Sharing Rare Attitudes Attracts

Abstract

People like others who share their attitudes. Online dating platforms as well as other social media platforms regularly rely on the social bonding power of their users’ shared attitudes. However, little is known about moderating variables. In the present work, I argue that sharing rare compared with sharing common attitudes should evoke stronger interpersonal attraction among people. In five studies, I tested this prediction for the case of shared interests from different domains. I found converging evidence that people’s rare compared with their common interests are especially potent to elicit interpersonal attraction. I discuss the current framework’s theoretical implications for impression formation and impression management as well as its practical implications for improving online dating services.

Keywords

People like others who share their attitudes. Social psychology has found that similarity regarding opinions, values, and interests is one of the strongest determinants of interpersonal attraction (Montoya, Horton, & Kirchner, 2008). Although different explanations have been offered for the similarity effect, little is known about whether attitudes differ regarding their bonding potential. In the present work, I suggest that the frequency of a given attitude is an important, yet overlooked determinant of interpersonal attraction. Specifically, I argue that people particularly like others who share their rare attitudes, that is, attitudes not shared by many others. This prediction can be derived from several theoretical perspectives assuming that rare compared with common attitudes are more informative, more diagnostic, and elicit a strong need for social validation.

Shared Attitudes and Interpersonal Attraction

The phenomenon that people like others who are similar to themselves has been described as “one of the most robust relationships in all of the behavioral sciences” (Berger, 1973, p. 281). The similarity effect has been demonstrated for personality traits (Steele & McGlynn, 1979), physical attributes (Stevens, Owens, & Schaefer, 1990), and most extensively for attitudes (Byrne, 1961; Byrne & Blaylock, 1963; Tan & Singh, 1995), such as values (Lewis & Walsh, 1980) or hobbies (Jamieson, Lydon, & Zanna, 1987). There are two main explanations for the similarity effect that have generated most of the existing empirical findings (Montoya & Horton, 2013, but see Rosenbaum, 1986). According to Byrne’s (1971) reinforcement model, people have a need for a logical and consistent view of the world and this effectance motive is satisfied by others who share their attitudes. According to an information processing explanation, interpersonal attraction results from an inference process. People infer other positive (negative) attributes based on attitude similarity (dissimilarity) (Ajzen, 1974; Kaplan & Anderson, 1973; Tesser, 1971). Likewise, Aronson and Worchel (1966) have argued that people anticipate to be liked by a person who holds similar attitudes, and that the anticipation of being liked itself elicits attraction. Although debates about the underlying processes are still ongoing, the similarity effect has found its way into many real-world applications aimed at connecting people. Social media outlets such as Facebook (FB) host interest groups where people can connect with others who share their interests. Likewise, online dating heavily relies on matching dating partners based on attitude similarity regarding political orientation, values, or interests.

Despite the large body of research on the similarity effect, little is known about moderating variables. One exception is the notion that the similarity effect is stronger for attitudes that are important to their holders (Bahns, Crandall, Gillath, & Preacher, 2017; Montoya & Horton, 2013). Other research has found that negative compared with positive attitudes elicit stronger interpersonal attraction (Bosson, Johnson, Niederhoffer, & Swann, 2006; Weaver & Bosson, 2011). Apart from that, it remains largely unclear what characteristics of attitudes may function as moderators of the similarity effect. In the present work, I propose attitude frequency as a moderator and argue that sharing rare attitudes elicits more interpersonal attraction than sharing common attributes. In the following, I explain how this prediction can be derived from different theoretical perspectives.

Unit Relationship

Research has shown that interpersonal attraction is elicited not only by sharing meaningful attitudes or attributes but also by incidental similarity. For example, when people learn that another person shares their birthday, name, or fingerprint type, they report increased liking for that person (Cialdini & Goldstein, 2004; Finch & Cialdini, 1989; see also Jiang, Hoegg, Dahl, & Chattopadhyay, 2009). The research on incidental similarity implies an important role of rarity. Treatments that are used to demonstrate the influence of incidental similarity on attraction typically feature some kind of rare attributes (e.g., birthday, fingerprint, etc.). Burger, Messian, Patel, del Prado, and Anderson (2004) directly tested this assumption and found that rare but not common incidental similarity increased attraction and compliance. For example, when participants learned that a person shared their fingerprint type, they only reported liking for that person when they were told it was rare to share this type, but not when they were told it was common to share the type. The authors explained this based on Heider’s (1958) concept of unit relationship.

People are said to form unit relationships when they are aware of a common attribute not shared by those around them. For example, two people from California will see themselves as belonging to a unit relationship if they meet in North Carolina but not if they meet in Los Angeles. (Burger et al., 2004, p. 36)

Although the authors argued that a unit relationship is sensitive to the frequency of the shared attribute, they did not provide a precise explanation for this. In the following, I explain how the stronger impact of sharing rare compared with common attributes and attitudes can be predicted more precisely from three different theoretical perspectives.

Informativeness

Interpersonal attraction, as any judgment, is relative in nature. When people judge the likability of a person, this judgment always reflects a comparison. Saying that I strongly like person A means that I like Person A better than I like most other people. The comparative nature of judgments is reflected in Fiske’s (1980) concept of informativeness. Accordingly, person-related information (e.g., attributes, behaviors) is more informative the less frequent it is in the general population. If a person tells the truth, this is not as informative as when a person lies, because most people tell the truth most of the time. Hence, lying is more informative as it sets apart the liar from a large part of the population. Crucially, the more informative a piece of information is, the more weight it receives during impression formation. Fiske (1980) used the informativeness concept to explain the dominance of negative over positive information in impression formation.

Applying this concept to interpersonal attraction, sharing rare compared with common attitudes should evoke stronger interpersonal attraction. Consider you meet a person and you learn this person shares your interest for The Beatles. This attitude is not very informative because many people like The Beatles and it probably does not greatly affect your liking for that person. If you meet a person and you learn that this person shares your interest for Tom Waits, you might instantly like that person assuming that you rarely meet people who like Tom Waits.

Diagnosticity

The second perspective is related to the information processing explanation of the similarity effect (Ajzen, 1974; Kaplan & Anderson, 1973; Tesser, 1971). Accordingly, when people are provided with information about a person, they infer additional characteristics of that person based on the information available. When people learn that a person has similar attitudes or attributes, they infer that this person also has other similar attitudes or attributes. Because people typically like their own attitudes and attributes, they thereby infer a number of positive qualities. Research on category diagnosticity has shown that rare qualities (e.g., behaviors) are more diagnostic for category membership (e.g., trait) than common qualities (Reeder & Brewer, 1979; Skowronski & Carlston, 1987). Consequently, people should draw more inferences about a person based on the person’s rare compared with the person’s common qualities. If we apply this reasoning to the information processing perspective of the similarity effect, rare compared with common shared attitudes and attributes can be expected to trigger stronger inferences regarding a target person’s additional qualities. If you meet someone who shares your liking for The Beatles, you may not infer as many other shared attitudes as when you meet someone who shares your liking for Tom Waits.

Social Validation

Byrne (1971) argued that similarity leads to attraction because people have a need for a logical and consistent view of the world. Hence, they seek social validation of their beliefs, values, and ultimately of themselves. Borrowing from the concept of cognitive dissonance and social comparison theories (Festinger, 1954, 1957), Byrne argued that individuals who are similar to the self serve as positive reinforcements as they validate the consistency and the logic of our worldviews. Similar people are therefore associated with positive feelings, which lead to attraction. Even though Byrne’s concept did not address the frequency of shared attitudes and attributes, it seems plausible that people should be especially motivated to validate attitudes that are rarely validated. If a person likes The Beatles, this attitude is often validated by the social environment and people should therefore not be strongly motivated to seek additional validation of their attitude by a given person. However, when a person likes Tom Waits, an attitude that is not often, people should feel a stronger need for validation of this attitude and should therefore feel more strongly attracted to someone who satisfies this need.

A similar prediction follows from the fundamental human need to affiliate with and be accepted by members of a group (Baumeister & Leary, 1995). People’s rare attitudes are a threat to this need because they may lead to ostracism. Therefore, learning that a person shares a rare attitude may elicit a feeling of belongingness where it is strongly needed.

Scope of the Present Work

A stronger influence of shared attitudes on interpersonal attraction can be predicted from different theoretical perspectives. Of course, the described mechanisms are not mutually exclusive and are likely to coexist. Given the consensus among the theoretical perspectives in predicting attitude frequency as a moderator of the similarity effect, I sought to empirically test this prediction. Given the many practical implications and applications of the similarity effect, I aimed for high ecological validity in the following empirical studies. I asked participants to name things they liked from different domains such as movies, music, hobbies, food, among others. The studies thereby featured meaningful stimuli that are likely to be the subject of conversation when meeting a stranger in real life or on some social media platform (e.g., online dating). In five studies, I tested whether people’s rare compared with their common interests elicit more interpersonal attraction toward a target person. All data and materials are available online (https://osf.io/4vkqy/?view_only=cb49804c2a2e4f8fa3e2a7d6fb88bb16).

Study 1

Study 1 tested the relation between perceived frequency of a personal interest and interpersonal attraction toward a person sharing that interest. Participants first named movies, musicians, and hobbies they liked before they rated for each interest how much they would like a person who shares that interest and how common/rare the interest is in the population.

Method

Participants and design

As power calculations for a correlational relationship at the within-participants level are not straightforward, I aimed at collecting data from a rather large participant sample of 200. I collected data from 202 participants (89 female, 113 male) who were recruited via the Mechanical Turk (MTurk) online platform and participated for US$0.70. All participants were located in the United States.

There was no experimental manipulation in the strict sense, but all participants generated interests from three different domains, which I treated as a random factor. The independent variable was perceived interest frequency and the dependent variable was perceived likability of a person who shares an interest.

Procedure

The study used the Qualtrics software. Instructions informed participants about confidentiality of the study and about their right to quit at any point.

After answering demographical questions, participants were asked to provide the names of seven movies, seven musicians, and seven hobbies they liked. The order of the three domains was randomized. For each domain, participants were provided with seven text boxes in which they typed their interests.

Next, participants were asked to provide interpersonal liking ratings and frequency ratings for each interest they had generated. The order of these two blocks was randomized. For the liking ratings, participants were repeatedly asked to “image you meet a person and you learn that this person also likes [name of interest]. How likable would you find that person?” Participants indicated likability using a 9-point scale that was labeled not at all likable (1), neutral (5), and extremely likable (9). Participants repeated this for all of their 21 interests. For the frequency ratings, participants were asked, “How common/rare is it to like [name of interest]?” and were provided a 7-point scale labeled extremely rare (1) and extremely common (7). Participants repeated this for all of their 21 interests. At the end of the study, participants were thanked, debriefed, and given a code necessary to receive their compensation via MTurk.

Results

Due to a programming error, the frequency estimates for one music interest were not collected. As the following analysis revealed robust results, the resulting loss in power seemed not to be a problem. Linear mixed effects model analysis was conducted using the lme4 package in R. I specified a most restricted model that included participant and domain as random factors with random error components for intercepts and slopes. This model revealed the predicted effect of frequency on likability (Figure 1), b = −0.11 (β = −0.13), t(12.34) = −4.92, p < .001, 95% confidence interval (CI) = [–0.19, –0.08].¹ That is, the less common an interest, the greater was the perceived likability of a target person sharing that interest. Regarding the random effects, there was significant variation across participants regarding slopes and intercepts, but no significant variation of intercept or slopes across the three different interest domains.

Figure 1.

Jitter plot illustrating the relation between interest frequency and liking at the level of individual observations.

I then specified an additional model that included a fixed effect for the order in which participants had provided likability and frequency judgments (1 = likability first, –1 = frequency first), as well as the interaction between frequency and order. The most restricted model that converged included the random factor subject with variance components for intercepts and frequency slopes, and the random factor domain, with variance components for intercepts as well as slopes for frequency, order, and their interaction. This model again confirmed the predicted effect of frequency on likability, b = −0.11 (β = −0.13), t(11.28) = −4.95, p < .001, 95% CI = [–0.18, –0.07]. There was no significant effect of order, b = 0.28 (β = 0.06), t(3.26) = 1.47, p = .230, 95% CI = [–0.03, 0.15], nor of the interaction between frequency and order, b = −0.04 (β = −0.05), t(3.34) = −1.11, p = .340, 95% CI = [–0.13, 0.04]. Hence, the relation of attitude frequency and target likability did not depend on which of the ratings participants provided first.

Discussion

As predicted, interest frequency was negatively related to the perceived likability of imaginary target persons sharing an interest. This effect was stable across the three different interest domains, movies, musicians, and hobbies, as there was no significant variation across the domains’ regression slopes. In addition, the effect did not depend on whether participants provided likability ratings before or after the frequency ratings. Thus, targets sharing rare interests were judged as more likable even when participants were not made aware of interest frequency before, ruling out a simple demand effect. Hence, Study 1 rendered first empirical support for the idea that the prevalence of a shared attitude influences interpersonal attraction.

However, the observed relationship between perceived interest frequency and likability may result from confounded variables. Specifically, people may hold stronger attitudes toward their rare compared with their common interests, that is, they may simply like their rare interests better. Shared rare interest would then naturally elicit stronger liking. Another confounded variable could be attitude importance. That is, rare attitudes may be more important to people than common attitudes and therefore also elicit stronger interpersonal attraction. Studies 2a and 2b addressed these possible confounds. In addition to measuring perceived likability of target persons, I measured attitude strength for common and rare interests (2a) as well as attitude importance (2b). In addition, Studies 2a and 2b moved from a purely correlational design to a quasi-experimental design.

Study 2a

Study 2a tested the relation between target likability and interpersonal attraction in a quasi-experimental design in which participants generated common and rare interests. To test whether this relationship could be explained by stronger attitudes that participants may hold toward rare interests, participants were asked to rate how much they liked each of their interests. Because the two dependent variables target likability and interest attitude strength may influence each other in a within-participants design, Study 2a varied these dependent variables between participants.

Method

Participants and design

I aimed at collecting data from 110 participants for each dependent variable (total = 220), providing sufficient statistical power to detect medium-sized effects in a within-participants design (Cohen, 1992). I collected data from 227 students of a large university (164 female, 63 male) who participated for candy or course credit.

The two dependent variables (likability, attitude strength) varied between participants, and the independent variable (interest frequency) varied within participants.

Procedure

The experiment used the Microsoft Visual Basic software. Participants were seated in front of a computer by a research assistant. Instructions informed participants about confidentiality of the study and about their right to quit at any point.

After answering demographical questions, participants were asked to provide the names of four movies they liked and “that many other people also like” (common interests) and the names of four movies they liked that “only a few other people also like” (rare interests). The order of these two blocks was randomized and each block presented participants with four text boxes in which they could type the movie names. The same procedure was repeated for musicians and hobbies.

Next, one half of participants were asked to provide likability ratings for persons who share their interests similar to Study 1. The other half of participants were asked to rate the attitude strength for the generated interests. Specifically, participants were told that “we would like to know how much you like the interest you have listed.” On the next screens, participants were then presented with their interests one by one in random order and asked “How much do you like [name of interest]?” Participants responded on a 9-point scale labeled not at all (1) and extremely much (9).

At the end of the experiment, participants were thanked, debriefed, and received their compensation.

Results

Likability

I first analyzed data from 112 participants who rated target likability. I first compared the mean likability ratings for targets sharing the common interest and for targets sharing the rare interests across interest domains. Participants produced significantly larger likability ratings for target persons who shared their rare compared with their common interests, M_rare = 6.77 (SD_rare = 0.91) versus M_common = 6.37 (SD_common = 0.93), t(111) = 3.77, p < .001, 95% CI = [0.19, 0.61], d_z = 0.36. As illustrated in Figure 2, this effect was evident in each of the three domains.

Figure 2.

Mean likability ratings for target persons who share participants’ common or rare interests in the three interest domains.

Attitude strength

Next, I analyzed the remaining data from 115 participants who rated attitude strength for each interest. Participants expressed stronger attitudes (more liking) toward their common compared with their rare interests, M_common = 7.52 (SD_common = 0.98) versus M_rare = 7.00 (SD_rare = 1.16), t(114) = 5.11, p < .001, 95% CI = [0.31, 0.73], d_z = 0.48. As illustrated in Figure 3, this effect was evident in each of the three domains.

Figure 3.

Mean attitude strength ratings for common and rare interests in the three interest domains.

Discussion

Results from Study 2a confirmed the predicted association between perceived interest frequency and imagined likability of a target person sharing that interest. Importantly, this effect is unlikely to be caused by attitude strength, as participants indicated to like their common interests better than their rare interest. The fact that participants have more positive attitudes toward their common interest is in line with the notion of social proof (Cialdini, 1993; Kelman, 1958). Accordingly, people’s evaluations are influenced by their social environment (Wooten & Reed, 1998). For example, people perceive a comedy show as more funny when it includes canned laughter (Cialdini, 1993). Hence, people may hold more positive attitudes toward their common interests because these attitudes are often reinforced by the social environment. Alternatively, attitude frequency may reflect some kind of “objective” quality. That is, a movie that many people like may, on average, be of better quality than a movie that only a few people like.

Yet, another possible confound is attitude importance. That is, people may feel that their rare interests are more important to them than their common interests. Importantly, attitude importance may be unrelated to attitude strength. A person may like Michael Jackson (common) better than Tom Waits (rare) but still consider his attitude toward Tom Waits as more important to him. That attitude importance is a potent predictor in the domain of interpersonal attraction has been argued before (e.g., Festinger et al., 1950; Newcomb, 1943, 1961), and recent research found that friendship pairs were especially similar regarding attitudes that they both considered important to themselves (Bahns et al., 2017). Hence, if attitude frequency and attitude importance are confounded, the present findings may not result from frequency-related processes like informativeness, diagnosticity, or social validation, but from the fact that rare attitudes are more important to people.

Study 2b

Study 2b aimed at replicating results from Study 2a and also tested whether the relationship between interest frequency and likability could be explained by attitude importance. While half of the participants again rated the likability of imaginary target persons sharing their interests, the other half of participants were asked to rate how important their interests are to them.

Method

Participants and design

Similar to Study 2a, I again aimed at collecting data from 220 participants. I collected data from 218 online participants whom I recruited via MTurk.

The two dependent variables (likability, attitude importance) varied between participants and the independent variable (interest frequency) varied within participants.

Procedure

The study used the Qualtrics software. Instructions informed participants about confidentiality of the study and about their right to quit at any point. The experimental procedure was similar to Study 2a except that half of the participants were asked to rate attitude importance instead of attitude strength. Specifically, for each of their interests, these participants were asked, “How important is it to you that you like [name of interest]?” Participants responded on a 9-point scale labeled not at all (1) and extremely much (9).

Results

Likability

I first analyzed data from 110 participants who rated target likability. Replicating the findings from Study 2a, participants produced significantly larger likability ratings for target persons who shared their rare compared with their common interests, M_rare = 6.95 (SD_rare = 1.12) versus M_common = 6.57 (SD_common = 1.13), t(109) = 4.57, p < .001, 95% CI = [0.22, 0.55], d_z = 0.44. As illustrated in Figure 4, this effect was evident in each of the three domains.

Figure 4.

Mean likability ratings for target persons who share participants’ common or rare interests in the three interest domains.

Attitude importance

Next, I analyzed the remaining data from 108 participants who rated attitude importance for each interest. Participants indicated that their common interest were more important to them than their rare interest, M_common = 6.26 (SD_common = 1.38) versus M_rare = 5.98 (SD_rare = 1.41), t(107) = 2.06, p = .042, 95% CI = [0.02, 0.54], d_z = 0.20. As illustrated in Figure 5, this effect was descriptively present in all three domains but statistically significant only in the movies domain.

Figure 5.

Mean attitude importance ratings for common and rare interests in the three interest domains.

Discussion

Results from Study 2b again confirmed the predicted association between interest frequency and likability of imaginary target persons. Importantly, Study 2b ruled out that this was simply an effect of attitude importance. Participants rated their common interests as more important than their rare interests. Taken together, findings from Studies 2a and 2b constitute a remarkable paradox. On one hand, the association between attitude frequency and interpersonal attraction is robust and generalizes across the three interest domains, movies, music, and hobbies. One the other hand, this relation exists despite the fact that participants hold stronger attitudes toward their common interests and consider these more important.

Yet, Studies 1, 2a, and 2b feature an important limitation. That is, attitude frequency ratings in Study 1, as well as attitude frequency instructions in Studies 2a and 2b addressed perceived frequency. Thus, participants could make use of some lay theory about the relation between attitude frequency and likability. To address this, Study 3 used an objective measure of interest frequency. In addition, the observed relation between attitude frequency and interpersonal attraction so far rests on stimuli from three domains only (movies, music, hobbies). Study 3 did not restrict interest to these domains and therefore tested the generalizability of the observed effects.

Study 3

FB conveniently provides objective measures of attitude frequencies. It hosts a large number of interest “pages” and users can express their positive attitudes by “liking” the page. The like count is displayed on the respective page and constitutes a good approximation of attitude frequency. The singer Shakira, for example, is among the most popular artists on FB with a total like count of more than 100,000,000. According to my hypothesis, learning that someone shares an interest for Shakira should not elicit much liking for that person because many people like Shakira. I therefore predict that the less likes an interest page has, the greater is the perceived likability of a person sharing that interest. To test this, Study 3 asked participants to provide their 15 most recent FB page likes, which served as interest stimuli, while the like counts of the FB pages served as an objective measure of interest frequency. The page likes were not restricted to certain domains and, thereby, provided a more representative sample of people’s interests.

Method

Participants and design

Similar to the previous studies, I aimed at collecting data from 200 participants. I collected data from 202 participants (105 female, 97 male) who were recruited via the MTurk online platform and participated for US$0.80.

The independent variable was page like counts, the dependent variable was likability of target persons, and interest page attitude ratings served as a control variable.

Procedure

The study used the Qualtrics software. The study’s description on MTurk specified the existence of an FB account as a requirement for participating in the study.

Instructions informed participants about confidentiality of the study and about their right to quit at any point. After answering demographical questions, participants were presented with a five-step instruction on how to find their 15 most recent FB page likes. Participants were then told to type the 15 most recent pages they had liked into provided text boxes. In addition, they were instructed to not enter pages that they can either not remember liking or that they do not actually like.

Similar to the previous studies, participants were then asked to provide likability ratings for imaginary target persons who shared one of their FB page likes, while this was repeated for all 15 pages. In a next step, participants were instructed how to look up the like counts for all 15 pages and to type them into 15 text boxes. Next, participants were asked about their attitudes toward the FB pages they had listed. For each page, they were asked, “How much do you like [name of page]?,” and given a 9-point scale that was labeled not at all (1), neutral (5), and very much (9).

At the end of the study, participants were thanked, debriefed, and given a code necessary to receive their compensation via MTurk.

Results

Prior to analysis, I checked the provided ratings, page names, and page counts for plausibility. If page counts looked odd, for example, when very similar numbers were provided, or when the numbers were very small or included zero, I looked up the provided FB pages of that participant and checked whether the numbers matched. This procedure led to a total exclusion of 24 participants. Specifically, one observation was removed because it stemmed from a participant who had already participated (same Internet Protocol address), seven participants did not provide liked FB pages, 15 participants provided unrealistic or no like counts, and one participant provided always the same ratings for all 15 likability and all 15 attitude ratings. The remaining dataset consisted of 178 participants.

Next, I analyzed descriptive statistics of the attitude ratings to ensure that participants held positive attitudes toward the pages they listed. There were missing page count values for five pages. It turned out that among the remaining total number of 2,665 pages, there were 322 pages that participants indicated to not have a positive attitude toward (rating <6). As these pages should not evoke interpersonal liking in the first place, they were excluded from the following analyses. However, results including these 322 observations were not greatly different from the ones following and are provided in the footnotes. The like count variable varied from 1 to more than 490 million; hence, its scaling was highly different from that of the dependent variable likability. As such large scaling differences do not allow model estimation in lmer4, I divided the like count variable by 10 million. Note that this does not change any of the interval relations between different count values.

Likability

I conducted linear mixed effects model analysis using the lme4 package in R. I specified a model that predicted likability by the fixed factor like count and that included subject as random factor with random error components for intercepts and slopes. This model revealed the predicted effect of like count on likability, b = −0.07 (β = −0.05), t(329.70) = −2.97, p = .003,² 95% CI = [–0.02, –0.06]. Regarding the random effects, there was significant intercept variation across participants, but no significant slope variation.

The like count variable included several outliers with extremely large numbers. To ensure that the obtained effects were not driven by these outliers, I capped like counts at the value corresponding to the upper limit of 1.5 times the interquartile range (Tukey, 1977). I then repeated the previous analyses and the effect of like count on likability remained significant and was even somewhat larger, b = −0.09 (β = −0.08), t(638.80) = −3.36, p < .001,³ 95% CI = [–0.12, –0.03]. Again, there was significant intercept variation across participants but no significant slope variation.

Attitude strength

I then specified a model that predicted attitude strength, that is, how much participants indicated to like the FB pages, from the fixed effect like counts and the random factor subject with random error components for intercepts and slopes. Similar to results from Study 2a, there was a positive relationship between attitude strength and attitude frequency (like counts), which was, however, only marginally significant, b = 0.03 (β = 0.03), t(137.80) = 1.65, p = .098,⁴ 95% CI = [–0.01, 0.07]. There was significant intercept variation across participants but no significant slope variation. When I included attitude strength to the previous model predicting likability from like counts (including all random intercept and slope components), the effect of like count on likability increased, b = −0.09 (β = −0.08), t(246.77) = −4.45, p < .001,⁵ 95% CI = [–0.11, –0.04].

Discussion

Participants expressed greater likability for imaginary target persons who shared rare rather than common FB page likes. The use of an objective frequency measure ruled out a possible influence of participants’ lay theories or demand characteristics. Instead, the present results suggest that sharing rare compared with common interests indeed elicits greater interpersonal attraction. Again, this effect was not due to a possible stronger attitude that participants may hold toward their rare interests. Descriptively, participants again expressed stronger attitudes toward their common attitudes. Consequently, when controlling for the influence of attitude strength, the influence of attitude frequency on target likability increased. In a final study, I tested the present predictions in the more applied context of online dating.

Study 4

Online dating services regularly rely on shared attitudes including interests when estimating the compatibility of dating partners. Tinder, for example, presents its users with their shared FB likes. In light of the present work, online dating services may be improved if more weight is given to rare shared attitudes. To test whether sharing rare interests elicit stronger dating preferences, I recruited singles for an alleged online dating study, in which they provided common and rare interests before being presented with profiles of matching dating partners. I measured two dependent variables relevant to online dating, namely, interpersonal attraction and intention to meet.

Method

Participants and design

I preregistered Study 4 (https://aspredicted.org/un4mu.pdf). The study was carried out as part of an experimental methods class assisted by three undergraduate students. I initially aimed at collecting data from 150 participants to achieve power >0.90 for a within-participants t test assuming a small to medium effect size. Hence, I preregistered Study 4 with a desired sample size of 150. However, as the online data collection went surprisingly fast, I decided to increase the sample size and collected data from 213 students of a large university (159 female, 52 male, 2 other) who participated online for course credit. Please note that analyzing only the first 150 participants did not substantially change any of the following results.

Study 4 featured two main dependent variables, interpersonal attraction and intention to meet the target person; both variables were measured for each participant. Attitude strength was measured as a control variable. The independent variable was frequency of interest, which was realized as a quasi-experimental manipulation within participants as each participant generated common and rare interests (cf. Study 2). I increased the number of interest domains from three (Studies 1, 2a, 2b) to six to further test the generalizability of the present results beyond specific interest domains (Westfall, Kenny, & Judd, 2014).

Procedure

The study used the Qualtrics software and was distributed among FB student groups of a large university. Among them were groups of psychology students and groups of students with other majors. Psychology students could earn course credit, students from other majors received no compensation. The study was advertised as an online dating study in which only singles could participate.

Study instructions informed participants that the aim of the study was to investigate the different factors that lead to interpersonal attraction in online dating. Participants were then informed about confidentiality of the study and about their right to quit at any point. After answering demographic questions, participants indicated whether they were primarily interested in dating a man or a woman.

Next, participants were asked to provide one common and one rare interest from the domains movies, musicians, hobbies, travel destinations, reads, and food/drinks. The order of the six domains was randomized. Instructions then informed participants that a large online dating database would now be searched for profiles of users who shared one of their interests.

Next, participants were told to wait a few seconds for the search process to be completed. After 10 s, the continue button was enabled and participants were guided to another instruction page. For each of their interests, participants were then presented with one profile of a person who allegedly shared this interest. The profiles consisted of either a male or a female silhouette along with the description, “the online dating database includes a Lisa. Lisa also likes [name of interest].” Whether participants saw male or female profiles was determined based on their indicated dating preferences. Each profile featured a distinct male or female name. Below each profile, there were four questions that participants were asked to answer on 9-point scales.

The first two questions addressed interpersonal attraction and asked, “How likable do you find Lisa?” (1 = not at all likable, 5 = neutral, 9 = extremely likable), and “How interesting do you find Lisa?” (1 = not at all interesting, 5 = neutral, 9 = extremely interesting). The other two questions addressed participants’ intentions to meet the target person and asked, “Would you be interested in getting to know Lisa?” (1 = not at all interested, 5 = neutral, 9 = extremely interested), and “How much would you like to meet Lisa?” (1 = not at all, 5 = neutral, 9 = extremely much). Participants answered these four questions for all 12 profiles, while each profile shared one of their interests.

Next, participants were asked to provide attitude strength ratings for each of their interests. Specifically, for each interest, they were asked, “How much do you like [name of interest]?” and provided ratings on a 9-point scale (1 = not at all, 5 = neutral, 9 = extremely).

Finally, participants were asked to indicate whether they indeed were single before they were debriefed and provided instructions on how to claim their course credit.

Results

Twenty-two participants indicated that they were not single at the moment. However, the following results remain almost unchanged if these participants are excluded from the analysis.

Interpersonal attraction

I first computed interpersonal attraction ratings by collapsing the two ratings, likability and interest, which were highly correlated across all stimuli and participants (r = .84). I then calculated mean interpersonal attraction ratings for targets sharing the rare interests and for targets sharing the common interests. As predicted, participants expressed significantly stronger interpersonal attraction for target persons who shared their rare compared with their common interests, M_rare = 5.58 (SD_rare = 1.46) versus M_common = 5.24 (SD_common = 1.26), t(212) = 4.91, p < .001, 95% CI = [0.20, 0.48], d_z = 0.34. As illustrated in Figure 6, the magnitude of this effect varied across the six interest domains.

Figure 6.

Mean interpersonal attraction ratings for target persons who share common or rare interests in the six interest domains.

Intention to meet

Next, I collapsed the two intention to meet items, which were also strongly correlated across all stimuli and participants (r = .94). I then repeated the previous analysis with the collapsed intention to meet measure. Participants expressed significantly stronger intention to meet target persons who shared their rare compared with their common interests, M_rare = 5.25 (SD_rare = 1.61) versus M_common = 4.92 (SD_common = 1.50), t(212) = 4.96, p < .001, 95% CI = [0.20, 0.46], d_z = 0.33. Again, the magnitude of this effect showed some variance across the different interest domains as illustrated in Figure 7.

Figure 7.

Mean intention to meet ratings for target persons who share common or rare interests in the six interest domains.

Attitude strength

Finally, I compared the rated attitude strength of common and rare interests. Replicating the finding from Study 2, participants expressed more liking for their common compared with their rare interests, M_common = 7.18 (SD_common = 1.31) versus M_rare = 6.87 (SD_rare = 1.35), t(212) = 4.44, p < .001, 95% CI = [0.17, 0.45], d_z = 0.31.

Regression analysis

As evident from Figures 6 and 7, there was variation in the differences between common and rare interests across the six different interest domains regarding the two crucial dependent variables, interpersonal attraction and intention to meet. These variations could reflect random stimulus variation; alternatively, the significant differences found when all domains are collapsed could reflect artifacts of individual domains. To test whether the present effects generalize across domains, I performed a series of linear mixed-model analyses that treated domains and participants as random factors (Judd, Westfall, & Kenny, 2017).

I first predicted interpersonal attraction from the fixed effect interest frequency (1 = common, –1 = rare) and the two random effects of domain and participants with random error components for intercepts and slopes. The analysis confirmed the significant effect of interest frequency on interpersonal attraction, b = −0.17 (β = −0.09), t(11.19) = −4.40, p = .001, 95% CI = [–0.14, –0.05], while there was significant variation across participants regarding intercepts and slopes, but no significant variation across interest domains, χ²(2) = 0.36, p = .834. Hence, the relation between interest frequency and interpersonal attraction generalizes across domains and was not an artifact of specific domains.

A similar model predicting intention to meet also confirmed the significant influence of interest frequency, b = −0.17 (β = −0.08), t(11.40) = −4.77, p < .001, 95% CI = [–0.12, –0.05]. Again, there was significant intercept and slope variation across participants but not across domains, χ²(2) = 0.04, p = .980. Thus, the relation between interest frequency and intention to meet also generalizes across the different domains.

In a final step, I aimed at determining the influence of interest frequency on the dependent variables beyond the influence of attitude strength. I therefore added the attitude strength variable to both models as a fixed effect. I first tested for the relation between attitude strength and the two dependent variables. I specified two models of which one predicted interpersonal attraction and the other predicted intention to meet by attitude strength and the random effects of domain and participants with random error components for intercepts and slopes. The analysis revealed that attitude strength strongly predicted interpersonal attraction, b = 0.47 (β = 0.48), t(16.62) = 16.01, p < .001, 95% CI = [0.42, 0.53], as well as intention to meet, b = 0.44 (β = 0.41), t(24.08) = 15.13, p < .001, 95% CI = [0.36, 0.46]. Thus, how much participants liked a given interest strongly determined liking for a person sharing that interest. Consequently, when attitude strength was added to the previous models, the influence of interest frequency on interpersonal attraction increased, b = −0.24 (β = −0.13), t(8.50) = −6.94, p < .001, 95% CI = [–0.17, –0.09], and so did the influence of interest frequency on intention to meet, b = −0.23 (β = 1.12), t(8.10) = −7.45, p < .001, 95% CI = [–0.15, –0.09]. Hence, when controlling for attitude strength, interest frequency is an even stronger predictor of interpersonal attraction and intention to meet.

Discussion

Study 4 tested the influence of interest frequency on dating preferences in an online dating setting. Results confirmed that sharing rare compared with common interests elicits stronger interpersonal attraction and intention to meet with a target person. Again, this effect occurred even though participants indicated to like their common interests better than their rare interests. In addition, Study 4 shows that the association between attitude frequency and likability generalizes across several different interest domains.

General Discussion

The present research identified a robust moderator of the similarity effect. People like others who share their attitudes, particularly when these attitudes are not shared by many others. Study 1 asked participants to name several interests from different domains and judged likability of target persons was a function of judged frequency of the shared interests. Studies 2a and 2b replicated this effect in a quasi-experimental design in which participants named common and rare interests before judging likability of imaginary target persons sharing their interests. Ruling out possible confounds, two other groups of participants expressed weaker liking for their rare interests and also found them less important than their common interests. Study 3 used an objective frequency measure (FB likes) and found that participants liked target persons better who share rare compared with common FB page likes. Finally, Study 4 tested the influence of attitude frequency in the more applied context of online dating. Participants expressed greater interpersonal attraction as well as greater intention to meet for target profiles that shared rare compared with common interests.

The moderating role of attitude frequency can be predicted from different theoretical perspectives that are not mutually exclusive. From an information integration point of view, rare attitudes should receive greater weighting in impression formation because they are more informative (Fiske, 1980). From a motivational perspective (Byrne, 1971), rare attitudes may elicit an especially strong need for social validation. From an information-inference perspective (Ajzen, 1974; Kaplan & Anderson, 1973; Tesser, 1971), people may draw more inferences about a person’s additional characteristics when that person shares a rare compared with a common attitude.

Given the different theoretical perspectives’ consensus in predicting the bonding potential of rare attitudes, it is likely that different processes simultaneously contribute to the present findings. While I did not aim at differentiating different process-level explanations for the moderating role of interest frequency in the present work, future research may address to what extent it reflects information weighting, inference processes, or motivational factors.

Beyond the question of underlying psychological processes, the present findings also relate more broadly to a biological perspective on the relation between similarity and attraction. Recently, Bahns and colleagues (2017) provided an interpretation of the similarity–attraction effect as reflecting social niche construction. Accordingly, people not only reactively adapt to their social environment but also shape their environment according to their traits and values and thereby construct their own social niches. People select similar others as their peers and mates, which enhances cooperation, reciprocity, and coalition formation. Following this perspective, people’s rare attitudes are most challenging for niche construction because it is difficult to find matching others. Therefore, finding another person who shares rare attitudes is especially valuable for the construction of one’s social niche.

Constraints on Generality

Several constraints may limit the conclusions that can be drawn from the present work. First, while I used different methodological approaches and measures to test my predictions, the current work does not include an experimental manipulation in the strict sense. The experimental designs were either correlational or quasi-experimental, which limits the inferences that can be drawn regarding a causal influence of attitude frequency on the similarity–attraction effect.

Second, the current studies did not measure actual interpersonal attraction among individuals but rather imagined attraction. That is, participants were always asked to imagine they would meet a stranger who shared one of their interests and to indicate how likable they would find this stranger. Study 4 was an exception as it presented participants with allegedly real dating profiles, but participants never actually met a real person.

Third, I operationalized attitudes as interests because they are often the subject of interaction when meeting strangers and are frequently shared on dating profiles. However, attitudes are not limited to interests. That is, interests refer to likes, but attitudes also include dislikes. Although the different theoretical accounts that predict attitude frequency to moderate the similarity–attraction effect would make the same prediction for shared dislikes as well, the present work did not test this.

Fourth, the question remains whether the current findings can be generalized to the domain of shared attributes. According to the theoretical perspectives presented in this work, the frequency of a given attribute should also influence the degree of interpersonal attraction among people who share that attribute. However, research has shown that people’s rare attributes are often negative ones (Alves, Koch, & Unkelbach, 2016, 2017, 2018). Hence, it is possible that the negativity of rare attributes overrules the otherwise boosting effect of rarity in interpersonal attraction.

Finally, the present work does not rule out that there are other characteristics of shared attitudes that lead to a particularly strong sense of interpersonal attraction. Another possible moderator may be the variability of attitudes in the population. That is, there are some attitude objects that people either love or hate (e.g., political preferences) while others do not cause such strong polarization (e.g., liking for ice cream). Sharing an attitude regarding a polarized attitude object may elicit stronger attraction. Future research should investigate the role of attitude variability and other possible moderators of the similarity–attraction effect to arrive at a more complete picture.

Implications

Despite the mentioned limitations, the current findings have several important implications for impression formation processes and their applications. Social media platforms such as online dating services might consider attitude frequency as a way to improve their matching algorithms. Greater initial interpersonal attraction can be expected when people are matched based on their rare compared with their frequent attitudes. As attitude similarity is also predictive of relationship length and quality (e.g., Cupach & Metts, 1995; Gonzaga, Campos, & Bradbury, 2007), future research should address a possible influence of attitude frequency regarding these long-term outcomes as well. Finally, frequency of shared attitudes can be expected to influence other outcomes related to interpersonal attraction such as trust, compliance, or purchase intentions (Burger et al., 2004; Jiang et al., 2009).

The greater attraction resulting from rare attitudes poses a strategic challenge for people’s impression management. When meeting strangers (e.g., during a date), should one reveal his or her rare attitudes or should one focus on expressing common attitudes? The present research suggests that there are two strategies that a person can apply in this situation. The safe strategy would be to express common attitudes because the probability that the other person shares these attitudes is high, but the effect on attraction is weak. The more risky strategy is to express rare attitudes, because the sharing probability is low but the attraction effect is strong. Different situations may call for different strategies. When the negative consequences of being disliked are stronger than the positive consequences of being liked, one should apply the safe strategy. In contrast, when it does not matter whether one is disliked or merely facing indifference, that is, when only strong attraction matters (searching a romantic partner), one should apply the riskier strategy of revealing rare attitudes. If two dating partners match regarding a number of rare attitudes, the present work suggests that they are very likely to see each other again.

From an interpersonal difference perspective, people’s general preferences for common or rare attitudes may have implications for their social relations. A person who holds many minority attitudes and has a number of exotic interests may have less but deeper friendships than someone who always holds common attitudes.

Conclusion

Attitude similarity is one of the strongest determinants of interpersonal attraction. But attitudes systematically differ regarding their potential to elicit attraction as a function of their prevalence. Even though people do not feel as strongly about their rare compared with their common attitudes, they particularly like other people who share their rare attitudes.

Supplementary Material

Supplementary Material, alves_online_appendix – Sharing Rare Attitudes Attracts

Supplementary Material, alves_online_appendix for Sharing Rare Attitudes Attracts by Hans Alves in Personality and Social Psychology Bulletin

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Supplemental Material

Supplementary material is available online with this article.

References

Ajzen

(1974). Effects of information on interpersonal attraction: Similarity versus affective value. Journal of Personality and Social Psychology, 29, 374-380.

Alves

Koch

Unkelbach

(2016). My friends are all alike—The relation between liking and perceived similarity in person perception. Journal of Experimental Social Psychology, 62, 103-117.

Alves

Koch

Unkelbach

(2017). The “common good” phenomenon: Why similarities are positive and differences are negative. Journal of Experimental Psychology: General, 146, 512-528.

Alves

Koch

Unkelbach

(in press). An ecological explanation of intergroup biases. Psychological Science.

Aronson

Worchel

(1966). Similarity versus liking as determinants of interpersonal attractiveness. Psychonomic Science, 5, 157-158.

Bahns

A. J.

Crandall

C. S.

Gillath

Preacher

K. J.

(2017). Similarity in relationships as niche construction: Choice, stability, and influence within dyads in a free choice environment. Journal of Personality and Social Psychology, 112, 329-355.

Baumeister

R. F.

Leary

M. R.

(1995). The need to belong: Desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117, 497-529.

Berger

C. R.

(1973). Task performance and attributional communication as determinants of interpersonal attraction. Communications Monographs, 40, 280-286.

Bosson

J. K.

Johnson

A. B.

Niederhoffer

Swann

W. B.

(2006). Interpersonal chemistry through negativity: Bonding by sharing negative attitudes about others. Personal Relationships, 13, 135-150.

10.

Burger

J. M.

Messian

Patel

del Prado

Anderson

(2004). What a coincidence! The effects of incidental similarity on compliance. Personality and Social Psychology Bulletin, 30, 35-43.

11.

Byrne

(1961). Interpersonal attraction and attitude similarity. The Journal of Abnormal and Social Psychology, 62, 713-715.

12.

Byrne

(1971). The attraction paradigm. New York, NY: Academic Press.

13.

Byrne

Blaylock

(1963). Similarity and assumed similarity of attitudes between husbands and wives. The Journal of Abnormal and Social Psychology, 67, 636-640.

14.

Cialdini

(1993). The psychology of influence. New York, NY: William Morrow.

15.

Cialdini

R. B.

Goldstein

N. J.

(2004). Social influence: Compliance and conformity. Annual Review of Psychology, 55, 591-621.

16.

Cohen

(1992). A power primer. Psychological Bulletin, 112, 155-159.

17.

Cupach

W. R.

Metts

(1995). The role of sexual attitude similarity in romantic heterosexual relationships. Personal Relationships, 2, 287-300.

18.

Festinger

(1954). A theory of social comparison processes. Human Relations, 7, 117-140.

19.

Festinger

(1957). A theory of cognitive dissonance. Oxford, UK: Row, Peterson.

20.

Festinger

Schachter

Back

K. W.

(1950). Social pressures in informal groups. New York, NY: Harper.

21.

Finch

J. F.

Cialdini

R. B.

(1989). Another indirect tactic of (self-) image management: Boosting. Personality and Social Psychology Bulletin, 15, 222-232.

22.

Fiske

S. T.

(1980). Attention and weight in person perception: The impact of negative and extreme behavior. Journal of Personality and Social Psychology, 38, 889-906.

23.

Gonzaga

G. C.

Campos

Bradbury

(2007). Similarity, convergence, and relationship satisfaction in dating and married couples. Journal of Personality and Social Psychology, 93, 34-48.

24.

Heider

(1958). The psychology of interpersonal relations. New York, NY: John Wiley.

25.

Jamieson

D. W.

Lydon

J. E.

Zanna

M. P.

(1987). Attitude and activity preference similarity: Differential bases of interpersonal attraction for low and high self-monitors. Journal of Personality and Social Psychology, 53, 1052-1060.

26.

Jiang

Hoegg

Dahl

D. W.

Chattopadhyay

(2009). The persuasive role of incidental similarity on attitudes and purchase intentions in a sales context. Journal of Consumer Research, 36, 778-791.

27.

Judd

C. M.

Westfall

Kenny

D. A.

(2017). Experiments with more than one random factor: Designs, analytic models, and statistical power. Annual Review of Psychology, 68, 601-625.

28.

Kaplan

M. F.

Anderson

N. H.

(1973). Information integration theory and reinforcement theory as approaches to interpersonal attraction. Journal of Personality and Social Psychology, 28, 301-312.

29.

Kelman

H. C.

(1958). Compliance, identification, and internalization three processes of attitude change. Journal of Conflict Resolution, 2, 51-60.

30.

Lewis

K. N.

Walsh

W. B.

(1980). Effects of value-communication style and similarity of values on counselor evaluation. Journal of Counseling Psychology, 27, 305-314.

31.

Montoya

R. M.

Horton

R. S.

(2013). A meta-analytic investigation of the processes underlying the similarity-attraction effect. Journal of Social and Personal Relationships, 30, 64-94.

32.

Montoya

R. M.

Horton

R. S.

Kirchner

(2008). Is actual similarity necessary for attraction? A meta-analysis of actual and perceived similarity. Journal of Social and Personal Relationships, 25, 889-922.

33.

Newcomb

T. M.

(1943). Personality and social change. New York, NY: Dryden Press.

34.

Newcomb

T. M.

(1961). The acquaintance process. New York, NY: Holt, Reinhart, & Winston.

35.

Reeder

G. D.

Brewer

M. B.

(1979). A schematic model of dispositional attribution in interpersonal perception. Psychological Review, 86, 61-79.

36.

Rosenbaum

M. E.

(1986). The repulsion hypothesis: On the nondevelopment of relationships. Journal of Personality and Social Psychology, 51, 1156-1166.

37.

Skowronski

J. J.

Carlston

D. E.

(1987). Social judgment and social memory: The role of cue diagnosticity in negativity, positivity, and extremity biases. Journal of Personality and Social Psychology, 52, 689-699.

38.

Steele

M. E.

McGlynn

R. P.

(1979). Effects of affective value and similarity on attraction. The Journal of Social Psychology, 108, 111-119.

39.

Stevens

Owens

Schaefer

E. C.

(1990). Education and attractiveness in marriage choices. Social Psychology Quarterly, 62-70.

40.

Tan

D. T. Y.

Singh

(1995). Attitudes and attraction: A developmental study of the similarity-attraction and dissimilarity-repulsion hypotheses. Personality and Social Psychology Bulletin, 21, 975-986.

41.

Tesser

(1971). Evaluative and structural similarity of attitudes as determinants of interpersonal attraction. Journal of Personality and Social Psychology, 18, 92-96.

42.

Tukey

J. W.

(1977). Exploratory data analysis. Reading, MA: Addison-Wesley.

43.

Weaver

J. R.

Bosson

J. K.

(2011). I feel like I know you: Sharing negative attitudes of others promotes feelings of familiarity. Personality and Social Psychology Bulletin, 37, 481-491.

44.

Westfall

Kenny

D. A.

Judd

C. M.

(2014). Statistical power and optimal design in experiments in which samples of participants respond to samples of stimuli. Journal of Experimental Psychology: General, 143, 2020-2045.

45.

Wooten

D. B.

Reed

(1998). Informational influence and the ambiguity of product experience: Order effects on the weighting of evidence. Journal of Consumer Psychology, 7, 79-99.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.60 MB