Moral Conviction As Social Identity

1 We’re Good People

Moral Conviction as Social Identity

Abstract

Moral convictions—attitudes that people construe as matters of right and wrong—have unique effects on behavior, from activism to intolerance. Less is known, though, about the psychological underpinnings of moral convictions themselves. I propose that moral convictions are social identities. Consistent with the idea that moral convictions are identities, I find in two studies that attitude-level moral conviction predicts (1) attitudes’ self-reported identity centrality and (2) reaction time to attitude-related stimuli in a me/not me task. Moral conviction also predicted participants’ scores on a latent identity-centrality variable derived from those two measures.

Consistent with the idea that moral convictions are social identities, I find evidence that participants used their moral convictions to perceive, categorize, and remember information about other individuals’ positions on political issues, and that they did so more strongly when their convictions were more identity-central.

Keywords: Moral Conviction, Politics, Identity, Intergroup Relations

Word count: 9,988 words

2 We’re Good People: Moral Conviction as Social Identity

The attitudes that people imbue with moral significance are uniquely powerful. These moralized attitudes are resistant persuasion, elicit strong emotions, and inspire political action and intolerance. Moreover, attitude moralization—or, to use Skitka and colleagues’ term, moral conviction—has effects distinct from other attitude-strength variables, suggesting that there is something psychologically unique about regarding an attitude as a matter of right and wrong

(e.g., Skitka, 2014).

Although this work has produced useful insights regarding the many consequences of construing attitudes in moral terms, it provides fewer clues about what moral convictions are.

That is, in what ways does the psychological representation of a moral conviction differ from the representation of a non-moral attitude? I propose that, unlike non-moral attitudes, moral convictions are represented as social identities—as a key element of individuals’ self-concept that guides how they perceive and relate to others.

Moral Conviction

Skitka and colleagues have defined moral convictions as attitudes that individuals attribute to their “fundamental beliefs about right and wrong” rather than to social conventions or to personal preferences (Skitka, 2010). Moral convictions have a number of special properties.

For example, people tend to believe that their moral convictions are universally and objectively correct—injunctions by which all people ought to abide. People who hold moral convictions on a given issue are also uniquely resistant to persuasion, even controlling for the certainty and personal importance of their attitude, and they manifest strong positive and negative emotions about these issues (Skitka & Wisneski, 2011). In addition, moral convictions mobilize; people with issue-specific moral convictions were especially likely to vote in line with their attitudes in

3 the 2004 presidential election (Skitka & Bauman, 2008) and appear to be more likely to vote overall (Morgan, Skitka, & Wisneski, 2010). Finally, moral convictions have a potent influence on how people evaluate other social actors. People prefer greater social distance from others who disagree with them in moral domains than they do from others who disagree with them in non- moral domains. Moreover, where political norms favoring civil liberties are not especially strong, people are even willing to subject moral dissenters to government surveillance and censorship (Skitka, Liu, Yang, Chen, Liu, & Xu, 2012). In sum, moral convictions are resistant to change and have a strong impact on individuals’ reasoning, their emotions, and their behavior, with important consequences for how they evaluate and interact with others.

That said, the rich body of literature documenting the consequences and correlates of moral conviction has done less to clarify the psychological bases of moral convictions. The construct is operationalized as the average score across 2-4 interrelated self-report items. One item, for example, asks respondents to indicate “the extent to which [their] position on [an attitude object]” is “a reflection of [their] core moral beliefs and convictions.” All of the questions ask respondents directly whether their attitude is a moral one. And although moral- conviction research has shown time and again that this “moral” label is affixed to a heady psychological cocktail with important effects, the ingredients of that cocktail remain unclear.

What psychological characteristic distinguishes a moral conviction from other strong attitudes?

The current study tests the hypothesis that moral convictions are unique in that they delineate social identities—that is, unlike other attitudes, they help individuals define who they are, who is like them, and who is not.

Morality and the Self-Concept

4 Psychologists have long assumed that individuals’ sense of self is connected to their beliefs about what is good, desirable, and right, without formally testing this connection. In his

“functional approach” to attitudes, for example, Katz (1960), proposed that value expression is one of the key functions that attitudes might serve—that people might hold an attitude because it helps them to express their beliefs, and in turn, something about who they are. In a similar vein, scholars of attitude strength have examined attitudes’ value-relevance as a type of “personal involvement” in an issue (Boninger, Krosnick, Berent, & Fabrigar, 1995; Thomsen, Borgida, &

Lavine, 1995). In short, attitudes researchers have at times assumed that individuals’ beliefs and values are connected to the self.

This assumption is not unreasonable. When people realize that they have violated some widely-held value, they behave as though their positive self-image is threatened (Pagliaro,

Ellemers, Barretto, Di Cesare, 2016). This evidence suggests a motive to be “moral” as a means to feel good about oneself. This motive (and the ease with which it can be satisfied) is evident in moral licensing—when people indulge in less-than-ethical behavior after performing some token good (Merritt, Effron, & Monin, 2010). Evidence suggests that this pattern is at least partly attributable to a private desire to be moral, not just to impress others (Khan & Dhar, 2006;

Monin & Miller, 2001). The motive to be moral is also evident in “do-gooder-derogation”—the tendency for perceivers to resent others who are more moral than themselves. The fact that this tendency is attenuated when perceivers have the opportunity to affirm themselves prior to evaluating the “do-gooder” (Minson & Monin, 2012) underscores the role of identity in the phenomenon; others’ morality is only threatening when it reflects poorly on the self. Taken together, these findings suggest individuals’ moral beliefs provide a rubric for self-evaluation.

5 That said, most of the evidence that morality structures self-evaluation comes from traits and behaviors that are consensually viewed as moral or immoral. Lying, cheating, and overt discrimination are bad. Helping others and being honest are good. People may generally define themselves by how well they uphold these broadly shared values, but if moral convictions are social identities, then individuals’ more idiosyncratic, concrete moral beliefs (e.g., that legal access to abortion is a moral necessity vs. a moral wrong) should also help constitute their self- concept. No existing research of which I am aware directly tests the idea that individuals’ moral attitudes toward specific issues are uniquely central to identity.1 This is the first hypothesis I test in the current study—that an attitude’s identity centrality (i.e., the connection between an attitude and an individual’s self-concept) is positively related to the moral conviction that individuals ascribe to that attitude.

Morality and Social Groups

Morality is no less important in social perception than it is in self-perception. For instance, people can classify targets as trustworthy or untrustworthy within milliseconds of exposure to the target’s face (Todorov, Pakrashi, & Oosterhof, 2009; Willis & Todorov, 2006).

Moreover, information about targets’ morality tends to dominate evaluations of those targets, trumping information about competence (Fiske, Cuddy, & Glick, 2007) and sociability (Landy,

Piazza, & Goodwin, 2016). Beyond its relevance for how people perceive individuals, however, morality also plays an important role in how people perceive and interact with social groups.

1 Aquino and Reed’s (e.g., 2002) program of research on moral identity is related but distinct from this line of inquiry. They focus on traits widely agreed upon to be morally desirable (e.g., honesty, generosity), and find that the identity centrality (or “self-importance”) of these traits predicts desirable behaviors (e.g., volunteer behavior, charitable giving). I focus on attitudes with no clear moral consensus (e.g., attitudes toward gun control and the death penalty) and whether variance in moral conviction predicts variance in identity centrality.

6 First, the moral attitudes that prevail within a group govern intragroup dynamics by defining what behavior is appropriate among members of that group (Ellemers & van den Bos,

2012). The idea that people conform to group norms is not new (e.g., Asch, 1956) but moral norms are especially influential. Ellemers, Pagliaro, Barreto, and Leach (2008) found that norms elicited more compliance when compliance was framed as indicative of morality than when compliance was framed as indicative of intelligence. In a related study, Van Nunspeet, Derks,

Ellemers, and Nieuwenhuis (2015) found that individuals invest more effort in appearing moral before an ingroup audience than before an outgroup audience. Ellemers and colleagues (e.g.,

Ellemers & van den Bos, 2012) argue that group norms create moral values from “guidelines that other groups treat as mere social conventions or even consider immoral” (p. 882). These shared moral values provide a sense of shared reality, which in turn is an essential foundation of group members’ shared identity (see also Haidt, 2013).

Second, moral attitudes shape intergroup relations. By defining who is part of a group, shared moral attitudes also define who is not. Rokeach and Mezei (1966), for example, report behavioral evidence that people are even more likely to interact with others who express similar values to them than they are to interact with others of the same race. Skitka et al. (2012) find evidence for similar preferences for social distance using self-report questionnaires and their moral conviction scale. Moral disagreement can inspire conflict as well as avoidance. Skitka,

Baumann, and Sargis (2005) find that among those who disagree, political discussions are more defensive and less productive when they focus on participants’ moral convictions than when they focus on non-moral attitudes. Moreover, studies show that people are willing to overtly discriminate against groups (and individuals within them) whose presumed ideological

7 worldview differs from their own (Brandt, Reyna, Chambers, Crawford, & Wetherell, 2014). The moral attitudes that distinguish groups from one another can exacerbate divisions between them.

This evidence suggests that moral attitudes can define strong social categories; moral attitudes can help bind groups together and shore up the boundaries between them. But existing research has not yet traced these social categories to individuals’ sense of moral conviction about specific issues. Do people use their moral convictions to place others into social categories? My second hypothesis is that individuals’ moral convictions define subjectively meaningful social categories—to the extent that those convictions are identity-central. For example, people who regard gun control as a moral issue should be especially likely to perceive and encode others as opponents or proponents of gun control, and they should manifest this tendency to the extent that their gun control attitude is central to their identity.

Study Overview

In sum, existing research on moral attitudes is consistent with the idea that moral convictions are social identities, but current empirical support for this idea is only indirect. To test this idea more directly, I decompose this broad thesis into two more concrete hypotheses:

Hypothesis 1: The more that individuals perceive their attitude to be a moral conviction, the more identity-central that attitude will be.

Hypothesis 2: Moral convictions define social categories to the extent that they are identity- central.

I test these hypotheses in two studies. In Study 1, participants reported their attitudes toward 10 political issues, then indicated the identity centrality and moral conviction that they ascribed to each attitude. Participants then completed a modified “Me/Not Me” task designed to measure the cognitive association between their attitudes and their self-concept. If Hypothesis 1

8 is correct, then within-subject variance in moral conviction should predict both within-subject variance in attitudes’ identity centrality and within-subject variance in me/not me reaction time.

Study 2 combines the measures from Study 1 with a “Who said what?” task designed to assess participants’ perceptions of social categories—including those that participants might infer from others’ attitudes. If Hypothesis 2 is correct, then participants’ performance on this task should follow a specific pattern, indicating that they used their moral convictions to categorize others to the extent that those convictions were identity-central. If moral convictions are relatively central to identity, and if the identity centrality of moral convictions predicts their use as social categories, then it would seem fair to conclude that moral convictions are social identities—a part of how people define themselves that shapes the groups they perceive in the social world.

Study 1

To test my hypothesis that moral convictions comprise a relatively central element of individuals’ self-concept, Study 1 measured moral conviction across several attitudes and included three complimentary operationalizations of attitudes’ identity centrality—a self-report measure, a reaction time task, and a latent variable estimate derived from these two measures.

Method

Participants

Participants in Study 1 were students at a large midwestern university, recruited from the subject pools of the Psychology Department and of the School of Journalism and Mass

Communication. Participants were compensated with extra course credit. The resulting sample

(N = 288) was relatively young (M = 20 years, SD = 2.86), mostly comprised of women (202 women, 84 men, 1 “other,” and 1 missing), and disproportionately likely to identify as White or

9 Asian (62% White, 29% Asian/Asian American, 6% Black/African American, 4% Latino; 6 participants selected more than one category).2

Measures and Materials

Verbatim wording for all Study 1 measures and materials is provided in online

Appendices A and B. All continuous variables were scaled to range from 0 to 1 unless otherwise noted.

Predictors. I assessed participants’ attitudes toward 10 political issues: government spending, legal abortion, defense spending, gay rights, gun control, affirmative action, free market economics, protecting the environment, tightening border security, and the death penalty.

Participants used a 9-point bipolar scale from “strongly oppose” to “strongly favor” to respond to the question “Generally speaking, would you say you support or oppose [this political issue]?”

Participants’ moral convictions were also assessed via self-report. Participants reported the extent to which they considered their position on each political issue to be a matter of right and wrong using Skitka and colleagues’ (see Skitka & Morgan, 2014) 4-item questionnaire.

Attitude importance is related to but distinct from moral conviction. Although moral convictions are nearly always important, some attitudes that individuals regard as personally important are not moral convictions (e.g., music preferences). I include attitude importance in analyses below to distinguish the effects of moral conviction per se from attitude strength more generally. I measured importance using three items. Participants indicated how “personally

2 I also assessed party identification and citizenship. 60% of respondents were Democrats or leaned left, 26% were Republicans or leaned right, and 14% were non-leaning independents. Sixty participants were not U.S. citizens. I included these participants in analyses in both studies. Non-citizens may know as much about American politics as citizens do, and regardless, lack of knowledge is no obstacle to political opinion formation. More important, I had no a priori expectation that national origin would moderate hypothesized effects.

10 important,” each issue was to them, how much they “care[d] about” that issue, and how deeply it

“concern[ed]” them (Boninger, Krosnick, Berent, & Fabrigar, 1995).

Because I measured the importance and moral conviction that participants ascribed to each of the 10 issue attitudes, I assessed the reliability of these scales by computing Cronbach’s

α for each scale for each issue (e.g., moral conviction regarding government spending, 4 items; moral conviction regarding gun control, 4 items; importance of government spending attitude, 3 items). The average α across the 10 issues was 0.95 for the moral conviction items and 0.93 for the importance items. I used the same procedure to estimate the average α for identity centrality across political issues.

Outcomes. Attitudes’ centrality to the self-concept were measured in two ways. First, participants simply responded to three 7-point, bipolar items indicating by self-report the extent to which their position on that issue was central to how they view themselves. These items have been used in previous research to assess the centrality of various social identities (e.g., Luhtanen

& Crocker, 1992; average α for political issues: 0.83).

Second, participants completed a modified version of Markus’s (1977) me/not-me task.

The original task presents participants with a number of traits and asks them to indicate as quickly as possible whether those traits are or are not characteristic of themselves. The quicker the reaction time, the more central that trait is presumed to be to the subject’s self-concept. I included three types of stimuli for participants to identify as characteristic or not characteristic of themselves—issue positions, trait adjectives, and values. The speed with which issue positions were categorized served as the primary dependent variable. As the traditional target stimuli in the me/not me task, trait adjectives were included to verify that my modified version of the me/not me task was still a valid indicator of self-schema content. Including traits as stimuli also helped

11 to contextualize the task as a self-descriptive exercise rather than an opportunity to recite opinions. Finally, I included values as target stimuli to support exploratory analyses ancillary to the current paper.

All three sets of stimuli were presented, intermingled, in a computer-based task. Each trial included two category labels in the upper corners of the computer screen: “I am someone who…” and “I am not someone who…” Under these labels, brief phrases of the form “supports the death penalty” or “opposes same-sex marriage” appeared for participants to categorize as characteristic or not characteristic of themselves, one phrase at a time. The left/right position of the category labels and their corresponding response keys were randomly assigned across participants. Participants had the opportunity to identify themselves (or not) as “someone who supports” and as “someone who opposes” each issue from the self-report questionnaire. Stimuli were presented in random order, with no response deadline, in 5 blocks of 50 trials each. All stimuli were presented in each block. Stimuli are provided in online Appendix B.

Demographics and additional variables. Demographic variables included race, gender, income, education, age, and citizenship status. I also measured the self-descriptiveness and identity centrality of participants’ traits in order to assess the validity of the me/not me task, given my modification to the original paradigm. Participants in Study 1 also answered questions about their party identification, left/right ideology, and the identity centrality, moral conviction, importance, and degree of endorsement that they ascribed to values. These variables were included for exploratory purposes only, but are nonetheless included in online Appendix A.

Procedure

Study 1 consisted of two parts to separate measures of moral conviction from outcome measures. The first was an online survey that included the self-report items assessing attitudes

12 and their associated moral conviction. Participants also rated themselves on each trait adjective.

The second part of the study took place in the lab no sooner than two nights after the initial survey. In the in-person session, participants indicated the personal importance and identity centrality of their attitudes, as well as the identity centrality of each trait adjective. Following this questionnaire, participants completed the modified me/not-me task described above and, finally, the demographic questions.

Results & Discussion

The purpose of Study 1 was to test the hypothesis that moral convictions are a relatively central element of the self-concept, even when controlling for attitude importance.3 I tested this hypothesis with three sets of analyses.

Self-reported Identity Centrality

First, I estimated a multi-level linear mixed model, in which moral conviction and attitude importance were entered as predictors of the self-reported identity centrality of participants’ issue attitudes. Attitudes were the level-1 units and participants were the level-2

(grouping) unit. I treated the effects of moral conviction and attitude importance as fixed but allowed the intercept to vary randomly across participants. See Table 1 for estimates. Because moral conviction, attitude importance, and identity centrality are all scaled to range from 0 to 1, the coefficients for moral conviction and importance can be interpreted as the proportion change in self-reported identity centrality as each predictor increases from its minimum to its maximum.

3 Moral conviction and attitude importance were empirically distinguishable. The bivariate correlation (which does not take the multi-level structure of the data into account) was r = 0.56 (p < .001). Online Appendix E also illustrates the correlations between moral conviction and importance across stimuli within each individual subject. These within-subject correlations were nearly uniformly positive, but they were not so strong (for most participants) as to imply a 1:1 correspondence between the two.

13 Consistent with my prediction, moral conviction did predict self-reported identity centrality (b =

0.217, p < .001), even controlling for the effect of importance (which, admittedly, was larger; b =

0.386, p < .001). Figure 1 illustrates this clear positive relation between moral conviction and self-reported identity centrality.

14 Table 1. Self-reported identity centrality as a function of moral conviction Study 1 Model b 95% CI p Fixed Effects Moral Conviction 0.217*** [0.190, 0.244] <0.001 Importance 0.386*** [0.358, 0.415] <0.001 Intercept 0.172*** [0.150, 0.193] <0.001 Random Effects σ2 (Intercept) 0.013 [0.011, 0.026 ] -2 x log-likelihood -2 X 962.56 Wald χ2 (degrees freedom) 1868.43 (2)*** N Level-1 units 2856 N Level-2 units 288 Note. Entries are coefficients from a multilevel linear mixed model. (†p < 0.10. *p < 0.05. **p < 0.01. ***p < 0.001.)

Figure 1. Self-reported identity centrality as a function of moral conviction.

Note. Figure is based on the fixed portion of the model presented in Table 1. Error bands indicate 95% confidence intervals.

15 Cognitive associations as measured by the me/not me task

Next, I tested the effect of moral conviction (net attitude importance) on me/not me reaction times.4 Before analyzing the reaction time data, I excluded reaction times that were either incredibly fast (< 300 ms) or unusually slow ( > 6.5 s),5 then performed a log (base 10) transformation on the remaining data to mitigate its strong positive skew, following the recommendations of existing literature (Bargh & Chartrand, 2000; Wentura & Degner, 2010).

All analyses presented below use this transformed variable. I also repeated these analyses using an alternative (reciprocal-square-root) transformation. Because these results were substantively identical, I do not report them here.

I then estimated a multi-level linear mixed model to test the effect of stimulus-level moral conviction on log-transformed me/not me reaction times. Me/not me responses were the level-1 units. Participants were the level-2 (grouping) units. The model included an intercept that varied randomly across participants and fixed effects for moral conviction and attitude importance. I also included a fixed linear effect for the block in which participants saw the stimulus (1 through

5), the number of characters in the stimulus, a binary indicator for whether the stimulus included the word “opposes” (rather than “supports”), and a binary indicator for whether the stimulus was consistent versus inconsistent with the participant’s position on the target stimulus.6 Moral conviction and importance were scaled to range from 0 to 1. The number of characters and the

4 Given the modifications from the original, I tested the validity of my me/not me task by replicating Markus’s (1977) findings. These analyses provided evidence that my modified task is a valid measure of identity centrality; see online Appendix F for details. 5 The 6.5 s upper bound was chosen because this latency was 3 interquartile ranges above the 75th percentile for reaction times to issue stimuli (see Wentura & Degner, 2010). 6 I defined stimulus’ consistency with the participant’s attitude in seven different ways, all yielding substantively identical results. See online Appendix C for details on each definition of consistency.

16 block in which the stimulus occurred were kept in their “natural” units, with number of characters ranging from 7 to 35 and block number ranging from 1 to 5.

Consistent with predictions, I found that attitude-level moral conviction predicted swifter me/not me responses (b = -0.068, p < .001), even controlling for attitude importance. See Table 2 for estimates.

When interpreting these estimates, recall that the outcome variable is the base 10 logarithm of reaction time in milliseconds. To illustrate the substantive effects of moral conviction on reaction time, I computed the average participant’s predicted reaction time in milliseconds by using the fixed portion of the model estimated in Table 2 to compute fitted values, ŷ, for each respondent. Each respondent’s predicted reaction time in milliseconds is 10ŷ.

Figure 2 depicts these predicted reaction times as a function of moral conviction, assuming an issue stimulus of moderate attitude importance (0.5 of 1) that is consistent with the participant’s attitude, includes the word “supports,” is of average length (26 characters), and was presented in the third block of the task. As can be seen from the figure, the effect of moral conviction is small—on the order of tenths of a second—but quite statistically reliable.

17 Table 2. Me/Not Me reaction time as a function of moral conviction Study 1 b 95% CI p Fixed Effects Moral Conviction -0.068*** [-0.078,-0.057] <0.001 Importance -0.091*** [-0.102,-0.079] <0.001 Number of Characters 0.003*** [0.003,0.003] <0.001 Consistency w Attitude -0.014*** [-0.019,-0.009] <0.001 Stimulus Contains "Oppose" 0.090*** [0.085,0.095] <0.001 Block Number -0.031*** [-0.033,-0.030] <0.001 Intercept 3.294*** [3.274,3.314] <0.001 Random Effects σ2 (Intercept) 0.011 [0.010, 0.013] -2 x log-likelihood -2 X 6624.16 Wald χ2 (degrees freedom) 4351.75 (6)*** N Level-1 units 27,774 N Level-2 units 287 Note. Entries are coefficients from multilevel linear mixed models. (†p < 0.10. *p < 0.05. **p < 0.01. ***p < 0.001.)

Figure 2. Me/Not Me reaction time as a function of moral conviction.

Note. Figure is based on the fixed portion of the model presented in Table 2. Error bands indicate 95% confidence intervals.

18 Identity centrality as a latent variable

Finally, I examined whether moral conviction predicted identity centrality as a latent variable indicated by self-reported identity centrality and by me/not me reaction times. Moral conviction significantly predicted each of these outcomes in the predicted direction; however, each finding could be attributable to theoretically uninteresting variables confounded with moral conviction scores (e.g., acquiescence bias, attitude accessibility).

To rule out these alternative explanations for the results described above, I estimated a multi-level structural equation model in which I entered latent variables for within-participant moral conviction and attitude importance as predictors of a within-participant latent identity centrality variable. This latent variable approach also helps correct for measurement error. The between-participant component of the model included 4 latent factors (for moral conviction, importance, reaction time, and self-reported identity centrality), to account for average differences in response tendencies across participants. I excluded from this model any me/not me responses that were inconsistent with respondents’ attitudes, which might have been errors. The latent within-participant identity centrality variable included the (up to) 10 reaction times observed for each issue and the three self-report identity centrality items. My primary goal was to test whether, in the within-participants portion of the model, the latent variable for moral conviction predicted the latent identity centrality variable, controlling for attitude importance.

This test indicates whether within-subject variance in moral conviction predicts the shared variance among the self-reported identity centrality and reaction time items—which, I argue, offers an index of identity centrality.

Figure 3 depicts the results of this structural equation model, including the standardized factor loadings for each latent variable and the standardized path estimates. Apart from a highly

19 significant χ2 statistic (unsurprising given the many observations), model fit was adequate (CFI =

0.904; RMSEA = 0.051; SRMR = 0.089 for Within, 0.080 for Between; χ2(331) = 2764, p <

0.001). As predicted, within-participant moral conviction significantly predicted within- participant identity centrality (β = 0.27, p < 0.001).

In sum, I find converging evidence for my prediction that moral convictions are tied to the self-concept. The moral conviction that participants ascribed to any given attitude was significantly related to how central they reported that attitude to be to their self-image, to how quickly they identified that attitude as characteristic of themselves, and to the variance shared across these two very different indices of identity centrality.

Figure 3. Figure depicts a multi-level structural equation model, with latent variables for moral conviction, attitude importance, and identity centrality (with this last variable indicated by me/not me reaction times and self-reported identity centrality). Path estimates are standardized. CFI = 0.908; RMSEA = 0.050; χ2(331)= 2522, p < 0.001. †p < 0.10. *p < 0.05. **p < 0.01. ***p < 0.001.

Study 2

Results from Study 1 were consistent with my first hypothesis: Moral convictions comprise a part of individuals’ identities. The purpose of Study 2 was to replicate Study 1 and to test my second hypothesis: Perceivers’ moral convictions will guide social categorization of others to the extent that moral convictions are identity-central.

In this study, I combined the procedures from Study 1 with the “Who said what?”

(WSW) task—a recall task that measures the extent to which participants tend to confuse members of a given social category with one another (e.g., by misattributing one White man’s statement to another White man). The task uses participants’ recall errors to quantify this confusion, providing an index of social categorization that varies both across and within participants.

Hypothesis 2 predicts that moral convictions form the basis of social categories to the extent that they are uniquely identity central. For the purposes of Study 2, this broad hypothesis can be decomposed into three more specific, operational predictions to be tested using the categorization effect size from WSW task:

Hypothesis 2a. Targets’ moral convictions are encoded as social categories. That is, the average WSW categorization effect size for moral convictions should be positive and significantly different from 0.

Hypothesis 2b. Moral convictions are encoded more strongly as social categories than are non-moral attitudes. Within-subject moral conviction should predict within-subject social categorization.

22 Hypothesis 2c. The tendency to encode moral convictions as a social category is attributable to the tendency to identify with those beliefs. That is, within-subject variance in moral conviction should indirectly predict the WSW categorization effect size via self-reported identity centrality and me/not me reaction time.7

Method

Participants

Participants in Study 2 were recruited from the Psychology Department’s and the business school’s subject pools at a large midwestern university. Psychology participants were given the option of participating for either course credit or $10, whereas all business-school participants were paid. The vast majority of participants were students, but non-students occasionally participate in the business school pool.

The initially recruited sample was screened to facilitate the construction of the WSW task. To test Hypothesis 2, targets in the WSW task varied in their position on two issues, with one issue implicating a moral conviction for any given participant and the other implicating a non-moral attitude. Participants were therefore only eligible for full participation in Study 2 if, in the initial online survey (see Procedure), they reported high moral conviction for one of the following issues and low moral conviction for another: the death penalty, gun control, and free market economics. See online Appendix D for why these issues were chosen. Of the 991

7 I predict these indirect effects as a pattern of shared variance rather than of true mediation. My prediction is not a causal chain from moral conviction, to identification, to social categorization. Instead, I contend that moral convictions are social identities. If an identity is social, its identity centrality and its role in social categorization should be related. People can categorize fruits as apples and oranges without identifying as any kind of fruit. A person can identify as creative without categorizing everyone she meets as an artist or a philistine. But if perceivers use their moral convictions to categorize others to the extent that their convictions are identity central, then moral convictions are not just different kinds of fruit or a personal trait, but a sign of “us” or “them.”

23 participants who completed the screening questionnaire, 540 (54.5%) were eligible. Of those eligible participants, 407 attended their scheduled lab session.

As in Study 1, analyzed participants were relatively young (M = 20 years, SD = 3.35), mostly women (n = 296 women, n = 103 men, 4 other, 4 declined to answer), and most likely to identify as White or Asian (65% White, 23% Asian/Asian American, 2% Black/African

American, 2% Latino, < 1% Native American, and 6% identifying with multiple racial groups).8

All 407 analyzed participants were used to replicate findings from Study 1. Only a subset of these (n = 296) could be used to test Hypothesis 2, due to the experimental manipulation in the

WSW task (see below). The subset of respondents used to test Hypothesis 2 did not significantly differ from the full analyzed sample in their age, race, sex, political orientation, or average degree of moral conviction.

Measures and Materials

Measures from Study 1. Participants’ moral convictions (average α = 0.92), issue positions, and the importance (average α = 0.88) and identity centrality (average α = 0.81) of these attitudes were assessed in the same manner as in Study 1. So too were trait adjectives and their identity centrality (average α = 0.75), and demographics. Values and their characteristics were not assessed, though political orientation items were retained for exploratory purposes.

Participants also completed the same me/not me task as in Study 1, except that value stimuli were omitted from the task.

“Who said what?” task. The statements included in the WSW task were experimentally manipulated to test hypotheses beyond the scope of the current paper. In the baseline condition,

8 71% of analyzed participants were Democrats or leaned left, 18% were Republicans or leaned right, and 10% were non-leaner independents. Sixty-three participants were non-citizens.

24 targets varied only in terms of their race and in the actual statements that they made. No information concerning their attitudes was included. These participants (n = 111) are therefore omitted from tests of Hypothesis 2, but were included in analyses replicating Study 1. In the attitude condition, targets’ race and their supposed attitudes on two issues were presented. One of these issues represented a moral conviction of the participant’s (i.e., its moral conviction score was above the midpoint; the moral issue). The other issue did not put at stake a moral conviction

(i.e., its moral conviction score was below the midpoint; the non-moral issue).

During the task’s initial presentation phase, participants were presented with a series of digital “slides.” On each slide, a picture of an individual (a “target”) was paired with a brief statement supposedly spoken by that individual. Each target appeared three times, each time paired with a unique statement. Statements either had nothing to do with politics or gave the target’s stance on the moral issue or on the non-moral issue. By the end of this phase, participants had seen all 8 targets take a stance on the moral issue and on the non-moral issue.

After the presentation phase, participants performed a brief distractor task. A surprise recall phase immediately followed. Participants were provided with photos of all of the target individuals and asked to indicate which of those people said each of the 24 statements, one statement at a time. Additional details about the WSW task, including all target images and statements, are provided in online Appendix D.

Respondents are inferred to have categorized targets if they are more likely to misattribute statements from one category member to another member of the same category. For example, a participant may misattribute statements made by African-American targets to another

African-American target—a within-category error. A participant might, though, misattribute that statement to a European-American target—a between-category error. If participants strongly

25 encode targets’ social category membership, within-category errors will be more common than between-category errors. The same logic holds for other categories: opponents vs. proponents of gun control, opponents vs. proponents of the death penalty, etc. I predicted that when perceivers held a moral conviction about a political issue, they would encode WSW targets’ positions on that issue and categorize them accordingly.

The WSW task permits estimates of social categorization both within participants and across participants. Within participants, I computed the difference between the within- and between-category error rates for each potential social category (i.e., race, moral issue position, non-moral issue position), following the procedure outlined by Kurzban et al. (2001). See online

Appendix D. The average of this difference score across participants yielded an aggregate effect size for each social category (Kinzler, Shutts, DeJesus, & Spelke, 2009; Kurzban, Tooby, &

Cosmides, 2001; Petersen, 2012; Pietraszewski, Curry, Petersen, Cosmides, & Tooby, 2015).

Procedure

Like Study 1, Study 2 included two parts. The first, which was again administered online, included measures of attitudes, trait adjectives, moral conviction, and attitude importance. The second part, which was administered in the lab, included the questions assessing identity centrality, then the me/not me task and WSW task in random order, then a final demographic survey.

Individuals who participated in the second session of the study before November 15,

2017 were randomly assigned to either the baseline or the attitude condition of the WSW task.

All participants after this date were assigned to the attitude condition, given that Hypotheses 2a,

2b, and 2c could only be tested among attitude-condition participants.

26 Unlike Study 1, Study 2 assessed attitude importance in the first (online) session.

Compared to Study 1, Study 2 likely yields a more conservative estimate of the association between importance and identity centrality and a more liberal estimate of the association between importance and moral conviction. Both strategies yield conservative tests of the unique association between moral conviction and identity centrality.

Results & Discussion

Replicating Study 1

Because Study 2 included the same key measures as Study 1, I was able to replicate the results of the first study in this second sample. Using the exact same models described in Study

1, I tested moral conviction as a predictor of self-reported identity centrality, me/not me reaction time, and the latent variable derived from these two identity-centrality measures. All results were substantively identical to those in Study 1. These results are provided in online Appendix G.

Testing Hypothesis 2

Study 1 and Study 2 provide strong evidence that moral convictions are central to individuals’ identities. I next tested whether moral convictions are social identities—that is, whether individuals’ moral convictions guide both how they think about themselves and how they perceive and relate to others.

Are moral convictions encoded as social categories? (Hypothesis 2a) Recall that the strength of social categories in the WSW task is quantified by the frequency of within-category errors relative to between-category errors for a given dimension of categorization (i.e., targets’ race, position on the moral issue, and position on the non-moral issue). The procedure for

27 estimating this quantity is described in Appendix D; the final result is a continuous effect-size measure for each categorization dimension.

With respect to targets’ stances on the moral issue, participants tended to make significantly more within-category errors (M = 13.03, SD = 2.54) than between-category errors

(Mcorrected = 8.22, SDcorrected = 1.91; Muncorrected = 10.96, SDuncorrected = 2.54, t(295) = 18.63, p <

0.001). The aggregate measure of this effect size, r, ranges from 0 to 1 and can be interpreted much the same as a correlation coefficient. For the moral issue, r was 0.61—a large effect.

Participants used their moral convictions to categorize targets.

Are moral convictions encoded more strongly than non-moral attitudes?

(Hypothesis 2b) I estimated the same categorization effect size r for race and for the non-moral issue. Contrary to my predictions, r for the non-moral issue (r = 0.55) was quite similar to that for the moral-issue. For comparison, the categorization effect size for race was the same as that for the non-moral issue; r = 0.55.

However, this aggregate effect size does not account for within-participant variance in moral conviction. It also dichotomizes moral conviction, despite the availability of a continuous measure. I therefore estimated a multi-level linear mixed model that predicts within-participant variance in social categorization from within-participant variance in moral conviction. The model is similar to those reported in Table 1, except that those included variance in moral conviction across ten issues; only two issues (the moral and the non-moral issue) can be modeled here because the WSW task only included those two issues. Again, I allow for a random intercept across participants but treat the effects of moral conviction and importance as fixed.

These results are shown on the left side of Table 3. Contrary to my predictions, the moral

28 conviction that participants ascribed to an issue did not predict the extent to which they categorized targets by their position on that issue (b = 0.362, p > 0.25).

Given the strong relation observed between moral conviction and identity centrality in both Study 1 and Study 2, these null effects raise an important question: Does identity centrality predict social categorization in the WSW task? If the two are totally unrelated, then the identities defined by individuals’ attitudes may not be relevant in social categorization. I therefore estimated the same multi-level linear model as above, adding self-reported identity centrality as a predictor of social categorization. These results are shown on the right side of Table 3. The coefficient for identity centrality was positive and significant (b = 2.426, p = 0.010), suggesting

(1) that the strength of social categories in the WSW task is related to participants’ identification with those categories and (2) that moral conviction may exert an indirect effect on social categorization via its effect on identity centrality.

Table 3. Social Categorization by Issue as a Function of Moral Conviction

Model 1 Model 2

b 95% CI p b 95% CI p

Fixed Effects Moral Conviction 0.362 [-0.950, 1.674] 0.589 -0.269 [-1.652, 1.115] 0.703 Importance 0.378 [-1.127, 1.884] 0.622 -0.122 [-1.666, 1.422] 0.877 Identity Centrality - - - 2.426** [0.581, 4.272] 0.010 Intercept 4.231*** [3.430, 5.031] <0.001 3.567*** [2.607,4.527] <0.001 Random Effects σ2 (Intercept) 4.896 [2.951, 8.123] 4.692 [2.786, 7.902] -2 x log-likelihood -2 X -1735.72 -2 X -1729.33 Wald χ2 (degrees freedom) 1.10 (2) 17.57 (3)† N Level-1 units 592 591 N Level-2 units 296 296 Note. Entries are coefficients from a multilevel linear mixed model. (†p < 0.10. *p < 0.05. **p < 0.01. ***p < 0.001.)

Does identification explain categorization by moral conviction? (Hypothesis 2c)

Despite the absence of a direct effect of moral conviction on social categorization, moral convictions—compared to non-moral attitudes—might still indirectly define stronger social categories by virtue of their greater identity centrality. I employ a structural equation model to test for the presence of this indirect effect. Note that although the resulting diagram implies a mediation model, I am less interested in causal order than in patterns of shared variance.

The structural equation models predicting identity centrality draw on participants’ attitudes toward 10 political issues. Because the WSW task included only 2 of these 10 issues, the models I use to examine variance in social categorization have far less within-participant variation upon which to draw. This poses problems for the estimation of complicated measurement models like those I employed in Study 1. I therefore refrain from estimating latent

30 variables in this model and instead estimate path models using attitude-level scale means (for moral conviction, importance, identity centrality, and reaction time) as observed variables.9 I continue to estimate two-level models with random intercepts across participant for each observed variable.

The multi-level structural equation model depicted in Figure 4 fit the data quite well (CFI

= 1.00; RMSEA = 0.000; SRMR = 0.024 for Within, 0.253 for Between; χ2(11)= 10.67, p =

0.47). Results also revealed the predicted indirect effect of moral conviction on social categorization (β = 0.05, p = 0.024). This indirect effect was channeled almost entirely through self-reported identity centrality.

Summary: Are moral convictions social identities?

I find clear evidence that moral convictions defined meaningful social categories in the

WSW task, but so too did non-moral attitudes. That said, multi-level models suggest that the relative identity centrality of moral convictions does matter; to the extent that moral convictions are more identity-central than non-moral attitudes, they also tended to reflect a uniquely strong category. The identities that moral convictions comprise appear to have a social component.

9 As in the other SEMs, only reaction times that were consistent with the participant’s overall attitude were used, in order to avoid analyzing responses made in error. These reaction times were averaged to form an acceptably reliable scale (α = 0.81).

Figure 4. Figure depicts a multi-level structural equation model that tests the direct and indirect effects of attitude-level moral conviction on participants’ tendency to use that attitude to perceive categories in the WSW task. Path estimates are standardized. 2 CFI = 1.00; RMSEA = 0.000; χ (11)= 10.67, p = 0.47. (†p < 0.10. *p < 0.05. **p < 0.01. ***p < 0.001.)

General Discussion

Moral Convictions as Social Identities

In two studies, I find strong support for the idea that the political issue positions that individuals construe in moral terms are also those most central to their identities. Moral conviction predicted self-reported identity centrality, reaction time in a task designed to tap the association between attitudes and the self-concept, and the shared variance between these measures. I also find support for the idea that moral convictions play a role in how individuals parse their social world, given participants’ strong tendency to categorize WSW targets according to their positions on the moral issue. In fact, my results indicate that targets’ positions on the moral issue were encoded as strongly (if not more so) than their race.

These results are complicated by the fact that non-moral attitudes also mattered in the

WSW task—to a degree comparable, on average, to moral convictions. This was inconsistent with my predictions, but does not imply that moral convictions are not social identities. Rather, it raises the question of whether the identities defined by moral convictions are any more powerful than those defined by other attitudes. I find some evidence that they are; moral convictions did define stronger social categories to the extent that they were also more identity central than non- moral attitudes. However, this indirect effect was small in magnitude.

The safest conclusion is that moral convictions are more likely to be central to the self- concept than non-moral attitudes are, and that people are especially likely to perceive others in terms of their positions on these identity-central issues. That said, people do not need to ground their self-concept in an attitude domain to use others’ attitudes in social perception; like height or hair style, a target’s attitude can provide a useful categorical mnemonic device for distinguishing among individuals. The indirect effect of moral conviction on social categorization via identity

33 centrality does imply, however, that for at least some individuals some of the time, moral convictions are social identities that divide the world into “us” and “them.” If my position on the death penalty is a moral one, and is relatively important to who I am, then I notice others’ position on the death penalty and cognitively connect them to like-minded others. Identities that guide social categorization in this way are social identities.

Social identities are known to vary in their salience across time and situations. There may be some contexts (e.g., primary elections, political debates) in which moral convictions about political issues and the identity centrality of these positions play a stronger role in how individuals make sense of their social world. On balance, my results are more consistent with the idea that moral convictions are social identities than that they are not, but I cannot claim that these groups are more powerful or universally salient than other social categories like race or gender.

Limitations

Relying solely on college students may limit the generalizability of my findings. On the one hand, college-aged students’ political attitudes are less crystalized than older adults’

(Valentino & Sears, 1998). Because students’ political identities are still developing, they may ground their identities less firmly in their moral convictions about political issues than do older adults. On the other hand, college-educated individuals tend to be more individualistic than non- college-educated individuals (e.g., Stephens, Markus, & Townsend, 2007). Consequently, students in my sample might be more likely than the average person to construe their moralized attitudes as highly central to their identity. In short, the demographic characteristics of my sample may either inflate or deflate the effects I estimate. Although existing research on moral conviction has found reliable effects in non-student samples (see Skitka et al., 2015), future work

34 is necessary to test whether my specific predictions would hold outside of this relatively narrow group.

Another limitation of my study is that my analyses are strictly correlational. I have tried to eschew causal language and inferences, but my research leaves open the question of how moral convictions come to be in the first place. Social identities delineated by religion, race, and education might predict unique patterns of moral conviction—toward abortion, affirmative action, or same-sex marriage, to name a few plausible examples. Conversely, people may come to identify with groups whose only shared characteristic is a moralized attitude (e.g., single-issue political activist groups). Moral conviction does seem to promote the development of politicized identities (van Zomeren, Postmes, & Spears, 2012). Or, people may identify more strongly with the groups whose moral outlook matches their own. For example, Catholics’ degree of identification with their religion may vary as a function of their personal moral beliefs and convictions and those expressed by the Church during their lifetime. I leave these questions to future work. My argument is simply that moral convictions take on the properties of a social identity. They form an important component of how individuals see themselves and those around them.

Finally, a brief caveat. In making the claim that moral convictions are social identities, I do not mean to suggest that moral convictions are not personal or private identities. It is hard to imagine how a person could have a social identity that was not also part of how she construed herself as an individual. Race and gender, for example, are both social categories and important parts of the self-concept. My contention is that moral convictions define identities that are social as well as personal.

Implications and Future Directions

The primary goal of the current research was to clarify what it means when people call their attitudes “moral.” Results suggest that these moralized attitudes are those that people use to define themselves as individuals and to draw boundaries around like-minded groups of others.

This characterization of moralized attitudes can help explain political conflict and may suggest ways to mitigate it.

For example, Hetherington and Weiler (2009) argue that increased polarization in

American politics is partly because partisan conflict in this context tends to focus on values- based issues (rather than how to achieve ends that both parties desire). Moral convictions’ high identity centrality and power to define social categories might help account for this trend. The identity centrality of moral conviction may also help predict which issues divide political coalitions across time and context: those that citizens construe as matters of right and wrong and that offer the possibility of positive distinctiveness within a given environment. For example, the

“social issues” that drive contemporary Americans’ allegiance to the Republican or Democratic parties (Johnston, Lavine, & Federico, 2017) are both morally charged and the subject of disagreement across large segments of the electorate. Perhaps issues that are most easily moralized are also those that serve as the most effective foundations for political coalitions. If so, political practitioners may (knowingly or not) use citizens’ moral beliefs to broaden their coalition at the expense of their opponents’ or to shift the focus of political debate. When party elites use “wedge issues” to pry apart old coalitions and construct new ones, they are essentially making salient an identity that a subset of their opponents share. It is telling that historically, many such wedge issues (e.g., abortion, gay marriage) have centered on questions of morality

(Hillygus & Shields, 2008).

Finally, my results suggest some ways to attenuate political conflict—even conflict that surrounds moral convictions. For decades, social psychologists have studied means of mitigating intergroup conflict (Allport, 1954; Dovidio & Gaertner, 2010). If, in the end, moral convictions amount to social identities, then methods previously used to reduce prejudice and discrimination between racial and ethnic groups should help to reduce prejudice and discrimination among people who do not share one another’s moral convictions. Given the centrality of moral convictions to individuals’ self-concepts, common moral beliefs might have effects similar to those of common ingroup identities. By highlighting widely shared moral beliefs and values, careful interventions might reduce the acrimony that arises from the smaller differences in opinion that abound within larger moral communities.

All that said, more research is needed to determine the most effective ways to frame such interventions. For example, are moral convictions psychologically interchangeable with other social identities? If so, conflict-reducing interventions that emphasize any meaningful shared identity (e.g., partisanship, race, gender) could suffice to reduce the vitriol surrounding moral disagreement. Alternatively, moral convictions might define unique, domain-specific identities with effects that can only be overridden by morality-related information. As we learn more about the similarities and differences between the identities defined by attitudes, demographic characteristics, and moral convictions, we can better chart the ways in which people’s most cherished beliefs can be twisted to foment division or leveraged to build bridges across such divides.

References

Allport, G. W. (1954). The Nature of Prejudice. Cambridge, MA: Perseus Books.

Aquino, K., & Reed, A. (2002). The self-importance of moral identity. Journal of personality

and social psychology, 83, 1423-1440.

Asch, S. E. (1956). Studies of independence and conformity: A minority of one against a

unanimous majority. Psychological Monographs, 70.

Bargh, J. A. & Chartrand, T. L. (2000). The mind in the middle: A practical guide to priming and

automaticity research. In H. T. Reis & C. M. Judd (Eds.), Handbook of Research

Methods in Social and Personality Psychology, Cambridge University Press, 253 – 285.

New York, NY.

Boninger, D. S., Krosnick, J. A., Berent, M. K., & Fabrigar, L. R. (1995). The causes and

consequences of attitude importance. In Petty, R. & Krosnick, J. A. (Eds.) Attitude

strength: Antecedents and consequences, 159-189.

Brandt, M. J., Reyna, C., Chambers, J. R., Crawford, J. T., & Wetherell, G. (2014). The

ideological-conflict hypothesis: Intolerance among both liberals and conservatives.

Current Directions in Psychological Science, 23, 27-34.

Dovidio J. F. & Gaertner, S.L. (2010). Intergroup Bias. In Fiske, Susan T., Daniel T. Gilbert, and

Gardner Lindzey (Eds.), Handbook of social psychology. Vol. 2. Hoboken, NJ:Wiley.

Ellemers, N., & van den Bos, K. (2012). Morality in Groups: On the Social‐Regulatory

Functions of Right and Wrong. Social and Personality Psychology Compass, 6, 878-889.

Ellemers, N., Pagliaro, S., Barreto, M., & Leach, C. W. (2008). Is it better to be moral than

smart? The effects of morality and competence norms on the decision to work at group

status improvement. Journal of personality and social psychology, 95, 1397-1410.

Fiske, S. T., Cuddy, A. J., & Glick, P. (2007). Universal dimensions of social cognition: Warmth

and competence. Trends in cognitive sciences, 11, 77-83.

Haidt, J. (2013). The Righteous Mind: Why Good People Are Divided by Politics and Religion.

Vintage Books.

Hetherington, M. J., & Weiler, J. D. (2009). Authoritarianism and polarization in American

politics. Cambridge University Press.

Hillygus, D. Sunshine, and Todd G. Shields. 2008. The Persuadable Voter. Princeton, NJ:

Princeton University Press.

Johnston, C. D., Lavine, H. G., & Federico, C. M. (2017). Open versus closed: Personality,

identity, and the politics of redistribution. Cambridge University Press.

Katz, D. (1960). The functional approach to the study of attitudes. Public opinion quarterly, 24,

163-204.

Khan, U., & Dhar, R. (2006). Licensing effect in consumer choice. Journal of Marketing

Research, 43, 259-266.

Kinzler, K. D., Shutts, K., DeJesus, J., & Spelke, E. S. (2009). Accent trumps race in guiding

children's social preferences. Social cognition, 27, 623-634.

Kurzban, R., Tooby, J., & Cosmides, L. (2001). Can race be erased? Coalitional computation and

social categorization. Proceedings of the National Academy of Sciences, 98, 15387-

15392.

Landy, J. F., Piazza, J., & Goodwin, G. P. (2016). When it’s bad to be friendly and smart: The

desirability of sociability and competence depends on morality. Personality and Social

Psychology Bulletin, 42, 1272-1290.

Luhtanen, R., & Crocker, J. (1992). A collective self-esteem scale: Self-evaluation of one's social

identity. Personality and Social Psychology Bulletin, 18, 302-318.

Markus, H. (1977). Self-schemata and processing information about the self. Journal of

personality and social psychology, 35, 63-78.

Merritt, A. C., Effron, D. A., & Monin, B. (2010). Moral self‐licensing: When being good frees

us to be bad. Social and personality psychology compass, 4, 344-357.

Minson, J. A., & Monin, B. (2012). Do-gooder derogation disparaging morally motivated

minorities to defuse anticipated reproach. Social Psychological and Personality Science,

3(2), 200-207.

Monin, B., & Miller, D. T. (2001). Moral Credentials and the Expression of Prejudice. Journal of

Personality and Social Psychology, 81, 33-43.

Morgan, G. S., Skitka, L. J., & Wisneski, D. C. (2010). Moral and religious convictions and

intentions to vote in the 2008 presidential election. Analyses of Social Issues and Public

Policy, 10, 307-320.

Pagliaro, S., Ellemers, N., Barreto, M., & Di Cesare, C. (2016). Once dishonest, always

dishonest? The impact of perceived pervasiveness of moral evaluations of the self on

motivation to restore a moral reputation. Frontiers in psychology, 7, 586.

Petersen, M. B. (2012). Social welfare as small‐scale help: evolutionary psychology and the

deservingness heuristic. American Journal of Political Science, 56, 1-16.

Pietraszewski, D., Curry, O. S., Petersen, M. B., Cosmides, L., & Tooby, J. (2015). Constituents

of political cognition: Race, party politics, and the alliance detection system. Cognition,

140, 24-39.

Rokeach, M., & Mezei, L. (1966). Race and shared belief as factors in social choice. Science,

151, 167-172.

Skitka, L. J. (2010). The psychology of moral conviction. Social and Personality Psychology

Compass, 4, 267-281.

Skitka, L. J. (2014). The psychological foundations of moral conviction. In J. Wright & H.

Sarkissian (Eds.), Advances in Moral Psychology, Bloomsbury Academic Press, 148 –

166. New York, NY.

Skitka, L. J., & Bauman, C. W. (2008). Moral conviction and political engagement. Political

Psychology, 29, 29-54.

Skitka, L. J., Bauman, C. W., & Sargis, E. G. (2005). Moral conviction: Another contributor to

attitude strength or something more? Journal of Personality and Social Psychology, 88,

895 – 917.

Skitka, L. J., Liu, J., Yang, Y., Chen, H., Liu, L. & Xu, L. (2012). Exploring the cross-cultural

generalizability and scope of morally motivated intolerance. Social Psychological and

Personality and Science

Skitka, L. J. & Morgan, G. S. (2014). The social and political implications of moral conviction.

In H. Lavine (Ed.), Advances in Political Psychology, 35, 95 – 110.

Skitka, L. J. & Wisneski, D. C. (2011). Moral conviction and emotion. Emotion Review, 3, 328 -

330.

Stephens, N. M., Markus, H. R., & Townsend, S. S. (2007). Choice as an Act of Meaning: The

Case of Social Class. Journal of Personality and Social Psychology, 93, 814-830.

Thomsen, C. J., Borgida, E., & Lavine, H. (1995). The causes and consequences of personal

involvement. In Petty, R. & Krosnick, J. A. (Eds.) Attitude strength: Antecedents and

consequences, 191-214.

Todorov, A., Pakrashi, M., & Oosterhof, N. N. (2009). Evaluating faces on trustworthiness after

minimal time exposure. Social Cognition, 27, 813-833.

Wentura, D., & Degner, J. (2010). A practical guide to sequential priming and related tasks.

Handbook of implicit social cognition: Measurement, theory, and applications, 95-116.

Willis, J., & Todorov, A. (2006). First impressions making up your mind after a 100-ms

exposure to a face. Psychological science, 17, 592-598.

Valentino, N. A., & Sears, D. O. (1998). Event-driven political communication and the preadult

socialization of partisanship. Political Behavior, 20, 127-154.

Van Nunspeet, F., Derks, B., Ellemers, N., & Nieuwenhuis, S. (2015). Moral impression

management: Evaluation by an in-group member during a moral IAT affects perceptual

attention and conflict and response monitoring. Social Psychological and Personality

Science, 6, 183-192. van Zomeren, M., Postmes, T., & Spears, R. (2012). On conviction's collective consequences:

Integrating moral conviction with the social identity model of collective action. British

Journal of Social Psychology, 51, 52-71.