This is an open access article distributed under the terms of the Creative Commons
Attribution License (Although the adage “opposites attract” may characterize lay understandings of attraction, empirical evidence consistently finds that people tend to like similar others over dissimilar others (Byrne, 1971; Byrne, Griffitt, & Stefaniak, 1967; Dustin & Alfonsin, 1971; Skitka, Bauman, & Sargis, 2005). People especially like those who share their attitudes and values (Byrne, 1971; Chambers, Schlenker, & Collisson, 2013; Clore & Gormly, 1974; Henry & Reyna, 2007) and dislike those whose beliefs oppose their own (e.g., political attitudes; Brandt, Reyna, Chambers, Crawford, & Wetherell, 2014; Chambers et al., 2013; Crawford, 2014). Consequently, similarity plays a major role in how we evaluate and form impressions of other people (Smeaton, Byrne, & Murnen, 1989). The present studies examined whether initial person impressions change after discovering value dissimilarity with people.
Attitudinal dissimilarity is a particularly strong interpersonal repellant (Brandt et al., 2014; Byrne, 1969; Smeaton et al., 1989). One reason for this is that individuals derive personal worth from their attitudes, and consequently dislike when those attitudes are challenged or undermined (Henry & Reyna, 2007; Katz, 1960). Further, people care about maintaining their values and worldviews, and thus seek to associate themselves with others who support and validate these ideas (Brandt et al., 2014; Chambers & Melnyk, 2006; Motyl, Iyer, Oishi, Trawalter, & Nosek, 2014). Thus, avoiding dissimilar others serves the function of preventing deeply held attitudes from being contested.
Markers of Value Similarity and Dissimilarity
Before individuals form opinions of others, they must consider factors indicating that the other person is similar or dissimilar. Political ideology is a particularly salient marker of value dissimilarity (Motyl, 2014; Motyl et al., 2014), as attitudes and values are closely tied to an individual’s identity (Hitlin, 2003). Group memberships such as political affiliations comprise social identities (Simon & Klandermans, 2001; van Zomeren, Postmes, & Spears, 2008), which are integral to a person’s sense of self (Brewer, 1991). Further, individuals have a strong need to belong, and tend to gravitate toward environments containing ideologically similar others to satisfy this need (Baumeister & Leary, 1995; Motyl et al., 2014). For example, one study found that following President Obama’s victory in the 2012 presidential election, conservatives felt a reduced sense of belonging to the United States and consequently expressed increased intentions to migrate to Canada compared to liberals (Motyl, 2014). This suggests that individuals prefer surrounding themselves with others who share their values, and political ideology is a salient marker of value dissimilarity.
Additionally, some evidence suggests that political dissimilarity is more important to disliking than other forms of dissimilarity. Chambers, Schlenker, and Collisson (2013) found that racial prejudice was eliminated when accounting for political ideology. That is, liberals and conservatives disliked targets of differing political ideologies to roughly the same extent, regardless of whether the target was White or African American. In this case, political similarity was more important to disliking than was racial dissimilarity. Further, Koch, Imhoff, Dotsch, Unkelbach, and Alves (2016) demonstrated that agency/socioeconomic success and conservative-progressive beliefs are the two most important components people use when developing stereotype content, more so than warmth and competence. The researchers argue this is because political ideology informs people of others’ intentions, as well as the way they think, feel, and behave. This suggests that political information is among the most important aspects of interpersonal perception, and specifically demonstrates that it is more important than other social information (e.g., warmth, competence, power). Lastly, Brandt (2017) showed that perceived ideology was the best predictor of expressions of prejudice, more so than other perceptions of group membership. Thus, taken together, the existing evidence comparing political ideology to other forms of group membership suggest that it is perhaps the most important aspect of interpersonal liking and disliking.
Similarity is especially important in attraction and close relationships (Pilkington & Lydon, 1997). As individuals learn that a stranger is similar to them, their attraction to the stranger increases (Smeaton et al., 1989). Specifically, shared political attitudes have been shown to be particularly essential in intimate relationships, even more so than shared personality or physical characteristics (Alford, Hatemi, Hibbing, Martin, & Eaves, 2011). For example, Alford et al. (2011) found evidence that romantic partners have similar attitudes because of initial preferences in mate selection, not because homogeneity develops over time. This suggests that political ideology is an important consideration in impression formation, especially of those with whom one anticipates forging a close relationship. Additionally, people in established relationships tend to overestimate the extent to which their partner is similar to themselves, and couples who discover value dissimilarity with their partner are motivated to mitigate this threat by enhancing their positive evaluations of the relationship (Auger, Hurley, & Lydon, 2016). Given this potency of political ideology and dissimilarity, the present work examines how political information changes initial impressions of a potential romantic partner. Because we consider initial impressions (rather than established romantic partner evaluations as did Auger and colleagues, 2016), we expect that people will mitigate the threat of value dissimilarity by downgrading the potential romantic partner.
Additionally, people experience unpleasant affective and cognitive reactions upon encountering dissimilar or unfamiliar others (Cottrell & Neuberg, 2005; Neuberg, Kenrick, & Schaller, 2011; Stephan & Stephan, 1985). Stephan and Stephan (1985) found that people experienced intergroup anxiety when they stereotyped outgroup members and assumed they were dissimilar, both of which occur between political groups (Graham, Nosek, & Haidt, 2012). This suggests that people likely experience discomfort when interacting with someone from a different political group. Further, according to evolutionary theories, dissimilar and unfamiliar others pose a threat to one’s ingroup, and thus humans evolved unpleasant emotional responses that motive actions to attenuate these threats (Cottrell & Neuberg, 2005). In particular, different value systems are an indicator of a dissimilar outgroup, and this threatens the ingroup’s own value system. This has implications for factors important to the ingroup, such as group coordination. As a result, people evolved emotional responses to such value threats, including disgust, anger, and fear. These are uncomfortable affective states, and it is likely that such affective responses are cognitively represented by underlying psychological discomfort, such as feeling uneasy, uncomfortable, and bothered. Indeed, some qualitative research finds that people feel uncomfortable discussing politics, and at times refrain from discussing it in order to avoid potential conflict (Conover, Searing, & Crewe, 2002).
Changes in Initial Impressions and Value Dissimilarity
Some attitudes change more easily than others. In particular, unfavorable information tends to hold more weight in impression formation than does favorable information (Hamilton & Zanna, 1972). Additionally, unfavorable first impressions are more resistant to change than are favorable impressions; that is, after receiving information inconsistent with previous information, favorable first impressions are more susceptible to become unfavorable than vice versa (Briscoe, Woodyard, & Shaw, 1967). If deeply rooted attitudes such as political ideology are markers of dissimilarity that lead to disliking (Brandt et al., 2014; Motyl, 2014), it stands to reason that an initially favorable impression of a person is particularly susceptible to become more negative after dissimilarity with the target is discovered. Given the importance of self-consistency, if dissimilar values are considered unfavorable, individuals will likely quickly change their initial impressions of another person to be less favorable after discovering value dissimilarity.
Additionally, although some work suggests that conservatives express more antipathy than liberals (e.g., Jost, Glaser, Kruglanski, & Sulloway, 2003; Sibley & Duckitt, 2008), other work indicates that political prejudice is symmetrical across both sides of the political spectrum (Brandt et al., 2014). Sibley and Duckitt (2008) found in a meta-analysis that right-wing authoritarianism and social dominance orientation (two constructs that are positively correlated with political conservatism) are related to increased prejudice, suggesting that conservatives are more prejudiced than liberals. However, prejudice research historically (such as the research meta-analyzed by Sibley & Duckitt, 2008) has focused on traditionally left-wing groups. Considering this, Brandt et al. (2014) found that prejudice emerges equally among both liberals and conservatives when considering both right-wing and left-wing groups. They show that liberals are just as prejudiced against right-wing groups (e.g., Christian fundamentalists, the military, elderly people) as conservatives are against left-wing groups (e.g., atheists, African Americans, people on welfare). This suggests that liberals and conservatives do not fundamentally differ in expressing prejudice, but rather express the same amount of prejudice toward ideologically dissimilar groups. Thus, in the present studies we expected antipathy to be symmetrical across liberals and conservatives.
(Dis)similarity and Attractiveness
Further, implications of value dissimilarity extend beyond the realm of liking or disliking. Particularly, physical attractiveness is the largest determinant of how much a person likes, wants to get to know, and attempts to date a potential target (Walster, Aronson, Abrahams, & Rottman, 1966). Highlighting its importance, Nicholson, Coe, Emory, and Song (2016) examined impressions of an opposite-gender target person’s attractiveness, varying whether the person was an Obama or a Romney supporter. They found that individuals rated targets as less physically attractive when they were politically dissimilar compared to when they were politically similar or when their political affiliation was unspecified, indicating that not only do people hold negative attitudes toward those with dissimilar values, but they also devalue other factors important to relationship formation, such as attractiveness (Walster et al., 1966). However, participants in their study were not given any context for the evaluations, but were simply presented with a picture of an opposite-gender person and information about them. Given the aforementioned importance of political similarity in close relationships (Alford et al., 2011), it stands to reason that the relationship between value similarity and perceived physical attractiveness should be especially relevant in a mating context. Since we are using a dating paradigm, a context in which physical attractiveness is highly relevant, we chose to focus on changes in physical attractiveness as one of our key variables of interest. Additionally, rated attractiveness was the only variable of interest in Nicholson et al.’s (2016) study, and they did not consider attitude change upon discovering information about political similarity/dissimilarity. Although a between-subjects comparison yielded expected results, the question remains whether evaluations of a target change after discovering target political similarity/dissimilarity. Past work on impression formation in the face of extreme and negative information has also utilized within-subjects designs (Fiske, 1980; Mende-Siedlecki, Cai, & Todorov, 2013), and thus we took such an approach to examine changes in impressions in the face of political information.
Further, perceived attractiveness might differ from rated attractiveness. Just as attitudes can be the result of motivated reasoning, so can perceptions (Balcetis & Dunning, 2006; Dunning & Balcetis, 2013; Riccio, Cole, & Balcetis, 2013). Some work shows that motivated participants view ambiguous images in line with their expectations (Balcetis & Dunning, 2006). Using a visual task that morphed faces of targets into less and more attractive versions, Cole, Trope, and Balcetis (2016) found that people in relationships (who felt threatened by an attractive potential romantic partner) perceived attractive individuals as less attractive than did single participants. They argued that people in relationships, particularly those who were highly satisfied with their relationships, altered their perceptions of a potentially threatening person in order to mitigate the threat the person posed. Thus, people’s perceptions may be biased in ways that attenuate threats. Perceiving an ideologically dissimilar other as attractive might be a potential source of threat, as there would be a conflict between feeling physical attraction and feeling antipathy toward the person due to their ideology. As a result, it is possible that people would perceive an ideologically dissimilar person as less attractive in order to resolve the conflict between physical attraction and disliking.
The Present Studies
The present work had five primary objectives. First, we sought to examine how attitudes toward dissimilar others change. Past work in this area has explored isolated impressions using between-subject designs rather than tracking how people dynamically update their impressions when provided with value information (Nicholson et al., 2016). Second, we aimed to examine attitudes toward politically similar or dissimilar others in a specific context – a dating paradigm. Past work has not specified a context for impressions of physical attractiveness and value similarity/dissimilarity, and we draw upon the idea that attractiveness is highly relevant to mating. Third, we considered changes in both self-reported physical attractiveness as well as perceived physical attractiveness, assessed using a visual perception task. Much work highlights differences between people’s self-reported attitudes and their actual perceptions or more subtle attitudes (Balcetis & Dunning, 2006; Cole, Balcetis, & Dunning, 2013; Greenwald & Banaji, 1995). Thus, we wished to examine not only if people change how attractive they rate a potential target to be, but how attractive they visually perceive them to be. Fourth, we examined changes in both romantic interest and liking. Romantic interest taps into intentions to pursue a relationship with a person, in addition to physical attractiveness and liking. Finally, we considered how psychological comfort changed after discovering political information. This would highlight whether it is simply attitudes that change in the face of political information, or if emotional experiences change as well.
The present studies used an online dating paradigm to determine whether individuals’ impressions of a potential romantic partner become more negative after discovering that the person holds dissimilar political beliefs, and whether psychological discomfort partially explains this attitude change. Specifically, we hypothesized that after discovering that a potential romantic partner is politically dissimilar, liberals and conservatives will like the target less, become less romantically interested in the target, find the target less attractive, and become more psychologically uncomfortable than before they discovered political dissimilarity with the target. We expect to find the reverse pattern for politically similar targets. Importantly, we expect that these effects will be roughly symmetrical among liberals and conservatives, although we will explore any differences in their responses.
We attempt to replicate important effects on similarity and liking, as well as provide further support for the idea that the link between ideological dissimilarity and disliking is roughly symmetrical across the political right and left. The studies also attempt to expand on past work by considering changes in attitudes after receiving political information, implicit evaluations of physical attractiveness, and psychological discomfort, which could provide a more nuanced understanding of the well-established relationship between similarity and liking.
Study 1
Method
Participants
We recruited 303 U.S. residents via Amazon’s Mechanical Turk in the fall of 2014 who were paid 50 cents for their participation. Sixty-one participants incorrectly identified the political ideology of the target they viewed,ii leaving a sample of 242 participants (Mage = 33; 53% male; 72% White; 58.7% liberal, 22.7% moderate, 18.6% conservative; 93% heterosexual; 36% single). To determine the proportion of liberals, moderates, and conservatives in our sample, we considered anyone who answered 1 (very liberal) to 3 (somewhat liberal) on our ideology measure to be liberal, 4 (moderate/middle of the road) to be moderate, and 5 (somewhat conservative) to 7 (very conservative) to be conservative. The 20% attrition rate in our sample is within the typical range for MTurk studies (Harms & DeSimone, 2015; Oppenheimer, Meyvis, & Davidenko, 2009).
Materials and Procedure
Participants learned they would view a person’s online dating profile. They first indicated whether they were in a romantic relationship, and they were instructed not to consider their own relationship status when evaluating the person. They then reported their gender and romantic preferences. Based on their indicated romantic preferences, they were directed to a male or female dating profile, and were randomly assigned to a liberal (N = 82) or conservative (N = 79) target condition.ii Participants received information about targets in four phases. The target’s photo was presented during each phase. We selected pictures of one male face and one female face with slightly above average attractiveness ratings from HotorNot.com, a method used in previous research to obtain images normed for attractiveness (Baxter & Walker, 2008; Devcic, Karimi, Popenko, & Wong, 2010; Lee, Loewenstein, Ariely, Hong, & Young, 2008). Phase 1 disclosed basic demographic information about the target (targets were described identically other than the fact that male and female targets were described as being of different heights to reflect average sex differences in height), Phase 2 included hobbies and a personality description, Phase 3 stated personal and career goals as well as pastimes, and Phase 4 noted desired traits in a partner. Phases 1, 2, and 4 were identical across conditions. The information provided in Phase 3 varied between conditions: the liberal target indicated that they had volunteered for Obama’s 2012 Presidential campaign, whereas the conservative target indicated that they had volunteered for Romney’s campaign.
After each phase, participants rated the extent to which they would like to meet the target, go on a date with the target, and thought the target would be a good romantic partner (1 = Not at all; 7 = Very much). These items were averaged to form a measure of romantic interest (α = .96). They next indicated how attractive they found the target, which was embedded among 14 other traits presented in random order (e.g., intelligent, bright; 0 = Not at all; 10 = Extremely). Participants then indicated how similar they saw the target to themselves and how much they liked the target (1 = Not at all; 7 = Very much). In Phase 4 only, they reported the extent to which they felt uneasy, uncomfortable, and bothered (McGregor & Little, 1998; Monin, Norton, Cooper, & Hogg, 2004; 1 = Not at all; 7 = Very much); these items were averaged to form a measure of psychological discomfort (α = .91). Participants also indicated the extent to which they saw the target as holding dissimilar political or social beliefs (Brandt, Chambers, Crawford, Wetherell, & Reyna, 2015; 1 = Not at all different from me; 7 = Very different from me).
Lastly, participants reported their age, race/ethnicity, political ideology on a single item self-report scale (1 = Very liberal; 7 = Very conservative), party affiliation, religious beliefs, education level, and socioeconomic status. At the end of the study, we asked participants the political ideology of the target they read about. Participants who reported the incorrect political ideology were not included in analyses.
Results and Discussion
We calculated attitude change for overall liking, romantic interest, and attractiveness by subtracting participants’ average ratings from Phases 1 and 2 (before discovering target political information) from Phases 3 and 4 (after discovering target political information), such that positive values indicate increased liking and negative values indicate decreased liking. Table 1 reports correlations, means, and standard deviations for study variables.
Table 1
Study 1: Correlations and Descriptive Statistics
| Study Variable | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| 1. Participant Ideology | |||||
| 2. Overall Liking Difference | .07 | – | |||
| 3. Romantic Interest Difference | .02 | .85*** | – | ||
| 4. Attractiveness Difference | .21** | .55*** | .51*** | – | |
| 5. Discomfort | -.17* | -.28*** | -.10 | -.28*** | – |
| M | 3.14 | -.18 | -.41 | -.11 | 1.68 |
| SD | 1.51 | 1.36 | 1.26 | 1.15 | 1.18 |
*p < .05. **p < .01. ***p < .001.
To ensure that participants who correctly identified the target truly were paying attention, we conducted a t-test to see whether participants who viewed an ideologically similar target (matched participants) rated the target as generally more similar to themselves after receiving political information than did participants who viewed an ideologically dissimilar target (mismatched participants). This check by necessity excluded moderates (as there was no moderate target). Before receiving political information, matched and mismatched participants did not differ in their ratings of their similarity to the target, t(111.23) = -.19, p = .848. After receiving political information, matched participants rated targets as significantly more similar to themselves than did mismatched participants, t(119) = 6.48, p < .001. When asked specifically to rate their political dissimilarity to the target, mismatched participants rated the target as significantly more dissimilar than did matched participants, t(117) = -17.31, p < .001. This provides another check of our manipulation and suggests that participants included in our sample attended to the target’s political information.
Primary Analysesiii
We hypothesized that participants’ impressions of a potential romantic partner would become more negative after discovering that the target held dissimilar political beliefs and more positive after discovering that the target held similar political beliefs. We used moderated multiple regression to test the study hypotheses and mean-centered the participant ideology and target ideology variables. Since target ideology was a binary variable, we added or subtracted the proportion of participants in each condition to probe the interactions that included that variable (Enders & Tofighi, 2007). For all analyses, we entered target ideology (0 = liberal, 1 = conservative) and participant ideology (both mean centered) in Step 1, and entered the Target Ideology × Participant Ideology interaction in Step 2 (Aiken & West, 1991).iv
Overall Liking
Table 2 Panel A reports results for change in liking. A significant negative main effect of target ideology emerged in Step 1, such that participants’ liking of the target in the conservative condition decreased compared to participants in the liberal condition. This was qualified by the predicted Target Ideology × Perceiver Ideology interaction in Step 2 (p < .001; see Figure 1 Panel A). Probing the interaction within each condition, conservatism was significantly related to decreased liking of the liberal target and increased liking of the conservative target.
Table 2
Study 1: Results for Moderated Multiple Regression Analyses for Dependent Variables With Semi-Partial r2
| Predictor | Step 1
|
Step 2
|
||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| b | SE | β | t | r2 | b | SE | β | t | r2 | |
| Panel A: Overall Liking Difference | ||||||||||
| Constant | -0.18 | 0.10 | -1.71† | -.17 | .09 | -1.90† | ||||
| Target Ideology | -.75 | .21 | -.28 | -3.60*** | .08 | -.74 | -.18 | -.27 | -4.03*** | .07 |
| Participant Ideology | .06 | .07 | .06 | .82 | .00 | .06 | .06 | .07 | 1.06 | .01 |
| Participant × Target | .83 | .12 | .46 | 6.82 | .21 | |||||
| R2 | .08** | .29*** | ||||||||
| Panel B: Romantic Interest Difference | ||||||||||
| Constant | -.41 | .10 | -4.18*** | -.38 | .08 | -4.61*** | ||||
| Target Ideology | -.74 | .20 | -.30 | -3.80*** | .09 | -.77 | .17 | -.31 | -4.60*** | .09 |
| Participant Ideology | .00 | .07 | .00 | .05 | .00 | .01 | .06 | .01 | .15 | .00 |
| Participant × Target | .87 | .11 | .51 | 7.67*** | .26 | |||||
| R2 | .09** | .33*** | ||||||||
| Panel C: Attractiveness Difference | ||||||||||
| Constant | -.11 | .09 | -1.30 | -.11 | .08 | -1.33 | ||||
| Target Ideology | -.56 | .17 | -.25 | -3.28** | .06 | -.56 | .17 | -.25 | -3.35** | .06 |
| Participant Ideology | .16 | .06 | .21 | 2.83** | .05 | .16 | .06 | .22 | 2.92** | .05 |
| Participant × Target | .33 | .11 | .21 | 2.91** | .05 | |||||
| R2 | .11*** | .15*** | ||||||||
| Panel D: Psychological Discomfort | ||||||||||
| Constant | 1.68 | .09 | 18.29*** | 1.68 | .09 | 18.43*** | ||||
| Target Ideology | .23 | .18 | .10 | 1.25 | .01 | .23 | .18 | .10 | 1.26 | .01 |
| Participant Ideology | -.13 | .06 | -.17 | -2.19* | .03 | -.13 | .06 | -.17 | -2.22* | .03 |
| Participant × Target | -.22 | .12 | -.14 | -1.85† | .02 | |||||
| R2 | .03* | .04* | ||||||||
†p < .10. *p < .05. **p < .01. ***p < .001.
Figure 1
Target Ideology × Perceiver Ideology interactions on dependent variables for Study 1. Standardized regression coefficients are reported for simple slopes and numbers in parentheses represent standard error.
We tested the intercepts for each condition to examine whether impressions changed significantly (i.e., whether attitude change was significantly different from zero). Conservative participants liked the target marginally less after discovering the target was liberal (b = -.34, SE = .18, t = -1.84, p = .068) but did not differ in their liking of the target after discovering the target was conservative (b = .18, SE = .18, t = .96, p = .336). Also in line with expectations, liberal participants liked the target significantly less after discovering the target was conservative (b = -1.25, SE = .18, t = -6.88, p < .001) and liked the target more after discovering the target was liberal (b = .74, SE = .18, t = 4.05, p < .001).
Romantic Interest
Table 2 Panel B reports results for change in romantic interest. A significant negative main effect of target ideology emerged in Step 1, such that participants in the conservative condition became less romantically interested in the target compared to participants in the liberal condition. This was qualified by the predicted Target Ideology × Perceiver Ideology in Step 2 (p < .001; see Figure 1 Panel B). Probing the interaction within each condition, conservatism was related to decreased romantic interest in the liberal target and increased romantic interest in the conservative target.
Tests of the intercepts provided support for hypotheses. Conservative participants were significantly less romantically interested in the target after discovering the target was liberal (b = -.65, SE = .17, t = -3.92, p < .001) but did not differ in romantic interest in the target after discovering the target was conservative (b = -.11, SE = .17, t = -.63, p = .532). Liberal participants were significantly less romantically interested in the target after discovering the target was conservative (b = -1.41, SE = .17, t = -8.57, p < .001) and were more romantically interested in the target after discovering the target was liberal (b = .66, SE = .18, t = 3.77, p < .001).
Attractiveness
Table 2 Panel C reports results for change in rated attractiveness. A significant positive main effect of perceiver political ideology and a significant negative main effect of target political ideology emerged in Step 1. Greater conservatism was associated with increased attractiveness ratings, and participants in the conservative condition decreased their attractiveness ratings of the target compared to participants in the liberal condition. This was qualified by the expected Target Ideology × Perceiver Ideology interaction in Step 2 (p = .004; see Figure 1 Panel C). Probing the interaction within each condition, conservatism was unrelated to changes in attractiveness ratings of the liberal target, but was related to increased attractiveness ratings of the conservative target.
Testing the intercepts, conservatives’ changes in rated attractiveness were not significant for either the liberal (b = .17, SE = .17, t = 1.01, p = .314) or conservative (b = .10, SE = .17, t = .59, p = .556) targets. However, liberals rated the target as significantly less attractive after discovering the target was a conservative (b = -.88, SE = .17, t = -5.23, p < .001), but the targets’ rated attractiveness did not change for the liberal target (b = .18, SE = .17, t = 1.05, p = .294).v
Discomfort
Table 2 Panel D reports results for psychological discomfort. A significant negative main effect of perceiver ideology emerged in Step 1. The expected Target Ideology × Perceiver Ideology interaction was marginally significant (p = .067; see Figure 1 Panel D). Probing the interaction within each condition, conservatism was unrelated to discomfort regarding the liberal target, but was negatively related to discomfort regarding the conservative target. Since discomfort was only measured at one time point and was not converted into a difference score, we probed the interaction at 1 SD above and below the mean of perceiver ideology. Liberal participants were significantly more uncomfortable in the conservative target condition than in the liberal target condition (b = .57, SE = .26, β = .24, t = 2.19, p = .030), whereas conservative participants did not differ in their discomfort between target conditions (b = -.11, SE = .26, β = -.05, t = -.42, p = .673).
Study 1 revealed qualified support for our hypotheses. As expected, the effects of participant ideology on changes in overall liking, changes in romantic interest, changes in perceived attractiveness, and general psychological discomfort were moderated by target ideology. Consistent with our hypotheses, both liberals and conservatives liked ideologically dissimilar targets less and were less romantically interested in them. However, reduced rated attractiveness emerged among liberals but unexpectedly not conservatives. Additionally, only liberals experienced increased discomfort when presented with an ideologically dissimilar target.
There are two limitations to Study 1 which may explain the observed qualified support for our hypotheses. First, we did not measure changes in psychological discomfort, but rather only measured psychological discomfort toward the end of the experiment. Second, the perceived attractiveness item did not specify “physical” attractiveness. This left open to interpretation whether “attractiveness” referred to the target’s appearance or the target’s general, global attractiveness (e.g., personality). We address both these limitations in Study 2.
Study 2
The aim of Study 2 was to replicate Study 1, address some of its limitations, and extend our hypotheses to perceived physical attractiveness. First, to address the fact that psychological discomfort was only measured at the end of Study 1, we measured discomfort after all four phases of Study 2. Second, we addressed the measurement of physical attractiveness in two ways: 1) specifically measuring “physical attractiveness” rather than the more general “attractiveness” measured in Study 1; and 2) using a well-validated visual matching task to examine changes in perceived physical attractiveness of the target rather than just self-reported evaluations of the target (Cole et al., 2016). As in previous work by Cole et al. (2016), using Abrasoft Fantamorph software, we morphed original images with a highly attractive face (a composite image of several dozen faces) and a highly unattractive face (a person with craniofacial syndrome). The morphing procedure produces a series of images of the target’s face that vary in attractiveness. The measure has been validated in Cole et al.’s (2016) work. Specifically, in their first pilot study, the researchers found that participants did not realize that the morphed faces differed in terms of attractiveness. However, when asked to choose the most flattering face in the array, participants overwhelmingly chose faces morphed with the attractive exemplar. This suggests that although participants did not realize that the faces differed in terms of attractiveness (supporting the indirect/implicit nature of the measure), they did choose more attractive faces when instructed to do so (supporting the construct validity of the task). Combined with the successful use of the measure in their work, this provides evidence for both face and construct validity of the measure.
Method
Participants
We recruited 365 participants through MTurk in the winter of 2014. Forty participants incorrectly identified the political ideology of the target they viewedvi and 17 participated in Study 1 and were thus excluded from analyses, leaving a sample of 311 (39.2% female; 64% White; Mage = 38; 58.4% liberal, 24.2% moderate, 17.4% conservative; 93% heterosexual; 28% single).
Materials and Procedure
Materials and procedures closely followed those of Study 1. Participants first reported their sex, age, race/ethnicity, romantic relationship status, and romantic preferences, as well as the youngest and oldest ages of people they would consider dating.
Participants then learned they would view and evaluate a person’s dating profile, and were instructed not to consider their own relationship status when evaluating the person. They were randomly assigned to a liberal (N = 176) or conservative target condition (N = 149). Target photos were taken from Google images and were pretested with MTurk workers to ensure they were slightly above average attractiveness (i.e., approximately 6 on a 10-point scale). As in Study 1, target profiles included a picture of the target with information presented in four phases, each followed by a series of questions. The information presented in each phase was identical to that of Study 1, with political information (i.e., whether the target volunteered for Obama or Romney in the 2012 presidential election) varying in Phase 3.
After each phase, participants encountered the same measures used in Study 1 to rate how similar the target was to themselves, their liking of the target, and their own psychological discomfort (α = .98). The same-self report measure was used to assess the target’s attractiveness, except that the item specified physical attractiveness.
After Phases 1 and 3 only, participants then answered questions testing their memory about the target. During this memory test, participants completed a visual matching task to assess their perceptions of the target’s attractiveness (Cole et al., 2016). Participants saw the original picture from the dating profile as a reference image, along with 11 faces morphed into more or less attractive variations of the target. Five faces were gradually more attractive than the original, five faces were gradually less attractive than the original, and one face was identical to the original. Participants were instructed to select which of the eleven faces was the actual face that matched the reference image. Their selection indicated perceived target attractiveness. Scores varied by increments of seven, with negative scores indicating that participants selected a less attractive version of the face (i.e., -35, -28, -21, -14, -7), positive scores indicating they selected a more attractive version of the face (i.e., +7, +14, +21, +28, +35), and zero indicating they selected the true face.
After all phases, participants then completed the same romantic interest measure as in Study 1 on a separate page (α = .97). After Phase 4 only, they indicated political dissimilarity with the target. Lastly, they reported the same demographic information as in Study 1, including political ideology.
Results and Discussion
As in Study 1, change scores were calculated for overall impressions, romantic interest, rated and perceived attractiveness, and psychological discomfort by subtracting average ratings during Phases 1 and 2 from average ratings during Phases 3 and 4. Thus, positive difference scores indicate more liking after learning the target’s political ideology, whereas negative scores indicate less liking. Table 3 reports correlations among and descriptive statistics for study variables.
Table 3
Study 2: Correlations and Descriptive Statistics
| Study Variable | 1 | 2 | 3 | 4 | 5 | 6 |
|---|---|---|---|---|---|---|
| 1. Participant Ideology | ||||||
| 2. Overall Liking Difference | .10† | – | ||||
| 3. Romantic Interest Difference | .01 | .79*** | – | |||
| 4. Rated Attractiveness Difference | .08 | .38*** | .34*** | – | ||
| 5. Perceived Attractiveness Difference | .13* | .02 | .03 | -.08 | – | |
| 6. Discomfort Difference | -.09 | -.25*** | -.28*** | -.15* | .00 | – |
| M | 3.18 | .02 | .05 | .13 | -1.18 | -.06 |
| SD | 1.54 | 1.09 | 1.01 | .77 | 21.07 | .65 |
*p < .05. **p < .01. ***p < .001.
We again compared similarity ratings between participants who had matching or mismatching political ideologies with the target they viewed (again, moderates could not be included in this check). Before receiving political information about the target, matched and mismatched participants only marginally differed in their ratings of general similarity to the target, t(230.68) = 1.88, p = .06. After receiving political information, matched participants rated the target as significantly more similar to themselves than did mismatched participants, t(210.49) = 6.97, p < .001. When asked specifically to rate their political dissimilarity to the target, mismatched participants rated themselves as significantly more dissimilar than did matched participants, t(234.27) = -12.41, p < .001. This again suggests that participants included in analyses attended to the provided political information.
Primary Analysesvii
We hypothesized that participants would like the target less, become less romantically interested in the target, rate and perceive him/her as less physically attractive, and become more uncomfortable after finding out he/she was ideologically dissimilar, but would like the target more, become more romantically interested in him/her, rate and perceive him/her as more physically attractive, and become less uncomfortable after finding out he/she was ideologically similar. To test these hypotheses, we conducted moderated multiple regression analyses on overall liking, romantic interest, rated attractiveness, perceived attractiveness, and psychological discomfort. For all analyses, target ideology (0 = liberal, 1 = conservative) and participant ideology (both mean centered) were entered in Step 1, and the Target Ideology × Perceiver Ideology interaction was entered in Step 2 (Aiken & West, 1991; see Table 4). Additionally, we tested the intercepts to determine if changes in impressions across target conditions differed from zero.
Table 4
Study 2: Results of Moderated Multiple Regression Analyses for Dependent Variables With Semi-Partial r2
| Predictor | Step 1
|
Step 2
|
||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| b | SE | β | t | r2 | b | SE | β | t | r2 | |
| Panel A: Overall Liking Difference | ||||||||||
| Constant | .02 | .06 | .38 | .03 | .06 | .52 | ||||
| Target Ideology | -.44 | .12 | -.20 | -3.59*** | .04 | -.46 | .11 | -.21 | -4.21*** | .04 |
| Participant Ideology | .07 | .04 | .10 | 1.75† | .01 | .09 | .04 | .12 | 2.38** | .01 |
| Participant × Target | .57 | .07 | .40 | 7.82*** | .16 | |||||
| R2 | .05*** | .21*** | ||||||||
| Panel B: Romantic Interest Difference | ||||||||||
| Constant | .05 | .06 | .80 | .05 | .05 | 1.01 | ||||
| Target Ideology | -.29 | .12 | -.14 | -2.47* | .02 | -.28 | .10 | -.14 | -2.67** | .02 |
| Participant Ideology | .01 | .04 | .01 | .14 | .00 | .02 | .03 | .03 | .50 | .00 |
| Participant × Target | .57 | .07 | .43 | 8.27*** | .18 | |||||
| R2 | .02* | .20*** | ||||||||
| Panel C: Rated Attractiveness Difference | ||||||||||
| Constant | .13 | .04 | 3.00** | .13 | .04 | 3.11** | ||||
| Target Ideology | -.28 | .09 | -.18 | -3.19** | .03 | -.28 | .09 | -.18 | -3.21** | .02 |
| Participant Ideology | .04 | .03 | .07 | 1.29 | .01 | .04 | .03 | .08 | 1.47 | .00 |
| Participant × Target | .16 | .06 | .16 | 2.82** | .02 | |||||
| R2 | .04** | .06** | ||||||||
| Panel D: Perceived Attractiveness Difference | ||||||||||
| Constant | -1.14 | 1.20 | -.96 | -1.15 | 1.20 | -.96 | ||||
| Target Ideology | .46 | 2.40 | .01 | .19 | .00 | .44 | 2.41 | .01 | .18 | .00 |
| Participant Ideology | 1.85 | .78 | .14 | 2.37* | .02 | 1.83 | .78 | .13 | 2.34* | .02 |
| Participant × Target | -.53 | 1.58 | -.02 | -.33 | .00 | |||||
| R2 | .02† | .02 | ||||||||
| Panel E: Psychological Discomfort Difference | ||||||||||
| Constant | -.06 | .04 | -1.75† | -.07 | .04 | -1.79† | ||||
| Target Ideology | .15 | .07 | .11 | 1.96† | .01 | .14 | .07 | .11 | 1.95† | .01 |
| Participant Ideology | -.04 | .02 | -.09 | -1.63 | .01 | -.04 | .02 | -.10 | -1.80† | .01 |
| Participant × Target | -.14 | .05 | -.17 | -2.96** | .03 | |||||
| R2 | .02* | .05** | ||||||||
*p < .05. **p < .01. ***p < .001.
Overall Liking
Table 4 Panel A reports results for differences in overall liking. There was a significant positive main effect of condition and a significant negative main effect of participant ideology in Step 1. Participants in the conservative condition increased their liking of the target compared to participants in the liberal condition, and greater liberalism was associated with increased liking. This was qualified by the hypothesized Target Ideology × Perceiver Ideology interaction in Step 2 (p < .001; see Figure 2 Panel A). Probing the interaction within each condition, conservatism was related to decreased liking of the liberal target and increased liking of the conservative target.
Figure 2
Target Ideology × Perceiver Ideology interactions on dependent variables for Study 2. Standardized regression coefficients are reported for simple slopes and numbers in parentheses represent standard error.
Testing the intercepts, conservative participants did not differ in their liking of the target after discovering the target was liberal (b = -.08, SE = .11, t = -.70, p = .482), but liked the target significantly more after discovering the target was conservative (b = .36, SE = .11, t = 3.33, p = .001). In line with expectations, liberal participants liked the target significantly less after discovering the target was conservative (b = -.70, SE = .11, t = -6.53, p < .001) and liked the target significantly more after discovering the target was liberal (b = .60, SE = .11, t = 5.27, p < .001).
Romantic Interest
Table 4 Panel B reports results for differences in romantic interest. There was a significant negative main effect of target ideology in Step 1, such that participants in the conservative condition became less romantically interested in the target. This was qualified by the expected Target Ideology × Perceiver Ideology interaction in Step 2 (p < .001; see Figure 2 Panel B). Probing the interaction within each condition, conservatism was related to decreased romantic interest in the liberal target condition and increased romantic interest in the conservative target condition.
Testing the intercepts, as predicted, conservative participants were significantly less romantically interested in the target after discovering he/she was liberal (b = -.24, SE = .11, t = -2.24, p = .026), but were significantly more romantically interested in the target after discovering the target was conservative (b = .35, SE = .10, t = 3.42, p < .001). In line with expectations, liberal participants were significantly less romantically interested in the target after discovering he/she was conservative (b = -.50, SE = .10, t = -4.99, p < .001) and were significantly more romantically interested in the target after discovering he/she was liberal (b = .65, SE = .11, t = 5.98, p < .001).
Rated Attractiveness
Table 4 Panel C reports results for differences in self-reported attractiveness. There was a significant negative main effect of target ideology in Step 1, such that participants in the conservative condition decreased their attractiveness ratings of the target. This was qualified by the predicted Target Ideology × Perceiver Ideology interaction in Step 2 (p = .009; see Figure 2 Panel C). Probing the interaction within each condition, conservatism was unrelated to differences in attractiveness ratings in the liberal target condition, but was related to increased attractiveness ratings in the conservative target condition.
Testing the intercepts, conservative participants unexpectedly rated the target as significantly more attractive after discovering he/she was liberal (b = .22, SE = .09, t = 2.45, p = .015) and significantly more attractive after discovering the target was conservative (b = .18, SE = .08, t = 2.13, p = .034). As expected, liberal participants rated the target as significantly less attractive after discovering the target was conservative (b = -.17, SE = .08, t = -2.02, p = .044) and as significantly more attractive after discovering the target was liberal (b = .35, SE = .09, t = 3.94, p < .001).
Perceived Attractiveness
Table 4 Panel D reports results for differences in perceived attractiveness. There was a significant positive main effect of participant ideology in Step 1, such that participant conservatism was associated with perceiving the target face as more attractive after discovering political information. However, the expected Target Ideology × Perceiver Ideology interaction did not emerge in Step 2 (p = .656), indicating that participants’ perceptions of the targets’ attractiveness did not differ depending on their political ideology.
Discomfort
Table 4 Panel E reports results for changes in psychological discomfort. There was a marginal positive main effect of target ideology in Step 1, such that participants in the conservative condition became marginally more uncomfortable. This was qualified by the expected Target Ideology × Perceiver Ideology interaction in Step 2 (p = .001); see Figure 2 Panel D). Probing the interaction within each condition, participant ideology was unrelated to differences in discomfort in the liberal target condition, but liberalism was related to increased discomfort in the conservative target condition.
Testing the intercepts, conservative participants did not differ in discomfort after discovering the target was liberal (b = -.09, SE = .08, t = -1.19, p = .233) but were significantly less uncomfortable after discovering the target was conservative (b = -.16, SE = .07, t = -2.30, p = .022). Liberal participants were significantly more uncomfortable after discovering the target was conservative (b = .16, SE = .07, t = 2.35, p = .020) and significantly less uncomfortable after discovering the target was liberal (b = -.19, SE = .08, t = -2.59, p = .010).viii
Study 2 found that people’s impressions of potential mates change after discovering their political ideology. In particular, liberals’ impressions of the target tended to become more favorable after discovering the target was also liberal, and more unfavorable after discovering the target was conservative. Conservatives’ impressions tended to become more favorable after discovering the target was also conservative, but generally did not change after discovering the target was liberal (aside from becoming less romantically interested in the target). Unexpectedly, conservatives’ ratings of target attractiveness increased after discovering the target was liberal. These reactions to ingroup and outgroup targets differ slightly from those found in Study 1, as conservatives in Study 1 liked the liberal target less and become more romantically interested in the target but did not change their liking of the conservative target, whereas in Study 2, the reverse occurred.
General Discussion
Across two studies, we demonstrated that initial impressions of other people change after discovering political information about the target. Using a dating paradigm, we found that both liberals and conservatives liked targets more and became more romantically interested in them after discovering that a target was politically similar, but liked targets less and became less romantically interested in them after discovering political dissimilarity with the target. Unexpectedly, conservatives did not change their physical attractiveness ratings of politically similar or dissimilar targets in Study 1, and surprisingly rated both liberals and conservatives as more attractive after discovering political information in Study 2. In contrast, liberals rated politically dissimilar targets as less physically attractive and politically similar targets as more attractive. Further, liberals became more uncomfortable after discovering a target held a dissimilar political ideology. In contrast, both liberals and conservatives became less uncomfortable after discovering a target held a similar political ideology. Some of these findings among conservatives could in part be due to the limited number of conservatives in our samples.
In Study 2, the perceived attractiveness task required participants to select the true face from an array of more or less attractive faces. This task is a more subtle measure of attractiveness intended to capture actual perceptions rather than self-reports. The self-reported measure of physical attractiveness was surprisingly uncorrelated with the perceived attractiveness task, suggesting responses on this measure of perceived attractiveness do not necessarily cohere with self-reports. While participants’ self-reported ratings of the target’s attractiveness changed after discovering the target’s political ideology, their actual perceptions of the target’s attractiveness did not. Thus, it appears that perceived target attractiveness was unaffected by the political ideology manipulation. Though unexpected, these findings are consistent with evidence suggesting that subtle measures of attitudes (e.g., implicit attitudes) are less susceptible to change than are explicit ones (DeCoster, Banner, Smith, & Semin, 2006; Rydell & McConnell, 2006; Rydell, McConnell, Strain, Claypool, & Hugenberg, 2007). Although caution should be exercised when interpreting null results, it seems that the self-report measures may be tapping into a more deliberate, explicitly motivated reasoning process, whereas the perceptual measure may be capturing a more automatic, implicitly motivated perception process. This suggests that while people change their attitudes about others’ physical attractiveness, their actual perceptions of attractiveness might not change. In this case, political dissimilarity might have changed deliberate but not automatic impressions of the targets. That is, people may have rated targets as less attractive to convey their dislike because of their support for a particular political candidate, but this may not have affected their perceptions of the targets since the photos they viewed were of physically attractive people. It appears that motivated perception of the target as less attractive to attenuate the ideological threat did not occur.
These studies go beyond other recent findings regarding political dissimilarity and physical attractiveness (Nicholson et al., 2016) in a number of ways. We specified a context (a dating paradigm) to give our studies more mundane realism and found similar effects across a variety of dependent variables (overall liking, romantic interest, and rated physical attractiveness). Additionally, we found that psychological discomfort also became elevated after discovering political dissimilarity and decreased after discovering political similarity. This suggests that there might be emotional as well as attitudinal effects of political dissimilarity.
Moreover, these studies suggest that liberals and conservatives might form impressions of others in roughly similar ways, in that both tend to become positively biased toward political ingroup members and negatively biased toward political outgroup members. Additionally, liberals, but not conservatives, became more uncomfortable after discovering that a target was ideologically dissimilar. This is somewhat inconsistent with work on negativity bias, which suggests that conservatives are particularly responsive to threats and negativity in their environment (Hibbing, Smith, & Alford, 2014). Although the results of our studies could in part be due to the small number of conservatives in our samples, our results are in line with recent findings that liberals and conservatives are similarly responsive to threats to meaning, defined as threats such as those to identity, purpose, and belongingness, which are likely elicited by ideologically dissimilar others (Crawford, 2017). Thus, our findings support work suggesting that liberals and conservatives might be similarly responsive to negativity posed by politically dissimilar others (Brandt et al., 2014; Crawford, 2017). However, conservatives were biased against outgroup members in Study 1 but were biased toward ingroup members in Study 2, making interpretation in terms of ingroup or outgroup bias amongst conservatives difficult.
Again, these ambiguous results are possibly due to the fact that there were smaller numbers of conservatives (Study 1, N = 26; Study 2, N = 57) and moderates (Study 1, N = 40; Study 2, N = 78) in both studies than there were of liberals (Study 1, N = 95; Study 2, N = 189). However, when we folded across our political ideology measure to consider extremity (1= Somewhat liberal/conservative, 3 = Very liberal/conservative), conservatives in our sample tended to be slightly more conservative than liberals were liberal (Study 1: Liberals – M = 2.07, SD = .75, Conservatives – M = 2.49, SD = .66, t(185) = -3.35, p = .001; Study 2: Liberals – M = 2.11, SD = .71, Conservatives – M = 2.32, SD = .76, t(244) = -1.92, p = .056). This could in part be due to higher sensitivity to extreme scores in the smaller sample of conservatives. In addition, when we examined the interaction between participant ideology and matched/unmatched targets for perceived similarity, liberals and conservatives did not differ in their ratings of how similar they were to ideologically dissimilar targets (Study 1: F(1, 117) = .63, p = .431; Study 2: F(1, 242) = .89, p = .345).ix Thus, it seems the effects of dissimilarity were generally similar amongst both liberals and conservatives.
Despite some inconsistencies, our findings are largely in line with recent research suggesting that negative intergroup attitudes are similarly expressed on both sides of the political spectrum (Brandt et al., 2014; Collins, Crawford, & Brandt, 2017; Crawford, Mallinas, & Furman, 2015). Brandt et al. (2014) developed the ideological conflict hypothesis (ICH), which posits that people on both the political right and left tend to express antipathy towards one another, and to equal degrees. Consistent with this, we found that both liberals and conservatives form impressions in roughly similar ways, in that both change their impressions of a target after discovering value similarity or dissimilarity with the target person. These studies are also in line with the growing body of evidence that political ideology is a potent factor in interpersonal evaluations (Brandt, 2017; Chambers et al., 2013; Koch et al., 2016). By simply manipulating the political campaign for which our fictional target volunteered, we were able to change participants’ initial impressions. This suggests that political dissimilarity strongly contributes to disliking, which manifests in a variety of related attitudinal domains, including overall liking, romantic interest, and rated attractiveness.
These findings provide a more nuanced understanding of the relationship between political ideology and disliking. Replicating past work (Brandt et al., 2014), we demonstrate that liberals and conservatives express roughly equal amounts of disliking of politically dissimilar targets. We also show that romantic interest and psychological discomfort are associated with discovering political (dis)similarity with another person, which builds on work that only examines disliking based on political dissimilarity. Additionally, we replicated Nicholson et al.’s (2016) findings that people rate politically dissimilar others less attractive compared to politically dissimilar others, and build on their work by showing that these ratings actually change after discovering political information. We also build on their work by showing that actual perceptions of attractiveness do not seem to change as a result of discovering political dissimilarity, although future work should examine this more thoroughly. It is possible that people still find others to be attractive regardless of their ideology due to the stability of their attractive physical features, but state that ideologically dissimilar others are less attractive as a way to express their general disliking of them. Together, these findings build on existing work on impression formation, political dissimilarity, and disliking.
Practically, these findings have implications for intergroup relations, particularly between political groups. As reviewed earlier, political ideology is one of the strongest sources of disliking (Brandt, 2017; Koch et al., 2016) and often leads to prejudice against politically dissimilar others (Chambers et al., 2013). These studies show that political information colors initial impressions and has the potential to both increase and decrease liking. This provides insight into the strength of political similarity and dissimilarity and suggests that political prejudice emerges soon after meeting someone. Although this is not necessarily encouraging for improving social relations, knowing that such disliking can occur so quickly can inform those wishing to promote more productive interactions between politically dissimilar others. Since contact is one of the most effective forms of prejudice reduction (Pettigrew & Tropp, 2006), and our findings suggest that contact between known politically dissimilar others could be difficult once political ideology is known, this information could assist in finding ways to encourage contact despite these negative reactions in order to reduce political prejudice. Additionally, even though our studies did not examine how impressions updated after discovering additional information after finding out a target’s political ideology, future work should examine whether this negativity could be undone. That is, future work could consider whether finding out more positive or neutral information about the person after discovering their political ideology decreases the negativity.
Limitations and Future Directions
It is possible that participants did not conceptualize the study as a true mating paradigm, but rather as simply an impression formation task. Sixty three percent of the sample in Study 1 and 73% of the sample in Study 2 reported being in romantic relationships, and it might have been less potent of a paradigm for those participants (although we did instruct them to imagine they were not in a relationship for the purposes of the study). While this is certainly possible, our results were not moderated by relationship status.x We also found differences in romantic interest and perceived attractiveness, suggesting that participants’ attitudes on these romantically-relevant factors did change, which indicates that they likely bought into the dating paradigm. However, future research could compare the impact of political ideology in a dating paradigm with only single individuals with a general impression formation setting unrelated to dating.
Although MTurk samples have been shown to be more representative than student samples, the means of ideology in both studies (Ms = 3.14 and 3.19, respectively, on 1-7 scales) and main effects suggesting greater dislike of conservatives than liberals among all participants indicate that the samples in these studies leaned to the political left. Methodologically, the studies would have benefited from a larger proportion of conservatives. It is possible that we might have observed greater effects for liberals due to the small sample of conservatives. If our sample included an equal distribution of liberals and conservatives, it is possible that we could have found conservatives to be more sensitive to dissimilarity. Future work could examine these changes in a sample with a more even distribution of liberals and conservatives to determine if the same pattern of results would also emerge more clearly on the political right.
Lastly, our samples were limited in their diversity. We sampled from MTurk, and thus participants might not have devoted their full attention to completing our studies. Further, although MTurk samples are often more diverse than student samples, the issue of generalizability still remains. For example, our samples only included participants in the United States. The United States has a highly charged political climate, and thus the effects we observed are likely driven by the strength of the ideological divide in American culture. Notably, these data were collected before the polarizing 2016 Presidential Election, and it is possible that our observed effects might have been even stronger if collected during or after that time. Future work should examine whether such effects of political (dis)similarity emerge in other cultures, particularly those in which political ideology is not particularly divisive within the population. Although research on worldview conflict has found similar effects in non-U.S. samples (e.g., Karpov & Lisovskaya, 2008; Kremer, Barry, & McNally, 1986; van der Noll, Poppe, & Verkuyten, 2010), it would be interesting to examine whether effects of ideological dissimilarity are stronger in nations that have a high degree of political polarization compared to those that do not.
Conclusion
We tested the idea that political partisans change their initial impressions of others after discovering value similarity or dissimilarity. Extending past work, we demonstrated that attitudes toward dissimilar others change, specifically when considering impression formation in a dating paradigm. Additionally, we found that people like and are less romantically interested in targets after discovering they hold dissimilar values. They also rate them as less attractive, although they do not appear to actually perceive them as such. By demonstrating that both liberals’ and conservatives’ initial impressions of another person were similarly affected by political information, our findings suggest that value dissimilarity is important in impression formation.