Meta-Accuracy and Relationship Quality: Weighing the Costs and Benefits of Knowing What People Really Think About You

Abstract

People use metaperceptions, or their beliefs about how other people perceive them, to initiate and maintain social bonds. Are accurate metaperceptions associated with higher quality relationships? In four studies, the current research answers this question but considers the possibility that the self might not experience the same relational benefits of accurate metaperceptions, or meta-accuracy, as the people who form judgments about the self. For example, people tend to like individuals who have accurate self-perceptions, yet individuals tend to enjoy their own relationships more with people they believe see them in desirable ways. To test whether meta-accuracy is linked to relationship quality and whether this link differs for the self and others, meta-accuracy for personality traits as well as metaperceiver- and judge-reported relationship quality were assessed among new acquaintances (N = 184), peers (N = 228), friends (N = 273), and romantic partners (N = 401). Results suggested that judges enjoyed relationships more with metaperceivers who knew the impression they made, regardless of whether judges’ impressions were desirable (i.e., positive or self-verifying). Initial meta-accuracy also predicted greater relationship quality over time, suggesting that accurate metaperceptions have positive effects on relationships. In contrast, rather than enjoying relationships more when they were accurate, metaperceivers enjoyed relationships more when they believed judges perceived them in positive or self-verifying ways. Thus, meta-accuracy seems to be a virtue in the eyes of judges, but metaperceivers do not seem to reap the same benefits of knowing what others really think. Implications for improving meta-accuracy are discussed.

“True friendship can afford true knowledge. It does not depend on darkness and ignorance.” (Henry David Thoreau)

“If most of us remain ignorant of ourselves, it is because self-knowledge is painful and we prefer the pleasure of illusion.” (Aldous Huxley)

People are inherently interested in what other people think about them. Arguably, one of the many reasons why people form metaperceptions, or beliefs about how others perceive them, is to initiate and maintain relationships. Indeed, people’s metaperceptions shape the ways in which they present themselves to others (Schlenker & Weigold, 1992) and are used to monitor the status of their social bonds (Leary, Tambor, Terdal, & Downs, 1995; Srivastava & Beer, 2005). Thus, metaperceptions are the implicit map people use to develop and preserve relationships with other people.

Given the pivotal role of metaperceptions in relationships, a logical assumption is that accurate metaperceptions, or meta-accuracy (Kenny & DePaulo, 1993), is linked to higher quality relationships. Take the example of Meg, a metaperceiver, and Jon, a judge of Meg’s personality. Meta-accuracy allows Meg to enjoy her good reputation if Jon forms a desirable impression of her, but meta-accuracy also affords her an opportunity to take corrective action if Jon sees her in an undesirable way. In contrast, if Meg fails to realize that Jon sees the best in her, she might react in ways that undermine their relationship, or if she fails to realize Jon sees her in an undesirable way, she might continue to engage in behavior that exacerbate Jon’s negative view of her.

It seems intuitive that, as Henry David Thoreau suggests, insight into what acquaintances really think might be the foundation of high quality relationships. Yet, link between meta-accuracy and relationship outcomes has not been systematically investigated, and there are reasonable arguments that meta-accuracy might not foster relationship quality. A large body of work suggests that people prefer self-enhancing or self-verifying information about themselves rather than accurate information (Sedikides, 1993). As such, rather than knowing what other people really think, people’s evaluations of their relationships might hinge on the degree to which they feel valued or understood. Metaperceptions powerfully shape how people feel about their relationships and relationship partners (e.g., Lemay & Dudley, 2009; Murray, Holmes, MacDonald, & Ellsworth, 1998), and assuming the best might be more important than knowing the truth. As Aldous Huxley suggests, self-knowledge can be painful, and knowing what other people really think of us might be difficult to accept.

The main goal of the current research is to empirically test whether meta-accuracy for core personality traits is associated with relationship quality. To do so, the link between meta-accuracy and quality is assessed among new acquaintances, peers, friends, and romantic partners. To test whether the benefits of meta-accuracy are experienced equally within a relationship, both metaperceiver- and judge-reports of relationship outcomes are measured in each study. To my knowledge, the current research is the first comprehensive investigation of if, when, and for whom meta-accuracy is linked to higher quality relationships. In addition to answering the basic question of whether meta-accuracy is linked to positive relationship outcomes, results also have implications for whether this form of self-knowledge should be improved.

The Paradoxical Link Between Meta-Accuracy and Relationship Quality

There are reasons to predict a paradoxical relationship between meta-accuracy and relationship quality such that meta-accuracy has positive effects on other people but might come at a cost to the self. For example, Jon likely enjoys Meg when she knows how he sees her, regardless of whether his impression is positive or negative, whereas Meg likely enjoys Jon more when she thinks he sees the best in her. The basic prediction that the self and others experience consequences of accuracy differently is supported by research exploring the consequences of other types of perceptual accuracy. For example, research exploring the degree to which people see themselves as others do (i.e., self-other agreement) tends to show that people whose self-perceptions of their personality align with their peers’ perceptions of them are liked more and are seen as more psychologically adjusted (Colvin, Funder & Block, 1995; Human, Sandstrom, Biesanz, & Dunn, 2012); however, self-enhancers, or people who see themselves more positively than others do, tend to like themselves more and report being more psychologically adjusted (Kurt & Paulhus, 2008; Paulhus, 1998). Self-other agreement indexes the relationship between self-views and others’ perceptions whereas meta-accuracy explicitly tests people’s understanding of how a specific person experiences them; thus, self-other agreement and meta-accuracy are distinct phenomena (Carlson, Vazire, & Furr, 2011). However, the same paradox might be observed for meta-accuracy, such that the adaptiveness of accurate metaperceptions depends on who reports on relationship quality, the judge or the metaperceiver.

The Accuracy Hypothesis: Judges Enjoy Accurate Metaperceivers

The first prediction made by the Accuracy Hypothesis is that judges enjoy relationships more with metaperceivers who know how they are perceived (i.e., meta-accurate). The Accuracy Hypothesis is based on a growing body of research showing that people tend to like individuals who have insight into their behavior. Meta-accuracy requires knowledge of one’s own behavior (Carlson & Kenny, 2012; Kenny & DePaulo, 1993), and as such, meta-accuracy might have similar interpersonal consequences as self-knowledge of behavior. People tend to like individuals more who know how they behave in everyday life, a finding that remains when controlling for the positivity of that individual’s personality (Tenney, Vazire, & Mehl, 2013). Likewise, compared to people who do not, people who do make explicit disclaimers about their flaws are liked more, suggesting that self-knowledge can attenuate the negative effects of a bad impression (Ward & Brenner, 2006). This work suggests that self-knowledge can be a virtue in the eyes of others, regardless of whether that knowledge is about desirable attributes. Given the links between self-knowledge of behavior and meta-accuracy (Carlson & Kenny, 2012; Kenny & DePaulo, 1993), judges might also like people who are aware of the impressions they make, regardless of how desirable those impressions are. Indeed, meta-accuracy might explain the observed relationships between self-knowledge of behavior and liking, given that acknowledging flaws requires some awareness of how others might experience a behavior in a negative way.

In sum, people tend to enjoy individuals who have self-knowledge of their behavior, regardless of whether that knowledge is about positive or negative attributes. Thus, the Accuracy Hypothesis predicts that judges enjoy relationships more with metaperceivers who know how they are seen, or who share their social reality. This hypothesis also predicts that judge-reports of relationship quality will be linked to meta-accuracy, regardless of whether judges’ impressions are desirable. These predictions are outlined in Table 1.

Table 1.

Predictions for the Accuracy and Subjective Reality Hypotheses

Accuracy Hypothesis	Subjective Reality Hypothesis
Judges enjoy relationships with people who have insight into how judges perceive their characteristic pattern of traits (meta-accuracy), regardless of whether impressions are positive or self-verifying.	Metaperceivers enjoy relationships with people they think see them in positive or self-verifying ways.
Role of Positive Impressions
Judges enjoy relationships with people who have insight into the distinctive and potentially negative impressions they make (distinctive meta-accuracy).	Metaperceivers enjoy relationships with people they think see them in positive ways (positivity).
Role of Self-Verifying Impression
Judges enjoy relationships with people who have insight into the ways in which judges see them differently from how they see themselves (meta-insight).	Metaperceivers enjoy relationships more with people they believe see them as they see themselves (transparency).

Note. Bold and italic terms refer to effects that are modeled in Models 1–3. Please see the supplemental materials for more detail about the modeling of effects.

The Subjective Reality Hypothesis: Metaperceivers Enjoy Feeling Valued and Understood

There are several reasons to predict that metaperceivers’ relationship quality depends on whether they believe they make desirable impressions rather than on knowing what other people really think of them, a prediction made by the Subjective Reality Hypothesis. There are at least two pathways by which this might occur: positivity, which is the belief that a judge perceives the self in positive ways (i.e., feeling valued), and transparency, which is the belief that a judge shares one’s self-perception (i.e., feeling understood). As such, the Subjective Reality Hypothesis makes two distinct predictions about positivity and transparency.

The Positivity Hypothesis: Metaperceivers prefer to feel valued

People typically expect and desire to be seen in positive ways (Carlson & Kenny, 2012; Hepper, Hart, Gregg, & Sedikides, 2011) and are happier in romantic relationships when their partner sees the best in them or idealizes them (Boyes & Fletcher, 2007; Murray, Holmes, & Collins, 2006; Murray, Holmes, & Griffin, 1996). Given the importance of positive regard, metaperceivers’ reports of relationship quality might hinge on whether they think others see them in a positive light rather than knowing what others actually think. Indeed, people who expect to be seen in positive ways by their romantic partner report greater satisfaction in their relationship, whereas questioning a partner’s positive regard fosters negative evaluations towards one’s partner (Murray et al., 1996, 1998). Thus, Meg’s meta-accuracy about a positive impression likely fosters positive feelings towards Jon but her meta-accuracy about a negative impression likely comes at the cost of her enjoying Jon less. Given the importance of positive expectations, the Positivity Hypothesis predicts that, for metaperceivers, greater relationship satisfaction is associated with positive metaperceptions but not with accuracy about a negative impression.

The Transparency Hypothesis: Metaperceivers prefer to feel understood

Self-verification theory argues that people prefer relationships with people who see them as they see themselves because feeling and being understood eliminates unrealistic expectations and attenuates fears of being rejected for hidden flaws that eventually come to light (Carnelley, Ruscher, & Shaw, 1999; Kwang & Swann, 2010). Indeed, people generally expect to be seen as they see themselves (Kenny & DePaulo, 1993) and are often more satisfied in relationships when they feel understood (Murray, Holmes, Bellavia, Griffin, & Dolderman, 2002; Weger, 2005). A large body of research supports the prediction that people enjoy their relationships more with people who share their self-views but are less committed to and tend to discontinue relationships with people who do not share their self-views (Swann & Ely, 1984; Swann, Hixon, & De La Ronde, 1992). Self-verification is operationalized as self-other agreement, but some of its positive effects are likely explained by a subjective sense of feeling understood, specifically a strong relationship between self-perceptions and metaperceptions. Thus, the Transparency Hypothesis predicts that metaperceivers enjoy relationships when they think they are seen as they see themselves but not when they realize they are seen differently than how they see themselves.

In sum, people are strongly motivated to believe that others see the best in them or see them as they see themselves. As outlined in Table 1, rather than knowing what others actually think, the Subjective Reality Hypothesis predicts that metaperceivers likely enjoy their relationships when they think they are seen in positive or in self-verifying ways. This hypothesis also predicts that metaperceivers’ relationship quality will not be linked to accuracy for less positive or less self-verifying impressions; that is, accuracy will be beneficial to metaperceivers to the extent to which impressions are actually positive or self-verifying.

Level of Acquaintanceship

Past work exploring other types of perceptual accuracy suggests that accuracy varies across levels of acquaintanceship (e.g., Funder, Kolar, & Blackman, 1995; Vazire, 2010; Watson, Hubbard, & Wiese, 2000). For example, self-other agreement tends to be weaker among friends and dating partners than among spouses (Watson et al., 2000). Furthermore, the effect of accuracy also varies across contexts and over time. For example, self-other agreement among peers becomes a more important factor in relationship quality over time, suggesting that initially sharing the same social reality has positive, cascading effects on relationships (Human et al., 2012; Paulhus, 1998). In romantic relationships, the link between positively biased relationship evaluations and relationship quality decreases over time, suggesting that initial effects of positivity can fade (Fletcher & Kerr, 2010), whereas being understood (i.e., self-verification) seems to be more important in established relationships than in new relationships, suggesting that transparency can become more important over time (Campbell, Lackenbauer, & Muise, 2006; Letzring & Noftle, 2010; Swann, De La Ronde, & Hixon, 1994). To determine if effects of accuracy, positivity and transparency also vary across levels of acquaintanceship, the current research tests hypotheses among new acquaintances (Study 1), peers who have known each other for a few months (Study 2), friends who have known each other for several years (Study 3), and romantic partners who have known each other for a couple of years (sample 1) and nearly 30 years (sample 2; Study 4). Going further, to explore whether accuracy, positivity and transparency have longitudinal consequences on relationship outcomes, peers’ perceptions in Study 2 are assessed twice, specifically after a first impression and again after a few months of acquaintanceship. This cross-sectional and longitudinal approach will reveal if effects generalize across levels of acquaintanceship. Furthermore, the four studies also represent different types of social contexts, which tests the degree to which effects generalize across types of relationships.

Research Overview

Rather than asking if meta-accuracy is linked to higher quality relationships, the current research explores for whom meta-accuracy has positive effects by measuring both judge- and metaperceiver-reports of relationship quality. In each study, meta-accuracy is indexed as the profile correlation between an individual’s metaperception for core personality traits (i.e., the Big Five) and the actual impression he or she makes on another person. This indicates whether, for example, Meg knows how Jon perceives her characteristic patterns of traits (e.g., does she realize that he sees her as more outgoing than dependable and more kind than anxious?). To test the general prediction that judges enjoy people who form accurate metaperceptions regardless of whether their insight is about positive or self-verifying impressions (Accuracy Hypothesis), meta-accuracy is also indexed when controlling for positivity and self-perceptions, respectively. To test the general prediction that metaperceivers enjoy relationships with people they feel see them in desirable ways (Subjective Reality Hypothesis), positivity (i.e., the degree to which people think they make a positive impression) and transparency (i.e., the degree to which people overestimate how much others share their self-perception) are indexed, controlling for actual levels of meta-accuracy. Taken together, this research provides an ecologically valid and robust test of the Accuracy and Subjective Reality Hypotheses and will reveal if and when it is adaptive to know what people really think about us.

Study 1: New Acquaintances

Study 1 tests the hypotheses in a first impression context where participants were randomly assigned to meet a new acquaintance, an approach that eliminates the possibility that effects are driven by self-initiated relationships. Furthermore, this design likely increases the variability in the positivity of impressions given that participants did not select their judge (Leising, Erbs, & Fritz, 2001).

Method

Participants

Undergraduate participants (N = 184; M_age = 19.56, SD = 1.17; 33% male; 64% Caucasian, 21% Asian-American, 10% Black, 2% Latin-American, 1%, Native American, .5% Middle Eastern; 2% other) were awarded course credit for participation. While a total of 218 participants completed the first impression activity, those who knew each other prior to the study were dropped from the analyses.¹ As part of a larger study, participants came to the lab in unacquainted dyads and provided self-perceptions of their personality in separate rooms. A research assistant (RA) escorted participants to an interaction room where they were instructed to talk about whatever they wished for five minutes. The RA left the room for the conversation, and after five minutes, returned participants to their original rooms where they provided metaperceptions, impressions of their partner’s personality, and liking ratings.

Measures

Self-perceptions (“I see myself as someone who…”), impressions (“I believe that my interaction partner is someone who…”), and metaperceptions (“Person X sees me as someone who…”) were provided for the 44-item Big Five Inventory (BFI; John & Srivastava, 1999) using a 1 (strongly disagree) to 7 (agree strongly) scale. These 44-item profiles were the basis for the meta-accuracy analyses described below. As found in past work (e.g., Wood & Furr, 2015), the normative impression profile for the BFI was associated with ratings of social desirability (r = .92, p < .001), which were reliably coded by 12 research assistants who rated each item (i.e., “How socially desirable is it to have this trait?”), ICC(2,1) = .808, ICC(2,12) = .981. Liking was rated on a 1 (not at all) to 7 (extremely) scale on the following two items: “How much do you like your partner?” and “Would you like to get to know your partner after today’s session?” (M = 4.52, SD = 1.01, α = .79; self-other agreement r = .23, p = .03).

Analyses

Models were based on the Social Accuracy Model (SAM; Biesanz, 2010), an approach that indexes profile agreement in a multilevel model. Profile agreement for each pair of metaperceivers and judges was modeled at level 1, and an overall index of the typical person’s meta-accuracy was modeled at level 2 (β₁₀). This model also controls for the fact that each participant rated and was rated by his or her partner. All effects were modeled in R (lme4 package; Bates & Sarkar, 2007). Please see the supplemental materials for specific details.

Model 1: Meta-accuracy

In Model 1, meta-accuracy was indexed as the profile correlation between an individual’s metaperception for core personality traits (i.e., the Big Five) and the actual impression he or she makes on another person. This indicated whether Meg knows how Jon perceives her characteristic patterns of traits (e.g., if she realizes that he sees her as more outgoing than dependable and more kind than anxious). This approach does not index mean-level differences between profiles (i.e., differences in elevation).

Profile meta-accuracy alone cannot test the Accuracy and Subjective Reality Hypotheses because the index often conflates accuracy with positivity (i.e., normative impressions) and transparency (i.e., self-perceptions; Carlson & Furr, 2013; Gallrein et al., 2013; 2016). For this reason, both judge- and metaperceiver-reports of relationship quality can be linked to raw meta-accuracy, but for different reasons. To provide a more rigorous test of the hypotheses, effects of positivity and transparency were separated from effects of meta-accuracy in two additional models.

Model 2: Identifying the role of positivity

In keeping with past work, a positive impression is operationalized as the average impression profile of the sample (i.e., the normative profile), which is highly socially desirable and yields nearly identical results when replaced with item social desirability ratings (Biesanz, 2010 Biesanz, 2010; Borkenau & Zaltauskas, 2009; Human et al., 2012; Lorenzo, Biesanz, & Human, 2010; Wood & Furr, 2015). Please see the supplemental materials for results when the normative profile was replaced by the social desirability profile. Isolate the roles of accuracy and positivity, the normative impression profile was removed from each judge’s impression profile to generate a distinctive impression profile for a given judge (i.e., Jon’s belief about what makes Meg different from the average person). Distinctive impression profiles and the normative impression profile were then simultaneously regressed onto metaperception profiles. The level 2 distinctive impression slope (β₁₀) indexed distinctive meta-accuracy for the typical person (i.e., the degree to which people detected what made their characteristic pattern of traits distinctive in their partner’s eyes) and the normative slope (β₂₀) indexed positivity (i.e., the degree to which people thought their partner saw them in normative ways, above and beyond the distinctive impression they made). Given the socially desirable properties of the normative profile, distinctive impressions are devoid of social desirability and are often neutral in valence (Wood & Furr, 2015), but there is some evidence that profile agreement among distinctive impressions can reflect a shared understanding of potentially negative characteristics (Leising et al., 2001). Thus, distinctive meta-accuracy reflects an understanding of the neutral but also potentially negative impressions ones makes. As shown in Table 1, evidence for the Accuracy Hypotheses will be observed if judges enjoy metaperceivers higher in distinctive meta-accuracy, and evidence for the Positivity Hypothesis, a component of the Subjective Reality Hypothesis, will be observed if metaperceivers enjoy their relationships more with people they think see them in a positive light.

Model 3: Identifying the role of transparency

Self-perceptions are a primary source of information people use when forming metaperceptions (Carlson & Kenny, 2012; Kenny & DePaulo, 1993). Thus, a link between relationship quality and meta-accuracy might also be driven by the fact that people who are in high quality relationships feel understood. To tease apart effects of accuracy from self-perception, Model 3 simultaneously regressed judges’ impression and metaperceivers’ self-perception onto metaperceptions. The level 2 impression slope (β₁₀) indexed meta-insight for the typical person (i.e., the degree to which people realized how their partner perceived their pattern of traits differently then how they saw their own pattern of traits) and the self-perception slope (β₂₀) indexed transparency (i.e., the degree to which people overestimated the degree to which their partner saw them as they saw themselves). Please see the supplemental materials for results in an alternative model that controls for both positivity and self-perceptions in a single model.

For each model, grand-mean centered judge-rated liking of the metaperceiver or metaperceiver-rated liking of the judge was entered as a level 2 predictor, or cross-level moderator, of slopes. These interaction terms test the Accuracy and Subjective Reality Hypotheses. Simple effects for significant interactions are reported for 1 standard deviation (SD) above and below the mean level of liking and reflect profile agreement for high or low liking. For example, if metaperceivers’ liking moderated meta-accuracy, a low b represents the strength of meta-accuracy when liking is one SD below the mean and a high b represents the strength of meta-accuracy when liking is one SD above the mean. While liking is modeled as a moderator of profile correlations (i.e. slopes), effects are interpreted as associations, specifically in terms of whether accuracy, positivity, or transparency are associated with liking (Human et al., 2012). Conceptually, this approach is akin to exporting each individual’s slope (e.g., meta-accuracy) as a profile correlation score and correlating those scores with each person’s liking ratings. Thus, the discussion of results in the current study and in Studies 2–4 refers to moderation results in terms of associations.

Results

As shown in Table 2, people were aware of how a new acquaintance perceived their characteristic pattern of traits (meta-accuracy), what made them distinctive in the eyes of their new acquaintance (distinctive meta-accuracy), and how their new acquaintance perceived their characteristic pattern of traits differently from how they perceived their own pattern of traits (meta-insight). The typical person also assumed they were seen in positive ways (positivity) and overestimated the degree to which they were seen as they saw themselves (transparency). Thus, the typical person had insight into the first impression they made but also assumed they were seen in positive ways and were more transparent than they really were to their partner.

Table 2.

Effects of Accuracy, Positivity, and Transparency for the Big Five Traits (Studies 1–4)

	Model 1: Accuracy		Model 2: Role of Positive Impressions				Model 3: Role of Self-Verifying Impressions
	Meta-Accuracy		Distinctive Meta-Accuracy		Positivity		Meta-Insight		Transparency
	β10 (SE)		β10 (SE)		β20 (SE)		β10 (SE)		β20 (SE)
Study 1 Acquaintance	.398 (.021)		.151 (.021)		.903 (.041)		.250 (.018)		.372 (.018)
Study 2 Peers	Time 1 .428 (.017)	Time 2 .471 (.020)	Time 1 .169 (.015) .	Time 2 212 (.017)	Time 1 .911 (.037)	Time 2 .918 (.043)	Time 1 .168 (.013)	Time 2 .195 (.014)	Time 1 .446 (.019)	Time 2 .494 (.021)
Study 3 Friends	.511 (.014)		.287 (.012)		.948 (.027)		.254 (.010)		.566 (.012)
Study 4	Sample 1	Sample 2	Sampl e 1	Sample 2	Sample 1	Sample 2	Sampl e 1	Sample 2	Sample 1	Sample 2
Romantic Partner	.571 (.022)	.520 (.012)	.378 (.028) .	367 (.013)	.990 (.038)	.943 (.023)	.270 (.020)	.225 (.010)	.612 (.024)	.562 (.012)

Note. All effects are significant (p < .05). Study 1: N = 184; Study 2: N = 228, Study 3: N = 273; Study 4: Sample 1 N = 77, Sample 2 N = 324. Meta-Accuracy = impression slope β₁₀; Model 2: Distinctive Meta-Accuracy = impression slope β₁₀; Positivity = normative impression slope β₂₀; Model 3: Meta-Insight = impression slope β₁₀; Transparency = self-perception slope β₂₀.

Accuracy Hypothesis: Do judges enjoy accurate metaperceivers?

As shown in Table 3, relative to people who were less aware, people who were more aware of how their new acquaintance perceived their characteristic pattern of traits were liked more (meta-accuracy: b = .060, p = .002; see Model 1). This pattern provides some support for the hypothesis that judges enjoy people who knew how they were perceived but does not reveal if judges enjoyed accuracy for impressions that were less positive or self-verifying. Did judges like metaperceivers who were more aware of the distinctive and potentially negative impressions they made? In contrast to the Accuracy Hypotheses, judge liking was not associated with distinctive meta-accuracy; instead, a marginal effect suggested that judges liked metaperceivers more who thought they were seen in positive ways (positivity: b = .071, p = .086; see Model 2 in Table 3). In other words, rather than enjoying people who knew the impressions they made, there was some evidence that judges tended to like people who thought they made positive first impressions. Did judges like metaperceivers who were more aware of how they saw them differently from how metaperceivers saw themselves? As predicted by the Accuracy Hypothesis, judge liking was associated with meta-insight (b = .036; p = .036; see Model 3 in Table 3). Judge liking was not associated with transparency.

Table 3.

Relationship Quality as a Moderator of Accuracy, Positivity, and Transparency for a New Acquaintance (Study 1), Friends (Study 3), and a Romantic Partner (Study 4)

	Study 1 New Acquaintance		Study 3 Friends		Study 4 Romantic Partner
	Metaperceiver Liking	Judge Liking	Metaperceiver Quality	Judge Quality	Metaperceiver Quality	Judge Quality	Metaperceiver Quality	Judge Quality
					Sample 1		Sample 2
	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]
Model 1: Accuracy
Meta-	.031 (.021)	.060* (.020)	.033* (.010)	.033* (.010)	.093* (.049)	.062 (.045)	.051* (.014)	.067* (.013)
Accuracy	[367, .430]	[.337, .459]	[.483, .535]	[.485, .535]	[.521, .615]	[.540, .601]	[.468, .562]	[.453, .581]
Model 2: Role of positive impressions
Distinctive	−.051* (.021)	.002 (.021)	.020† (.011)	.022* (.011)	.083 ( .062)	.049 (.056)	.004 (.014)	.029* (.014)
Meta-Accuracy	[.204, .101]	[.151, .156]	[.269, .301]	[.269, .303]	[.335, .416]	[.354, .402]	[.360, .368]	[.337, .393]
Positivity	.148* (.040)	.071† (.041)	.062* (.016)	−.008 (.016)	.153† (.082)	.027 (.076)	.112* (.025)	.047 (.026)
	[.753, 1.053]	[.830, .972]	[.898, .997]	[.953, .940]	[.911, 1.062]	[.976, 1.003]	[.830, 1.038]	[.894, .984]
Model 3: Role of self-verifying impressions
Meta-Insight	−.010 (.018)	.036* (.017)	−.010 (.009)	.021* (.009)	.002 (.044)	.025 (.040)	−.007 (.011)	.015 (.011)
	[.260, .239]	[.213, .286]	[.260, .244]	[.236, .269]	[.269, .270]	[.256, .281]	[.230, .217]	[.209, .237]
Transparency	.068* (.017)	.005 (.018)	.066* (.009)	.017† (.009)	.105* (.052)	.035 (.048)	.044* (.012)	.013 (.012)
	[.303, .440]	[.366, .377]	[.514, .619]	[.553, .580]	[.559, .661]	[.596, .630]	[.517, .599]	[.549, .573]

Note.

^*p < .05

^†p < .10. Judge or metaperceiver relationship quality ratings were moderators of the following slopes: Meta-Accuracy (Model 1), Distinctive Meta-Accuracy and Positivity (Model 2), and Meta-Insight and Transparency (Model 3). [low, high] = simple slope values when quality ratings were 1 standard deviation below (low) or above (high) the mean.

Subjective Reality Hypothesis: Do metaperceivers enjoy feeling valued and understood?

Did metaperceivers like judges more when they thought they were seen in positive or self-verifying ways? As predicted by the Positivity Hypothesis, results revealed that relative to people who thought they made a less positive impression, people who thought they were seen in positive ways tended to like their acquaintance more (b = .148, p < .001; see Model 2 in Table 3). Interestingly, while metaperceiver liking was not linked to meta-accuracy (see Model 1 in Table 3), metaperceivers who were aware of the distinctive impression they made tended to like their judge less (b = −.051, p = .014), suggesting that accuracy about a distinctive first impression was detrimental to metaperceiver liking. As predicted by the Transparency Hypothesis, people who tended to overestimate how transparent they were to their new acquaintance tended to like their new acquaintance more (b = .068, p < .001) and liking was not linked to meta-insight (see Model 3 in Table 3).

Discussion

Results provided some support for the Accuracy and Subjective Reality Hypotheses in early acquaintanceship. As predicted, judges tended to like individuals who knew the first impression they made, but metaperceivers did not reap the same relational benefits from their own insight. Instead, metaperceivers tended to like their new acquaintance more when they thought that acquaintance saw the best in them (positivity) or saw them as they saw themselves (transparency).

Results mostly supported the Accuracy Hypothesis, but judges did not necessarily like individuals who were aware of the distinctive impressions they made. One explanation for the null effect of distinctive meta-accuracy is that metaperceivers who were liked made and realized they made positive impressions on judges, leaving little distinctive information to be detected (Wood & Furr, 2015). Interestingly, metaperceivers tended to dislike judges when they realized what made them distinctive in their eyes. Given that agreement about a distinctive impression can be negative (Leising et al., 2001), this finding suggests that people liked a new acquaintance less when they realized that they started off on the wrong foot with that person.

Study 2: Longitudinal Impressions among Peers

Study 2 was designed to replicate Study 1 by measuring effects in a first impression context as well as explore whether the observed pattern extends beyond a first impression. Initially unacquainted peers were assigned to groups, and impressions and liking were assessed after individuals met for the first time and again after a few months of acquaintanceship. In addition to testing whether effects extend beyond a first impression, the longitudinal approach provides a test of whether initial accuracy or subjective reality predicts greater liking over time. Specifically, the current study controls for initial liking and tests whether knowing the first impression one makes leads to being liked more and whether assuming peers perceive the self in positive or self-verifying ways leads to liking others more over time.

Method

Participants (Time 1 N = 228, Time 2 N = 184; M = 19.79, SD = .98 years of age of 201 provided; 38% male of 205 provided; 67.5% Caucasian, 22.5% Asian-American, 5.5% African-American, 3% Hispanic, 1% Middle Eastern, .7% other, or did not indicate) were undergraduates in two sections of a personality course who took part in the current study as part of a larger set of class activates.

Procedures

As part of a class activity, participants in sections of a psychology course were randomly assigned to unacquainted groups of 3–8 people (M = 4.43 people) during the first week of class and met with the same group in class once a week for approximately 20 minutes each week until the end of the semester (14 weeks). Participants were asked to avoid joining a group if they knew another classmate in that group. For the first meeting, participants played the ice-breaking game “Two Truths and a Lie” where each person makes three statements about themselves, two of which are true, one of which is a lie, and the other members try to identify which statement was the lie. After this game, participants described their own personality, rated each group member’s personality, provided metaperceptions for each person in their group, and rated how much they liked each person in their group. Each subsequent week, group members engaged in discussions about the course material (e.g., their thoughts about how best to measure personality), and at the end of the course, participants completed the same ratings as in the first week.²

Measures

Using a 1 (Strongly Disagree) to 15 (Strongly Agree) scale, participants provided self-perceptions, metaperceptions (“How Person X sees you”) and impressions (“How you see Person X”) for the Ten Item Personality Inventory (TIPI; Gosling, Rentfrow, & Swann, 2003) and a two-item measure of liking (“How much do you like Person X” and “How likeable is Person X?” Time 1 liking M = 10.99, SD = 1.73, α = .90; Time 2 liking M = 11.02, SD = 1.59, α = .84; self-other agreement Time 1 r = .33; Time 2 r = .31 p < .001). As in Study 1, the normative profile was strongly, positively associated with ratings of social desirability (r = .986, p < .001), which were reliably coded by 12 research assistants ICC(2,1) = .879, ICC(2,12) = .989.

Analyses

Effects were indexed using the SAM (Biesanz, 2010) in R (lme4 package; Bates & Sarkar, 2007). The current study tested the three models described in Study1, but both metaperceivers and judges were modeled as random to account for the interdependence of the ratings among groups. Effects were computed separately for Time 1 and Time 2 to explore the concurrent links between impressions and liking.

Longitudinal effects were tested using the approach outlined by Human and colleagues (2012), which explored the link between self-other agreement and liking over time using the SAM framework. Specifically, grand-centered Time 1 and Time 2 liking scores (judge-rated or metaperceiver-rated) were entered as simultaneous level 2 predictors, or cross-level moderators, of slopes. This analysis reveals, for example, if Meg’s initial meta-accuracy predicted Jon’s liking of Meg at the end of the study (Time 2 liking), above and beyond Jon’s initial liking of Meg (Time 1 liking). Similar to Human and colleagues, concurrent Time 1 liking effects in this model are referred to as associations (e.g., Time 1 meta-accuracy is associated with Time 1 liking) whereas Time 2 liking effects are referred to as predictions (e.g., Time 1 meta-accuracy predicts Time 2 liking). Simple effects for significant interactions are reported for 1 SD above and below the mean and reflect profile agreement slopes for high or low liking.

Results

As shown in Table 2, at the beginning and end of the study, people tended to know how their pattern of traits were perceived by peers (meta-accuracy), what made them distinctive in the eyes of their peers (distinctive meta-accuracy) and how peers saw them differently from how they saw themselves (meta-insight). They also tended to think they were seen in positive ways (positivity) and overestimated how understood they were (transparency).

Accuracy Hypothesis: Do judges enjoy accurate metaperceivers?

As shown in Table 4, meta-accuracy was associated with being liked more by judges at both time points (concurrent effects: Time 1 b = .028, p < .001; Time 2 b = .037, p < .001). Did initial meta-accuracy have longitudinal effects on being liked by judges? Judges tended to like metaperceivers at Time 2 who were more aware of the first impression they made (b = .023, p < .001), an effect that remained when controlling for Time 1 liking (Time 2 liking b = .011, p = 002; see Model 1 in Table 4). Thus, initial meta-accuracy had positive, longitudinal effects on being liked by judges.

Table 4.

Liking as a Moderator of Accuracy, Positivity, and Transparency for Peers over Time (Study 2)

	Concurrent Effects of Liking				Longitudinal Effects of Time 1 Accuracy, Positivity, and Transparency
	Time 1		Time 2		Metaperceiver Liking		Judge Liking
	Metaperceiver Liking	Judge Liking	Metaperceiver Liking	Judge Liking	Time 1 Liking	Time 2 Liking	Time 1 Liking	Time 2 Liking
	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]	b (SE) [low, high]
Model 1: Accuracy
Meta-	.027* (.004)	.028* (.004)	.023* (.005)	.037* (.004)	.024* (.006)	.004 (.005)	.023* (.005)	.011* (.005)
Accuracy	[.358, .490]	[.352, .488]	[.411, .522]	[.367, .552]	[.356, .492]		[.330, .498]
Model 2: Role of positive impressions
Distinctive	.007 (.004)	.008* (.003)	.005 (.005)	.012* (.004)	.007 (.005)	−.004 (.005)	.000 (.005)	.009† (.005)
Meta-Accuracy	[.146, .181]	[.153, .190]	[.196, .219]	[.100, .245]	[.147, .165]		[.135, .180]
Positivity	.047* (.006)	−.005 (.005)	.036* (.007)	−.012† (.007)	.044* (.008)	−.002 (.007)	−.005 (.007)	.002 (.007)
	[.792, 1.024]	[.918, .894]	[.824, 1.005]	[.942, .883]	[.823, 1.029]		[.925, .913]
Model 3: Role of self-verifying impressions
Meta-Insight	.007* (.003)	.009* (.003)	.012* (.003)	.015* (.003)	.006 (.004)	−.001 (.004)	.005 (.004)	.008* (.004)
	[.147, .181]	[.143, .188]	[.161, .222]	[.154, .229]	[.150, .174]		[.129, .191]
Transparency	.020* (.003)	−.001 (.003)	.006* (.003)	−.004 (.003)	.012* (.004)	.007* (.003)	−.003 (.004)	.002 (.003)
	[.398, .494]	[.448, .442]	[.477, .509]	[.501, .482]	[.405, .498]		[.452, .445]

Note.

^*p < .05

^†p < .10. Judge or metaperceiver liking ratings were moderators of the following slopes: Meta-Accuracy (Model 1), Distinctive Meta-Accuracy and Positivity (Model 2), and Meta-Insight and Transparency (Model 3). Concurrent Effects: Time 1 liking ratings were entered as moderators of Time 1 slopes, Time 2 liking ratings entered as moderators of Time 2 slopes. Longitudinal Effects: Time 1 and 2 liking ratings were entered as moderators of Time 1 slopes; thus, Time 2 liking reflect the degree to which Time 1 meta-accuracy, positivity or transparency predicted Time 2 liking. [low, high] = simple slope values when quality ratings were 1 standard deviation below (low) or above (high) the mean.

As predicted by the Accuracy Hypothesis, judges liked people who were more aware of the distinctive impressions they made at both time points (distinctive meta-accuracy: Time 1 b = .008, p = .030; Time 2 b = .012, p = .004; see Model 2 in Table 4). Interestingly, a marginal effect at Time 2 suggested that judges liked people less who thought they were seen in positive ways (Time 2 b = −.012, p = .072). Did initial distinctive meta-accuracy foster liking over time? Time 1 distinctive meta-accuracy predicted being liked more by judges at Time 2 (b = .010, p = .025), an effect that remained marginally significant when controlling for how much people were liked at Time 1 (Time 2 b = .009, p = .052), suggesting that insight into distinctive and potentially negative first impressions had positive effects on relationship quality over time.

Did judges like individuals who were aware of the ways in which they were seen differently from how they saw themselves? As predicted by the Accuracy Hypothesis, judges liked people more who had higher meta-insight at both time points (Time 1 b = .009, p = .001; Time 2 b = .015, p < .001; see Model 3 in Table 4). There were no effects for transparency. Time 1 meta-insight also positively predicted judge liking at Time 2 (b = .010, p = .002), a pattern that remained when controlling for Time 1 liking (Time 2 b = .008, p = .046; Table 4), suggesting that initial meta-insight had positive effects on judges’ perceptions of relationship quality over time.

Subjective Reality Hypothesis: Do metaperceivers enjoy feeling valued and understood?

As predicted by the Positivity Hypothesis, metaperceivers tended to like peers more when they thought they were seen in positive ways at both time points (Time 1 b = .047, p < .001, Time 2 b = .036, p < .001; see Model 2 in Table 4). However, Time 1 positivity did not predict Time 2 liking when controlling for Time 1 liking suggesting that people who assumed their peers initially saw them in a positive light did not necessarily like their peers more over time.

As predicted by the Transparency Hypothesis, transparency was positively associated with metaperceiver liking of peers at both time points (Time 1 b = .020, p < .001; Time 2 b = .006, p = .029; see Model 3 in Table 4). Time 1 transparency also predicted metaperceivers’ liking of their peers at Time 2 (b = .012, p < .001) a pattern that remained when controlling for Time 1 liking (Time 2 b = .007, p = .028; Table 4). Thus, feeling understood early on fostered liking over time. Interestingly, in contrast to predictions, people with higher meta-insight also tended to like their peers more at both time points (Time 1 b = .007, p = .037; Time 2 b = .012, p < .001), although Time 1 meta-insight did not predict Time 2 liking of peers (see Table 4), suggesting that the ability to make valid distinctions between one’s self-perception and judges’ perceptions of the self did not necessarily foster liking over time. Similarly, as shown in Table 4, meta-accuracy was associated with liking judges more at both time points (Time 1 b = .027, p < .001; Time 2 b = .023, p < .001), but metaperceivers who were more aware of the first impression they made did not necessarily like others more at Time 2 when controlling for Time 1 liking.

Discussion

Results provided additional support for the Accuracy and Subjective Reality Hypotheses and revealed that the predicted effects generalized past a first impression. In line with the Accuracy Hypothesis, during the first few months of acquaintanceship, judges tended to like peers who were aware of how their characteristic pattern of traits were perceived (meta-accuracy), what made them distinctive in their judges’ eyes (distinctive meta-accuracy) and how judges saw them differently from how they saw themselves (meta-insight). Thus, there was a consistent link between accuracy and being liked by other people over the first few months of acquaintanceship. Going further, initial meta-accuracy, distinctive meta-accuracy, and meta-insight predicted being liked more by judges over time, suggesting that people who were more aware of the first impressions they made, regardless of whether those impressions were positive or self-verifying, were liked more over time.

In line with the Subjective Reality Hypothesis, at the beginning and end of the study, metaperceivers liked their peers more when they thought they were seen in positive ways (Positivity Hypothesis) and when they felt more transparent to their peers (Transparency Hypothesis). This pattern suggests a consistent link between subjective reality and liking others over time. Interestingly, initial transparency, but not positivity, fostered greater liking over time suggesting that feeling understood early on played a powerful role in who people enjoyed in the long run. This finding is in line with past work showing that people prefer to interact with strangers who see them as they see themselves (Swann, Pelham, & Krull, 1989). Yet, in contrast to predictions, meta-insight also played a positive role in metaperceivers’ liking. One explanation for this finding is that people might expect to be seen differently than how they see themselves in professional contexts (i.e., among classmates), and as such, making valid distinctions between global self-views and the impression one made on a classmate was functional. However, transparency rather than meta-insight seemed to have long term effects on liking, suggesting that feeling understood might ultimately be more important than accuracy.

Study 3: Friends

Results from Studies 1–2 provided some support for the Accuracy and Subjective Reality Hypotheses in the early phase of acquaintanceship. To test whether effects generalize to individuals in established relationships, judges in Study 3 were friends. Results will reveal if effects are observed among people who have known each other for several years and among peers who are already close.

Method

Participants

Undergraduate participants (N = 306; M_age = 19.77, SD = 1.61; 40% male; 58.3% Caucasian, 27.6% Asian-American, 10.4% African American, 1.6% Latin American, .6% Native American, .9% Middle Eastern, .6% other or did not report) were awarded course credit or $20 for participation. Participants provided self-perceptions of their personality as well as metaperceptions of personality and relationship quality ratings for up to four friends, a parent, and a romantic partner if applicable. Analyses for parents are presented in the supplemental section and results for romantic partners are presented in Study 4. Acquaintances were contacted via email and asked to provide impressions of the targets but were not compensated for their participation (Vazire, 2006). Participants nominated 1122 friends (M = 3.67, SD = 1.20 per participant), and at least one friend responded for 273 participants. Participants knew responding friends (N = 638; M = 2.34, SD = 1.10 per participant) for about four and half years (M_months = 58.69; SD = 53.31).

Measures

Self-perceptions, metaperceptions for each friend, and friends’ impressions for the BFI were provided on a 1 (strongly disagree) to 7 (agree strongly) scale. Participants and friends described their relationship quality with one another on a five-item measure using a 1 to 7 scale (e.g., “How close are you and Person X” “How would you rate the quality of your relationship with Person X”; self-reported M = 6.01, SD = .80, α = .89; friend-reported: M = 5.97, SD = .77, α = .86; self-other b = .42, p < .001, estimated r = .43). As found in Studies 1–2, the normative profile was strongly associated with the socially desirable profile (r = .903, p < .001), which was reliably coded by the same 12 coders as reported in Study 1.

Analyses

Models were the same as those in Studies 1 and 2; however, given the one-with-many design, judges were nested within metaperceivers.

Results

As shown in Table 2, the typical person achieved meta-accuracy, distinctive meta-accuracy, and meta-insight for friends. The typical person also thought their friends saw them in positive ways and assumed their friends shared their self-perceptions more so than their friends actually did. Relationship length was not linked to meta-accuracy (b = .0001, p = .353), distinctive meta-accuracy (b = .0002, p = .264), meta-insight, (b = .0003, p = .066), positivity (b = −.0002, p = .476), or transparency (b = −.00005, p = .739); thus, acquaintanceship length was not entered as a covariate.

Accuracy Hypothesis: Do judges enjoy accurate metaperceivers?

As shown in Table 3, relative to people who were less aware, people who were more aware of how their friends perceived their characteristic pattern of traits had friends who reported greater relationship satisfaction with them (meta-accuracy: b = .033, p = .001). Also as predicted by the Accuracy Hypothesis, Table 3 shows that friends reported greater relationship satisfaction with metaperceivers who were more aware of the distinctive impressions they made (distinctive meta-accuracy: b = .022, p = .043; Model 2) and who were more able to make valid distinctions between their self-views and their friend’s view of them (meta-insight: b = .021, p = .021; Model 3) than metaperceivers who were less aware of the impressions they made. There were no significant effects for positivity, but there was a marginal effect for transparency (b = .017, p = .070) providing some evidence that friends liked metaperceivers who overestimated their transparency.

Subjective Reality Hypothesis: Do metaperceivers enjoy feeling valued and understood?

As predicted by the Positivity Hypothesis, metaperceivers who thought they were seen more positively tended to like their friends more (b = .062, p < .001) than metaperceivers who thought they were seen less positively. However, in contrast to predictions, Table 3 also shows a marginal effect for distinctive meta-accuracy, suggesting that metaperceivers tended to report higher quality relationships with their friends when they realized the distinctive impressions they made (b = .020, p = .062; see Model 2). As predicted by the Transparency Hypothesis, metaperceivers who tended to overestimate how transparent they were to their friends tended to like their friends more (b = .066, p < .001; Model 3), and there was no effect for meta-insight.

Discussion

Results provided additional support for the Accuracy and Subjective Reality Hypotheses in established friendships. With respect to the Accuracy Hypothesis, judges reported greater relationship satisfaction with metaperceivers who were aware of how their friends perceived their characteristic pattern of traits (meta-accuracy), the distinctive and potentially negative impressions they made on friends (distinctive meta-accuracy), and the degree to which friends perceived them differently from how they perceived themselves (meta-insight). Thus, judges seemed to enjoy friends who knew the impressions they made, regardless of whether impressions were positive or self-verifying. As predicted by the Subjective Reality Hypothesis, metaperceivers enjoyed friends more who saw them in positive ways and who saw them as they saw themselves more so than they really did.

Study 4: Romantic Partners

The main goal of Study 4 was to test whether the hypothesized effects generalize to romantic couples. In addition, the current study explores dating and well-established relationships given past work showing differences between new and established romantic couples in other forms of perceptual accuracy, positivity biases and transparency (e.g., Campbell et al., 2006; Fletcher & Kerr, 2010; Letzring & Noftle, 2010; Swann et al., 1994; Watson et al., 2000). For example, self-other agreement is higher among married couples than dating couples (Watson et al., 2000), the effects of positive impressions on relationship quality seem to fade over time (Fletcher & Kerr, 2010), and self-verification seems to be especially important in established relationships rather than in new relationships (Campbell et al., 2006). As such, the current study provides a robust test of hypotheses by assessing effects among romantic couples who have known each other, on average, for a couple of years (sample 1) and for nearly three decades (sample 2).

Method

Participants

Sample 1 participants (N = 77; 34% male) were a subset of participants from Study 3 who nominated a responding romantic partner as an informant. As described in Study 2, participants provided self-perceptions, metaperceptions and relationship quality ratings while partners provided impressions and quality ratings. Participants knew their partner for approximately two years (months: M = 26.31, SD = 15.76).

Sample 2 participants from the St. Louis Personality and Aging Network (SPAN) study (N = 324; age M = 62.21, SD = 2.74; 60% male; 81.5% Caucasian, 17.6% African American, .6 % Latino, .3% other) who were paid $60 for their participation in a larger study. Participants provided self-perceptions of their personality, metaperceptions and relationship quality ratings for their romantic partner. Their romantic partner, whom they knew for more than 30 years (years: M = 31.53, SD = 12.09) described the target’s personality and provided relationship quality ratings.

Measures

Sample 1 participants completed the same measures as reported in Study 3. Metaperceiver-reported quality (M = 6.65; SD = .45) and partner-reported quality (M = 6.54, SD = .50) were positively associated (r = .57, p < .001).

For sample 2, self-perceptions and partner impressions of the metaperceiver were completed on the 240-item NEO-PI-R (Costa & McCrae, 1992). Metaperceptions were measured on a 30-item subset of the NEO Five Factor Inventory (NEO-FFI; Costa & McCrae, 2009). Analyses were based on the 30 items that appeared on all reports, but all items were rated on a 5-point Likert scale (strongly disagree to strongly agree). The normative profile was strongly correlated with social desirability ratings (r = .873, p < .001), which were coded by 13 research assistants ICC(2,1) = .542, ICC(2,13) = .939.

Metaperceiver- and partner-reported relationship quality was described on a six-item scale. Two items were: “How close are you to this person?” (1 = not at all to 5 = Closer than anyone else) and “How much do you like this person” (1 = not at all to 5 = more than anyone else; metaperceiver M = 4.45, SD = .58, α = .88; partner M = 4.40, SD = .65; α = .88). The remaining four items were the 4-item version of the Dyadic Adjustment Scale (DAS-4; Sabourin et al., 2005): “How often do you discuss or have you considered divorce, separation, or terminating your relationship?”, “In general, how often do you think that things between you and your partner are going well?”, “Do you confide in your mate?”, and “Please circle the dot which best describes the degree of happiness, all things considered, of your relationship.” The DAS-4 was completed with respect to how people felt about their relationship during the past four weeks (metaperceiver-reported M = 16.92, SD = 2.94, α = .83; partner-reported M = 16.22, SD = 3.17, α = .85). The six items were rated on different scales; thus, they were standardized prior to aggregation (metaperceiver scale α = .82; partner scale α = .87). Metaperceivers and partners agreed about the quality of their relationship (r = .54, p < .001).

Analyses

Participants in both samples were rated by and provided metaperceptions for their partner but their partner did not complete reciprocal ratings; thus, judges were not modeled as random. With this exception, models paralleled those from Studies 1–3.

Results

As shown in Table 2, in both samples, people had insight into how their romantic partner perceived their characteristic pattern of traits (meta-accuracy), how others perceived what made them distinctive in their partner’s eyes (distinctive meta-accuracy), and how their partner saw them differently from how they saw themselves (meta-insight). However, people also assumed they were seen in positive ways and assumed they were seen as they saw themselves more so than they really were. Length of acquaintanceship was not linked to meta-accuracy (sample 1: b = .0003, p = .846; sample 2: b = .000, p = .976), distinctive meta-accuracy (sample 1: b = −.001, p = .776; sample 2: b = −.001, p = .508), positivity (sample 1 b = .001, p = .838; sample 2: b = −.002, p = .404), meta-insight (Sample 1: b = .002, p = .294; Sample 2: b = .000, p = .736), or transparency (sample 1 b = −.003, p = .151; sample 2: b = .001, p = .482); thus, length of acquaintanceship was not entered as a covariate.

Accuracy Hypothesis: Do judges enjoy accurate metaperceivers?

As shown in Table 3, sample 1 meta-accuracy was not associated with partner-reported quality (b = .062, p = .173; Model 1), but in sample 2, relative to people who were less aware, people who were more aware of how their partner perceived their pattern of personality traits tended to have a partner who reported higher relationship quality with them (b = .067, p < .001; Model 1). Thus, in established couples, judges enjoyed relationships more with a metaperceiver who knew how he or she was perceived.

Judge-reported quality in sample 1 was not associated with distinctive meta-accuracy, but as predicted, partners in sample 2 reported higher quality relationships with metaperceivers who were more aware of the distinctive impression they made (b = .029, p = .033; see Model 2 in Table 3). Thus, in established couples, judges enjoyed relationships more with a metaperceiver who knew the distinctive impression he or she made.

In contrast to predictions, judge-reported quality was not associated with meta-insight in either sample, suggesting partners did not necessarily enjoy their relationship with metaperceivers who were aware of how they were seen differently from how they saw themselves.

Subjective Reality Hypothesis: Do metaperceivers enjoy feeling valued and understood?

As predicted by the Positivity Hypothesis, metaperceivers in both samples enjoyed relationships more when they thought their partner saw them in positive ways (sample 1 b = .153, p = .063; sample 2 b = .112, p < .001; see Model 2 in Table 3), although this effect was marginally significant for sample 1. As predicted by the Transparency Hypothesis, metaperceivers in both samples also enjoyed relationships more when they felt more transparent (sample 1 b = .105, p = .005; sample 2 b = .044, p < .001; see Model 3 in Table 3). While metaperceiver-reported liking was not linked to distinctive meta-accuracy or to meta-insight, as shown in Table 3, meta-accuracy was positively associated with metaperceiver-reported quality in both samples (sample 1 b = .093, p = .049; sample 2 b = .051, p < .001).

Discussion

Overall, results provided mixed support for the Accuracy and Subjective Reality Hypotheses in romantic relationships. On one hand, as predicted, partners in long-term relationships (sample 2) tended to enjoy their relationships more with metaperceivers who had insight into the impressions they made (meta-accuracy) regardless of whether the impressions they made were positive (distinctive meta-accuracy), and metaperceivers in shorter and longer term relationships (samples 1 and 2) enjoyed their relationships more when they thought they were seen in positive or self-verifying ways. On the other hand, in contrast to predictions, judges in less established couples did not necessarily reap relational benefits of meta-accuracy. Likewise, judges in longer-term couples did not reap the relational benefits of meta-insight.

Why did judges in a romantic context benefit less from accuracy? The links between romantic relationship satisfaction, positive biases and self-verification are some of the most robust findings in romantic relationships, such that partners are happier in relationships when they idolize and are idolized by their partners (Murray et al., 1996) and when they are and feel understood (Swann et al., 1994; Weger, 2005). Forming or detecting negative impressions or less verifying impressions might be especially threatening to judges and metaperceivers, respectively. Yet, the links between judge-reports of relationship quality and accuracy were not negative. The null associations between judge-reports of quality and distinctive meta-accuracy (sample 1) and meta-insight might reflect the fact that this type of knowledge has mixed effects on relationships rather than a particularly negative effect. Notably, judges in well-established relationships did seem to reap some benefit from distinctive meta-accuracy, a finding that might reflect the fact that positive biases become weaker predictors of relationship quality over time (Fletcher & Kerr, 2010). However, a longitudinal approach can shed more light on how these effects change over time given that the samples differed in age as well as length of relationships.

General Discussion

Do individuals who know what other people really think about their personality have higher quality relationships? Results from four ecologically valid studies testing over a thousand participants (N = 1086) suggest that the answer largely depends on who reports on relationship quality. In support of the Accuracy Hypothesis, new acquaintances, peers, friends, and romantic partners generally enjoyed their relationships more with individuals who knew the impressions they made, specifically how acquaintances perceived their characteristic pattern of traits (meta-accuracy), regardless of whether these individuals were aware of potentially negative (distinctive meta-accuracy) or less self-verifying (meta-insight) impressions. Going further, people who knew the first impressions they made, including less positive or self-verifying impressions, tended to be liked more over time, suggesting that accuracy has positive, cascading effects on relationships. In support of the Subjective Reality Hypothesis, metaperceivers did not necessarily enjoy their relationships when they detected potentially negative or less self-verifying impressions they made. Instead, they enjoyed relationships more when they thought they were seen positively (Positivity Hypothesis) and when they overestimated the degree to which others saw them as they saw themselves (Transparency Hypothesis).

Findings supported hypotheses in multiple social contexts and across levels of acquaintanceship, suggesting that judges’ tendency to enjoy accuracy and metaperceivers’ tendency to enjoy feeling valued and understood are robust effects. Indeed, this pattern was observed from the very first interaction (Studies 1–2) to relationships spanning over 30 years (Study 4). In addition to spanning time, effects were also observed across qualitatively different types of relationships (e.g., strangers versus romantic partners), although Study 4 suggested that judge-reported relationship quality in romantic relationships could be independent of meta-accuracy or a metaperceiver’s subjective reality. Effects also generalized across contexts that varied in terms of whether relationships were self-initiated (i.e., friends, romantic partner; Studies 3–4) or not (i.e., new acquaintances, peers; Studies 1–2). This distinction is important because people tend to perceive individuals they know and like in a more normative (i.e., positive) way than people they know but do not necessarily like (Leising et al., 2010). Arguably, new acquaintances and peers held less positive impressions than did friends and romantic partners, which provided a strong test of the prediction that judges enjoy individuals more who know how they are seen, regardless of the valence of the impressions they make.

Why did judges value meta-accuracy? One possibility is that people who are aware of how others see them use their knowledge to guide their behavior in interpersonally adaptive ways. Given the robust, positive effect of accuracy on judges’ experience of relationship quality, another possibility is that meta-accuracy has a conspicuous interpersonal signature. Indeed, meta-accuracy might correspond to an interpersonal style such as greater self-disclosure. Another possibility is that metaperceivers who know how they are seen convey their knowledge to judges in ways that make judges feel understood. For example, Meg might see herself as conscientious but realize that Jon does not agree with her (i.e., meta-insight); as such, he might feel as though Meg values his perspective when she acknowledges his impression directly or indirectly (e.g., by deferring organizational tasks). Future work that explicitly measures metaperceivers’ behavior in and outside of the laboratory might shed light on whether certain behaviors or interpersonal styles explain the link between meta-accuracy and judge-reported quality.

Metaperceivers enjoyed relationships more with people they thought saw them in a positive or self-verifying light. However, with the exception of a new acquaintance, it was not the case that metaperceivers’ insight into potentially negative or less self-verifying impressions was detrimental to their own relationship quality. Rather, results suggested that metaperceivers’ subjective reality was more important to their experience of quality than was accuracy. One possible explanation for this pattern is that people are motivated to enhance and to verify their self-concept (Sedikides, 1993; Sedikides & Strube 1995; Swann, 1990) and one way of doing so is through social relationships. Another explanation is that positivity and transparency provide a sense of unconditional acceptance and understanding, which are arguably key components of high quality relationships. Assuming another person perceives the self in a positive light is likely experienced as unconditional positive regard (i.e., being accepted for one’s flaws; Campbell et al., 2006) and feeling as though another person understands the self eliminates unrealistic expectations and fears of being rejected for hidden flaws (Carnelley et al., 1999; Kwang & Swann, 2010; Letzring & Noftle, 2010). Interestingly, transparency rather than positivity fostered relationship quality over time (Study 2), suggesting that feeling understood might be more important in the long run than expecting to be seen in positive ways or knowing what others really think.

In addition to demonstrating links between accuracy and relationship quality, results also add to the growing body of work showing that people have self-knowledge of how they are experienced by other people across a broad range of social contexts. Self-knowledge of personality can as been defined as knowledge about one’s patterns of thinking, feeling, and behaving as well as knowledge about how others perceive these patterns (Carlson, 2013; Vazire & Carlson 2010). As such, evidence for meta-accuracy, distinctive meta-accuracy and meta-insight across contexts and levels of acquaintanceship provides more evidence that people have an impressive level of social acuity. Evidence for the Accuracy Hypothesis also adds to the growing body of work suggesting that self-knowledge is a virtue in the eyes of others (Colvin et al., 1995; Tenney, Spellman, & MacCoun, 2008; Tenney et al., 2013; Ward & Brenner, 2006). Given that meta-accuracy requires insight into one’s own behavior as well as how other people might interpret that behavior differently (Carlson & Kenny, 2012; Kenny & DePaulo, 1993), meta-accuracy might explain why people with more self-knowledge tend to be liked more. Specifically, meta-accuracy likely allows people to temper the negative effects of their aversive behaviors and continue to behave in ways that have desirable effects on other people.

Unlike other forms of self-knowledge that indicate someone’s awareness of their global traits (e.g., self-knowledge of intelligence or agreeableness), meta-accuracy is relational and tests people’s knowledge for a specific person’s experience of them. As such, poor meta-accuracy likely has greater consequences on relationship outcomes than does self-knowledge of general tendencies. Results for meta-insight provide some evidence for this hypothesis given that the index partials out the overlap between self-perceptions and actual impressions (e.g., self-other agreement), or what some might consider self-knowledge of social reality. For example, Meg might see herself as dependable while Jon thinks she is unreliable (i.e., poor self-other agreement); however, Meg’s insight into Jon’s impression seems to offset some of the negative consequences of this disagreement, at least for Jon.

The finding that metaperceivers benefited more from their subjective reality than from accuracy raises the question of whether meta-accuracy should be improved. With the exception of new acquaintances (Study 1), metaperceivers did not necessarily suffer from knowing what other people really thought, but there might be ways of improving accuracy in ways that allow metaperceivers to reap the same rewards of meta-accuracy that judges enjoy. How can meta-accuracy be improved such that metaperceivers can experience the same positive effects as judges? The answer does not seem to be direct feedback, given that feedback about others’ impressions is not internalized (Shechtman & Kenny, 1994) or has negative outcomes. For example, in business settings, people receive feedback about others’ impressions of them in performance reviews (i.e., 360° feedback), but this information often backfires and undermines rather than improves work performance and satisfaction (Brett & Atwater, 2001). One possibility is that they way people learn about others’ impressions affects their relationship quality in key ways. Specifically, feedback delivered in ways that reduce defensiveness or hurt feelings might improve both accuracy and relationship quality. There are reasons to predict that self-compassion and mindfulness might improve self-knowledge without increasing defensiveness (Carlson, 2013). For example, a series of studies found that, relative to people with less, people with more self-compassion were more accurate about their performance on a self-introduction task and tended to take greater responsibility for their shortcomings without experiencing negative emotional reactions (Leary, Tate, Adams, Allen, & Hancock, 2007). Likewise, positive affirmations in the face of threatening feedback can also reduce defensiveness (Critcher, Dunning, & Armor, 2010), suggesting that feedback mixed with affirmations might successfully improve accuracy and quality.

Limitations and Future Directions

The current research has many strengths, including high-powered, ecologically valid samples of over a thousand participants spanning multiple contexts and levels of acquaintanceship. However, there are also important limitations that might be addressed with future research. First, meta-accuracy was operationalized as profile agreement, but there are many ways to conceptualize and measure meta-accuracy that might yield different results (Carlson & Kenny, 2012). The profile approach is a statistically powerful tool that avoids multiple comparisons across traits, but indices such as distinctive meta-accuracy are quite conservative and can be difficult to interpret (Wood & Furr, 2015). For example, if Meg makes and realizes she makes normative impressions, her distinctive meta-accuracy would appear to be zero (i.e., there is no distinctive information to agree upon). Future work that measures meta-accuracy for specific traits (e.g., agreeableness) might reveal important trait-specific links between meta-accuracy and relationship quality. Indeed, some work suggests that the link between relationship quality and other interpersonal perceptions differs for global versus specific traits (Neff & Karney, 2005) and for traits that are important versus irrelevant to the specific relationship (Boyes & Fletcher, 2007; Campbell, 2005).

Second, future work examining the interpersonal benefits of meta-accuracy in different social contexts (e.g., the workplace) and for different outcomes (e.g., status) might reveal that metaperceivers benefit from accuracy more than a positive subjective reality. For example, rather than affecting metaperceivers’ experiences of their dyadic relationships, meta-accuracy might be a skill that affects broader social outcomes such as overall popularity and social status in social networks, which in turn, yield a host of emotional and material resources. The current research explored contexts where metaperceivers and judges were of equal status, but in contexts where a judge’s opinion of the metaperceiver has a direct impact on his or her life (e.g., a job interview), meta-accuracy could impact people’s lives in important ways that go beyond social evaluations (e.g., job performance ratings and subsequent monetary rewards). Low-status individuals tend to have higher meta-accuracy, arguably because it is especially important for low-status individuals to predict the outcomes that high-status individuals control (Fiske, 1993; Snodgrass, Hecht, & Ploutz-Snyder, 1998), suggesting that meta-accuracy might have especially positive effects on low-status versus high-status metaperceivers. Given these possibilities, future work might explore broader outcomes of meta-accuracy for metaperceivers across a wider range of social contexts and for different outcomes to determine if meta-accuracy has positive effects for metaperceivers on outcomes or in contexts that were not included in the current research.

Third, there may be personality characteristics of metaperceivers or judges that affect the extent to which meta-accuracy is beneficial. For example, meta-accuracy might be especially beneficial for people who tend to form overly negative metaperceptions, such as social anxiety, (Christensen, Stein, & Means-Christensen, 2003), because expecting negative evaluations seems to hinder relationship quality. Likewise, accuracy might attenuate the negative impact of personality problems (e.g., personality disorders) on relationship quality given that judges seem to enjoy people who know the impressions they make regardless of whether impressions are desirable (Carlson & Oltmanns, 2015). With respect to individual differences in judges, some people are especially positive or especially critical judges of other people (Srivastava, Guglielmo, & Beer, 2010). For example, harsh perceiver effects tend to covary with especially difficult personality traits (e.g. narcissism; Lukowitsky & Pincus, 2013; Wood, Harms, & Vazire, 2010). Thus, meta-accuracy for harsh judges might be especially painful and hinder relationship quality for metaperceivers whereas meta-accuracy for positive judges might be especially beneficial.

Conclusion

Accurate metaperceptions are functional and seem to foster better relationships with other people. Indeed, new acquaintances, peers, friends, and romantic partners enjoyed relationships more with people who had insight into the impressions they made. As Henry David Thoreau suggests, good relationships seem to involve knowing the truth. Yet, subjective reality also plays a powerful role in shaping people’s social experiences. Rather than knowing what others really think, people’s beliefs that others saw them in positive or self-verifying ways played a strong role in how much they enjoyed their relationships. In sum, while our ability to understand the impressions we make on others seems to be a virtue in the eyes of other people, we might ultimately have to choose between feeling better about our relationships and facing the truth about what other people really think about us. As Aldous Huxley suggests, self-knowledge might be hard to accept.