An Informational Influence Perspective on (Non) conformity: Perceived Knowledgeability Increases Expression of Minority Opinions

Normative Versus Informational Influence Processes of Opinion Conformity

The tendency for minorities to express their opinions less readily than those who are in the majority has been demonstrated across a range of measures, including likelihood of voicing one’s opinion in a variety of scenarios (Hornsey, Majkut, Terry, & McKimmie, 2003; Noelle-Neumann, 1993), hypothetical or retrospective willingness to enter hypothetical “dangerous discussions” with people whose opinions differ from one’s own (Eveland & Shah, 2003; Matthes et al., 2010), and stating one’s opinion in a group discussion with confederates (Hayes, Uldall, & Glynn, 2010). These effects are smaller but still robust in computer-mediated settings (Ho & McLeod, 2008), and in situations where the expression is less overtly public (e.g., as with the “minority slowness effect” whereby the time people take to report their opinions is measured on the computer; Bassili, 2003; Huge & Glynn, 2013; Rios & Chen, 2014).

Why is opinion conformity so prevalent? Since the mid-20th century, two explanations—which are theoretically distinct but difficult to disentangle empirically—have emerged. The first, normative influence, posits that people seek to fit in with their peers and to avoid social sanctions. In other words, they conform because of the potential repercussions of not doing so (e.g., “If my opinion is not shared by others, I will face embarrassment and/or ostracism”). The second, informational influence, posits that people are motivated to accept others’ opinions as valid representations of reality. Thus, they conform because they look to the majority for evidence that their opinions are true (i.e., “If my opinion is shared by others, it must have some basis in reality”; Deutsch & Gerard, 1955; Price & Oshagan, 1995).

Recent research has uncovered several moderators of opinion conformity—that is, conditions under which individuals are more versus less reticent to express their minority opinions. Most of these moderators provide evidence for normative rather than informational influence processes. For instance, people are especially susceptible to the “minority slowness effect” when they are high in fear of negative evaluation (i.e., concerns that they will be judged harshly by others; Huge & Glynn, 2013) or high in need to belong (i.e., motives to be accepted by and have close interpersonal relationships with others; Rios & Chen, 2014). Relatedly, people high in fear of social isolation (i.e., the fear of being excluded or shunned by others) are particularly likely to report that they would keep their minority opinions to themselves in hypothetical situations (Matthes et al., 2012). By contrast, people low in fear of negative evaluation, need to belong, or fear of social isolation are equally willing to express or report minority and majority opinions.

Most relevant to the present studies, opinion conformity can also be influenced by properties of the attitude itself—namely, attitude strength, or the durability and impact of one’s attitude toward an issue (Petty & Krosnick, 1995). In one set of studies, minorities who held their opinions with high levels of certainty—that is, those who were confident in their opinions and believed their opinions were valid (see Tormala & Petty, 2002)—reported a greater likelihood of voicing these opinions to others than did minorities who held their opinions with low or moderate levels of certainty (Matthes et al., 2010). Similarly, Hornsey and colleagues (2003) found that individuals with a strong (relative to weak) moral basis for their attitudes reported being more willing to express minority opinions. However, as we explain below, certainty and moralization do not clearly map onto either normative or informational influence processes. A primary goal of our studies is to more effectively demonstrate the role of informational influence in opinion (non)conformity tendencies, using a yet untested moderator of minority-majority opinion expression: perceived knowledgeability about an issue. Specifically, we predict that individuals who perceive themselves as high (relative to low) in knowledgeability will more readily express their minority opinions.

Metacognitive Predictors of Opinion Expression

Both attitude certainty and moralization are considered metacognitive properties of attitude strength. According to the metacognitive model (Petty et al., 2007), attitudes are defined by both primary evaluations (i.e., positive, negative) and secondary metacognitions, or the extent to which the attitudes are endorsed (e.g., “I am certain/uncertain that I feel positively about X”). Attitudes that are endorsed weakly—that is, those with weak metacognitive properties—are more susceptible to change and less likely to influence behavior. Recent theoretical extensions of this model to communication-related outcomes have suggested that, in deciding to adopt or change an opinion, people who hold weak attitudes are especially likely to rely on cues about whether their opinions are different from others (e.g., whether they are in the minority or majority; Song & Ewoldsen, 2015). Thus, one possible implication of the metacognitive model for opinion expression is that people with weakly held attitudes are especially likely to conform to the majority, and that strongly held attitudes may serve as a buffer enabling people to express their opinions regardless of majority or minority status.

Attitude certainty and moralization each involve conviction in beliefs (Luttrell, Petty, Briñol, & Wagner, 2016; Matthes et al., 2010; Tormala & Petty, 2002). However, attitude strength is multidimensional and encompasses other metacognitive indicators, such as those that reflect integration of the attitude into one’s self-concept or knowledge structure (Pomerantz, Chaiken, & Tordesillas, 1995). One of these indicators, perceived knowledgeability (i.e., beliefs that one has expertise in and knows the relevant facts about the issue at hand), has yet to be tested as a predictor of minority-majority opinion expression and may be especially relevant to informational influence processes. In the case of attitude certainty or moralization, it is difficult to determine whether normative influence and/or informational influence is at play. If people who lack conviction in their opinions are most likely to keep their minority opinions to themselves, this could mean they are uncertain that their minority opinions would enable them to fit in with others (normative influence), uncertain that their minority opinions represent reality (informational influence), or a combination. However, if people withhold their minority opinions to the extent that they see themselves as less knowledgeable about an issue, this would directly demonstrate their motives to express only those opinions that are factually accurate and valid.

Perceived knowledgeability and attitude certainty, despite being correlated with one another (Krosnick, Boninger, Chuang, Berent, & Carnot, 1993), differ in their correlations with demographic variables (e.g., education, gender, race; Visser & Krosnick, 1998) and can be considered distinct psychological constructs. For instance, it is possible for people to feel they know a lot about an issue but still not be certain of their attitude—or alternatively, to be certain of their attitude even when they do not claim to know very much about the issue. Thus, one goal of the present studies was to test whether perceived knowledgeability, not just attitude certainty, moderates conformity processes. Doing so is critical for determining when informational influence—not just normative influence, as has been uncovered previously (e.g., Huge & Glynn, 2013; Matthes et al., 2012; Rios & Chen, 2014)—operates in decisions to express versus withhold minority opinions. Although people report being more likely to voice their opinions toward issues on which they feel as though they are more knowledgeable (Dalisay, Hmielowski, Kushin, & Yamamoto, 2012; Salmon & Neuwirth, 1990; Shamir, 1997), it remains to be seen whether perceived knowledgeability has stronger effects on the expression of minority versus majority opinions, as well as whether the effects of perceived knowledgeability extend beyond hypothetical to actual expression of opinions.

The Present Research

In two studies, we manipulated whether participants believed they held a minority (uncommon) or majority (common) opinion in society as a whole on a controversial issue (several issues in Study 1 and gun control in Study 2). Next, we assessed participants’ decisions to express their opinions in ostensibly real online discussions with another person, using two different dependent measures: speed of opinion expression in Study 1 (i.e., the “minority slowness effect”; Bassili, 2003; Huge & Glynn, 2013; Rios & Chen, 2014) and decisions to enter a “safe” (same-opinion) versus “dangerous” (opposite-opinion) discussion (Eveland & Shah, 2003). Across both studies, we predicted the following hypotheses:

Study 1

In Study 1, we measured perceived knowledgeability and certainty about one of several controversial social issues in the United States. (We did not predict any differences between issues; rather, we wanted to determine whether our hypothesized results would generalize to a variety of topics.) We selected these issues because on many of them, the distribution of public opinion in the general population is not well-known. Consequently, we could convincingly manipulate the percentages of minority versus majority opinion holders. With some widely publicized issues (e.g., same-sex marriage, preferences for political candidates), we were concerned that most participants would have been regularly exposed to opinion polls prior to our study, and hence that they would find our “background information” less plausible.

We then gave participants ostensibly real “background information” about most other Americans’ opinions on their assigned issue, suggesting that they themselves were in either the majority or minority relative to society as a whole. Finally, we measured the amount of time participants took to report their own opinion prior to a supposed discussion with another person. The tendency for minorities to take longer to report their opinions than majorities is known as the “minority slowness effect” (Bassili, 2003; Huge & Glynn, 2013; Rios & Chen, 2014). This measure is more sensitive than many other means of assessing opinion expression, as it captures subtle variations in hesitancy (i.e., on a continuum) rather than simply gauging whether people do or do not choose to express their opinions (Bassili, 2003). In addition, because people are often unaware of precisely how much time passes before they report their opinions, such a measure minimizes the social desirability concerns involved in directly asking what one would do in a hypothetical situation.

Results

We predicted that participants who learned they were in the minority would be slower than those who learned they were in the majority to report their opinions if they were low in perceived knowledgeability or attitude certainty, but that this effect would be reduced or eliminated among participants who were high in perceived knowledgeability (Hypothesis 1) or attitude certainty (Hypothesis 2). We did not expect any moderating effects of the issue to which participants were assigned. To test these hypotheses, we conducted separate regression analyses with opinion condition (0 = minority, 1 = majority), each potential moderator (perceived knowledgeability and attitude certainty), and the relevant two-way interaction term as predictors and reporting time as the outcome, collapsing across the four issues and using the PROCESS macro in SPSS (Hayes, 2013).

Perceived knowledgeability

There were no significant overall effects of knowledgeability or opinion condition (ps > .218, total R² = .005). However, the hypothesized interaction between knowledgeability and opinion condition was significant (b = 0.45, SE = .19, β = .43), t(343) = 2.36, p = .019 (total R² = .021; see Figure 1). To decompose this interaction, we used the Johnson-Neyman technique (Hayes & Matthes, 2009), which is advantageous to “pick-a-point” methods (e.g., testing significance at 1 SD above and below the mean value of the moderator) in that it identifies entire regions of significance—in other words, a range of values of the moderator at which the independent variable produces a statistically significant effect.

OPEN IN VIEWER

Figure 1.

Time (seconds) taken to report opinion as a function of condition (minority vs. majority opinion) and perceived knowledgeability, Study 1.

Analyses using the Johnson-Neyman technique revealed that minorities were slower to report their opinions than majorities when their perceived knowledgeability was at or below 3.85 (out of 9), or 0.95 SD below the mean (b = −0.93, SE = .47, t = −1.97, p = .05, 95% confidence interval [CI] = [−1.87, 0]). In other words, minorities and majorities did not significantly differ from one another in speed of reporting their opinions when their perceived knowledgeability was above 3.85. There was no level of perceived knowledgeability at which minorities were faster to report their opinions than majorities. In addition, perceived knowledgeability predicted faster reporting times among those in the minority opinion condition (b = −0.37, SE = .14, β = −.21), t(343) = −2.59, p = .010, whereas it did not predict reporting times among those in the majority opinion condition (b = 0.08, SE = .13, β = .04), t(343) = −0.62, p = .538. Hypothesis 1 was thus supported.

Attitude certainty

There were no significant overall effects (ps > .180, total R² = .006), but the Certainty × Opinion Condition interaction was significant (b = 0.45, SE = .21, β = .49), t(343) = 2.17, p = .030 (total R² = .019; see Figure 2). Decomposition of the interaction indicated that minorities were slower to report their opinions than majorities when their attitude certainty was at or below 4.70 (out of 9), or 1.19 SD below the mean (b = −1.04, SE = .53, t = −1.97, p = .05). There was no level of attitude certainty at which minorities were significantly faster to report their opinions than majorities. Additional analyses revealed that attitude certainty was negatively correlated with speed of reporting minority opinions (b = −0.37, SE = .15, β = −.19), t(343) = −2.51, p = .013, but not majority opinions (b = 0.07, SE = .14, β = .04), t(343) = 0.52, p = .603. These results support Hypothesis 2.

OPEN IN VIEWER

Figure 2.

Time (seconds) taken to report opinion as a function of condition (minority vs. majority opinion) and attitude certainty, Study 1.

Analyses with issue as a factor

We did not predict any differences in the above results based on the issue to which participants were assigned. Nevertheless, we conducted a set of exploratory analyses in which we created three issue dummy variables (airline seating regulations = 1, other = 0; gun control = 1, other = 0; capital punishment = 1, other = 0), with racial profiling as the reference category. We then regressed response time onto opinion condition (0 = minority, 1 = majority), the moderator (knowledgeability or certainty), the three issue dummy variables, and all two- and three-way interaction terms. As above, we conducted two separate regression analyses: one with knowledgeability and the other with certainty as a moderator. The main effects were interpreted in the first step, the two-way interaction terms in the second step, and the three-way interaction terms in the third step of the analyses (Cohen, Cohen, West, & Aiken, 2013).

When perceived knowledgeability was the moderator, there were no main effects (ps > .225). However, the two-way Opinion Condition × Knowledgeability remained significant (b = 0.39, SE = .20, β = .38), t(334) = 1.98, p = .049. Unexpectedly, a three-way interaction between opinion condition, knowledgeability, and issue (airline seating regulations vs. racial profiling) also emerged (b = 1.15, SE = .56, β = .62), t(331) = 2.06, p = .040, such that the predicted Opinion Condition × Knowledgeability effect was stronger for airline seating regulations (b = 0.93, SE = .39, β = .90), t(331) = 2.41, p = .016, than for racial profiling (b = −0.22, SE = .40, β = −.21), t(331) = −0.54, p = .589. There were no other significant three-way interactions (ps > .227).

When attitude certainty was the moderator, there were no main effects or three-way interactions with issue (ps > .209). The two-way interaction between opinion condition and attitude certainty, reported earlier, was marginally significant in this analysis (b = 0.41, SE = .25), β = .45, t(334) = 1.65, p = .101.

Study 2

In Study 2, we sought to conceptually replicate our findings using a new dependent measure of opinion expression: decisions to discuss the focal issue with someone who holds the opposing (rather than the same) opinion as oneself. Some prior research has employed measures of willingness to enter hypothetical “dangerous discussions” (e.g., Matthes et al., 2010) or has asked participants to retrospectively report how often they have discussed issues with agreeing versus disagreeing others (Eveland & Shah, 2003). To our knowledge, however, this is the first measure that requires people to choose their preferred partner in an ostensibly real discussion, thus minimizing social desirability concerns involved in hypothetical decisions or retrospective reports.

Results

We hypothesized that among participants who felt less knowledgeable about gun control, those in the minority opinion condition would choose a discussion partner with the opposite opinion as theirs less often than would those in the majority opinion condition. However, we hypothesized that this effect would be attenuated among participants who felt knowledgeable about gun control (Hypothesis 1). In addition, we predicted that participants low in attitude certainty would demonstrate the same pattern of results as those low in perceived knowledgeability (Hypothesis 2). To test these predictions, we first created a dummy variable for opinion condition (0 = minority, 1 = majority) as in Study 1. Next, we conducted two separate logistic regression analyses using the PROCESS macro (Hayes, 2013) with opinion condition, the moderator (knowledgeability or certainty), and the two-way interaction terms as the predictors and discussion partner preference (0 = same opinion, 1 = opposite opinion) as the outcome.

Perceived knowledgeability

There were no overall effects of perceived knowledgeability or opinion condition (ps > .657). However, the Opinion Condition × Knowledgeability interaction was significant (b = 0.28, SE = .12), z = 2.39, p = .018 (see Figure 3). Decomposition of the interaction using the Johnson-Neyman technique indicated that minorities were significantly less willing to choose a discussion partner with the opposite opinion when their perceived knowledgeability was at or less than 3.82 out of 9, or 0.96 SD below the mean (b = −0.59, SE = .30, z = −1.96, p = .05, 95% CI = [−1.19, 0]). Minorities and majorities were equally willing to choose an opposite-opinion discussion partner when knowledgeability was above 3.82; there was no level of knowledgeability at which minorities were more willing than majorities to choose an opposite-opinion discussion partner. These results support Hypothesis 1. Moreover, perceived knowledgeability was marginally positively correlated with choice of an opposite-opinion partner in the minority opinion condition (b = 0.14, SE = .09), Wald = 2.81, p = .094, but was marginally negatively correlated with choice of an opposite-opinion partner in the majority opinion condition (b = −0.13, SE = .08), Wald = 2.83, p = .093.

OPEN IN VIEWER

Figure 3.

Choice of same-opinion (1) versus opposite-opinion (0) discussion partner as a function of condition (minority vs. majority opinion) and perceived knowledgeability, Study 2.

Attitude certainty

Neither overall effect of certainty nor opinion condition reached significance (ps > .658); the only significant effect to emerge was the predicted Opinion Condition × Attitude Certainty interaction (b = 0.37, SE = .14), z = 2.65, p = .008 (see Figure 4). Decomposition of the interaction indicated that minorities were less likely than majorities to pick an opposite-opinion discussion partner when their attitude certainty was 5.70 or less (out of 9), or 0.75 SD below the mean. There was no level of attitude certainty at which minorities were more likely than majorities to choose an opposite-opinion discussion partner. Hypothesis 2 was therefore supported.

OPEN IN VIEWER

Figure 4.

Choice of same-opinion (1) versus opposite-opinion (0) discussion partner as a function of condition (minority vs. majority opinion) and attitude certainty, Study 2.

In addition, as with perceived knowledgeability, attitude certainty and choice of an opposite-opinion partner were marginally negatively correlated in the majority opinion condition (b = −0.18, SE = .09), Wald = 3.59, p = .058, and marginally positively correlated in the minority opinion condition (b = 0.20, SE = .11), Wald = 3.49, p = .062.

General Discussion

As noted earlier, one distinction between perceived knowledgeability and attitude certainty, from an attitude strength perspective, is that knowledgeability is most relevant to the embeddedness of the opinion in one’s self-concept and areas of expertise (Pomerantz et al., 1995). Attitude certainty, however, is most relevant to conviction in one’s opinion. Our studies thus extend the metacognitive model (see Petty et al., 2007) to an important communication process—decisions to express an uncommon opinion—by using two different metacognitive indicators of attitude strength as predictors. Whereas attitude certainty has been tested in only one previous set of studies using a single-item measure (Matthes et al., 2010), the present studies are the first to test knowledgeability as a moderator of minority-majority opinion expression.

Our studies also have implications for the psychological processes underlying opinion conformity. Attitude certainty could potentially demonstrate processes of either normative influence (i.e., people who are certain that their attitudes enable them to fit in with others will express their opinions regardless of minority status) or informational influence (i.e., people who are certain that their attitudes represent the truth will express their opinions regardless of minority status). Similarly, existing research on individual-difference predictors of opinion conformity has focused more on normative than informational influence processes. For instance, this research has shown that need to belong (Rios & Chen, 2014), fear of negative evaluation (Huge & Glynn, 2013), and fear of isolation (Matthes et al., 2012) moderate majority-minority opinion expression. By contrast, the moderating effects of knowledgeability on majority-minority opinion expression clearly point to informational influence as an underlying mechanism, as people only choose to express minority opinions when they feel they have expertise on an issue (i.e., when they feel their opinions are valid markers of reality). Thus, our studies shed light on when people will conform (or not) as a function of their motives to be knowledgeable.

Our research makes several methodological contributions to the existing literature as well. First, whereas the only prior demonstration of the effect of attitude certainty on minority opinion expression assessed certainty using just one question (Matthes et al., 2010), we use a multi-item, previously validated measure of attitude certainty (Petrocelli et al., 2007). Second, our four-item measure of knowledgeability, which we developed across two studies, is new and could be utilized by other researchers interested in the differences between metacognitive indicators of strength. Third, and perhaps most important, prior work has largely measured opinion expression in hypothetical situations—for example, by asking participants how likely they would be to voice their opinions to someone who disagreed with them, or how likely they would be to participate in a demonstration (e.g., Hornsey et al., 2003; Matthes et al., 2010). The present studies, by contrast, used supposedly real measures of opinion expression: speed of reporting one’s opinion for an upcoming debate (Study 1) and the decision to discuss an issue with someone who holds the opposite opinion (Study 2). As such, our results demonstrate that knowledgeability and attitude certainty do not simply predict opinion minorities’ claims that they would express their opinions; rather, knowledgeability and certainty predict the readiness with which minorities actually do so.

One limitation of the current research is that the effect sizes in both studies were relatively small. However, we argue that even small effects in this context are meaningful. Regarding Study 1, our effect sizes are actually larger than in previous minority slowness effect studies (e.g., Huge & Glynn, 2013), possibly because we assessed attitudes toward controversial social and political issues, whereas previous studies assessed attitudes toward ordinary objects (e.g., spiders, puppies). More critically, in real-life situations, there are many ways in which people may choose to express an opinion versus keep silent; thus, such decisions are rarely dichotomous (see Hayes, 2007). An advantage of assessing speed of opinion expression is that it is sensitive to variation, and so it can capture differences of a mere second or two that naturally occur in many conversational settings. Regarding Study 2, choosing to discuss an issue with an opposite-opinion (instead of a same-opinion) partner, however infrequently, could have downstream effects on open-mindedness and perhaps ultimately attitude change. Given people’s pervasive tendencies to affiliate and converse with like-minded others (Byrne, 1961), it is critical to determine when they might be motivated to override these tendencies. Across both studies, the fact that the results of just one opinion poll can change people’s patterns of opinion expression even slightly is notable. Moreover, our studies likely underestimated the strength of the moderating effects of knowledgeability and certainty on minority-majority opinion expression, as effect sizes tend to be less pronounced in computer-mediated than face-to-face settings (Ho & McLeod, 2008) and in smaller group than larger group settings (Glynn, Hayes, & Shanahan, 1997).

Another limitation of the present work is that only one issue (gun control) was examined in Study 2. Because the dependent measure (selection of a same-opinion vs. opposite-opinion discussion partner) required participants to make a dichotomous choice, we needed more power than in Study 1—in which the dependent measure (speed of reporting opinion) was continuous—to detect a significant effect. Thus, even with a sample size of 400 (which was our targeted number of participants in Study 2), we would have been underpowered to test for differences between issues. Encouragingly, past studies focusing on hypothetical or retrospective willingness to enter “dangerous discussions” have collectively demonstrated effects across a variety of social and political issues (e.g., Eveland & Shah, 2003; Matthes et al., 2010), and so we would also expect our results to apply to topics other than gun control. We see this as a fruitful direction for future research, especially given that our “dangerous discussion” measure is the first of its kind to assess actual choices prior to ostensibly real interactions.

In the future, it will be important for researchers to test the role of informational influence in conformity even more directly than was done in the present studies. Traditional conceptions of normative and informational influence focus on adoption, rather than expression, of majority (vs. minority) opinions (Deutsch & Gerard, 1955; Price, Nir, & Cappella, 2006). Although opinion adoption and expression are closely related, it remains to be seen whether individuals low (but not high) in perceived knowledgeability are more easily persuaded by, and hence more likely to hold, majority than minority opinions. This would provide additional evidence that people are motivated to espouse the views of most others because they feel these views best represent reality.

Additional research should also more thoroughly investigate possible differences across issues. In Study 1, we unexpectedly found that the moderating effect of knowledgeability on majority-minority opinion expression was stronger for airline seating regulations than for racial profiling at airports. As noted earlier, the issue of airline seating regulations was rated as less divisive and controversial, compared with the other issues in Study 1. It may be that relatively uncontroversial issues are less susceptible to social desirability concerns, and hence produce clearer and more consistent effects of minority/majority opinion status on measures of expression, than issues that people view as riskier to discuss with others. It would be interesting to follow up on this possibility. Again, however, we still observed significant moderating effects of perceived knowledgeability and attitude certainty collapsing across issues, consistent with our central hypotheses.

Conclusion

The present studies introduce one new factor (perceived knowledgeability) and shed more light on a previously tested factor (attitude certainty) influencing minorities’ decisions to voice their opinions in interpersonal contexts. By showing that perceived knowledgeability—a metacognitive indicator of attitude strength relevant to informational influence—can make people more willing to express their views to others regardless of their minority status, we fill a gap in the existing literature, which has primarily investigated normative influence processes of conformity. Whether individuals who perceive themselves as knowledgeable or individuals who feel certain in their attitudes are most successful in influencing others and ultimately enacting social change is still an open question. In any case, the presence of multiple types of opinion minorities in society may engender not only diversity in which opinions are expressed but also diversity in how opinions are expressed.