See attached document for literature review instructions. Use attached beginning paragraph for reference. Use the following citation, plus ones attached (4): Best, L. M., & Shelley, D. J. (2018).

Keywords:egocentrism, empathy, interpersonal accuracy, perspective taking, social cognition Supplemental materials:http://dx.doi.org/10.1037/pspa0000115.supp Understanding the minds of others is essential for social func- tioning, but another person’s mind is one of the most complicated systems that any person will ever think about. Just consider the numbers. The average human brain contains roughly one hundred billion neurons connected to anywhere between one thousand and 10 thousand other neurons through synapses that can be in a variety of excitatory or inhibitory states. Based on these figures, neuroscientists calculate that a human brain could be in more possible brain states than there are elementary particles in the known universe (Ramachandran, 2004, p. 3). Given the complex- ity of another person’s mind, what strategy should people use to understand the mind of another person more accurately?

One strategy is so routinely endorsed that its effectiveness seems taken for granted: perspective taking. That is, to understandanother person’s mind accurately you have to overcome your own egocentric perspective, “put yourself in another person’s shoes,” and try to perceive a situation from another person’s point of view.

This suggestion appears in politics, as when Barak Obama argued before the United Nations, “the deadlock [between the Israelis and Palestinians] will only be broken when each side learns to stand in each other’s shoes.” It appears in best-selling wisdom about human relations, as whenDale Carnegie (1936)lists the principles that will teach youHow to Win Friends and Influence People.Principle #8 is “a formula that will work wonders for you:....Tryhonestly to see things from the other person’s point of view.” And, accord- ing to a survey we conducted, it appears so routinely in people’s intuitions as to qualify as genuine common sense.

In this survey, 336 Amazon.com Mechanical Turk workers read about a series of experiments we conducted in which participants completed one of eight tests of interpersonal understanding (de- scribed later in detail). Four tests were taken from the existing scientific literature: The Mind in Eyes Test (Baron-Cohen, Wheel- wright, Hill, Raste, & Plumb, 2001), the Diagnostic Analysis of Nonverbal Behavior for faces (DANVA-faces,Nowicki & Duke, 1994), the Diagnostic Analysis of Nonverbal Behavior for postures (DANVA-postures,Nowicki & Duke, 1994), and the Fake Smiles Test (BBC science website; e.g.,Bernstein, Sacco, Brown, Young, Tal Eyal, Department of Psychology, Ben Gurion University of the Negev; Mary Steffel, D’Amore-McKim School of Business, Northeastern University; Nicholas Epley, Booth School of Business, University of Chicago.

Correspondence concerning this article should be addressed to Tal Eyal, Department of Psychology, Ben Gurion University of the Negev, Beer- Sheva 84105, Israel. E-mail:[email protected] This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Journal of Personality and Social Psychology © 2018 American Psychological Association2018, Vol. 114, No. 4, 547–571 0022-3514/18/$12.00http://dx.doi.org/10.1037/pspa0000115 547 & Claypool, 2010). Four additional tests were relatively routine social judgments: predicting a romantic partner’s consumer atti- tudes, movie preferences, activity preferences, and joke prefer- ences. Participants were randomly assigned to read a short descrip- tion of just one of these tests and were presented with one sample item. Participants then predicted which of two groups of people was more accurate in an experiment: people in a control condition who simply completed the test without further instruction, or people in a perspective taking condition who were asked to com- plete the test while “trying to adopt the perspective of the other person, putting yourself into the other person’s shoes as if you were that person.” Participants predicted the outcome of the ex- periment by choosing one of three options: “Condition 1 (Control) did significantly better,” “Condition 2 (Perspective Taking) did significantly better,” or “No significant difference between the two conditions in performance on the test.” As shown inFigure 1, most participants predicted that partici- pants in the perspective taking condition were more accurate than those in the control condition (67.75%). Few believed that partic- ipants in the control condition were more accurate (16%) or that participants in the two conditions differed in accuracy (16.25%).

Even those who are unlikely to have ever read Dale Carnegie’s book seem likely to believe in his “formula that will work wonders for you.” Despite a large scientific literature on the consequences of perspective taking in social interaction, whether perspective taking actuallyincreases accurate insight into the mind of another person is unclear. Many experiments test how perspective taking affects social cognition or interpersonal interaction. Very few measure the accuracy of interpersonal judgments. Those that do provide incon- sistent results, or confound egocentrism and accuracy, making it difficult to assess whether perspective taking merely shifts per- spective or actually increases accurate insight. Here we report a large number of experiments that test whether perspective taking increases interpersonal accuracy, using the very same tests from the pretest described above plus several others. These experiments are important because they are the first to systematically examine the validity of a widely endorsed strategy for increasing interper- sonal insight. They make an important theoretical advance byclarifying the nuanced consequences of a frequently studied topic in social psychology, thereby providing a better understanding of how perspective taking may affect interpersonal interactions.

These experiments also offer practical advice about how to under- stand the mind of another person more accurately. Perspective taking may indeed work wonders for you in social life. Is increas- ing accurate insight into the mind of another person one of them? Known Consequences of Perspective Taking Each person views the world from a potentially unique vantage point, collecting information through physical senses and inter- preting it through his or her own beliefs, attitudes, knowledge, experiences, and personality. Children become aware of their unique perspective as they age because they learn that others sometimes evaluate the world differently. This learning develops a highly sophisticated capacity to imagine another person’s unique perspective in adulthood, a capacity for social cognition that seems unmatched by any other species (Herrmann, Call, Hernández- Lloreda, Hare, & Tomasello, 2007).

Having a capacity and using that capacity, however, are two very different things. Considering another person’s perspective does not seem to be automatic and effortless, but instead requires time, motivation, and attentional resources to execute. Anything that reduces the time, inclination, or attention available for per- spective taking increases reliance on a relatively automatic ego- centric default in judgment (Epley, Keysar, Van Boven, & Gilov- ich, 2004;Epley, Morewedge, & Keysar, 2004;Karniol, 2003).

Likewise, explicitly encouraging perspective taking, by instructing people to “put themselves in another person’s shoes” and imagine another’s thoughts and feelings as if they were this other person, reliably affects people’s inferences and actions toward others com- pared with receiving no explicit encouragement.

Existing research on perspective taking typically does not assess interpersonal accuracy, but instead measures intrapersonal conse- quences that follow directly from being asked to shift from an egocentric perspective to an allocentric perspective. For instance, people who are explicitly instructed to attend to another’s perspec- tive are more likely to engage in deliberate thinking (e.g.,Epley et 0 10 20 30 4050 6070 80 Mind in Eyes DANVA FacesDANVA PosturesFake Smiles Consumer attitudesMovie preferencesActivity preferencesJoke preferences % of Participants Control Bette r PT Better The Same Figure 1.Percentage of participants who predicted that accuracy would be higher in the perspective taking condition, higher in the control condition, or equally the same in both conditions (Pretest). This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 548 EYAL, STEFFEL, AND EPLEY al., 2004;Todd, Bodenhausen, Richeson, & Galinsky, 2011;Todd, Galinsky, & Bodenhausen, 2012), mimic another person (Char- trand & Bargh, 1999;Genschow, Florack, & Wanke, 2013), report empathizing with another person’s emotional state (Batson, Early, & Salvarni, 1997;Davis, 1983;Maner et al., 2002; cf.Vorauer & Sasaki, 2009), take on another person’s stereotypic attributes (Ga- linsky, Wang, & Ku, 2008), and rely less on egocentric defaults in judgment (Caruso, Epley, & Bazerman, 2006;Savitsky, Van Boven, Epley, & Wight, 2005;Steffel & LeBoeuf, 2014;Wade- Benzoni, Tenbrunsel, & Bazerman, 1996;Yaniv & Choshen- Hillel, 2012;Zhang & Epley, 2009). Imagining oneself as another person also increases a sense of similarity with the other person (Davis, Conklin, Smith, & Luce, 1996), reduces the use of group- based stereotypes when evaluating others (Galinsky & Moskowitz, 2000), and reduces prejudice toward outgroups (Todd, Boden- hausen et al., 2011). In negotiations, perspective taking can in- crease coordination and cooperation, improving outcomes for both sides in contexts where a purely self-focused approach is detri- mental (Galinsky, Maddux, & Gilin, & White, 2008;Gilin, Mad- dux, Carpenter, & Galinsky, 2013;Trötschel, Huffmeier, Loschelder, Schwartz, & Gollwitzer, 2011).

All of these results suggest that being told to put oneself into another’s perspective may result in increased interpersonal accu- racy. First, deliberation increases accuracy on many decisions (e.g.,Payne, Bettman, & Johnson, 1988, but seeAmbady & Gray, 2002andHall et al., 2009). Second, mimicking another’s facial expression or body language could increase emotion recognition accuracy (e.g.,Niedenthal, Brauer, Halberstadt, & Innes-Ker, 2001;Oberman, Winkielman, & Ramachandran, 2007;Stel & van Knippenberg, 2008; cf.,Hess & Blairy, 2001;Cheung, Slotter, & Gardner, 2015). Third, shifting attention to another’s perspective (Yaniv & Choshen-Hillel, 2012;Zhang & Epley, 2009) may lead perceivers to focus on cues that yield more accurate judgment.

Fourth, perspective taking can create a merging of one’s cognition, emotion, motivation, and action with those of another person (Davis et al., 1996), thereby increasing the sense of similarity to that person and strengthening relational bonds (Galinsky, Ku, & Wang, 2005). Unknown Consequences of Perspective Taking At first glance, merging of self and another by reducing ego- centrism and decreasing stereotyping would seem to qualify as evidence of more accurate insight. However, most existing exper- imental research examines the psychological consequences of per- spective taking only in the mind of the perspective taker. Without measuring the mind of the person whose perspective was taken, researchers cannot tell whether perspective taking increases accu- racy in judgment or not. Perspective taking may increase the tendency to feel the pain another person ispresumedto be feeling (Batson et al., 1997), but does it increase the accuracy of recog- nizing how much pain another person isactuallyfeeling? Adopt- ing an adversary’s perspective in a negotiation could improve outcomes in some specific settings (Galinsky et al., 2008;Gilin et al., 2013), but does it do so by increasing insight into the other side’s sophisticated preferences or through some other mechanism (such as an increased willingness to cooperate, or incorporating another person’s known preferences into one’s own behavior)?Reducing a bias like egocentrism or stereotyping is not the same as increasing accuracy in judgment, even though evidence of the former could easily be mistaken for direct evidence of the latter. A bias in judgment is a systematic tendency that departs from a normative standard. The normative standard could be logical, rational, or moral, but it need not be accuracy. Likewise, reducing an egocentric bias or reliance on a stereotype could increase accurate insight into the mind of another person, but it does not have to. For instance, one of the most reliable egocentric biases in judgment is a tendency to assume that others’ attitudes and pref- erences are relatively similar to one’s own (Ross, Greene, & House, 1977). If two people actually have verydifferentattitudes, then reducing egocentrism through perspective taking should log- ically increase accuracy at predicting another person’s beliefs, not necessarily because a person has achieved genuinely greater in- sight into the mind of another person but rather because they have simply relied less on a known source of error. If, however, two people actually have verysimilarattitudes, then reducing egocen- trism to an equivalent degree through perspective taking could decrease accuracy. In one experiment consistent with the latter possibility, married couples would have been more accurate pre- dicting each other’s preferences if they simply projected their beliefs completely onto their partner (Hoch, 1987). Decreasing egocentrism in this experiment could have decreased accuracy because married couples tend to have very similar beliefs.

Reducing reliance on stereotypes also does not necessarily in- crease accuracy (Jussim, Crawford, & Rubinstein, 2015). When beliefs about a group (that is, a stereotype) contain some degree of accuracy, such as believing that tigers are dangerous but rabbits are not, then reducing reliance on it could decrease the overall accu- racy of judgments about a specific individual. For instance, per- spective taking in one experiment reduced the tendency of rela- tively young participants to rely on stereotypes about the elderly when evaluating a relatively old person (Galinsky & Moskowitz, 2000). However, age-related stereotypes appear to contain a large degree of accuracy, qualifying as perhaps the most accurate ste- reotype that psychologists have identified (Chan et al., 2012).

Reducing a young person’s use of an age-related stereotype when thinking about an elderly person does not necessarily mean that she will evaluate an elderly person more accurately. In fact, she might evaluate an elderly personlessaccurately. A person standing in front of a wild tiger who fails to consult his tiger stereotype is unlikely to gain more insight into the tiger’s likely behavior.

Without measuring the actual attitudes or experience of an elderly person, or the behavioral proclivities of a tiger, a researcher cannot tell whether reducing reliance on a stereotype increases the accu- racy of judgment or not.

Unfortunately, existing experiments often subtly confound a reduction in bias with an increase in accuracy because researchers purposely study contexts in which people’s perspectives are known to diverge. These include known perspective gaps between buyers and sellers (Galinsky et al., 2008), givers and receivers (Adams, Flynn, & Norton, 2012;Baskin, Wakslak, Trope, & Novemsky, 2014;Cavanaugh, Gino, & Fitzsimons, 2015;Flynn & Adams, 2009;Gino & Flynn, 2011;Teigen, Olsen, & Solås, 2005; Zhang & Epley, 2009,2012), Republicans and Democrats (Van Boven, Judd, & Sherman, 2012), speakers and listeners (Stinson & Ickes, 1992), actors and observers (Davis et al., 1996;Pronin & Ross, 2006), or negotiators with opposing incentives (Epley, This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 549 PERSPECTIVE MISTAKING Caruso, & Bazerman, 2006;Galinsky & Mussweiler, 2001; Trötschel et al., 2011). In these nonrandomly selected situations where there is a known egocentric bias that creates a systematic error in judgment, reducing an egocentric bias will increase accu- racy by necessity in much the same way that decreasing a bias to pick “tails” in a coin flip would increase accuracy in predicting a two-headed coin. For instance, if Republicans and Democrats are known to have opposing views on an issue, and researchers select only this issue to study instead of a randomly sampled set of issues that vary in the degree of opposition, then reducing egocentrism by encouraging perspective taking would increase accuracy by neces- sity even if a person does not actually gain any new insight into the mind of the opposition. Increased accuracy from perspective tak- ing would be reflected in an ability to differentiate between atti- tudes that truly differ and attitudes that do not. Testing whether explicit perspective taking actually increases understanding of another’s mind requires measuring sensitivity to the actual mental states of another person in cases where two minds are not already known to have systematically opposing viewpoints.

The surprisingly few published experiments that actually do measure interpersonal accuracy following perspective taking yield inconsistent results. In one involving a dictator game (Gilin et al., 2013), the authors report that encouraging perspective taking in- creased participants’ ability to accurately identify good potential game partners from bad ones (defined as partners who were likely to be generous vs. selfish) based on cognitive appeals, compared with participants who were asked to empathize with their partner.

However, the “partners” in the one experiment that measured accuracy (Study 4) were hypothetical rather than real, and accu- racy was defined as agreement with the authors’ assessment of these hypothetical appeals rather than agreement with actual be- havior of real people. Nevertheless, these results suggest that perspective taking might focus attention on cues that increase accuracy in judgment.

Other results suggest no increase in accuracy following perspec- tive taking or even a decrease in accuracy. In two different exper- iments, participants were asked to predict how attractive a member of the opposite sex would evaluate them. Being explicitly asked to adopt an observer’s perspective did not significantly increase people’s ability to accurately predict others’ evaluations of them (Eyal & Epley, 2010). In a series of competitive negotiations, adopting the perspective of an opponent led participants to over- estimate how selfish their partners would be compared with a control condition, suggesting less accuracy following perspective taking (Epley et al., 2006). In a study of close relationships, encouraging perspective taking increased the tendency to overes- timate how transparent one’s values, preferences, traits, and feel- ings were to a close relationship partner (Vorauer & Sucharyna, 2013).

These results do not invalidate the common wisdom and occa- sional experimental evidence that perspective taking increases accuracy in social judgment, but these results along with method- ological confounds and potential misinterpretations suggest that the common wisdom about putting oneself in another person’s shoes deserves systematic empirical attention. On the one hand, being explicitly asked to engage in perspective taking could in- crease accuracy in interpersonal judgment by highlighting accurate information that a person might otherwise overlook. On the other hand, being explicitly asked to engage in perspective taking mighthave no meaningful effect on accuracy if the information people consult is not systematically more accurate than the information they would have consulted without being asked to engage in perspective taking. In general, we would expect interpersonal accuracy to increase only when people get additional information about another person that is more accurate than what they would have consulted otherwise. Our current experiments test whether perspective taking does this reliably across many different con- texts, or not. Answering this question is essential for developing accurate theories of the consequences of perspective taking in social interactions. Preview of Current Experiments We report the results of a long process of testing many different methods and measures to examine whether or not perspective taking systematically increases interpersonal accuracy. We began by using direct tests of interpersonal accuracy taken from the empirical literature that both our pretest participants and existing psychological theory predict would increase accuracy. From our very first experiments, we identified reliable effects of perspective taking on some measures, including increased self-reported effort to take another person’s perspective, increased mental effort (e.g., greater response times), and occasionally, increased confidence in judgment. However, we found no reliable increases in accuracy. If anything, accuracy was somewhat worse (and sometimes signifi- cantly worse) among perspective takers than among control par- ticipants. These initial results led us on a long empirical trail of testing whether any theoretically relevant measure of accuracy would benefit from perspective taking. Our selection of experi- mental stimuli was guided by presumed mechanisms by which perspective taking could increase accuracy in an effort to be as comprehensive as possible, using both standardized measures from the existing literature as well as more naturalistic tests derived for our purposes. Our experiments tested accuracy among strangers, acquaintances, friends, and spouses. Our experiments found no evidence that the cognitive effort of imagining oneself in another person’s shoes, studied so widely in the psychological literature, increases a person’s ability to accurately understand another’s mind. Of course, it is always possible that our experiments failed to test just the right measure, or the precise context in which perspective taking could increase accuracy. We simply note that our experiments involved contexts in which we, and our pretest participants, expected that perspective taking could plausibly in- crease accuracy. Indeed, our pretest participants predicted signif- icantly more accurate judgments in the perspective taking condi- tion on every measure we asked them about.

Because of the large number of experiments we conducted, the main text will describe each of the 25 experiments in general terms and report only the primary analyses for the comparison between perspective taking and control conditions. The Supplemental Ma- terials describe details for each experiment including additional conditions, measures, and secondary analyses. All data are publi- cally available online (https://osf.io/4k7tv/).

We divide our experiments into three groups. The first group (Experiments 1–15) includes standard interpersonal accuracy tests between strangers taken largely from the existing experimental literature that could be affected by perspective taking based on current theorizing. The second group (Experiments 16 –24) in- This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 550 EYAL, STEFFEL, AND EPLEY cludesmore naturalistic tests between people who were familiar with each other, or who had meaningful information about another per- son’s potentially unique perspective. In these cases, someone engag- ing in perspective taking might have more information about another person’s perspective to guide their thinking and might therefore benefit more from considering another person’s perspective. These relationships included romantic partners, friends, spouses, or strangers following a get-acquainted conversation who were trying to predict another person’s attitudes, preferences, or beliefs in a variety of different domains. Because each of these experiments included a comparison between a perspective taking condition and a control condition, we report the primary results from these two sets of experiments in two meta-analyses (plus a meta-analysis of all exper- iments in the General Discussion). We report additional experiment- specific analyses in the Supplemental Materials.

A final experiment (Experiment 25) compares perspective tak- ing to a more direct approach to increasing accuracy, which we refer to asperspective getting(see alsoZhou, Majka, & Epley, 2017). This experiment demonstrates that it is indeed possible to increase interpersonal accuracy, tests the degree to which people are aware of the effectiveness of different prediction strategies, and suggests a subtle distinction that is critical for both scientific theorizing about the consequences of perspective taking and for attempting to understand the mind of another person more accu- rately in everyday life. Experiments 1–15: Standard Tests of Interpersonal Accuracy Participants completed standard tests that assess people’s ability to determine others’ mental states by viewing their eyes (the Mind in the Eyes Test,Baron-Cohen, et al., 2001), facial expressions (DANVA-faces,Nowicki & Duke, 1994; Fake Smiles,Bernstein, et al., 2010), or body postures (DANVA-postures;Nowicki & Duke, 1994). Participants also completed a test of lie detectionusing a standard experimental procedure. We chose commonly used tests from the existing interpersonal accuracy literature (e.g., Bernstein et al., 2010;Castelli et al., 2010;Galinsky, Magee, Inesi, & Gruenfeld, 2006;Nowicki & Duke, 1994;Ruben & Hall, 2013; Van Doesum, Van Lange, & Van Lange, 2013).

Participants were randomly assigned to one of several different conditions across these experiments. In each experiment, one group was asked to take the target’s perspective, following the standard instructional manipulation used in the existing psycho- logical literature (e.g.,Batson et al., 1997;Galinsky & Moskowitz, 2000). A control group in each experiment received no special instructions. Other groups across experiments were asked to apply a different strategy that we selected for a specific theoretical reason (e.g., encouraging participants to think hard, to mimic the target’s expressions, to rely on their own feelings or intuitions).

We chose comparison conditions that we believed would help explain our observed results, either because of patterns we ob- served in our data (e.g., perspective taking increased effort com- pared with control, so we encouraged participants in one condition to think hard) or because of assumptions in the literature about how perspective taking operates (e.g., via mimicry). We also included one test that measures egocentrism directly (and con- founds it with accuracy): a false-belief task (Birch & Bloom, 2007). We included this test simply to confirm, consistent with past research, that explicit perspective taking can reduce egocen- tric biases in judgment. It was not included in the meta-analysis on accuracy because it is confounded with a reduction in egocentrism. Method Participants.We sampled participants in Experiments 1–15 (N 1476) from a wide range of populations: Undergraduate students from a non–American university (non–US U.), under- graduate students from an American university (US U. #1), people in the community (Community), and MTurk workers (SeeTable 1 Table 1 Demographics and Meta-Analysis on Manipulation Check for Perspective Taking Versus Control Conditions, Experiments 1–15 Experiment Task Location (N) Mean age#of women ControlPT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 1 DANVA Faces Non-US U. (57) 23.23 (1.46) 46 4.14 (1.58) 5.55 (.99) 1.07 .28 .52 1.63 3.79 .001 2 DANVA Postures Non-US U. (124) 23.42 (1.65) 81 5.03 (1.45) 5.38 (1.28) .26 .18 .10 .61 1.42 .156 3 DANVA Faces Non-US U. (62) 23.78 (2.07) 72 5.19 (1.72) 5.45 (.85) .19 .26 .31 .69 .75 .452 4 DANVA Faces US U. #1 (88) 19.91 (3.01) 5.29 (2.91) 7.44 (1.89) .87 .22 .44 1.31 3.91 .001 5 DANVA Faces Non-US U. (80) 23.00 (.91) 72 5.03 (1.54) 5.55 (1.24) .37 .23 .07 .81 1.65 .099 6 DANVA Faces Non-US U. (61) 24.80 (1.69) 33 4.61 (1.84) 5.77 (94.) .79 .27 .27 1.31 2.97 .003 7 DANVA Faces MTurk (109) 29.61 (9.34) 33 6.74 (2.66) 8.62 (2.32) .75 .20 .36 1.14 3.79 .001 8 DANVA Faces Non-US U. (57) 24.19 (1.73) 33 4.89 (1.47) 5.69 (1.31) .57 .27 .04 1.10 2.12 .034 9 Mind in the Eyes Non-US. U (76) 24.77 (4.36) 39 5.24 (1.15) 5.63 (94.) .37 .23 .08 .83 1.61 .109 10 Mind in the Eyes Non-US. U (37) 25.11 (1.87) 21 5.42 (90.) 5.94 (73.) .63 .34 .03 1.29 1.88 .060 11 Mind in the Eyes Community (85) 38.35 (14.87) 7.64 (1.99) 8.19 (.207) .27 .22 .16 .70 1.24 .217 11 DANVA Faces Community (84) 38.35 (14.87) 6.02 (2.29) 7.24 (2.17) .55 .22 .11 .98 2.48 .013 12 Fake Smiles Community (70) 33.93 (13.09) 36 6.61 (2.60) 7.56 (1.85) .42 .24 .06 .89 1.73 .083 13 Fake Smiles Community (61) 34.52 (13.47) 33 6.39 (2.57) 7.13 (2.36) .30 .26 .21 .80 1.16 .245 14 Fake Smiles Non-US U. (55) 23.16 (1.53) 41 7.11 (2.52) 8.14 (1.11) .53 .27 .01 1.07 1.94 .052 15 Detecting Lies Community (81) 39.46 (15.32) 40 7.20 (2.53) 7.83 (2.00) .28 .22 .16 .71 1.24 .217 Total.49 .06 .37 .62 7.90 .001 Note. Participants made their ratings on a 7-point scale in Experiments 1–3, 5, 6, and 8 –10 and on an 11-point scale in Experiments 4, 7, and 11–15.

Participants did not report their gender in Experiments 4 and 11. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 551 PERSPECTIVE MISTAKING for sample sizes and demographics for each experiment). Our first experiments targeted sample sizes of typically 30 participants per cell, but we used larger sample sizes in later experiments to test the robustness of a null result. The only exception for this rule is a sample of 37 participants in Experiment 10 for which we stopped data collection before completion due to technical problems, and include it here for the sake of completeness. We sought to maxi- mize power by running multiple experiments with varied samples all utilizing the same experimental manipulation of perspective taking. We present results only for the perspective taking and control conditions in the main text (N 1103). Results of other conditions are presented in the Supplemental Materials.

Interpersonal accuracy measures.Participants completed the experiments individually in a laboratory, except for Experi- ment 7 that was conducted online. To measure interpersonal ac- curacy, participants completed standard tests in which they were asked to identify people’s feelings, thoughts, and intentions by watching a target’s picture or video. All participants completed one test, except for Experiment 11 in which participants completed both the DANVA-faces and the Mind in the Eye Test, Experiments 4, 5, and 8 in which participants completed the DANVA-faces and the false-belief task, and Experiment 13 in which participants completed both the Spot the Fake Smiles Test and the false-belief task. All tasks were computerized, except for the false-belief task.

We describe each test below.

Diagnostic analysis of nonverbal accuracy (DANVA,Now- icki & Duke, 1994).We used two subtests of the DANVA: faces and postures. The DANVA consists of 24 pictures of male and female faces (Experiments 1, 3– 8, 11) or body postures (Experi- ment 2) expressing one of four basic emotions: happiness, sadness, anger, or fear. Participants indicated the emotion the person in the picture feels.

Reading the mind in the eye (ME,Baron-Cohen et al., 2001, Experiments 9 –11).This test consists of 36 black and white pictures of the area around the eyes of males and females. The actual task was preceded by one practice trial. Participants indi- cated which of 4 words (e.g., serious, ashamed, alarmed, bewil- dered) described what the person in the picture was thinking or feeling.

Spot the fake smile (Experiments 12–14).This task was obtained from the BBC science website (http://www.bbc.co.uk/ science/humanbody/mind/surveys/smiles) and has been used pre- viously in experiments (e.g.,Bernstein et al., 2010). The test consists of 20 videos, approximately 4 seconds long, showing an individual (13 men and 7 women) with an initially neutral expres- sion that shifts into a smiling expression and then returns to a neutral expression (10 Duchenne and 10 non-Duchenne smiles of 20 different models trained to activate the Zygomaticus muscles involved in genuine smiles or not). For each video, participants indicated whether the smile was genuine or fake. Each video was shown only once.

Lie detection (Experiment 15).We created this task based on a standard procedure for testing lie detection in the existing ex- perimental literature (Bond & DePaulo, 2006). This test consisted of 10 videos of individuals (6 men and 4 women) answering a question posed by a research assistant about their experiences and preferences (e.g., “What is your happiest childhood memory?

Please describe it briefly,” “What celebrity would you most like to meet? What would you say to them?”). Following the video,participants were reminded of the question the participant in the video was asked and then indicated whether they thought the answer was true or false. The order of videos was fixed. Half of the answers were true and half were false.

False-belief task (Experiments 4 –5, 8, 13).To test whether perspective taking reliably reduces egocentric biases in judgment, consistent with considerable amounts of prior research, we used a modified version of the false-belief task designed byBirch and Bloom (2007). Participants were handed two pictures, one at a time. The first picture portrayed a girl playing the violin beside a sofa. There were four containers in different sizes and colors (red, purple, blue, and green) in front of her. Participants read: “This is Vicki. She finishes playing her violin and puts it in the blue container. Then she goes outside to play.” The second picture portrayed a different girl holding a violin beside a different array of the same containers. Participants read: “While Vicki is outside playing, her sister, Denise, moves the violin to the red container.

Then, Denise rearranges the containers in the room until the room looks like the picture below.” Participants indicated the likelihood that Vicki would first look for her violin in each of the four containers. The percentage participants assign to the red box is an indication of egocentrism, because participants know that the violin has been moved to the red box but Vicki in the scenario does not know this.

Independent variables.Each experiment included a perspec- tive taking condition and a control condition in which participants were given no specific instructions. This served as our primary comparison in each experiment, and the focus of this paper. Seven experiments included additional conditions that tested the impact of other strategies. We added these additional conditions based on the results of our initial experiments that found a negative impact of perspective taking on accuracy. Because they did not yield very meaningful insights, we will describe these conditions briefly in the main text and present them in detail in the Supplemental Materials. Participants in each condition received a brief descrip- tion of the experimental task, and then were given additional instructions depending on their experimental condition.

Perspective taking conditions.Our primary manipulation asked participants to engage in perspective taking using instruc- tions taken from the existing literature (e.g.,Batson et al., 1997; Galinsky & Moskowitz, 2000). In particular, participants in the perspective taking conditions read: “While watching the pictures [videos], please think about the person in the picture [video]. Try to adopt the perspective of the person in the picture [video] as if you were the person who is answering the question. Do your best to adopt his or her perspective, putting yourself into the other person’s shoes as if you were that person. Remember that the person in the picture [video] may have a different perspective than you do as the viewer of the picture [video].” The perspective taking instructions for the false-belief task (Experiments 4, 5, 8, 13) were adapted to fit details of the task: “When answering the question, we would like you to do your best to adopt Vicki’s perspective, putting yourself into Vicki’s shoes as if you were her.

Remember that Vicki may have a different perspective than you do.” Additional conditions.Our initial results from Experiments 1 and 2 suggested that perspective taking might diminish accuracy.

We therefore introduced several additional conditions across ex- periments to explore this potential negative effect in more detail. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 552 EYAL, STEFFEL, AND EPLEY These included instructions to consult one’s own feelings (Exper- iment 3), to think especially hard (Experiment 3), to rely on one’s intuitions (Experiments 4, 14), to personally display the facial expressions posed in the pictures before guessing the emotions expressed (Experiment 6), to empathize with the person in each photo (Experiment 7), to consider similarities or differences (Ex- periment 8), to predict how most people would answer the question (Experiment 11), and to mimic the target’s facial expressions while observing each picture (Experiment 13). None of these conditions significantly increased accuracy compared with the control condi- tion, but two conditions significantly decreased accuracy com- pared with the control condition (consult one’s own feelings in Experiment 3 and follow your intuition in Experiment 4). These experimental conditions did not prove to be especially informative.

We therefore highlight notable findings from these additional conditions in the General Discussion and report these results in full in the Supplemental Materials.

Additional measures.We also collected several additional measures to provide further tests of reliable consequences of perspective taking:

Manipulation check.Participants completed a manipulation check to assess how hard they tried to adopt the other person’s perspective. We used a 7-point scale (1 not at all,7 very much) in Experiments 1–3, 5, 6, and 8 –10. We used an 11-point scale (1 not at all,11 extremely) in Experiments 4, 7, and 11–15. Participants in some experiments completed additional manipulation checks consistent with the conditions we added to our basic perspective taking versus control comparison. These are described in the Supplemental Materials.

Difficulty.Participants reported how difficult they found the task to be. We used a 7-point scale (1 not at all,7 very much) in Experiments 1–3, 5, 6, and 8 –10 and an 11-point scale (1 not at all,7 extremely) in Experiments 4, 7, and 11–15.

Response times.We measured participants’ response times to complete the computerized tasks in all but Experiment 7 to provide an indirect measure of mental effort (with more time indicating more effort expended).

Confidence.Participants reported their confidence in judg- ment by indicating the number of responses they thought they predicted accurately. In Experiments 12–14, participants rated how confident they were with their answer after every one of the 20 predictions they made (1 just guessing,11 absolutely certain) and we computed an average confidence score.

Results Meta-analyses.Our primary interest was testing whether per- spective taking increases interpersonal accuracy. To ease presen- tation of such a large number of experimental results, we present only the primary comparisons between the perspective taking and control conditions on our primary outcomes: the manipulation check, accuracy, confidence, perceived difficulty, and response times. Because the experiments were run on diverse populations and used different tests of interpersonal accuracy, we conducted random effects meta-analyses using the Comprehensive Meta- analysis 2 software (Borenstein, Hedges, Higgins, & Rothstein, 2010) to identify the robust effects across all experiments. We did not observe reliable gender differences on accuracy or the impact of perspective taking in these 15 experiments, or those we reportin the remainder of this paper. We therefore do not discuss gender differences further. 1 Manipulation checks.Participants seemed to do as they were instructed. Those in the perspective taking conditions reported considering others’ perspective more than those in the control conditions across Experiments 1–15,d 0.49, 95% CI [0.37, 0.62],z 7.90,p .001 (seeTable 1). This significant result is important for understanding the consequences of perspective tak- ing on accuracy that we discuss next, because it demonstrates that participants were indeed attempting to follow the critical experi- mental manipulation.

Accuracy.Participants in the perspective taking conditions were not significantly more accurate across Experiments 1–15 than participants in the control conditions. In fact, participants in the perspective taking conditions were significantlylessaccurate over- all than participants in the control conditions,d 0.26, 95% CI [ 0.40, 0.12],z 3.74,p .001. As can be seen inTable 2, this negative effect of perspective taking on accuracy, compared with the control conditions, was not especially robust across indi- vidual experiments. It was statistically significant in 4 of 17 instances, but even these 4 significant results were not reliable across replications of the same procedure. 2The Fake Smiles Test, for instance, produced one of the five significant negative effects of perspective taking on accuracy (Experiments 12), but one rep- lication yielded a significant result in theoppositedirection (Ex- periment 13). Although these experiments do not provide espe- cially reliable evidence that perspective taking systematically 1We observed significant gender effects in only four of these 25 exper- iments, and even these effects were inconsistent across experiments. In Experiment 5 we observed a significant gender perspective taking interaction,F(1,71) 4.05,p .048, p 2 .05. There was a marginally significant gender effect in the perspective taking condition—women were more accurate than men, (Ms 17.58 and 13.50,t(38) 1.84,p .078, d 0.60), but no gender effect in the control condition (Ms 18.82 and 19.50,t(38) 0.62,p .54,d 0.20). This difference might be driven by the small number of men compared with women in this sample (8 vs. 72). In Experiment 7 we observed a marginally significant main effect for gender such that women were more accurate than men (Ms 19.98 and 19.15),F(1,155) 21.40,p .072, p 2 .02. In Experiment 20 there was a Marginally Significant Gender Perspective Taking Interac- tion,F(1, 81) 3.58,p .062, p 2 .04. Women were more accurate than men in the perspective taking condition (Ms 1.13 and 1.38,t(38) 1.74, p .090,d 0.56) but not in the control condition (Ms 1.20 and 1.11, t(43) 0.80,p .43,d 0.24). Finally, in Experiment 21 there was a marginally significant main effect for gender,F(1, 85) 3.61,p .061, p 2 .04, but this effect was qualified by a Significant Gender Perspec- tive Taking Interaction,F(1, 85) 6.08,p .016, p 2 .07. Women were more accurate than men in the perspective taking condition (Ms 2.13 and 2.68,t(42) 3.09,p .004,d 0.95) but not in the control condition (Ms 2.19 and 2.12,t(43) 0.40,p .69,d 0.12). Given that we observed no reliable gender differences across our experiments, we do not discuss it further. 2We conducted heterogeneity tests to examine whether the effect sizes for accuracy obtained in the meta-analyses are more variable than expected from normal sampling variation. We obtained nonsignificant effects of heterogeneity in Experiments 1-15,Q(15) 21.07,p .14,I 2 28.80, Experiments 16-24,Q(12) 14.88,p .25,I 2 19.34, and also in all 25 experiments,Q(29) 38.27,p .12,I 2 24.23, indicating that dispersion in the effects of perspective taking on accuracy across experiments is not due to real differences in the experiments other than random error. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 553 PERSPECTIVE MISTAKING decreases accuracy, they provide no evidence whatsoever that perspective taking systematically increases accuracy. 3 Perceived difficulty and response times.Those in the per- spective taking conditions reported that their task was more diffi- cult than those in the control conditions across Experiments 1–15, d 0.16, 95% CI [0.04, 0.27],z 2.69,p .007 (seeTable 3).

Participants in the perspective taking conditions were also slower in their responses compared with participants in the control con- ditions,d 0.40, 95% CI [0.28, 0.52],z 6.50,p .001 (see Table 4). Results on these two measures, along with the manipu- lation check, suggest that participants in the perspective taking conditions were indeed trying harder to consider another person’s perspective than participants in the control conditions.

Confidence and overconfidence.Not only were participants in the perspective taking condition less accurate, they also believed they predicted fewer of their partner’s responses accurately com- pared with participants in the control conditions across Experi- ments 1–15,d 0.20, 95% CI [ 0.32, 0.09],z 3.48,p .001 (seeTable 5). This negative effect of perspective taking on confidence may be because participants had no knowledge about the targets they tried to mind read that they could use when encouraged to take their perspective, beyond the minimal infor- mation that appeared in the picture or video.

Having both accuracy and predicted accuracy measures al- lows us to calculate whether participants were systematically overconfident in their evaluations. To assess overconfidence, we subtracted the number of accurate responses from the pre- dicted number of accurate responses. Overall, participants were underconfident in their performance on these measures, d 0.17, 95% CI [ 0.30, 0.03],z 2.39,p .017. This was the case in all experiments but one (Experiment 15), in which participants were significantly overconfident,d 0.80, 95% CI [0.55, 1.05],z 6.27,p .001. In addition, perspec- tive taking did not significantly affect overconfidence, d 0.03, 95% CI [ 0.14, 0.09],z 0.45,p .65.Reducing egocentrism: The false-belief test.Consistent with prior research, perspective taking reliably decreased egocen- tric biases in the four experiments that included the false-belief task (seeTable 6). Perspective taking participants indicated that it was significantly less likely for the protagonist to look in the location suggested by an egocentric perspective than participants in the control condition,d 0.28, 95% CI [ 0.51, 0.05], z 2.35,p .019. Discussion In a series of 15 experiments, using standard tests of interper- sonal accuracy, an explicit instruction to engage in perspective taking reliably altered judgments in a manner consistent with the explicit instruction to shift perspective from their own to another’s perspective. Consistent with past research (Todd et al., 2012), this shift in perspective leads to more deliberation reflected in our studies by increased response time and greater perceived difficulty.

This reliable shift in perspective, however, does not systematically increase accuracy except in cases where egocentrism and accuracy are necessarily confounded (such as in the false-belief task). These findings suggest that the benefits of perspective taking for increas- ing accuracy may be very circumscribed, increasing accuracy only when an egocentric bias is known to be producing error.

3In Experiments 12–14 (Fake Smiles Test) and Experiment 15 (Lie Detection) we also computed accuracy using a detection theory sensitivity measure (d-prime, representing the difference between the proportion of hits and false alarms). The analyses yielded similar results to those ob- tained perspective taking effects on number of correct responses. In Ex- periments 12 and 14 the d-prime was directionally lower in the PT condition compared with control (Experiment 12:t(68) 1.28,p .204; Experiment 14:t(53) 0.70,p .485. In Experiment 13 the d-prime was significantly higher in the PT condition compared with control, t(59) 2.62,p .011. For the Lie Detection Test (Experiment 15) perspective taking did not have an effect on d-prime,t(79) 0.53,p .60. Table 2 Meta-Analysis on Accuracy (Number of Correct Responses) for Perspective Taking Versus Control Conditions, Experiments 1–15 Experiment Task Location (N) ControlPT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 1 DANVA Faces Non-US U. (57) 19.14 (2.29) 17.89 (2.41) .53 .27 1.06 .00 1.97 .049 2 DANVA Postures Non-US U. (124) 16.54 (2.55) 15.94 (2.35) .25 .18 .60 .11 1.36 .174 3 DANVA Faces Non-US U. (62) 18.48 (2.71) 17.74 (2.75) .27 .26 .77 .23 1.06 .288 4 DANVA Faces US U. #1 (88) 18.89 (2.17) 17.91 (2.74) .40 .22 .82 .02 1.86 .065 5 DANVA Faces Non-US U. (80) 18.93 (2.47) 17.38 (3.14) .55 .23 1.00 .10 2.41 .016 6 DANVA Faces Non-US U. (61) 19.52 (1.63) 18.40 (2.81) .49 .26 1.00 .02 1.88 .060 7 DANVA Faces MTurk (109) 19.75 (1.95) 19.06 (3.20) .26 .19 .64 .11 1.37 .172 8 DANVA Faces Non-US U. (57) 18.38 (2.67) 18.17 (3.02) .07 .27 .59 .45 .28 .781 9 Mind in the Eyes Non-US U. (76) 23.60 (4.02) 22.58 (3.60) .27 .23 .72 .18 1.16 .246 10 Mind in the Eyes Non-US U. (37) 25.42 (4.77) 24.83 (4.56) .13 .33 .77 .52 .38 .701 11 Mind in the Eyes Community (85) 25.11 (4.46) 23.65 (5.58) .04 .22 .46 .39 .16 .873 11 DANVA Faces Community (84) 18.24 (2.43) 18.14 (3.27) .29 .22 .72 .14 1.32 .188 12 Fake Smiles Community (70) 13.72 (2.17) 12.24 (2.09) .69 .25 1.18 .21 2.82 .005 13 Fake Smiles Community (61) 12.42 (2.93) 14.13 (2.54) .62 .26 .11 1.14 2.38 .018 14 Fake Smiles Non-US U. (55) 13.63 (2.31) 12.68 (2.04) .44 .27 .97 .10 1.60 .110 15 Detecting Lies Community (81) 4.93 (1.62) 4.80 (1.51) .08 .22 .52 .35 .37 .709 Total .26 .07 .40 .12 3.74 .001 Note. There are 24 items in DANVA, 36 items in the Mind in the Eyes, 20 items in the Fake Smiles, and 10 items in Detecting Lies. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 554 EYAL, STEFFEL, AND EPLEY Experiment 1–15 tested the impact of perspective taking on inter- personal accuracy using standardized measures of interpersonal ac- curacy taken from the existing experimental literature. Although both existing theory and intuition (as indicated by our pretest) suggest that perspective taking could increase accuracy on these tests, our exper- iments indicate that perspective taking increased mental effort and decreased egocentrism but did not reliably increase accurate insight into the mind of another person. These standardized tests enable precise and reliable accuracy measurement, but they are also ab- stracted from everyday life in a way that makes it difficult to take the perspective of the targets being evaluated. For instance, participants knew nothing about the targets or about the thoughts, feelings, atti- tudes, or context that targets were actually in. Perspective takingmight have been especially ineffective in these contexts because there was no unique information that participants could access when they shifted attention to the targets’ perspective. We next explore whether the weak negative relationship between perspective taking and accu- racy generalizes to more naturalistic contexts. Experiments 16 –24: Naturalistic Tests of Interpersonal Accuracy Experiments 16 –24 tested the impact of perspective taking on interpersonal accuracy using judgments that people are more likely to make in everyday life, including predictions of others’ sense of humor, opinions, and preferences. They also involved predictions Table 3 Meta-Analysis on Perceived Difficulty for Perspective Taking Versus Control Conditions, Experiments 1–15 Experiment Task Location (N) Control PT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 1 DANVA Faces Non-US U. (57) 2.79 (1.55) 3.24 (1.83) .27 .27 .26 .79 1.00 .319 2 DANVA Postures Non-US U. (124) 3.56 (1.43) 3.46 (1.55) .07 .18 .42 .29 .37 .709 3 DANVA Faces Non-US U. (62) 3.35 (1.79) 3.65 (1.58) .18 .26 .32 .68 .70 .485 4 DANVA Faces US U. #1 (88) 5.20 (2.38) 5.51 (2.00) .14 .21 .28 .56 .66 .510 5 DANVA Faces Non-US U. (80) 2.78 (1.33) 3.60 (1.41) .60 .23 .15 1.05 2.62 .009 6 DANVA Faces Non-US U. (61) 2.74 (1.63) 3.50 (1.53) .48 .26 .03 .99 1.85 .064 7 DANVA Faces MTurk (109) 4.23 (2.40) 4.63 (2.38) .17 .19 .21 .54 .87 .384 8 DANVA Faces Non-US U. (57) 2.86 (1.41) 3.21 (1.54) .24 .27 .28 .76 .89 .373 9 Mind in the Eyes Non-US U. (76) 3.84 (1.41) 4.03 (1.67) .12 .23 .33 .57 .54 .592 10 Mind in the Eyes Non-US U. (37) 3.84 (1.43) 4.33 (1.68) .32 .33 .33 .96 .95 .341 11 Mind in the Eyes Community (85) 6.79 (2.20) 7.09 (2.40) .13 .22 .30 .56 .59 .555 11 DANVA Faces Community (84) 5.52 (2.11) 5.50 (2.30) .01 .22 .43 .42 .04 .967 12 Fake Smiles Community (70) 6.14 (2.36) 6.59 (2.12) .20 .24 .28 .68 .83 .409 13 Fake Smiles Community (61) 6.81 (2.27) 6.50 (2.30) .14 .30 .72 .45 .45 .649 14 Fake Smiles Non-US U. (55) 6.52 (2.17) 6.75 (2.19) .11 .28 .45 .66 .38 .707 15 Detecting Lies Community (81) 7.59 (2.10) 7.68 (2.19) .04 .22 .39 .48 .19 .850 Total.16 .06 .04 .27 2.69 .007 Note.Participants made their ratings on a 7-point scale in Experiments 1–3, 5, 6, and 8 –10 and on an 11-point scale in Experiments 4, 7, and 11–15.

Table 4 Meta-Analysis on Response Times for Perspective Taking Versus Control Conditions, Experiments 1–15 Experiment Task Location (N) Control PT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 1 DANVA Faces Non-US U. (57) 78.95 (21.35) 102.37 (37.50) .76 .27 .23 1.30 2.78 .005 2 DANVA Postures Non-US U. (124) 101.17 (45.36) 122.23 (52.48) .43 .18 .07 .79 2.36 .018 3 DANVA Faces Non-US U. (62) 91.53 (42.59) 126.43 (71.32) .59 .26 .09 1.10 2.29 .022 4 DANVA Faces US U. #1 (88) 63.64 (12.59) 91.14 (45.94) .83 .22 .39 1.26 3.71 .001 5 DANVA Faces Non-US U. (80) 115.67 (48.30) 142.89 (55.37) .52 .23 .08 .97 2.30 .021 6 DANVA Faces Non-US U. (61) 76.65 (19.72) 76.71 (23.97) .00 .26 .50 .51 .01 .991 8 DANVA Faces Non-US U. (57) 93.47 (32.28) 109.62 (38.32) .46 .27 .07 .98 1.70 .089 9 Mind in the Eyes Non-US U. (76) 263.42 (83.01) 288.11 (126.01) .23 .23 .22 .68 1.01 .315 10 Mind in the Eyes Non-US U. (37) 251.88 (87.60) 279.32 (89.23) .31 .33 .34 .96 .94 .348 11 Mind in the Eyes Community (85) 133.96 (42.91) 143.32 (54.25) .19 .22 .24 .62 .88 .382 11 DANVA Faces Community (85) 102.01 (32.45) 121.33 (42.88) .51 .22 .08 .94 2.30 .021 12 Fake Smiles Community (70) 154.87 (21.02) 161.28 (23.78) .29 .24 .19 .76 1.19 .234 13 Fake Smiles Community (61) 159.27 (21.21) 163.98 (31.12) .18 .26 .33 .68 .69 .489 14 Fake Smiles Non-US U. (55) 191.41 (59.15) 200.42 (53.54) .16 .27 .37 .69 .59 .554 15 Detecting Lies Community (81) 27.05 (12.28) 35.30 (23.71) .44 .23 .00 .88 1.95 .051 Total.40 .06 .28 .52 6.50 .001 Note. We report the sum of the response times in seconds across the task. We did not measure response time in Experiment 7. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 555 PERSPECTIVE MISTAKING of actual interaction partners. Some involved evaluations of targets who were generally well known to our participants, such as a friend or spouse, and others involved evaluations of strangers after a brief get-to-know-you activity. These experiments involved con- texts where participants were likely to have at least some knowl- edge of their partners’ preferences, either because of existing relationship knowledge (e.g., a spouse’s preferences for going bowling or doing dishes) or group-based stereotype knowledge (e.g., a male or female partner’s reactions to movies, videos, or jokes targeted toward a stereotypically male or female audience).

Each experiment used a perspective taking manipulation similar to those in Experiments 1–15, and several provided instructions asking participants to take their target’s perspective in a somewhat different way. These variants allowed us to test whether our results were restricted to simply the most common experimental approach for encouraging perspective taking. Finally, all experiments except for Experiments 21 and 24 included an “egocentric” condition in which participants were asked to assume that the other person perceived the world exactly as they did themselves. This conditioncould provide a more extreme test of whether or not considering another’s perspective increases accuracy by including a condition that does precisely the opposite. Because this is not our primary focus, we briefly discuss these results in the General Discussion and present these results in full in the Supplemental Materials. Method Participants.One thousand, one hundred thirty-two individ- uals participated in Experiments 16 –24. Participants were under- graduates from an American university (US U. #2), MBA students, and people in the community, from two locations (Community #1, Community #2; seeTable 7for sample sizes and demographics for each experiment). Targeted sample sizes were typically 30 partic- ipants per cell, but we increased our samples sizes in subsequent versions of our experiments to test the robustness of a null result.

We made seven exclusions from the analyses: three participants received the wrong verbal instructions (Experiment 16), three participants did not have a partner (Experiment 21), and one Table 5 Meta-Analysis on Confidence (Predicted Number of Correct Responses) for Perspective Taking Versus Control Conditions, Experiments 1–15 Experiment Task Location (N) ControlPT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 1 DANVA Faces Non-US U. (57) 17.75 (2.78) 17.17 (3.58) .18 .27 .70 .34 .68 .496 2 DANVA Postures Non-US U. (124) 16.48 (3.86) 15.97 (3.91) .13 .18 .48 .22 .73 .465 3 DANVA Faces Non-US U. (62) 17.61 (3.99) 16.39 (3.79) .31 .26 .81 .19 1.23 .220 4 DANVA Faces US U. #1 (88) 17.31 (4.21) 16.23 (4.02) .26 .21 .68 .16 1.22 .221 5 DANVA Faces Non-US U. (80) 18.38 (3.26) 16.83 (4.31) .41 .23 .85 .04 1.80 .073 6 DANVA Faces Non-US U. (61) 18.77 (2.49) 18.07 (3.05) .25 .26 .76 .25 .98 .327 7 DANVA Faces MTurk (109) 19.33 (3.13) 18.04 (3.31) .40 .19 .78 .02 2.07 .038 8 DANVA Faces Non-US U. (57) 18.68 (2.53) 17.24 (3.93) .44 .27 .96 .09 1.63 .103 9 Mind in the Eyes Non-US U. (76) 23.11 (5.78) 21.59 (6.15) .26 .23 .71 .20 1.11 .269 10 Mind in the Eyes Non-US U. (37) 25.37 (5.69) 22.39 (5.87) .52 .33 1.17 .14 1.54 .123 11 Mind in the Eyes Community (85) 21.40 (7.45) 21.16 (6.12) .04 .22 .46 .39 .16 .871 11 DANVA Faces Community (84) 16.81 (3.90) 17.27 (4.18) .11 .22 .31 .54 .52 .602 12 Fake Smiles Community (70) 12.88 (3.26) 11.41 (4.43) .38 .24 .85 .09 1.57 .116 13 Fake Smiles Community (61) 12.17 (4.28) 10.91 (5.12) .27 .27 .80 .26 .98 .325 14 Fake Smiles Non-US U. (55) 11.98 (3.65) 11.43 (3.00) .16 .26 .67 .34 .64 .522 15 Detecting Lies Community (81) 6.39 (1.72) 6.95 (1.69) .33 .22 .11 .77 1.47 .142 Total .20 .06 .32 .09 3.48 .001 Note. There are 24 items in DANVA, 36 items in the Mind in the Eyes, 20 items in the Fake Smiles, and 10 items in Detecting Lies. Table 6 Meta-Analysis on the Mean Ratings of the Likelihood That Vickie Will Search in the Red Box First in the False-Belief Task for Perspective Taking Versus Control Conditions, Experiments 1–15 Experiment Location (N) Control PT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 4 US U. #1 (88) 20.62 (23.87) 16.90 (20.51) .17 .22 .59 .26 .78 .438 5 Non-US U. (80) 38.79 (29.03) 27.55 (22.19) .88 .23 .88 .01 1.92 .054 8 Non-US U. (57) 36.30 (30.93) 30.25 (26.26) .21 .27 .73 .31 .79 .428 13 Community (61) 27.13 (28.54) 19.61 (19.64) .31 .26 .82 .20 1.18 .237 Total .28 .12 .51 .05 2.35 .019 Note. Larger percentages indicate smaller egocentric bias. Thus, a negative sign of d indicates smaller egocentric bias in perspective taking condition compared to control. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 556 EYAL, STEFFEL, AND EPLEY Table 7 Demographics and Meta-Analysis on Accuracy (Absolute Difference Between Predicted Responses and Actual Responses) for Perspective Taking VersusControl Conditions, Experiments 16 –24 Experiment Task Location (N)#of womenLength of relationship (months) Mean age ControlPT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 16 Activities (partners) Community #1 (74) 37 103.18 (108.31) 32.65 (9.17) 1.21 (.31) 1.15 (.34) .17 .23 .63 .29 .74 .461 17 Activities (partners) Community #1 (66) 33 133.70 (107.80) 37.06 (12.18) 1.13 (.26) 1.33 (.44) .57 .25 .08 1.07 2.28 .023 18 Movies (strangers) US U. #2 (80) 41 20.25 (1.65) 1.03 (.31) 1.20 (.41) .49 .23 .05 .93 2.16 .031 19 Jokes (strangers) Community #2 (78) 39 28.71 (11.35) 1.24 (.40) 1.46 (.50) .47 .23 .02 .92 2.06 .039 20 Videos (strangers) US U. #2 (85) 46 20.12 (1.57) 1.16 (.37) 1.26 (.48) .24 .22 .19 .66 1.08 .281 21 Art (strangers) US U. #2 (92) 42 20.12 (1.81) 2.15 (.59) 2.31 (.64) .26 .21 .15 .67 1.24 .214 22 Opinions (partners) Community #2 (82) 42 101.79 (117.44) 35.28 (11.68) 1.56 (.35) 1.52 (.42) .10 .22 .53 .34 .43 .664 23 Opinions (partners) Community #2 (80) 38 131.84 (151.97) 38.81 (14.00) 1.71 (.50) 1.71 (.36) .01 .23 .45 .44 .02 .984 23 Opinions (strangers) Community #2 (79) 37 38.57 (13.93) 1.97 (.42) 1.98 (.38) .02 .23 .43 .46 .08 .938 24Performance appraisal simulation Burke’s chance MBA (101) 21.08 (16.60) 27.89 (19.05) .38 .20 .02 .78 1.89 .059 24Performance appraisal simulation Stanley’s chance MBA (101) 22.04 (17.53) 24.92 (22.30) .14 .20 .25 .54 .72 .475 24Performance appraisal simulation Burke’s impression MBA (101) 1.89 (1.62) 1.67 (1.68) .13 .20 .53 .26 .67 .506 24Performance appraisal simulation Stanley’s impression MBA (101) 1.92 (1.91) 2.04 (1.70) .07 .20 .33 .46 .33 .740 Total.16 .07 .03 .30 2.44 .015 Note. Larger absolute differences indicate less accuracy. In Experiment 24 participants did not report their age and gender.This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 557 PERSPECTIVE MISTAKING participant’s predictions of a stranger because he or she predicted the preferences of a hypothetical stranger (Experiment 23). We were also unable to calculate accuracy in 24 instances: 10 in- stances in which participants did not make predictions or their partners did not report own responses (Experiments 16, 19, 22, and 23) and 14 instances in which we could not match participants to a partner because they mis-entered their Participant ID or because they were part of a triplet and it was unclear whose preferences they predicted (Experiment 20). Additionally, we could not calcu- late correlational accuracy in nine instances in which participants or their partners gave the same response to all items (Experiments 19 and 20). We present results only for the perspective taking and control conditions in the main text (N 825), as we did for Experiments 1–15.

Participants completed each experimental session in pairs where each person served as both a “predictor” and a “target.” The pairs varied in their relationship status across experiments. Some were romantic partners (Experiments 16, 17, and 22), and others were strangers who had a short introductory conversation (Experiments 18 –21, and 24). Pairs were strangers of the opposite sex in Experi- ments 19 and 20, and included both heterosexual romantic partners and strangers of the opposite sex in Experiment 23. There were a few exceptions to these rules: four participants were not of the opposite gender and six participants were not strangers (Experiment 19), 34 participants were not of the opposite gender or were part of a triplet and six participants could not be matched to a partner (Experiment 20), four participants were not romantic partners (Experiment 22), eight participants predicted a romantic partner who was not of the opposite gender, two participants predicted someone who was not a romantic partner, and eight participants predicted a stranger’s prefer- ences who was not of the opposite gender (Experiment 23). Excluding these additional participants does not change the results in any mean- ingful way, and so we report the results with these participants included.

Materials and procedure.Heterosexual romantic partners were recruited together in Experiments 16 –17 and 22. Individuals were recruited separately and paired with a stranger in Experi- ments 18 –21. Heterosexual romantic partners were recruited to- gether and paired with another couple that they did not already know in Experiment 23. Experiments involving pairs of strangers began with a get-acquainted session in which each participant introduced himself/herself guided by a series of questions: “Where are you from?,” “What are you doing in the lab/museum today?,” and “What are you doing when you are not at the lab/museum?” Participants in Experiments 16 –23 predicted their partner’s responses and stated their own responses. Participants in Experi- ments 18 –20 also predicted the responses of an average man and an average woman. Participants in Experiments 16, 18 –19, and 22–23 first predicted their partner’s responses for all of the items, and then stated their own responses for all of the items. We counterbalanced the order of these ratings in Experiments 17 and 21. Participants in Experiment 20 predicted their partner’s re- sponses, stated their own responses, predicted the responses of an average man, and predicted the responses of an average woman for each item before moving to the next. Participants in Experiment 24 first gave their own impressions based on the role they played and then predicted their partner’s response based on the partner’s role.

Interpersonal accuracy measures.These experiments as- sessed interpersonal accuracy on predictions of partners’ responsesto six different judgments: activities, movies, jokes, videos, art, opinions, and a performance appraisal simulation. We summarize each task below. All stimuli from questionnaires are publically available online athttps://osf.io/4k7tv/.

Activities (Experiments 16 & 17).Participants rated how much their partner liked or disliked 37 activities on 7-point scales (1 dislike very much,4 neutral or don’t know,7 like very much) using a measure taken fromSwann and Gill (1997; e.g., go to a bar or a pub, play tennis, visit with family, go bowling, do dishes).

Movies (Experiment 18).Participants saw posters for 16 mov- ies targeted for female audiences (e.g.,Pretty Woman,Legally Blonde) or male audiences (e.g.,Casino Royale,Transformers).

Participants rated how much they thought their partner would like each movie on 5-point scales (1 strongly dislike,5 strongly like).

Jokes (Experiment 19).Participants read 12 sexist jokes tar- geted for female audiences (e.g., “Why are men like strawberries?

Because they take a long time to mature and by the time they do most are rotten.”) or male audiences (e.g., “What is the difference between a battery and a woman? A battery has a positive side.”).

Participants rated how funny they thought their partner would rate each joke on 5-point scales (1 not at all funny,5 extremely funny).

Videos (Experiment 20).Participants watched eight 2–3 min videos with humorous dating advice targeted for female audiences (e.g., “How to survive shopping with your boyfriend,”) or male audiences (e.g., “How to tell her she looks terrible”). Participants rated how much their partner would like each video on 5-point scales (1 strongly dislike,5 strongly like).

Art (Experiment 21).Participants viewed 18 pieces of art (paintings and photographs). They rated how much their partner would like each piece of art on 10-point scales (1 strongly dislike,10 strongly like).

Opinions (Experiments 22 and 23).Participants read 21 opin- ion statements selected from Consumer Reports (taken fromHoch, 1987; e.g., “I would like to spend a year in London or Paris,” “I have somewhat old-fashioned tastes and habits,” “Police should use whatever force is necessary to maintain law and order”). They predicted how their partner would respond to each statement on 7-point scales (1 strongly disagree,” 4 neither agree nor disagree,7 strongly agree ).

Performance appraisal simulation (Experiment 24).MBA students were divided into pairs and assigned to the role of a partner in a firm (Stanley) or a manager being evaluated for promotion (Burke). In this simulation, Stanley has evaluated all the managers in the division and is prepared to give Burke his or her appraisal. Burke is certain that he or she outperforms the other managers and should be promoted to partnership in the firm.

Stanley believes Burke has many strong points, but he also has many concerns and estimates Burke has only a 10% chance of making partner in the next two years. Participants first received 10 min to read one-page long background about the person he or she was role playing. Each pair then conducted the performance eval- uation for 20 min. Finally, all participants answered four questions, according to their role. The first two questions were about their own impression: (1) “According to the materials you received and your performance appraisal, what do you think is the likelihood that you (manager Burke) [your manager Burke] will be promoted This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 558 EYAL, STEFFEL, AND EPLEY to partner at the end of the next two years?” (2) “What is your overall impression of yourself (Burke) [of Burke] as a manager?” Participants then made predictions regarding the thoughts of the other person: (3) “What do you believe Stanley (the partner) thinks is the likelihood that you will be promoted to partner at the end of the next two years [What do you believe Burke thinks is his or her likelihood of making partner]?” (4) “What do you believe is Stanley’s (the partner) overall impression of you as a manager [What do you believe Burke thinks is your overall impression of him or her as a manager]?” Answers on questions 1 and 3 were given on a line ending with a % sign. Ratings on questions 2 and 4 were given on an 11-point scale ( 5 very negative,0 neutral,5 very positive).

Independent variables.All experiments shared a basic de- sign of at least a perspective taking condition and a control condition. All experiments included a perspective taking condition that encouraged participants to imagine they were the other person (“partner’s shoes condition”). Two experiments (Experiments 16 and 22) also included an additional perspective taking condition that encouraged participants to focus on the other person’s thoughts and feelings (“partner’s perspective condition”). All but Experiments 21 and 24 also included a condition that encouraged participants to base their predictions of their partners’ responses on their own responses (“egocentrism condition,” see Supplemental Materials).

Control conditions.Participants in Experiments 17–21 and 23–24 were told: “We would like for you to use whatever strategy you think is best.” Participants in Experiments 16 and 22 received no instructions about how to predict the other person’s responses.

Perspective taking (partner’s shoes) conditions.Part- icipants in Experiment 16 (Activities) were told: When predicting your partner’s responses, it is very important that you put yourself in your partner’s shoes. Try to envision what your attitudes toward the following activities would be if you were your partner. Concentrate on how you would feel about each activity if you were your partner: i.e., whether you would like to do it or would not like to do it. Imagine how strongly you would feel. Circle the answers that best reflect the thoughts and feelings you would have about each activity if you were your partner.

Participants in Experiments 22 and 23 (Opinions) were told:

When predicting your partner’s responses, it is very important that you put yourself in your partner’s shoes. Try to envision how you would react to each of the statements if you were your partner.

Concentrate on what you would think if you were your partner, i.e., whether you would agree or disagree with each statement. Imagine how strongly you would feel. Circle the answers that best reflect the reactions, thoughts, and feelings you would have if you were your partner.

Participants in Experiments 17–21 (Activities, Movies, Jokes, Videos, Art) were told:

When rating how much your partner would like the following [activ- ities, movies, jokes, video clips pieces of art], it is very important that you put yourself in your partner’s shoes. Think carefully about what you know about your partner— consider their personality, their back- ground, and their tastes. Imagine what they would like and dislike about each [activity, movie, joke, video clip, piece of art], andconsider how that would influence their ratings of each activity [movie, joke, video clip, piece of art]. Participants in Experiment 24 (Performance Appraisal Simula- tion) were told:

Try to adopt Stanley’s perspective (the partner’s perspective) [Burke’s perspective (the manager’s perspective)] as if you were him or her. Do your best to put yourself into Stanley’s [Burke’s] shoes, trying to understand your interaction through Stanley’s [Burke’s] eyes— considering what Stanley [Burke] isthinking, and what Stanley’s interests and purposesare. Remember that Stanley [Burke] may have a different perspective than you do.

Perspective taking (partner’s perspective) conditions.Part- icipants in Experiment 16 (Activities) were told:

When predicting your partner’s responses, it is very important that you consider what you know about your partner. The best way to do that is to think about your partner’s behavior and visible reactions in the past. Try to think about which activities your partner has engaged in and how often he/she has engaged in those activities, or in activities that are similar to the ones below. Concentrate on what your partner has actually said to you about each activity: i.e. whether he/she has said that he/she likes to do it or does not like to do it. Circle the answers that best reflect what you think are your partner’s prefer- ences, based as much as you can on how your partner has behaved or responded in the past to these activities or to similar activities.

Participants in Experiment 22 (Opinions) were told:

When predicting your spouse’s responses, it is very important that you consider your spouse’s perspective. Try to envision his/her reactions to each of the statements. Concentrate on what your spouse thinks, i.e., whether your spouse agrees or disagrees with each statement.

Imagine how strongly he/she feels. Circle the answers that best reflect your spouse’s reactions, thoughts, and feelings.

Additional measures.Lastly, participants answered addi- tional questions:

Confidence.Participants in all but Experiments 18 and 24 predicted the number of responses they thought they predicted accurately. This provided us with a measure of participants’ con- fidence in the accuracy of their predictions.

Difficulty.Participants in Experiments 20, 21, and 23 rated how easy or difficult it was for them to predict their partner’s preferences using the strategy they did on a 10-point scale (1 very easy,10 very hard).

Measures about the relationship between partners included how well participants thought they knew their partners, how well they thought their partners knew them, how long they and their partner had known each other, how long they were romantically involved, whether they were married, and how long they were married.

These measures are reported in the Supplemental Materials.

Results Meta-analyses.Because Experiments 16 –24 used diverse populations and tests of interpersonal accuracy tests, we conducted random effects meta-analyses using the Comprehensive Meta- analysis 2 software (Borenstein et al., 2010) to identify the robust effects across experiments. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 559 PERSPECTIVE MISTAKING Accuracy.We conducted three meta-analyses to test the effect of perspective taking on accuracy. The first meta-analysis utilized the 13 comparisons in which we could calculate accuracy as the absolute difference between predicted opinions and preferences of others and their actual opinions and preferences (larger absolute differences indicate smaller accuracy). The second analysis was conducted for the eight comparisons in which we could calculate accuracy as the mean correlation between predicted opinions and preferences of others and their actual opinions and preferences.

The third analysis was conducted for the eight comparisons in which we could count the number of participants’ correct predic- tions. These meta-analyses yielded a significantnegativeeffect for accuracy when calculated as absolute differences:d 0.16, 95% CI [0.03, 0.30],z 2.44,p .015 (a positive sign indicates less accuracy in perspective taking condition compared with control, Table 7), and nonsignificant effects for accuracy when calculated as mean correlations:d 0.10, 95% CI [ 0.25, 0.05], z 1.36,p .17 (a negative sign indicates less accuracy in perspective taking condition compared with control,Table 8), and when calculated as the number of correct predictions:d 0.01, 95% CI [ 0.23, 0.20],z 0.12,p .90 (a negative sign indicates less accuracy in perspective taking condition compared with control,Table 9). The results of the three analyses indicate that perspective taking did not increase accuracy in predicting partners’ opinions and preferences. If anything, it reduced accu- racy as we also observed in Experiments 1–15.

These results did not change in a meaningful way when we replaced the “partner’s shoes” conditions with the “partner’s per- spective” conditions in Experiments 16 and 22. In addition, par- ticipants in the “partner’s shoes” condition were directionally more accurate than participants in the “partner’s perspective” condition.

This difference was marginally significant for the number ofcorrect predictions measure in Experiment 22,t(74) 1.84,p .070,d 0.43, but nonsignificant in all other measures in Exper- iment 22 and in all measures in Experiment 16.

Perceived difficulty.In a meta-analysis of the four compari- sons in which we measured perceived difficulty, we observed a nonsignificant difference between the perspective taking and con- trol conditions,d 0.03, 95% CI [ 0.30, 0.36],z 0.16,p .87 (seeTable 10). Note that this differs from Experiments 1–15 in which perspective takers reported experiencing more difficulty than those in the control condition. This different pattern of results could stem from participants’ increased familiarity with the targets of judgment in Experiments 16 –24.

Confidence and overconfidence.In a meta-analysis of the eight comparisons in which we measured confidence, we observed a nonsignificant difference between perspective taking and control conditions,d 0.08, 95% CI [ 0.09, 0.25],z 0.90,p .37 (see Table 11). In contrast to Experiments 1–15, perspective taking did not significantly reduce confidence. If anything, it directionally increased it. This may again be because participants in Experi- ments 16 –24 had more knowledge about the other person’s per- spective to rely on in the perspective taking condition. We calcu- lated overconfidence by subtracting the number of accurate responses from confidence scores (i.e., predicted number of accu- rate responses). Overall, participants were highly overconfident in their predictions,d 1.50, 95% CI [1.23, 1.78],z 10.58,p .001. This overconfidence was statistically significant in all eight comparisons. In a meta-analysis of the eight comparisons in which we measured both confidence and accuracy, there was no effect of overconfidence, indicating that perspective taking did not influ- ence overconfidence,d 0.04, 95% CI [ 0.16, 0.24],z 0.39, p .70.

Table 8 Meta-Analysis on Accuracy (Mean Correlations Between Predicted Responses and Actual Responses) for Perspective Taking Versus Control Conditions, Experiments 16 –24 Experiment Task Location (N) ControlPT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZP 16 Activities (partners) Community #1 (74) .64 (.15) .65 (.16) .07 .23 .38 .53 .32 .751 17 Activities (partners) Community #1 (66) .68 (.14) .62 (.16) .36 .25 .84 .13 1.44 .150 18 Movies (strangers) US U. #2 (80) .40 (.36) .33 (.37) .14 .22 .58 .30 .63 .528 19 Jokes (strangers) Community #2 (74) .16 (.36) .07 (.38) .24 .23 .69 .22 1.01 .314 20 Videos (strangers) US U. #2 (84) .19 (.31) .10 (.45) .24 .22 .67 .19 1.10 .270 21 Art (strangers) US U. #2 (92) .35 (.26) .30 (.25) .20 .21 .61 .21 .97 .331 22 Opinions (partners) Community #1 (82) .48 (.21) .46 (.23) .11 .22 .54 .33 .47 .635 23 Opinions (partners) Community #1 (80) .33 (.28) .35 (.20) .03 .23 .41 .47 .14 .887 23 Opinions (strangers) Community #1 (79) .07 (.22) .11 (.17) .24 .23 .21 .68 1.06 .291 24Performance appraisal simulation Burke’s chance MBA (101) .30 .21 24Performance appraisal simulation Stanley’s chance MBA (101) .21 .14 24Performance appraisal simulation Burke’s impression MBA (101) .10 .05 24Performance appraisal simulation Stanley’s impression MBA (101) .17 .04 Total .10 .08 .25 .05 1.36 .173 Note. The meta-analysis was conducted on a Fisher-transformation of the correlations. For ease of interpretation we report the mean Pearson correlations.

In Experiment 24, the correlations were calculated between two ratings rather than between mean ratings as in the other experiments. Therefore, it was not included in the meta-analysis. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 560 EYAL, STEFFEL, AND EPLEY Reducing egocentrism.We conducted two meta-analyses on the nine comparisons in which we measured participants’ own preferences that allow us to test whether or not perspective taking systematically decreases egocentric projection compared with a control condition. When calculating projection as the absolute difference between predicted opinions of others and self-opinions (larger absolute differences indicate smaller projection), we ob- served a nonsignificant reduction in egocentric projection in the perspective taking conditions compared with the control condi- tions,d 0.12, 95% CI [ 0.04, 0.27],z 1.49,p .14 (see Table 12). This was also the case when calculating projection as the mean correlation between predicted opinions of others and self-opinions (smaller correlations indicate smaller projection), d 0.14, 95% CI [ 0.34, 0.06],z 1.38,p .17, such that participants in the perspective taking condition were less egocen- tric than participants in the control condition (seeTable 13).

Discussion Across nine experiments consisting of naturalistic tests of inter- personal accuracy—predicting a partner’s preferences and opin- ions—we found that an explicit instruction to engage in perspec- tive taking did not increase accuracy. If anything, it decreased accuracy.

Experiments 16 –24 do not provide a clear explanation for why perspective taking failed to increase accuracy. Among pairs of participants who were encouraged to take the perspective of their partner, reading the mind of their partner was not perceived to bemore or less difficult, and did not yield more or less confidence, compared with control condition. Interestingly, unlike participants in Experiments 1–15 who were underconfident in their predictions, participants in Experiments 16 –24 were dramatically overconfi- dent. Participants in Experiments 16 –24 were more familiar with their targets’ perspectives and we therefore think it was likely that the judgment task was generally easier as a result, thereby increas- ing confidence. Participants’ overconfidence, however, did not differ systematically between perspective taking and control con- ditions. Perspective takers did seem to be less egocentric (i.e., projective) to some extent compared with control participants, but this effect was only marginally significant when measured as correlations between predictions and self-ratings but not when measured as absolute differences between predictions and self- ratings. Less projection, however, did not increase accuracy in the perspective taking condition compared with the control condi- tion.

We believe the collective results of all of the experiments presented so far (Experiments 1–24) are especially interesting because they stand in stark contrast to the survey we presented in the introduction, where respondents tended to predict that perspective taking would increase accuracy across many of these tests. Common sense indicates that perspective taking should increase interpersonal understanding.

Likewise, psychological theory predicts that perspective taking could increase interpersonal accuracy through a variety of different mech- anisms (such as behavioral mimicry, increased empathy, or reduced egocentrism). These mechanisms all presume that taking another’s Table 9 Meta-Analysis on Accuracy (Number of Correct Predictions) for Perspective Taking Versus Control Conditions, Experiments 16 –24 Experiment Task Location (N) ControlPT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 16 Activities (partners) Community #1 (74) 11.89 (2.81) 13.03 (3.67) .35 .23 .11 .81 1.48 .138 17 Activities (partners) Community #1 (66) 12.24 (4.02) 10.66 (3.92) .40 .25 .89 .09 1.60 .110 18 Movies (strangers) US U. #2 (80) 5.03 (2.20) 4.13 (2.44) .39 .23 .83 .06 1.72 .086 19 Jokes (strangers) Community #2 (78) 3.35 (1.72) 3.24 (1.94) .06 .23 .50 .38 .27 .791 20 Videos (strangers) US U. #2 (79) 1.89 (1.21) 1.98 (1.23) .07 .22 .35 .50 .34 .734 21 Art (strangers) US U. #2 (92) 2.50 (1.56) 2.26 (1.41) .16 .21 .57 .25 .77 .440 22 Opinions (partners) Community #1 (82) 4.89 (2.25) 6.00 (1.90) .53 .23 .08 .98 2.33 .020 23 Opinions (partners) Community #1 (80) 4.64 (2.22) 4.03 (1.80) .30 .23 .74 .14 1.32 .186 23 Opinions (strangers) Community #1 (79) 3.40 (2.03) 3.83 (1.77) .23 .23 .22 .67 .99 .321 Total .01 .11 .23 .20 .12 .904 Note. There are 37 items in Activities, 16 items in Movies, 12 items in Jokes, 8 items in Videos, 18 items in Art, and 21 items in Opinions.

Table 10 Meta-Analysis on Perceived Difficulty for Perspective Taking and Control Conditions, Experiments 16 –24 Experiment Task Location (N) Control PT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 20 Videos (strangers) US U. #2 (93) 4.72 (1.57) 5.19 (1.26) .33 .21 .08 .74 1.58 .113 21 Art (strangers) US U. #2 (88) 6.44 (2.15) 7.07 (2.19) .29 .21 .13 .71 1.36 .176 23 Opinions (partners) Community #1 (80) 4.39 (1.96) 3.86 (1.55) .30 .23 .74 .15 1.31 .189 23 Opinions (strangers) Community #1 (79) 7.84 (1.99) 7.36 (1.87) .25 .23 .69 .20 1.09 .274 Total.03 .17 .30 .36 .16 .870 Note. Participants made their ratings on 11-point scales. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 561 PERSPECTIVE MISTAKING perspective will lead people to consider new information that they would not have considered otherwise, and that this new information will provide a systematically more accurate guide to another person’s mental experience. Our results simply suggest that the information people consider when they shift perspective may not be systematically more accurate than the information they would have considered otherwise. Experiment 25: Perspective Getting If taking another person’s perspective does not systematically increase accuracy, is there anything one can do to reliably increase understanding? If so, are people who are using this more effective strategy aware of its usefulness?

In one final experiment, we compared the effectiveness of perspective taking against another approach that almost necessar- ily collects more accurate information directly from another per- son’s perspective, what we refer to asperspective getting.In particular, increasing insight into another person’s mind should require getting more accurate information about his or her per- spective. One obvious way to do this is by asking a person to report directly on his or her thoughts, beliefs, attitudes, and other mentalstates and then using that information as a guide, just as survey researchers do to assess public opinions with a relatively high degree of accuracy.

Of course, self-reports are sometimes inaccurate, such as when asking people to explain their own mental processes (Nisbett & Wilson, 1977), or when discussing topics with strong demand characteristics (Schwarz, 1999). However, self-reports of con- scious mental experiences, such as conscious beliefs, emotions, or attitudes, are still consistently the best predictors of behavior (Greenwald, Poehlman, Uhlmann, & Banaji, 2009;Oswald, Mitch- ell, Blanton, Jaccard, & Tetlock, 2013). More important, perspec- tive taking is often presumed to increase understanding of another person’s conscious experience, regardless of whether that experi- ence itself accurately reflects some objective reality or not. If you want to know whether your spousebelieveshe or she would prefer a weekend in London or Paris, or watchLove Actuallyrather than Iron Man 3,the most accurate strategy would likely be to get your spouse’s perspective byaskingwhat he or she prefers rather than trying to take his or her perspective and guess.

Although this approach to increasing accuracy seems obvious, we believe it is worth comparing its effectiveness against perspec- Table 11 Meta-Analysis on Confidence (Predicted Number of Correct Responses) for Perspective Taking Versus Control Conditions, Experiments 16 –24 Experiment Task Location (N) Control PT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 16 Activities (partners) Community #1 (74) 23.61 (6.21) 24.95 (4.66) .25 .23 .21 .70 1.05 .294 17 Activities (partners) Community #1 (65) 21.39 (7.38) 24.44 (7.06) .42 .25 .07 .91 1.68 .092 19 Jokes (strangers) Community #2 (75) 6.11 (2.22) 6.81 (2.59) .29 .23 .17 .75 1.25 .211 20 Videos (strangers) US U. #2 (93) 4.54 (1.53) 4.02 (1.44) .35 .21 .76 .06 1.68 .094 21 Art (strangers) US U. #2 (88) 7.60 (3.45) 7.74 (3.67) .04 .21 .38 .46 .19 .850 22 Opinions (partners) Community #1 (82) 12.94 (3.18) 12.42 (4.23) .14 .22 .58 .29 .63 .526 23 Opinions (partners) Community #1 (79) 13.43 (4.20) 13.77 (3.23) .09 .23 .36 .53 .40 .693 23 Opinions (strangers) Community #1 (79) 8.30 (4.07) 9.03 (4.07) .18 .23 .26 .62 .79 .428 Total.08 .09 .09 .25 .90 .370 Note. There are 37 items in Activities, 16 item in Movies, 12 items in Jokes, 8 items in Videos, 18 items in Art, and 21 items in Opinions. We did not measure confidence in Experiments 18 and 24.

Table 12 Meta-Analysis on Egocentric Projection (Absolute Difference Between Predictions of Partners’ Responses and Own Responses) for Perspective Taking Versus Control Conditions, Experiments 16 –24 Experiment Task Location (N) Control PT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 16 Activities (partners) Community #1 (74) 1.39 (.42) 1.43 (.38) .10 .23 .36 .56 .43 .667 17 Activities (partners) Community #1 (66) 1.50 (.48) 1.44 (.45) .13 .25 .61 .35 .52 .601 18 Movies (strangers) US U. #2 (80) 1.01 (.66) 1.03 (.47) .04 .22 .40 .47 .16 .876 19 Jokes (strangers) Community #2 (78) .93 (.50) 1.08 (.56) .28 .23 .16 .73 1.24 .214 20 Videos (strangers) US U. #2 (93) .91 (.42) .97 (.62) .11 .21 .29 .52 .55 .586 21 Art (strangers) US U. #2 (92) 1.57 (.67) 1.68 (.72) .16 .21 .25 .57 .76 .449 22 Opinions (partners) Community #1 (82) 1.21 (.54) 1.55 (.52) .64 .23 .20 1.09 2.82 .005 23 Opinions (partners) Community #1 (80) 1.55 (.46) 1.48 (.50) .15 .23 .59 .30 .65 .516 23 Opinions (strangers) Community #1 (79) 1.77 (.57) 1.75 (.55) .04 .23 .48 .41 .16 .875 Total.12 .08 .04 .27 1.49 .137 Note. Larger absolute values indicate less projection. We did not test projection in Experiment 24. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 562 EYAL, STEFFEL, AND EPLEY tive taking for three reasons. First, none of our experiments pro- vide concrete insight into how a person might actually increase interpersonal understanding above and beyond a control condition.

Indeed, perspective taking across our experiments tended to de- crease accuracy. Testing the effectiveness of perspective getting would test whether it is even possible to systematically improve interpersonal accuracy. It would also offer practical advice about exactly what kind approach a person should take to understand another’s mind more accurately. Second, we believe our perspec- tive taking results highlight an important subtlety that is often overlooked in the existing psychological literature. Getting anoth- er’s perspective directly through bottom-up processes of direct questioning is different than trying to take another’s perspective through top-down inferences. It is important to clearly distinguish between these processes because they may have very different implications for interpersonal understanding. This distinction can also serve to refine the theoretical concept of perspective-taking, which is sometimes used broadly to describe both top-down pro- cesses of inference and bottom-up processes of direct questioning or personal experience in another person’s situation. Finally, per- spective getting may seem like an obvious approach to increasing interpersonal accuracy, but it may not be so obvious to those in the midst of interpersonal interactions. By measuring participants’ confidence in judgment, we can assess the degree to which people are aware of which strategies provide better insight into the mind of another person than others.

Specifically, we conducted a replication of Experiments 22 and 23, in which participants predicted their romantic partner’s agree- ment or disagreement with a series of opinion statements. Partic- ipants in the perspective taking condition followed the same in- structions as in Experiments 22 and 23, whereas participants in the control condition were instructed to “use whatever strategy you think is best.” Each participant in the perspective getting condition, in contrast, was first given the chance to ask his or her partner either half or all of the opinion statements, listen to the partner’s verbal response, and then to later predict how his or her partner would respond on the numeric preference scale for each item.

Allowing participants to get their partners’ perspective on either a subset of the items, or the full set of items, enables us to more precisely assess the impact of this approach on accuracy.We predicted thatgettinga partner’s perspective would increase accuracy compared withtakinghis or her perspective and to the control condition. Because the survey items we used were de- signed so as to be uncorrelated with each other (Hoch, 1987), we expected that getting perspective would increase accuracy only on the items people discussed. Those in the partial perspective-getting conditions, who discuss only half the items, should therefore obtain accuracy rates somewhere in between the control and full perspective getting condition (where participants discuss all of the items). Obviously, these results would change if we used a set of highly intercorrelated survey items. Finally, given the tenuous relationship between confidence and accuracy in judgment, we expected to observe a smaller difference in participants’ confi- dence across conditions than in accuracy across conditions. Those in the perspective-getting conditions, we predicted, would not be fully aware of just how much their judgment improved compared with the other experimental conditions. Method Participants.One hundred four heterosexual romantic cou- ples were recruited in the community to complete a short survey.

Of these, 58% were married. Participants ranged in age from 19 to 72 (M 36), and were in a relationship between one month and 43 years (M 10 years).

Materials and procedure.Couples were invited to partici- pate in a study on how well people can gauge their partner’s opinions, using the same test as reported in Experiments 22 and 23.

One member of each couple (predictor) was asked to predict how their partner would respond to 20 opinion statements selected from Consumer Reports(Hoch, 1987) and then report his or her own opinions. The other member (target) was only asked to rate his or her own opinions. Predictors were randomly assigned to one of five strategies and read the following instructions:

Control condition.Participants read, “We would like for you to use whatever strategy you think is best.” Participants in the control condition received no further suggestions on what these strategies might be.

Perspective taking condition.Participants read, “We would like for you to take the perspective of your partner. Please imagine Table 13 Meta-Analysis on Egocentric Projection (Mean Correlations Between Predictions of Partners’ Responses and Own Responses) for Perspective Taking Versus Control Conditions, Experiments 16 –24 Experiment Task Location (N) Control PT: Other’s shoesMeta-analysis results dSECI lower limitCI upper limitZp 16 Activities (partners) Community #1 (74) .45 (.20) .40 (.21) .25 .23 .70 .21 1.06 .291 17 Activities (partners) Community #1 (66) .38 (.23) .39 (.28) .14 .25 .35 .62 .56 .578 18 Movies (strangers) US U. #2 (80) .44 (.51) .36 (.44) .25 .22 .69 .19 1.12 .262 19 Jokes (strangers) Community #1 (75) .31 (.39) .27 (.42) .01 .23 .47 .44 .06 .954 20 Videos (strangers) US U. #2 (92) .29 (.47) .15 (.59) .29 .21 .70 .13 1.36 .173 21 Art (strangers) US U. #2 (92) .52 (.31) .49 (.31) .14 .21 .55 .27 .69 .491 22 Opinions (partners) Community #1 (82) .57 (.26) .37 (.27) .76 .23 1.21 .31 3.33 .001 23 Opinions (partners) Community #1 (80) .36 (.29) .43 (.30) .29 .23 .15 .73 1.29 .198 23 Opinions (strangers) Community #1 (79) .19 (.32) .22 (.28) .06 .23 .39 .50 .25 .799 Total .14 .10 .34 .06 1.38 .168 Note. The meta-analysis was conducted on a Fisher-transformation of the correlations. For ease of interpretation we report the mean Pearson correlations. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 563 PERSPECTIVE MISTAKING a typical day in the life of your partner as if you were him/her, looking at the world through his/her eyes and walking through the world in his/her shoes. You should start from the beginning of your partner’s day to the end, focusing on his/her thoughts and feelings.

Please take approximately five minutes to write about a day in the life of your partner. Once you have done that, we would like for you to use this information to rate the extent to which your partner would agree or disagree with the following statements. Please use this strategy even if you think another strategy would be better.” Perspective getting (–all, – even, and – odd) conditions.

Participants read, Before you rate the extent to which your partner would agree or disagree with the following statements, we would like for you to ask your partner to tell you about their opinions. We will give you a list of statements. Please take approximately five minutes to ask your partner about the extent to which they agree or disagree with each of the statements on the list, trying to get a sense of the range of your partner’s opinions. Your partner might strongly agree with some statements, somewhat agree with others, and they may strongly dis- agree with others. Once you have done that, we would like for you to use the information you got from your partner to predict the extent to which your partner would agree or disagree with these statements.

Please use this strategy even if you think another strategy would be better. Participants in theperspective getting—allcondition received the full list of statements, and those in theperspective getting-evenand -oddconditions received a list with only the even- or odd- numbered statements, respectively.

Each participant in the perspective taking condition was given five minutes in which to write about a day in the life of his or her partner, a commonly used perspective-taking manipulation (adapted fromMacrae, Bodenhausen, Milne, & Jetten, 1994). Each participant in the perspective getting conditions was given five minutes in which to ask his or her partner about their opinions on the items provided. Note that perspective getting participants sim- ply discussed their opinions verbally, rather than putting them on the numeric scale that they would use later in the experiment. This is important because participants in the perspective getting condi- tion still had to infer their partner’s numeric response from their verbal answer, rather than simply remember the exact numeric answer that a partner provided. After this 5-min period, partners were moved to separate locations and given the full list of opinion statements. Targets reported their opinion on each item on a7-point scale ( 3 strongly disagree,3 strongly agree).

Predictors guessed how their target would respond to each item on the same scales.

We next measured participants’ confidence in the accuracy of their own or their partner’s judgment in two different ways. First, predictors rated how confident they were that their predictions of their partner’s opinions were correct, and targets rated how con- fident they were that their partner’s predictions of their opinions were correct on an 11-point scale (0 not at all confident,10 extremely confident). Second, predictors indicated the number of responses they thought they predicted exactly correctly, and targets also indicated the number of responses they thought their partner predicted exactly correctly.

Finally, participants reported how long they and their partner had been romantically involved and how long (if applicable) they had been married. Participants were then reunited with their part- ners and debriefed. Results Means, standard deviations, and correlations for the different dependent measures are presented inTable 14.

Accuracy.We assessed accuracy in three ways. First, we calculated the absolute mean difference between predicted and actual opinions (larger absolute differences indicate smaller accu- racy). Second, we calculated the correlation between predicted and actual opinions (larger correlations indicate greater accuracy), using a Fisher-transformation to correct for non-normality in Pear- son correlations (Fisher, 1915). For ease of interpretation, we report untransformed Pearson correlations in the tables and text.

Third, we calculated the number of items predictors guessed cor- rectly. All results are presented inTable 14. Across these mea- sures, perspective getting improved accuracy relative to the control condition. Perspective taking did not increase accuracy. If any- thing, it again decreased accuracy.

Accuracy as measured by the absolute mean difference between predicted and actual ratings significantly varied by experimental condition,F(4, 99) 14.61,p .001, p2 .37. Compared with the control condition, participants were significantly more accurate (reflected in smaller absolute differences between predicted and actual responses) in the perspective getting-full condition, t(99) 4.89,p .001,d 1.85, perspective getting-even condition,t(99) 2.62,p .010,d 0.75, and perspective Table 14 Results for Experiment 25 Measure ControlPT: Other’s shoesPerspective getting (even)Perspective getting (odd)Perspective getting (all) 1. Accuracy (ABS difference between predicted and actual opinions) 1.46 a(.31)1.71 b(.43)1.15 c(.49)1.21 c(.29).88 d(.32) 2. Accuracy (Mean correlation between predicted and actual opinions) .50 a (.15).39 a (.24).65 b(.23).66 b(.16).81 c(.12) 3. Accuracy (# of correct predictions) 4.90 a (1.70)3.95 a (1.94)7.43 b(2.79)6.35 bc(2.52)8.60 cd(3.02) 4. Projection (ABS difference between predicted and self opinions) 1.62 (.33) 1.56 (.60) 1.45 (.56) 1.47 (.42) 1.42 (.52) 5. Projection (Mean correlation between predicted and self opinions) .35 a (.24).37 ab (.30).49 ab (.24).44 ab(.24).53 c (.23) 6. Confidence (Ratings) 7.30 (1.13) 7.05 (1.63) 7.00 (1.26) 6.65 (2.01) 7.55 (1.36) 7. Confidence (# of estimated correct predictions) 12.60 a(3.21)13.48 ab(3.12)14.45 b(2.46)13.53 ab(3.36)14.80 b(2.75) 8. Overconfidence (Difference between #7 and #3) 7.60 a(3.73)9.67 b(4.25)6.65 ac(2.93)7.00 ac(3.90)6.20 ac(3.65) Note. There are 20 items in the Hoch questionnaire. Within each measure, numbers that do not share a superscript differ significantly atp .05. For numbers that share an identical number of asterisks the difference is marginally significant atp .10. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 564 EYAL, STEFFEL, AND EPLEY getting-odd condition,t(99) 2.04,p .044,d 0.82. How- ever, participants were significantlylessaccurate in the perspec- tive taking condition than in the control condition,t(99) 2.20, p .031,d 0.68.

Accuracy as measured by the correlation between predicted and actual ratings also varied by experimental condition,F(4, 99) 15.24,p .001, p2 .38. Compared with the control condition, participants were significantly more accurate in in the perspective getting-full condition,t(99) 5.23,p .001,d 2.28, perspec- tive getting-even condition,t(99) 2.66,p .009,d 0.77, and perspective getting-odd condition,t(99) 2.68,p .009,d 1.03. Participants in the perspective taking condition were not more accurate than those in the control condition,t(99) 1.89, p .061,d 0.55. If anything, they were again directionally less accurate.

The number of items predictors guessed exactly correctly also varied by experimental condition,F(4, 99) 12.41,p .001, p2 .33. Compared with the control condition, participants pre- dicted significantly more items correctly in the perspective getting- full condition,t(99) 4.86,p .001,d 1.50, and the perspec- tive getting-even condition,t(99) 3.36,p .001,d 1.09, and marginally more items in the perspective getting-odd condition, t(99) 1.90,p .060,d 0.68. Participants in the perspective taking condition again were not more accurate than those in the control condition,t(99) 1.28,p .20,d 0.52. They were directionally less accurate.

Notice that the two perspective getting conditions that discussed only half of the survey items yielded accuracy that fell midway between the perspective getting-full condition and the control condition on both absolute mean difference and correlational ac- curacy, the two accuracy measures for which we had item-level measures of predicted and actual accuracy. This moderate increase in accuracy compared with the control condition occurred because predictors’ accuracy significantly increased only on the items thatpredictors discussed explicitly with their partners. On those items, participants in the two partial perspective-getting conditions were as accurate as those in the perspective getting-all conditions, but they were no more accurate than the control condition on the items they did not discuss with their partner. We discuss the details of these secondary analyses in the Supplemental Materials. Again, we note that the items within this survey were designed so as to be independent from each other, and so these results simply reflect the nature of the survey items used in the experiment. Accurate insight gained from any strategy generalizes to other contexts only to the extent that those contexts are intercorrelated.

These results make it clear that participants gained insight into their partner’s opinions when they got the person’s perspective directly, in this case through a bottom-up process of directly asking him or her to report on an opinion. As in the preceding experi- ments, taking another’s perspective through a top-down process of inference did not increase accuracy compared with a control con- dition. If anything, perspective taking again decreased accuracy.

Confidence and overconfidence.Despite large differences in accuracy across conditions, the right panel ofFigure 2shows that confidence in the accuracy of their judgment (measured on 0 –10 scales) did not vary across conditions,F(4, 97) 1.01,p .41, p2 .04. These confidence ratings were also nonsignificantly correlated with absolute mean difference accuracy,r .08,p .44, correlational accuracy,r .08,p .44, and the number of items participants predicted correctly,r .12,p .21.

We also measured participants’ sense of their own accuracy by asking them to predict how many items they guessed exactly correctly. Again despite large differences in the actual accuracy, participants’ predictions of the number guessed exactly correctly did not vary by experimental condition,F(4, 95) 1.69,p .16, p2 .07. Predictors’ beliefs about the number they guessed exactly correctly was nonsignificantly correlated with the number they actually guessed correctly,r .15,p .13. It was also Figure 2.Confidence and overconfidence measures as a function of condition, Experiment 25. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 565 PERSPECTIVE MISTAKING nonsignificantly correlated with absolute mean difference accu- racy,r .13,p .19, and correlational accuracy,r .10,p .31.

Comparing the predicted number of items guessed correctly against the actual number of items guessed correctly provides a direct measure of overconfidence. As shown in the left panel of Figure 2, participants had very limited insight into how their prediction strategy affected their actual accuracy. A 5 (Condi- tion) 2 (Number correct: Actual vs. Predicted) ANOVA with repeated measures on the second factor indicated that participants across conditions were dramatically overconfident, believing they predicted more items correctly (M 13.71,SD 3.22) than they actually did (M 6.20,SD 2.92),F(1, 95) 408.76,p .001, p2 .81. The interaction was not significant,F(4, 95) 1.02,p .403, p2 .04, indicating that overconfidence did not vary by experimental condition.

Reducing egocentrism.We also calculated the correspon- dence between predictors’ own stated opinions and their predic- tions of partner’s stated opinions in two different ways. First, as the absolute mean difference between each predictor’s own opin- ions and predictions of his or her partner’s opinions. Second, as the correlation between each predictor’s own opinions and predictions of his or her partner’s opinions. We report the results of these analyses for the sake of consistency with the preceding experi- ments, but urge caution interpreting these results in the perspective getting conditions. In particular, predictors in the perspective get- ting conditions may have aligned their attitudes with their partner’s stated opinions, meaning that these measures may reflect social influence rather than projection. Indeed, partners in the perspective getting-full condition reported more similar preferences than those in the control condition: contrast analyses indicated that the abso- lute difference between own and partner’s opinions was margin- ally smaller in the perspective getting-full condition (M 1.54, SD .25) than in the control condition (M 1.77,SD .43), t(30.36) 2.00,p .054,d 0.65, and the correlation between own and partner’s opinions was significantly larger in the perspective getting—full condition (M .44,SD .24) than in the control condition (M .31,SD .17),t(98) 2.01,p .047, d 0.63. These results in the perspective getting conditions are therefore difficult to interpret.

With that concern in mind, the absolute difference between own opinions and predicted partner’s opinions did not vary by experi- mental condition,F(4, 98) .58,p .68, p2 .02, but the correlation between these two measures did,F(4, 97) 2.61,p .040, p2 .10. In contrast to the meta-analysis of Experiments 16 –24, we did not observe a significantly smaller correlation in the perspective taking condition (M .37,SD .30) than in the control condition (M .35,SD .24),t(97) 0.08,p .94,d 0.08, indicating that perspective taking did not significantly reduce egocentrism in this experiment. In contrast, compared with the control condition, the correlation was significantly larger in the perspective getting-full (M .53,SD .23),t(97) 2.38,p .019,d 0.77, and perspective getting-even conditions (M .49, SD .24),t(97) 2.13,p .036,d 0.58. The correlation between own opinions and predicted partner opinions did not differ significantly between the perspective getting-odd condition (M .44,SD .24) and the control condition,t(97) 1.17,p .25,d 0.37. Discussion Romantic partners, most of whom were married, and who had been together for an average of 10 years, presumably know a lot about their partner’s perspective. Nevertheless, trying to take a partner’s perspective again failed to increase insight into their partner’s mind. Instead, compared with predicting their partner’s responses without special instructions, perspective taking in- creased confidence but decreased accuracy, thereby increasing overconfidence. This is not the outcome that perspective taking is presumably intended to create.

This experiment also tested a more obvious strategy for increas- ing accuracy into the mind of one’s partner: getting another’s perspective by asking him or her directly. We referred to this as perspective-getting to contrast a bottom-up approach to understand another person against a top-down approach of trying to take another’s perspective by shifting cognitive attention to another’s point of view. Although asking one’s partner to state his or her preferences is an obvious way to increase understanding, perhaps the most important result from this experiment is that participants themselves did not seem to be aware of how this strategy actually affected their insight compared with the relatively ineffective strategies used in our other conditions. One might imagine that students who ask their teacher the answers to exam questions would be more confident when completing the exam than students who did not. Our romantic partners in the perspective-getting conditions did something conceptually similar and yet were not markedly more confident than those whose accuracy was some- times only slightly better than chance guessing in the control and perspective taking conditions. Taking perspective and getting per- spective are two obviously different approaches to understanding the mind of another person. The obvious benefit of one strategy compared with the other was not, however, so obvious to those who were actually using each strategy. General Discussion A survey of 1,020 Americans asked them to indicate which of 5 superpowers they would most like to possess: invisibility, telepor- tation, flight, time travel, or reading others’ minds (Marist, 2011).

Tied with time travel for the most desired superpower was the ability to read the minds of others. On the one hand, this is somewhat ironic as the ability to read the minds of others is arguably the only capacity among that list that people already possess. The human brain stands out in our primate lineage for its relatively large neocortex (Jerison, 1971;Herrmann et al., 2007), a feature that may be the product of natural selection to handle the cognitive requirements of living in large social groups (Dunbar, 1993). By the age of 2, human toddlers’ capacity to understand the minds of others has already surpassed that of our nearest primate relatives, the chimpanzee (Herrmann et al., 2007). On the other hand, being able to understand the mind of another person does not mean that one is able to do so perfectly. Studies of human social cognition routinely reveal accuracy rates in understanding others’ beliefs, attitudes, emotions, and intentions that are significantly better than random chance but also markedly worse than perfect (e.g.,Ambady, Bernieri, & Richeson, 2000;Funder, 1995;Hall & Schmid Mast, 2007;Ickes, 1997;Kenny, 1991;Swann & Gill, 1997;Todorov, Olivola, Dotsch, & Mende-Siedlecki, 2015). In many ways, everyday life would be much easier if people were This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 566 EYAL, STEFFEL, AND EPLEY able to understand exactly what others thought of them, could understand when others were lying versus telling the truth, could identify who really loved them and who was just pretending, and could anticipate others’ actions based on an accurate understand- ing of their intentions. It is therefore easy to understand why a person might want to make this potential super power work even better.

Here we reported the results of 25 experiments that tested one common sense strategy for enabling more accurate mind reading:

perspective taking. Across a wide variety of experimental tests, involving relationships that ranged from strangers to spouses, we found no evidence that perspective taking systematically increased one’s ability to accurately understand the mind of another person compared with a control condition. If anything, we found that perspective taking tended todecreaseinterpersonal accuracy. A meta-analysis on all 25 experiments (number of correct responses in Experiments 1–15 and absolute differences in Experiments 16 –25) yielded a statistically significant, albeit small,negative effect of perspective taking on accuracy,d 0.23, 95% CI [ 0.32, 0.13],z 4.72,p .001. This result does not change in a meaningful way if the meta-analysis includes other accuracy measures for Experiments 16 –25 (for correlations between pre- dicted and actual responses,d 0.21 and for number of items predicted correctly,d 0.17).Dale Carnegie (1936)suggested that “trying honestly to see things from the other person’s point of view” was a “formula that will work wonders for you.” Perspective taking may indeed work some interpersonal wonders, but our results suggest that increasing insight into the mind of another person is not among them.

It is worth noting that we began our research presuming, con- sistent with the common sense we observed in the pretest reported in the introduction and with existing psychological theory, that shifting perspective to another person’s point of view could in- crease interpersonal accuracy in many circumstances. We con- ducted such a large number of experiments across a wide variety of contexts and utilizing a variety of interpersonal understanding measures because we kept searching for contexts as well as mea- sures that might reveal, based on existing theory, circumstances in which perspective taking could increase accuracy. Because the scientific method is unable to confidently affirm the null hypoth- esis, our experiments are unable to confirm that perspective taking is ineffective for increasing interpersonal accuracy. They can only show the absence of positive evidence despite a concerted effort to test the most likely contexts where we, our pretest participants, and existing psychological theory presumed that perspective taking could increase accuracy.

That perspective taking failed to increase accuracy was not the product of ineffective experimental manipulations. In a manipula- tion check across Experiments 1–15, participants in the perspective taking conditions reported trying harder to adopt another’s per- spective than participants in the control conditions. In addition, in a meta-analysis across all experiments in which we measured perceived difficulty, perspective taking was perceived as more difficult,d 0.13, 95% CI [0.03, 0.23],z 2.54,p .011. This perception was stronger and more reliable in Experiments 1–15 than in Experiments 20 –23, perhaps because participants were less familiar with the judgmental tasks or the targets of judgment. We also found evidence that perspective taking tended to decrease egocentric biases in judgment. This effect was stronger and moreconsistent when measured by the false belief task (Experiments 4, 5, 8, and 13) than when measured incidentally as the correspon- dence between one’s own opinions and preferences and a target’s opinions and preferences (Experiments 16 –23). However, even when perspective taking reliably decreased egocentrism it did not reliably increase accuracy. Finally, we did not find a reliable effect of perspective taking on confidence. In a meta-analysis across all experiments in which participants estimated the number of correct responses as a measure of confidence, perspective taking de- creased confidence, although this effect was only marginally sig- nificant,d 0.10, 95% CI [ 0.20, 0.01],z 1.81,p .070. This reduction in confidence was mostly evident in Experi- ments 1–15. As reported before, in Experiments 16 –25, this effect was reversed, with perspective taking directionally increasing con- fidence. This non reliable effect of perspective taking on confi- dence across experiments may be because participants in the perspective taking condition in Experiments16-25 were more fa- miliar with and had more knowledge about the other person’s perspective to rely on than participants in Experiments 1–15.

However, even when perspective taking increased confidence it did not reliably increase accuracy. 4 Some of the additional conditions we included across experi- ments, discussed in more detail in the Supplemental Materials, were meant to address potential explanations for these negative results of perspective taking on accuracy. None of these conditions yielded what we believe is a clear explanation. For instance, it is possible that perspective taking caused people to thinktoo hard, leading them to overlook intuitive responses that might have been correct. However, an explicit instruction for participants to think hard in Experiment 3 did not significantly reduce accuracy com- pared with the control condition (Ms 18.83 vs. 18.48, respec- tively,t(119) .46,p .64,d 0.02). Perspective taking might also have led participants to distrust their intuitions and kept them from going with their first more accurate intuitive response, but an explicit instruction to rely on intuitions actuallydecreasedaccu- racy in Experiment 4 compared with the control condition (Ms 17.84 vs. 18.89,t(129) 2.03,p .044,d 0.46). We also tested whether perspective taking leads to greater mimicry, but an ex- plicit instruction to mimic the smile of the person in the video did not change accuracy in Experiment 13 compared with the condi- tion (Ms 13.23 vs. 12.42, respectively,t(116) 1.15,p .252, d 0.30). It could also be argued that perspective taking did not reduce egocentrismenoughto measurably improve accuracy. Our experiments suggest otherwise: Explicitly instructing participants 4We also examined whether accuracy was predicted by three variables that yielded interesting results— confidence, response time, and perceived effort. Across Experiments 1–25 there was an overall weak positive rela- tionship between confidence and accuracy. This was true for all three measures of accuracy in Experiments 16-25: mean absolute difference,r .10,CI(.04, .15),Z 3.42,p .001, correlations between predicted and actual responses,r .11,CI(.05, .16),Z 3.64,p .001, and number of correct predictions,r .14,CI(.07, .20),Z 4.11,p .001. None of these correlations differed by condition: mean absolute difference,Q(1) 0.41,p .52, correlations between predicted and actual responses,Q(1) 0.28,p .59, and number of correct predictions,Q(1) 1.96,p .16. In Experiment 1-15, accuracy was weakly predicted by response time, r .08,CI( .16, .01),Z 2.26,p .024, but this relationship between accuracy and response time did not differ by condition,Q(1) 0.16,p .69. Accuracy, however, was not predicted by effort,r .03, CI( .10, .04),Z 0.83,p .41. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 567 PERSPECTIVE MISTAKING to rely on their own perspectives in Experiments 16 –20 and Experiments 22–23 meaningfully increased egocentrism (when measured as absolute difference between predicted and self- responses) relative to control conditions,d 0.64, 95% CI [ 0.94, 0.33],z 4.08,p .001, but did not significantly decrease accuracy (when measured as absolute difference between predicted and actual responses),d 0.09, 95% CI [0.07, 0.25], z 1.08,p .28. Although our experiments do not provide an explanation for why perspective taking sometimesdecreasedac- curacy, they clearly demonstrate that perspective taking does not systematicallyincreaseaccuracy.

Of course, it is important to keep these results in perspective. In particular, to measure the accuracy of social judgment, participants in our control conditions also needed to be predicting others’ thoughts, beliefs, or mental states. This means that participants in the control conditions of our experiments were already making inferences about another person’s perspective. Perspective taking could increase the use of accurate information that people already possess about another person when making decisions if they would have overlooked this information otherwise. For instance, perspec- tive taking might increase the likelihood that a politician would consider what he or she already knows about citizen’s attitudes and beliefs before proposing a policy. Such a result would simply reflect an increased accessibility about others’ thoughts and feel- ings while making a decision. Our research suggests that perspec- tive taking would not systematically increase the accuracy of a politician’s inferences about a citizen’s attitudes and beliefs. How perspective taking affects the use of available social knowledge is distinct from how perspective taking affects the accuracy of avail- able social knowledge.

Interestingly, the negative effect of perspective taking on accu- racy that we observed was more pronounced for strangers (Exper- iments 18 –21, 23, 24) than for partners in a relationship (Exper- iments 16, 17, 22, 23, 25) when accuracy was measured as absolute differences between predicted and actual responses (Strangers:d 0.20, 95% CI [0.07, 0.34],z 2.89,p .004, Partners:d 0.16, 95% CI [ 0.17, 0.48],z 0.96,p .34). This decrease in accuracy following perspective taking for strangers compared with partners was also apparent, but weaker, when accuracy was measured as the number of predicted correct re- sponses (Strangers:d 0.06, 95% CI [ 0.26, 0.14],z 0.62, p .54, Partners:d 0.05, 95% CI [ 0.47, 0.37],z 0.24, p .81) and as correlations between predicted and actual re- sponses (Strangers:d 0.12, 95% CI [ 0.32, 0.08],z 1.21, p .23, Partners:d 0.10, 95% CI [ 0.31, 0.10],z 0.97, p .33). Thus, the difference between partners and strangers in the effect of perspective taking on accuracy was unreliable. It is worthwhile to note, that for both strangers and close others, we failed to find any evidence that perspective taking systematically increasedinterpersonal accuracy.

Our final experiment suggests that there are likely to be much more effective ways of gaining more accurate insight into another person’s mind. In particular, human beings have also evolved a sophisticated language whose primary purpose is to convey the contents of one conscious mind to another (Pinker & Bloom, 1990). Increasing interpersonal understanding may come most readily from becoming a more effective questioner and listener, like a skilled journalist or a survey interviewer, rather than by trying to become a more routine perspective taker. If you want toknow what another person is thinking, it may be best to put them in a situation where they can answer honestly and then ask them directly.

This may seem like an obvious solution to increasing interper- sonal insight, but our final experiment found little evidence that this was obvious to the participants who were actually using this strategy. Indeed, the most interesting aspect of Experiment 25 was the notable disconnect between confidence and accuracy. Despite large differences in accuracy that came from using different strat- egies, confidence in judgment did not vary meaningfully across conditions. This is important because it suggests that people may have little insight into which strategies are likely to increase interpersonal understanding and which are not. This result is consistent with a small body of emerging research that finds meaningful misunderstanding of effective versus ineffective strat- egies for improving social cognition. In one experiment (Zhou et al., 2017), participants were asked to guess another person’s emo- tional reactions to an evocative series of images. Participants made their predictions either byreadingthe target’s expression by watching a video of his or her facial expressing, or bybeingin the person’s situation by seeing the image the target was rating.

Results indicated that participants were dramatically more accurate when they saw the image the target was rating, and yet participants tended to dramatically overestimate how effectively they could read the target’s expressions. In another experiment (Gilbert, Kill- ingsworth, Eyre, & Wilson, 2009), participants attempted to pre- dict their own emotional experience in an unknown event either by learning about details of the event or by getting another person’s report of his or her experience. Participants tended to believe they would be more accurate if they learned about details of the event, when they were actually more accurate if they got another person’s report of the experience. Each of these three lines of research suggest that people may overestimate the effectiveness of top- down processes of inference for understanding the mind of another person compared with bottom-up process of direct experience or knowledge acquisition. Mistaken expectations about how best to understand the minds of others could lead people to choose inef- fective strategies, thereby increasingmisunderstanding. Learning the cause of these mistaken expectations, and identifying their consequences, are pressing issues for future research.

Of course, there are limits to the accuracy that can be gained by trying to get another person’s perspective through bottom-up pro- cesses. Others may not tell the truth or know their own minds, such that self-reports are inaccurate. Emotional reactions to an experi- ence may differ, such that one person’s experience is a poor simulation for another’s experience. Or a simulation may turn out to be a poor proxy for the situation being simulated, such as a sighted person trying to simulate the experience of lifelong blind- ness by walking around a room blindfolded (Silverman, Gwinn, & Van Boven, 2015). No strategy for interpersonal understanding is perfect. The useful comparison standard is therefore not perfec- tion, but rather the accuracy obtained from other available strate- gies, as well as people’s beliefs about the effectiveness of these strategies. Research on both of these comparison standards is currently limited, and offer promising opportunities for future research. When it comes to understanding the mind of others, existing evidence suggests that people may systematically misun- derstand what’s good for them. This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 568 EYAL, STEFFEL, AND EPLEY Finally, we believe our experiments may be of practical value to those who are trying to understand the most complicated system any of us will ever think about—another person’s mind—a little bit better. Engaging in active perspective taking appears to have a number of reliable interpersonal consequences: it increases empa- thy for another person, increases the sense of similarity and con- nection to others, and encourages cooperation in negotiations. One recent theoretical model argues that perspective taking’s main benefit, in fact, is to strengthen social bonds (Galinsky et al., 2005). Our experiments are not inconsistent with this perspective.

If a person is wanting to feel more connected to another person, then imagining oneself in another’s shoes is likely to be a useful strategy to adopt. But if a person is really trying to gain an accurate understanding of another person’s mind, then another approach seems to be called for. If you really want to know what’s on the mind of another person, it is hard to do better than getting their perspective by just asking them. References Adams, G. S., Flynn, F. J., & Norton, M. I. (2012). The gifts we keep on giving: Documenting and destigmatizing the regifting taboo.Psycholog- ical Science, 23,1145–1150.http://dx.doi.org/10.1177/09567976 12439718 Ambady, N., Bernieri, F., & Richeson, J. (2000). Towards a histology of social behavior: Judgmental accuracy from thin slices of behavior. In M. P. Zanna (Ed.),Advances in experimental social psychology(Vol.

201–272). Amsterdam, The Netherlands: Elsevier.http://dx.doi.org/10 .1016/S0065-2601(00)80006-4 Ambady, N., & Gray, H. M. (2002). On being sad and mistaken: Mood effects on the accuracy of thin-slice judgments.Journal of Personality and Social Psychology, 83,947–961.http://dx.doi.org/10.1037/0022- 3514.83.4.947 Baron-Cohen, S., Wheelwright, S., Hill, J., Raste, Y., & Plumb, I. (2001).

The “Reading the Mind in the Eyes” Test revised version: A study with normal adults, and adults with Asperger syndrome or high-functioning autism.Journal of Child Psychology and Psychiatry, 42,241–251.

http://dx.doi.org/10.1111/1469-7610.00715 Baskin, E., Wakslak, C. J., Trope, Y., & Novemsky, N. (2014). Why feasibility matters more to gift receivers than to givers: A construal-level approach to gift giving.Journal of Consumer Research, 41,169 –182.

http://dx.doi.org/10.1086/675737 Batson, C. D., Early, S., & Salvarni, G. (1997). Perspective taking: Imag- ining how another feels versus imaging how you would feel.Personality and Social Psychology Bulletin, 23,751–758.http://dx.doi.org/10.1177/ 0146167297237008 Bernstein, M. J., Sacco, D. F., Brown, C. M., Young, S. G., & Claypool, H. M. (2010). A preference for genuine smiles following social exclu- sion.Journal of Experimental Social Psychology, 46,196 –199.http:// dx.doi.org/10.1016/j.jesp.2009.08.010 Birch, S. A. J., & Bloom, P. (2007). The curse of knowledge in reasoning about false beliefs.Psychological Science, 18,382–386.http://dx.doi .org/10.1111/j.1467-9280.2007.01909.x Bond, C. F., Jr., & DePaulo, B. M. (2006). Accuracy of deception judg- ments.Personality and Social Psychology Review, 10,214 –234.http:// dx.doi.org/10.1207/s15327957pspr1003_2 Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2010).

A basic introduction to fixed-effect and random-effects models for meta-analysis.Research Synthesis Methods, 1,97–111.http://dx.doi.org/ 10.1002/jrsm.12 Carnegie, D. (1936).How to win friends and influence people. New York, NY: Simon & Schuster.Caruso, E., Epley, N., & Bazerman, M. H. (2006). The costs and benefits of undoing egocentric responsibility assessments in groups.Journal of Personality and Social Psychology, 91,857– 871.http://dx.doi.org/10 .1037/0022-3514.91.5.857 Castelli, I., Baglio, F., Blasi, V., Alberoni, M., Falini, A., Liverta-Sempio, O.,...Marchetti, A. (2010). Effects of aging on mindreading ability through the eyes: An fMRI study.Neuropsychologia, 48,2586 –2594.

http://dx.doi.org/10.1016/j.neuropsychologia.2010.05.005 Cavanaugh, L. A., Gino, F., & Fitzsimons, G. J. (2015). When doing good is bad in gift-giving: Mis-predicting appreciation of socially responsible gifts.Organizational Behavior and Human Decision Processes, 131, 178 –189.http://dx.doi.org/10.1016/j.obhdp.2015.07.002 Chan, W., Mccrae, R. R., De Fruyt, F., Jussim, L., Löckenhoff, C. E., De Bolle, M.,...Terracciano, A. (2012). Stereotypes of age differences in personality traits: Universal and accurate?Journal of Personality and Social Psychology, 103,1050 –1066.http://dx.doi.org/10.1037/a0 029712 Chartrand, T. L., & Bargh, J. A. (1999). The chameleon effect: The perception-behavior link and social interaction.Journal of Personality and Social Psychology, 76,893–910.http://dx.doi.org/10.1037/0022- 3514.76.6.893 Cheung, E. O., Slotter, E. B., & Gardner, W. L. (2015). Are you feeling what I’m feeling? The role of facial mimicry in facilitating reconnection following social exclusion.Motivation and Emotion, 39,613– 630.

http://dx.doi.org/10.1007/s11031-015-9479-9 Davis, M. H. (1983). Measuring individual differences in empathy: Evi- dence for a multidimensional approach.Journal of Personality and Social Psychology, 44,113–126.http://dx.doi.org/10.1037/0022-3514 .44.1.113 Davis, M. H., Conklin, L., Smith, A., & Luce, C. (1996). Effect of perspective taking on the cognitive representation of persons: A merging of self and other.Journal of Personality and Social Psychology, 70, 713–726.http://dx.doi.org/10.1037/0022-3514.70.4.713 Dunbar, R. I. M. (1993). Coevolution of neocortical size, group size and language in humans.Behavioral and Brain Sciences, 16,681–735.

http://dx.doi.org/10.1017/S0140525X00032325 Epley, N., Caruso, E., & Bazerman, M. H. (2006). When perspective taking increases taking: Reactive egoism in social interaction.Journal of Per- sonality and Social Psychology, 91,872– 889.http://dx.doi.org/10.1037/ 0022-3514.91.5.872 Epley, N., Keysar, B., Van Boven, L., & Gilovich, T. (2004). Perspective taking as egocentric anchoring and adjustment.Journal of Personality and Social Psychology, 87,327–339.http://dx.doi.org/10.1037/0022- 3514.87.3.327 Epley, N., Morewedge, C., & Keysar, B. (2004). Perspective taking in children and adults: Equivalent egocentrism but differential correction.

Journal of Experimental Social Psychology, 40,760 –768.http://dx.doi .org/10.1016/j.jesp.2004.02.002 Eyal, T., & Epley, N. (2010). How to seem telepathic: Enabling mind reading by matching construal.Psychological Science, 21,700 –705.

http://dx.doi.org/10.1177/0956797610367754 Fisher, R. A. (1915). The frequency distribution of the correlation coeffi- cient in samples from an indefinitely large population.Biometrika, 10, 507–521.

Flynn, J. F., & Adams, G. S. (2009). Money can’t buy love: Asymmetric beliefs about gift price and feelings of appreciation.Journal of Experi- mental Social Psychology, 45,404 – 409.http://dx.doi.org/10.1016/j.jesp .2008.11.003 Funder, D. C. (1995). On the accuracy of personality judgment: A realistic approach.Psychological Review, 102,652– 670.http://dx.doi.org/10 .1037/0033-295X.102.4.652 Galinsky, A. D., Ku, G., & Wang, C. S. (2005). Perspective-taking and self-other overlap: Fostering social bonds and facilitating social coordi- This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 569 PERSPECTIVE MISTAKING nation.Group Processes & Intergroup Relations, 8,109 –125.http://dx .doi.org/10.1177/1368430205051060 Galinsky, A. D., Maddux, W. W., Gilin, D., & White, J. B. (2008). Why it pays to get inside the head of your opponent: The differential effects of perspective taking and empathy in negotiations.Psychological Sci- ence, 19,378 –384.http://dx.doi.org/10.1111/j.1467-9280.2008.02096.x Galinsky, A. D., Magee, J. C., Inesi, M. E., & Gruenfeld, D. H. (2006).

Power and perspectives not taken.Psychological Science, 17,1068 – 1074.http://dx.doi.org/10.1111/j.1467-9280.2006.01824.x Galinsky, A. D., & Moskowitz, G. B. (2000). Perspective-taking: Decreas- ing stereotype expression, stereotype accessibility, and in-group favor- itism.Journal of Personality and Social Psychology, 78,708 –724.

http://dx.doi.org/10.1037/0022-3514.78.4.708 Galinsky, A. D., & Mussweiler, T. (2001). First offers as anchors: The role of perspective-taking and negotiator focus.Journal of Personality and Social Psychology, 81,657– 669.http://dx.doi.org/10.1037/0022-3514 .81.4.657 Galinsky, A. D., Wang, C. S., & Ku, G. (2008). Perspective-takers behave more stereotypically.Journal of Personality and Social Psychology, 95, 404 – 419.http://dx.doi.org/10.1037/0022-3514.95.2.404 Genschow, O., Florack, A., & Wänke, M. (2013). The power of movement:

Evidence for context-independent movement imitation.Journal of Ex- perimental Psychology: General, 142,763–773.http://dx.doi.org/10 .1037/a0029795 Gilbert, D. T., Killingsworth, M. A., Eyre, R. N., & Wilson, T. D. (2009).

The surprising power of neighborly advice.Science, 323,1617–1619.

http://dx.doi.org/10.1126/science.1166632 Gilin, D., Maddux, W. W., Carpenter, J., & Galinsky, A. D. (2013). When to use your head and when to use your heart: The differential value of perspective-taking versus empathy in competitive interactions.Person- ality and Social Psychology Bulletin, 39,3–16.http://dx.doi.org/10 .1177/0146167212465320 Gino, F., & Flynn, F. (2011). Give them what they want: The benefits of explicitness in gift exchange.Journal of Experimental Social Psychol- ogy, 47,915–922.http://dx.doi.org/10.1016/j.jesp.2011.03.015 Greenwald, A. G., Poehlman, T. A., Uhlmann, E. L., & Banaji, M. R.

(2009). Understanding and using the Implicit Association Test: III.

Meta-analysis of predictive validity.Journal of Personality and Social Psychology, 97,17– 41.http://dx.doi.org/10.1037/a0015575 Hall, J. A., Blanch, D. C., Horgan, T. G., Murphy, N. A., Rosip, J. C., & Schmid Mast, M. (2009). Motivation and interpersonal sensitivity: Does it matter how hard you try?Motivation and Emotion, 33,291–302.

http://dx.doi.org/10.1007/s11031-009-9128-2 Hall, J. A., & Schmid Mast, M. (2007). Sources of accuracy in the empathic accuracy paradigm.Emotion, 7,438 – 446.http://dx.doi.org/10 .1037/1528-3542.7.2.438 Herrmann, E., Call, J., Herna `ndez-Lloreda, M. V., Hare, B., & Tomasello, M. (2007). Humans have evolved specialized skills of social cognition:

The cultural intelligence hypothesis.Science, 317,1360 –1366.http://dx .doi.org/10.1126/science.1146282 Hess, U., & Blairy, S. (2001). Facial mimicry and emotional contagion to dynamic emotional facial expressions and their influence on decoding accuracy.International Journal of Psychophysiology, 40,129 –141.

http://dx.doi.org/10.1016/S0167-8760(00)00161-6 Hoch, S. J. (1987). Perceived consensus and predictive accuracy: The pros and cons of projection.Journal of Personality and Social Psychology, 53,221–234.http://dx.doi.org/10.1037/0022-3514.53.2.221 Ickes, W. (1997).Empathic accuracy. New York, NY: Guilford Press.

Jerison, H. J. (1971). More on why birds and mammals have big brains.

American Naturalist, 105,185–189. http://dx.doi.org/10.1086/282714 Jussim, L., Crawford, J. T., & Rubinstein, R. S. (2015). Stereotype (in)ac- curacy in perceptions of groups and individuals.Current Directions in Psychological Science, 24,490 – 497.http://dx.doi.org/10.1177/0963 721415605257Karniol, R. (2003). Egocentrism versus protocentrism: The status of self in social prediction.Psychological Review, 110,564 –580.http://dx.doi .org/10.1037/0033-295X.110.3.564 Kenny, D. A. (1991). A general model of consensus and accuracy in interpersonal perception.Psychological Review, 98,155–163.http://dx .doi.org/10.1037/0033-295X.98.2.155 Macrae, C. N., Bodenhausen, G. V., Milne, A. B., & Jetten, J. (1994). Out of mind but back in sight: Stereotype on the rebound.Journal of Personality and Social Psychology, 67,808 – 817.http://dx.doi.org/10 .1037/0022-3514.67.5.808 Maner, J. K., Luce, C. L., Neuberg, S. L., Cialdini, R. B., Brown, S., & Sagarin, B. J. (2002). The effects of perspective taking on motivations for helping: Still no evidence for altruism.Personality and Social Psychology Bulletin, 28,1601–1610.http://dx.doi.org/10.1177/014 616702237586 Marist Poll. (2011).McClatchy-Marist poll for public opinion. Retrieved fromhttp://maristpoll.marist.edu/28-holy-super-powers-batman-mind- reading-and-time-travel-top-list/ Niedenthal, P. M., Brauer, M., Halberstadt, J. B., & Innes-Ker, A. H.

(2001). When did her smile drop? Facial mimicry and the influences of emotional state on the detection of change in emotional expression.

Cognition and Emotion, 15,853– 864.http://dx.doi.org/10.1080/02699 930143000194 Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know:

Verbal reports on mental processes.Psychological Review, 84,231–259.

http://dx.doi.org/10.1037/0033-295X.84.3.231 Nowicki, S., Jr., & Duke, M. P. (1994). Individual differences in the nonverbal communication of affect: The Diagnostic Analysis of Non- verbal Accuracy Scale.Journal of Nonverbal Behavior, 18,9 –35.http:// dx.doi.org/10.1007/BF02169077 Oberman, L. M., Winkielman, P., & Ramachandran, V. S. (2007). Face to face: Blocking facial mimicry can selectively impair recognition of emotional expressions.Social Neuroscience, 2,167–178.http://dx.doi .org/10.1080/17470910701391943 Oswald, F. L., Mitchell, G., Blanton, H., Jaccard, J., & Tetlock, P. E.

(2013). Predicting ethnic and racial discrimination: A meta-analysis of IAT criterion studies.Journal of Personality and Social Psychology, 105,171–192.http://dx.doi.org/10.1037/a0032734 Payne, J. W., Bettman, J. R., & Johnson, E. J. (1988). Adaptive strategy selection in decision making.Journal of Experimental Psychology:

Learning, Memory, and Cognition, 14,534 –552.http://dx.doi.org/10 .1037/0278-7393.14.3.534 Pinker, S., & Bloom, P. (1990). Natural language and natural selection.

Behavioral and Brain Sciences, 13,707–727.http://dx.doi.org/10.1017/ S0140525X00081061 Pronin, E., & Ross, L. (2006). Temporal differences in trait self-ascription:

When the self is seen as an other.Journal of Personality and Social Psychology, 90,197–209.http://dx.doi.org/10.1037/0022-3514.90.2.197 Ramachandran, V. S. (2004).A brief tour of human consciousness: From impostor poodles to purple numbers. New York, NY: Pi Press.

Ross, L., Greene, D., & House, P. (1977). The “false consensus effect”: An egocentric bias in social perception and attribution processes.Journal of Experimental Social Psychology, 13,279 –301.http://dx.doi.org/10 .1016/0022-1031(77)90049-X Ruben, M. A., & Hall, J. A. (2013). “I know your pain”: Proximal and distal predictors of pain detection accuracy.Personality and Social Psychology Bulletin, 39, 1346 –1358.http://dx.doi.org/10.1177/0146 167213493188 Savitsky, K., Van Boven, L., Epley, N., & Wight, W. (2005). The unpack- ing effect in allocations of responsibility for group tasks.Journal of Experimental Social Psychology, 41,447– 457.http://dx.doi.org/10 .1016/j.jesp.2004.08.008 Schwarz. (1999). How the questions shape the answers.American Psy- chologist, 54,93–105.http://dx.doi.org/10.1037//0003-066X.54.2.93 This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 570 EYAL, STEFFEL, AND EPLEY Silverman, A., Gwinn, J., & Van Boven, L. (2015). Stumbling in their shoes: Brief experience simulations reduce judged competency of the disabled.Social Psychological and Personality Science, 6,464 – 471.

http://dx.doi.org/10.1177/1948550614559650 Steffel, M., & LeBoeuf, R. (2014). Over-individuation in gift giving:

Shopping for multiple recipients leads givers to choose unique but less preferred gifts.Journal of Consumer Research, 40,1167–1180.http:// dx.doi.org/10.1086/674199 Stel, M., & van Knippenberg, A. (2008). The role of facial mimicry in the recognition of affect.Psychological Science, 19,984 –985.http://dx.doi .org/10.1111/j.1467-9280.2008.02188.x Stinson, L., & Ickes, W. (1992). Empathic accuracy in the interactions of male friends versus male strangers.Journal of Personality and Social Psychology, 62,787–797.http://dx.doi.org/10.1037/0022-3514.62.5.787 Swann, W. B., Jr., & Gill, M. J. (1997). Confidence and accuracy in person perception: Do we know what we think we know about our relationship partners?Journal of Personality and Social Psychology, 73,747–757.

http://dx.doi.org/10.1037/0022-3514.73.4.747 Teigen, K. H., Olsen, M. V. G., & Solås, O. E. (2005). Giver-receiver asymmetries in gift preferences.British Journal of Social Psychology, 44,125–144.http://dx.doi.org/10.1348/014466604X23428 Todd, A. R., Bodenhausen, G. V., Richeson, J. A., & Galinsky, A. D.

(2011). Perspective taking combats automatic expressions of racial bias.

Journal of Personality and Social Psychology, 100,1027–1042.http:// dx.doi.org/10.1037/a0022308 Todd, A. R., Galinsky, A. D., & Bodenhausen, G. V. (2012). Perspective- taking undermines stereotype maintenance processes: Evidence from social memory, behavior explanation, and information solicitation.So- cial Cognition, 30,94 –108.http://dx.doi.org/10.1521/soco.2012.30.1.94 Todd, A. R., Hanko, K., Galinsky, A. D., & Mussweiler, T. (2011). When focusing on differences leads to similar perspectives.Psychological Science, 22,134 –141.http://dx.doi.org/10.1177/0956797610392929 Todorov, A., Olivola, C. Y., Dotsch, R., & Mende-Siedlecki, P. (2015).

Social attributions from faces: Determinants, consequences, accuracy, and functional significance.Annual Review of Psychology, 66,519 –545.

http://dx.doi.org/10.1146/annurev-psych-113011-143831 Trötschel, R., Hüffmeier, J., Loschelder, D. D., Schwartz, K., & Gollwit- zer, P. M. (2011). Perspective taking as a means to overcome motiva- tional barriers in negotiations: When putting oneself into the opponent’s shoes helps to walk toward agreements.Journal of Personality and Social Psychology, 101,771–790.http://dx.doi.org/10.1037/a0023801Van Boven, L., Judd, C. M., & Sherman, D. K. (2012). Political polariza- tion projection: Social projection of partisan attitude extremity and attitudinal processes.Journal of Personality and Social Psychology, 103,84 –100.http://dx.doi.org/10.1037/a0028145 Van Doesum, N. J., Van Lange, D. A., & Van Lange, P. A. (2013). Social mindfulness: Skill and will to navigate the social world.Journal of Personality and Social Psychology, 105,86 –103.http://dx.doi.org/10 .1037/a0032540 Vorauer, J. D., & Sasaki, S. J. (2009). Helpful only in the abstract?

Psychological Science, 20,191–197.http://dx.doi.org/10.1111/j.1467- 9280.2009.02265.x Vorauer, J. D., & Sucharyna, T. A. (2013). Potential negative effects of perspective-taking efforts in the context of close relationships: Increased bias and reduced satisfaction.Journal of Personality and Social Psy- chology, 104,70 – 86.http://dx.doi.org/10.1037/a0030184 Wade-Benzoni, K. A., Tenbrunsel, A. E., & Bazerman, M. H. (1996).

Egocentric interpretations of fairness in asymmetric, environmental so- cial dilemmas: Explaining harvesting behavior and the role of commu- nication.Organizational Behavior and Human Decision Processes, 67, 111–126.http://dx.doi.org/10.1006/obhd.1996.0068 Yaniv, I., & Choshen-Hillel, S. (2012). When guessing what another person would say is better than giving your own opinion: Using perspective-taking to improve advice-taking.Journal of Experimental Social Psychology, 48,1022–1028.http://dx.doi.org/10.1016/j.jesp.2012 .03.016 Zhang, Y., & Epley, N. (2009). Self-centered social exchange: Differential use of costs versus benefits in prosocial reciprocity. Journal of Person- ality and Social Psychology, 97,796 – 810.http://dx.doi.org/10.1037/ a0016233 Zhang, Y., & Epley, N. (2012). Exaggerated, mispredicted, and misplaced:

When “it’s the thought that counts” in gift exchanges.Journal of Experimental Psychology: General, 141,667– 681.http://dx.doi.org/10 .1037/a0029223 Zhou, H., Majka, E. A., & Epley, N. (2017). Inferring perspective versus getting perspective: Understanding the value of being in another per- son’s shoes.Psychological Science, 28,482– 493.http://dx.doi.org/10 .1177/0956797616687124 Received September 15, 2017 Revision received December 13, 2017 Accepted December 14, 2017 This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 571 PERSPECTIVE MISTAKING