Love is not blind: What romantic partners know about our abilities compared to ourselves, our close friends, and our acquaintances

Love is not blind: What romantic partners know about our abilities compared to ourselves, our close friends, and our acquaintances. Gabriela Hofer, Silvia Macher, Aljoscha Neubauer. Journal of Research in Personality, March 4 2022, 104211. https://doi.org/10.1016/j.jrp.2022.104211

Abstract: How much do our partners, close friends, and acquaintances know about our abilities, as compared to ourselves? This registered report aimed to investigate asymmetries in these perspectives’ knowledge of a person’s verbal, numerical, and spatial intelligence, creativity, and intra- and interpersonal emotional abilities. We collected self-estimates and performance measures of these abilities from 238 targets. Each target’s abilities were also rated by their romantic partner, a close friend, and an acquaintance. Results showed knowledge-asymmetries but also similarities between perspectives. People themselves were at least moderately accurate across all six domains. However, partners achieved similar accuracy and both partners and friends could provide unique insights into some abilities. We discuss these results with regard to Vazire’s self-other knowledge asymmetry model.

Introduction

“How am I doing?” This question is one that many of us likely ask themselves on a regular basis. Whether it concerns academic performance or everyday skills like driving ability, knowing how well we are doing is essential and sometimes our impression of our abilities shapes important life decisions (e.g., Ackerman & Wolman, 2007). It is, therefore, of little surprise that a lot of research has investigated the accuracy of self-estimates of abilities, reaching the conclusion that they are less accurate than one would imagine (Freund and Kasten, 2012, Zell and Krizan, 2014). Indeed, our self-estimates seem to be distorted by overestimation (e.g., Visser et al., 2008). Moreover, other people can also provide valuable information about our abilities and skills and their estimates might be similarly accurate or sometimes even slightly more accurate than our own (Denissen et al., 2011, Steinmayr and Spinath, 2009). To this date, however, hardly any research has directly compared the accuracy of self- and other-estimates of abilities (for an exception see Neubauer et al., 2018, who investigated accuracies of self- and peer-estimates in adolescents). The first main goal of this article was to provide such a comparison and to do so for an adult population and a wide range of abilities. When people make important decisions like vocational choices, they may ask several others for feedback. Close friends and romantic partners are probably common sources people turn to. Our second main goal was, therefore, to investigate, whether romantic partners and close friends have special insights or biases when it comes to assessing our abilities by comparing the accuracy of their judgments with those of acquaintances.

A considerable amount of research has focused on the accuracy of self- and other-perceptions of personality traits. Both types of perceptions can predict important outcomes like academic success or job performance and other-perceptions can provide incremental validity over self-perceptions (Connelly & Ones, 2010). However, neither perspective is without its biases. As an example, Anusic, Schimmack, Pinkus, and Lockwood (2009) found evidence for an evaluative bias factor in self- and other-ratings of the Big Five personality traits. In their truth and bias model of person perception, West and Kenny (2011) proposed that a perceiver’s rating of a target on a given trait does not only reflect the target’s true score (and measurement error) but is also affected by certain bias variables. John and Robins (1993) showed that self-other and other-other agreement for the Big Five are determined by a trait’s observability (i.e., its visibility to observers) and evaluativeness (i.e., its social desirability or undesirability). Both self-other and other-other agreement were highest for highly observable traits of low evaluativeness. High evaluativeness seemed particularly detrimental for self-other agreement. Earlier work by Paunonen (1989) had shown that not observability per se but the interaction between observability and acquaintance is related to self-other agreement: Low observability is only related to lower self-other agreement when the level of acquaintance between target and rater is low. More recently, Connelly and Ones (2010) confirmed this meta-analytically and showed that the interpersonal intimacy between perceiver and target might be even more important than acquaintance per se. They found that the most accurate ratings in terms of self-other correlations come from spouses and dating partners. A recent extension of the truth and bias model (Leising et al., 2015) found that ratings of a target were influenced by perceiver’s attitudes (liking) but only when items were high in evaluativeness. Finally, current work found that how much the perceiver likes the target and how well he/she knows the target have opposing effects on accuracy: Whereas higher knowing was associated with higher accuracy and lower positivity bias, higher liking was related to lower accuracy and higher positivity bias (Wessels et al., 2018). Overall, past research seems to agree that both characteristics of the trait to be judged and of the relationship between target and perceiver affect the accuracy of ratings. However, hardly any of these models have focused on mechanisms behind potential differences in accuracy between self- and other-estimates.

Simine Vazire’s (2010) self-other knowledge asymmetry (SOKA) model offers a framework for systematic comparisons of the accuracy of self- and other-estimates. The model builds on the Johari window (Luft & Ingham, 1955) and assumes that a person’s traits fall into one of four different quadrants, depending on how much the person themselves and others know about the respective characteristic: Traits in the ‘open area’ are judged accurately by both the self and others. If only others are accurate about a trait, it is in the ‘blind spot’, whereas traits only validly judged by oneself are in the ‘hidden area’. Lastly, traits that neither perspective can judge accurately are in the ‘unknown area’. Drawing on the research summarized in the past section, the model proposes that the position of a trait in the Johari window should be determined by two factors: observability and evaluativeness. Vazire argued that self-estimates of highly evaluative traits are often distorted, since these traits are relevant to the person’s self-esteem (see also John & Robins, 1993). At the same time, others can only make accurate estimates about observable traits. Taken together, others might have more accurate views of our observable and evaluative traits than we ourselves do. Vazire (2010) allocated traits to the positions within the SOKA model/Johari window based on differences in correlation coefficients between self- and peer-estimates and relevant behavior (for extraversion and neuroticism) or objective performance (for intellect). In this initial study, she found extraversion (high observability, low evaluativeness) to be in the open area, intellect (low observability, high evaluativeness) mostly in the blind spot, and neuroticism (low observability, low evaluativeness) in the hidden area.

Similar to some of the models discussed in section 1.1, Vazire (2010) also considered a third aspect that might influence a trait’s position within the SOKA model: the level of acquaintance. She discussed that, while well-acquainted others might have advantages compared to less acquainted others when it comes to judging low observability traits (see also Connelly and Ones, 2010, Paunonen, 1989), they might also share some of the self’s self-protective biases, leading to less accurate judgments. Unexpectedly, she found friends to be more accurate than strangers when judging the highly evaluative trait intellect. Thus, she proposed that distortions of other-estimates due to high evaluativeness might only occur in particularly emotionally invested known others like romantic partners. The emotional investment in friendships might have been too low for the negative effects of evaluativeness on accuracy to emerge. This would be in line with the negative association between liking and accuracy found by Wessels and colleagues (2018). John and Robins (1993) proposed that judgments by emotionally invested others might involve similar psychological processes as self-perception. On a similar note, it has been suggested that “in a close relationship, the person acts as if some or all aspects of the partner are partially the person's own” (Aron et al., 1991, p. 242). This is also in line with the self-evaluation maintenance model (Tesser, 1988), according to which the performance of a close other might affect one’s own self-esteem and do so negatively, if the domain in question is relevant for one’s self-definition. Vazire (2010) proposed that a direct comparison between ratings by romantic partners and similarly well-acquainted friends could provide valuable insight into this question. Surprisingly, such a study does not seem to exist until today. In general, only little research on the SOKA model has been conducted. At the time of writing, it has mainly been investigated for personality traits (e.g., Beer & Vazire, 2017) but pertinent research also exists for personality disorders (Carlson et al., 2013), and moral behaviors (Thielmann et al., 2017).

To this point, hardly any studies have investigated the SOKA model for different aspects of intelligence or other abilities, even though this line of research might provide valuable insights. When making important life decisions, people may rely on feedback about their abilities from different sources (e.g., self, parents, friends, partners or teachers; Neubauer et al., 2018). Thus, it seems essential to investigate which of these sources can provide accurate estimates for a given domain.

First evidence on self-other knowledge asymmetries for abilities comes from Vazire (2010), whose findings on intellect are based on creativity (originality in a divergent thinking task) and overall intelligence. Both abilities were measured with objective ability tests. Results showed that creativity is in the blind spot, with only friends but not the self providing accurate estimates. Findings for intelligence were similar but less clear-cut, since self-estimates showed at least some accuracy in this domain. Strangers were unable to make accurate estimates for either ability.

Only recently, Neubauer and colleagues (2018) have analyzed the position of a more diverse set of abilities within the SOKA model based on self-ratings and ratings of randomly assigned classmates in 14- and 18-year-old pupils (i.e., ages when important educational decisions have to be made). The following abilities were assessed: verbal, numerical, and spatial intelligence (as measured by a standardized intelligence test), creativity (originality in a divergent thinking task), and intra- and interpersonal emotional management abilities (as measured by a situational judgment test). In both age groups, numerical intelligence and creativity were open, verbal intelligence was in the blind spot, and intra- and interpersonal emotional abilities were hidden. Spatial intelligence was unknown in the younger group and hidden in the older one. Thus, there seems to be variation in the location of abilities within the SOKA model, even though most of those examined could be considered to belong to the concept of intellect investigated by Vazire (2010) and might, therefore, be expected to be located in the blind spot. Self-reported closeness to the rated peer did not moderate any of the effects, a finding that the authors mainly attribute to the random assignment of peer-raters.

The relevance of having an accurate view of one’s own abilities and those of one’s peers (e.g., in order to give them feedback) might be particularly high during adolescence, given that important (educational/vocational) decisions have to be made around this time (Neubauer et al., 2018). Nevertheless, accurate self- and other-assessments are probably also relevant later in life and maintaining self-insight over the course of life may prove increasingly difficult, since adults usually receive less regular feedback on their abilities than pupils in school do. The accuracy of self- and other-estimates of abilities can also be important in clinical contexts: Accurate perceptions of a person’s memory decline – which might, for example, be due to a cognitive disorder – could be essential to provide them with appropriate and timely care (Buelow et al., 2014). Even though self-reported memory complaints show a small (negative) correlation with objective cognitive function in the general aging population (Burmester et al., 2016), this association seems to disappear in individuals with mild cognitive impairment (Buelow et al., 2014, Fyock and Hampstead, 2015) or Alzheimer’s disease (Buelow et al., 2014). It has also been shown that informant-reports can outperform self-reports in terms of accuracy for individuals with mild cognitive impairment (Buelow et al., 2014, Fyock and Hampstead, 2015).

Providing a systematic comparison of the accuracy of self- and other-estimates of abilities in adults was one of the main goals of the present work. In view of the lack of literature that directly compares these perspectives, we summarized available work that focused on the accuracy of either self-estimates or other-estimates in the upcoming sections. In line with recent suggestions regarding the interpretation of effect sizes in individual difference research (Gignac & Szodorai, 2016), we classified correlations starting from .1 to indicate low accuracy and correlations starting from .2 to indicate medium or moderate accuracy. However, we used the conventional—and, thus, stricter—threshold (r ≥ .5) for high accuracy (Cohen, 1992; for a display of the practical importance of such a correlation see Table 2).

A considerable amount of research has focused on the accuracy of self-estimates of abilities, resulting in several meta-analyses (e.g., Freund and Kasten, 2012, Mabe and West, 1982, Ross, 1998) and even one metasynthesis (i.e., a combination of several meta-analyses; Zell & Krizan, 2014). According to this metasynthesis, overall accuracy of self-estimates is moderate (rmean = .29) with considerable variability of effects depending on the ability domain in question (rs ranging from .09 for interpersonal sensitivity to .63 for second language competence). Freund and Kasten (2012) focused their meta-analysis on verbal, numerical, spatial, and overall intelligence and also found moderate accuracy (rmean = .33). Additionally, they found greater accuracy of self-estimates of numerical intelligence as compared to overall intelligence, with no comparable differences in accuracy between overall and verbal or spatial intelligence.

Past results on the accuracy of self- and other- estimates of the domains that we investigated in the present study, that is verbal, numerical, and spatial intelligence, creativity, and intra- and interpersonal emotional management abilities, are summarized in Table 1. As can be seen, these results again point towards an accuracy advantage for self-estimates of numerical intelligence compared to those of other intelligence facets: In the majority of cases, very low to medium accuracy was reported for self-estimates of verbal and spatial intelligence, while medium to high accuracy was found for numerical intelligence (Furnham et al., 2001, Neubauer et al., 2018, Proyer and Ruch, 2009, Rammstedt and Rammsayer, 2002, Steinmayr and Spinath, 2009, Visser et al., 2008). Correlations between self-estimates of creativity and creative performance were found to range from slightly negative to .44, depending on the way creativity was assessed (Furnham et al., 2005, Neubauer et al., 2018, Pretz and McCollum, 2014, Vazire, 2010). For both inter- and intrapersonal emotional management abilities, correlations between self-estimates and performance were moderate to high (Freudenthaler and Neubauer, 2005, Neubauer et al., 2018). In addition to the results presented in Table 1, it seems noteworthy that Elfenbein, Barsade, and Eisenkraft (2015) reported low (r = .13) to medium (r = .3) accuracy of self-reported overall emotional management abilities in two studies, even though they did not differentiate between intra- and interpersonal aspects.

The predominant focus on correlation coefficients in this line of research has repeatedly been criticized (e.g., Dunning & Helzer, 2014) and some research has instead focused on the direction of misestimation. There is a large amount of indirect evidence for humans’ tendency to overestimate themselves. As an example, people were repeatedly shown to believe that they perform better than the average person (e.g., Dunning et al., 1989, Horrey et al., 2015, Kruger and Dunning, 1999), a phenomenon known as the above-average or better-than-average effect (Alicke & Govorun, 2005). A recent study found that 65 percent of Americans believe they are more intelligent than the average person, something that is logically impossible (Heck et al., 2018). Visser and colleagues (2008) showed that students judge their intelligence on all of Gardner’s eight intelligence domains to be above that of the average student at their university. Still, hardly any research has investigated people’s apparent tendency to overestimate themselves more directly by comparing self-estimated and objectively measured intellectual abilities (Gignac & Zajenkowski, 2019). In a rare exception, Reilly and Mulhern (1995) found that men, on average, overestimate their IQ by about 8 IQ points, while women’s self-estimates did not differ significantly from their measured IQ. In a recent study, Gignac and Zajenkowski (2019) reported that both men and women overestimate their IQ by on average 30 IQ points, which represents a large effect. Clearly, more research on this topic is needed before a definite conclusion can be made. Given the differences in accuracy correlations for different ability domains, investigating over-/underestimation in several domains seems particularly interesting.

Past work focusing on other-estimates of intelligence yielded moderate to high accuracy correlations but also overestimation by close others. Several correlational studies showed that others are already able to make reasonably accurate intelligence judgements after watching short standardized videos of a person (Borkenau et al., 2004; rs between .22 and .53 Borkenau and Liebler, 1993, Reynolds and Gifford, 2001). Denissen and colleagues (2011) investigated how intelligence-estimates by fellow students develop over the course of a semester and found accuracy correlations of .25 after one week, .27 after one month, and .22 after another 4 months of acquaintance. Borkenau and Liebler (1993) investigated intelligence estimates by a person’s cohabitant (in most cases their romantic partner) and reported a correlation of .29 with objectively measured intelligence. Recently, Gignac and Zajenkowski (2019) found that women’s estimates of their male romantic partner’s intelligence correlated at .30 with the partner’s actual intelligence, whereas men’s estimates only correlated at .19 with their female partner’s intelligence. Moreover, the authors found that both genders did not only overestimate their own but also their partner’s intelligence by around 30 IQ points, which again constitutes a large effect.

As shown in Table 1, only little research seems to have investigated accuracy of other-estimates for different ability domains. Steinmayr and Spinath (2009) reported that parents judged their adolescent sons’ and daughters’ verbal, numerical, and spatial intelligence with medium accuracy. Sommer, Fink, and Neubauer (2008) found that both teachers and parents estimated elementary school pupils’ intelligence with an accuracy of around .5, creativity with an accuracy of between .2 and .3, and social competence (consisting of inter- and intrapersonal parts) with an accuracy of only .1. Neubauer and colleagues (2018) found a similar pattern of results for peer-estimates in their older age group: Numerical and verbal intelligence as well as creativity were estimated with medium accuracy, whereas estimates of intra- and interpersonal emotional management abilities were of low or low to medium accuracy. Low accuracy was also reported for peer-estimates of spatial intelligence. More support for the comparatively low accuracy of other-ratings of emotional management abilities (again consisting of intra- and interpersonal aspects) comes from Elfenbein and colleagues (2015), with estimate*performance correlations between -.04 (student classmates) and .04 (work colleagues). Vazire (2010) found quite low accuracy of stranger-ratings of creativity and slightly higher accuracy for friend-ratings.

In the present study, we investigated the position of six abilities within the SOKA model in an adult sample. We aimed to:

(1)

investigate the accuracy of self- and other-estimates of abilities, with the latter stemming from the target’s romantic partner, their best or a very close friend, and an acquaintance. Thus, we collected data from two sources who knew the target considerably well but differed with regard to the expected closeness/intimacy of their relationship to the target (friends and partners; in line with the proposition by Vazire, 2010) and added a source that we expected to know the target less well and be less close to him/her (acquaintances).

(2)

determine for which domains the four perspectives (self, partner, friend, and acquaintance) differ in their accuracy.

(3)

investigate the unique insights of each perspective and the overall amount of variance all four perspectives can jointly explain.

(4)

determine the direction of misestimation by targets, friends, partners, and acquaintances.

We included verbal, numerical, and spatial intelligence, creativity, and inter- and intrapersonal emotional management abilities due to their relevance for important life outcomes. Verbal, numerical, and spatial intelligence form part of most modern models of intelligence (see Hunt, 2010) and several meta-analyses have determined that intelligence is an important predictor of professional and socioeconomic success (Hülsheger et al., 2007, Schmidt and Hunter, 2004, Schmidt and Hunter, 1998, Strenze, 2007). Creativity is seen as essential for solving key problems and has been connected with many essential aspects of life (Hennessey and Amabile, 2010, Plucker et al., 2004). A recent meta-analysis found that creativity is associated with academic achievement, although with only a small to medium effect (Gajda et al., 2017). Emotional management comprises the highest branch in one of the most influential models of emotional intelligence (Mayer & Salovey, 1997) and refers to the “ability to manage emotions and emotional relationships for personal and interpersonal growth” (Mayer et al., 2001, p. 235). Both intra- and interpersonal emotional management abilities are associated with life satisfaction and (lower) depressive tendencies (Freudenthaler, Neubauer, & Haller, 2008). Emotional intelligence as a broader ability exhibits small but meta-analytically stable associations with job performance (Joseph et al., 2015) and was found to predict academic and social success over and above personality and psychometric intelligence (van der Zee et al., 2002).

Perhaps one of the most important methodological considerations when conducting research on accuracy in person perception relates to the choice of accuracy criteria. When it comes to perceptions of a person’s abilities, the target’s performance in objective ability tests constitutes an obvious accuracy criterion. Intelligence tests, for example, have long been accepted as objective measures of cognitive abilities and their scores are widely used as accuracy criteria (Freund & Kasten, 2012). Thus, we used subscales of a well-established, standardized intelligence test battery to measure verbal, numerical, and spatial intelligence. For conceptually broader abilities like creativity and emotional competence, the choice of adequate accuracy criteria becomes less obvious. One widely accepted measure of creativity, or to be more precise creative potential, is originality in divergent thinking tasks, which shows good reliability and validities when scored adequately (Benedek et al., 2013, Diedrich et al., 2018) and has served as accuracy criterion in past research (Neubauer et al., 2018, Pretz and McCollum, 2014, Vazire, 2010). Therefore, we used originality in the widely applied alternative uses task (AUT; Guilford, 1967) as accuracy criterion for creativity. Emotional management abilities are typically measured by confronting individuals with hypothetical situations and asking them how they would change or maintain their emotions (Mayer et al., 2004). Hence, tests of emotional management abilities like the respective subscale of the Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT; Mayer et al., 2003), the Situational Test of Emotional Management (STEM; MacCann & Roberts, 2008) or the Typical-performance Emotional Management Test (TEMT; Freudenthaler & Neubauer, 2005) belong to the family of situational judgement tests (SJTs)1. Maximum performance tests of emotional management like the MSCEIT or STEM, which ask the individual to judge the most effective actions in each situation, have been criticized for measuring a person’s knowledge about how to behave in emotional situations instead of their actual regulative behavior (Freudenthaler & Neubauer, 2005; see also Brackett et al., 2006). Therefore, we used a typical performance situational judgment test comprised of subscales for intra- and interpersonal emotional management.

In line with past studies (Beer and Vazire, 2017, Neubauer et al., 2018, Vazire, 2010), we considered positive correlations between estimates and performance starting from .2 to indicate relevant levels of accuracy. A correlation of this size represents a typical effect in the individual differences literature (Gignac & Szodorai, 2016) and seems like a reasonable threshold, given average effects found in research looking at the accuracy of self- and other-estimates of abilities (e.g., Denissen et al., 2011, Zell and Krizan, 2014).

We considered estimate*performance correlation coefficients that differed in at least .15 to indicate relevant differences in accuracy between two perspectives. Vazire (2010) proposed that differences in accuracy correlations of more than .15 can be considered as substantial, given that this number is close to one standard deviation in effect size distributions in personality and social psychology (see Richard et al., 2003). It is also slightly higher than one standard deviation in the distribution of self-estimate*performance correlations for different abilities (Zell & Krizan, 2014). To illustrate the practical importance of a difference of .15, we show binomial effect size displays (BESDs) for correlations of various sizes in Table 2. BESDs are an intuitive method to evaluate the size of correlations (Rosenthal & Rubin, 1982; see Funder & Ozer, 2019 for a recent discussion) but have also sparked some controversy (see Hall et al., 2008). Nevertheless, if we think of both measured and estimated abilities as dichotomous constructs, BESDs can provide rough estimates of the proportion of individuals that are correctly characterized as high- or low-performers based on their own or someone else’s judgment (for a similar application of BESDs see Naumann et al., 2009). Table 2 shows four BESDs for estimate*performance correlations of .05, .20, .35, and .50 in a sample of 200 and, therefore, illustrates the impact of correlational differences of .15 for correlations of various strength. To provide an example, if the correlation between partner-estimated and measured verbal intelligence is .20, this indicates 60% correct predictions (e.g., both measured and estimated verbal intelligence is high). If the correlation for friend-estimates is .35, this relates to 67.5 % correct predictions. In this example, friends clearly have higher success at providing accurate feedback than romantic partners.