitudes of real, immediately present others to the influence of exposure to those same other’s attitudes without their immediate presence. Study 2 also assesses attitude strength, to determine whether weaker and non-attitudes respond differently.

Study 2 Method

Study 2 is an expanded version of study 1, which includes a between-subjects manipulation of normative influence in which participants were randomly assigned to a live or computerized group session (see below). Additionally, study 2 focused on 4 salient issues, and assessed attitude strength of each in the first phase of the study. Finally, a follow-up attitude questionnaire was sent to participants to track longer-term attitude change.