Similarity and Attractiveness 1 Running head: SIMILARITY AND ATTRACTIVENESS IN PERCEPTION Similarity and Attractiveness in Social Perception: Differentiating Between Biases for the Self and the Beautiful

An impressive empirical literature suggests that physical attractiveness is associated with interpersonal attraction and with specific personality traits, some of which are favorable and many of which are objectively unrelated to attractiveness. Interestingly, much of the research investigating the impact of attractiveness on liking and personality inference has not accounted for the role of the self in social perception. In this project I investigated the possibility that perceptual similarity between the self and target contributes to differential judgments of attractive and unattractive individuals. Participants completed self-ratings of physical attractiveness and evaluated a target who was high, medium, or low in physical attractiveness. As expected, participants rated themselves as relatively attractive and as similar on attractiveness to attractive targets. Target evaluations varied as a function of target attractiveness; however, selfratings on attractiveness moderated the impact of target attractiveness on liking for and personality inferences about the target. Attractiveness similarity also completely mediated the effect of target attractiveness on liking. The findings are discussed with regards to implicit personality theory and implicit egotism. Similarity and Attractiveness 3 Similarity and Attractiveness in Social Perception: Differentiating Between Biases for the Self and the Beautiful Beautiful people curry favor in extensive ways and in a variety of contexts. They receive lighter prison sentences for the same crime (Kulka & Kessler, 1978; Soloman & Schopler, 1978) and better grades for similar work (Landy & Sigall, 1974), are more persuasive (Parekh & Kanekar, 1994), and elicit more favorable interpersonal behavior (Snyder, Tanke, & Berscheid, 1977) and occupational evaluations (Chaikin, Gillen, Derlega, Heinen, & Wilson, 1978) than their less attractive counterparts. Attractive individuals are also perceived as having better mental health (Cash, Kehr, Polyson, & Freeman, 1977) and social aptitude, as being happier with life and more successful in their occupations (Dion, Berscheid, & Walster, 1972), and as more popular, intelligent, confident, assertive (see Eagly, Ashmore, Makhijani, & Longo, 1991 for a review), likable, and romantically desirable (Stroebe, Insko, Thompson, & Layton, 1971) than less attractive people. So pervasive are the emotional, evaluative, and inferential differences inspired by attractive and unattractive individuals that researchers originally subscribed to the notion that “what is beautiful is good.” Though we now understand this idiom as an oversimplification (see Dermer, & Thiel, 1974), the advantages afforded attractive individuals are substantial and varied. To explain the favor afforded attractive individuals, researchers have cited a halo effect (i.e., individuals who are gifted in one aspect are assumed to be gifted in other aspects as well; Nisbett, & Wilson, 1977; Wetzel, Wilson, & James, 1981), implicit personality theories created by cultural representation and direct observation of attractive individuals (Ashmore, 1981, Eagly, et. al., 1991), egoistic needs that are served via Similarity and Attractiveness 4 identification with attractive partners (Campbell, 1999; Sigall, & Landy, 1973), and evolutionary mechanisms that prioritize attractiveness in mate selection (Li, Bailey, Kenrick, & Linsenmeier, 2002). In this paper, I explore additional influences on judgments of attractive and unattractive individuals: self-perceptions and similarity to the self. Previous investigations of the judgmental effects of target attractiveness have proceeded, for the most part, without acknowledging the role of the evaluator’s selfbeliefs in the judgment process, a curious omission given the importance of the self in social perception (e.g., Green & Sedikides, 2001; Lewicki, 1983; Sedikides & Skowronski, 1993). In the current project I address this empirical void by investigating the role of evaluator physical attractiveness (the self-rated physical attractiveness of the individual evaluating the target) in the perception of and feelings about social targets who vary in physical attractiveness. In doing so, I merge relevant perspectives regarding (a) the impact of attractiveness on social judgment, (b) biased self-perceptions, and (c) the role of similarity in social perception. Attractiveness Effects on Social Judgment As discussed above, the physical attractiveness of a target influences a variety of evaluative and judgmental outcomes. The current project focused attention onto the impact of attractiveness on liking and personality inference. Liking. Physical attractiveness (PA) is a powerful predictor of interpersonal attraction; in fact, early investigations of attraction suggested that PA was the single potent predictor of liking and romantic interest (Walster, Aronson, Abrahams, & Rottman, 1966). More recent studies have identified additional predictors of attraction Similarity and Attractiveness 5 (e.g., reciprocity, similarity, etc.); however, the importance of PA for predicting liking, as well as romantic attraction, is well established (Byrne, London, & Reeves, 1968; Li, Bailey, Kenrick, & Linsenmeier, 2002; Snyder, Tanke, & Berscheid, 1977; Stroebe, Insko, Thompson, & Layton, 1971; Timmerman, & Hewitt, 1980). Personality Inferences. The physical attractiveness stereotype refers to a set of personality traits that are linked cognitively to attractiveness (Eagly, et. al., 1991). In an exhaustive meta-analysis of the physical attractiveness stereotype, Eagly and colleagues concluded that attractive, relative to unattractive, targets are consistently rated more favorably on social characteristics, such as popularity, social skills, and social confidence (also see Bassili, 1981; Dion, Berscheid, & Walster, 1972; Dushenko, Perry, Schilling, & Smolarski, 1978; Eagly, et. al., 1991; Johnson & Pittenger, 1984), that the effect sizes for intellectual competence, potency, and adjustment favor attractive individuals but are typically moderate, and that attractive and unattractive individuals are evaluated similarly on items related to integrity (honesty, concern for others, etc.). Interestingly, attractive individuals may be regarded as more vain and egotistical and as more likely than unattractive individuals to divorce or to engage in infidelity (Dermer & Thiel, 1975). Recent theorizing regarding the physical attractiveness stereotype suggests that attractiveness is implicitly, and automatically, associated with a particular set of personality traits (Eagly, et. al., 1991) and that these associations result from direct observation and cultural (i.e., media) representations of attractive and unattractive individuals. The current project investigated the possibility that self-perceptions contribute to and/or qualify the personality traits one associates with physical attractiveness. Consistently with this possibility, Pelham and colleagues found that Similarity and Attractiveness 6 participants high in self-rated PA evaluated attractive targets more positively than participants low in self-rated PA (Pelham, Ly, & Hetts, 2002). This result provides preliminary evidence that the physical attractiveness stereotype varies as a function of self-perception. Biased Self-Evaluations How might self-perceptions explain the favor attractive individuals enjoy? Such explanation is grounded in the chronic valence of selfrepresentations. Most individuals regard themselves positively (Schwartz, 1996), have high self-esteem (Taylor & Brown, 1988), and display attributional (Miller & Ross, 1975), memorial (Sedikides & Green, 2000) and evaluative biases that favor the self. People regard themselves as better than the average person (Alicke, 1985; Brown, 1986; Svenson, 1981) and more positively (1) than they are evaluated by others (Ruble, Eisenberg, & Higgins, 1994) and (2) than is warranted by objective criteria (Epley & Dunning, 2000; Gabriel, Critelli, & Ee, 1994). The pervasive tendency for self-promotion suggests that people will regard themselves as relatively attractive (as more attractive than the average person) and thus, as similar, on attractiveness, to attractive targets. The perceptual similarity between the self and attractive targets that is a logical result of inflated self-perceptions of attractiveness creates ambiguity in the interpretation of research that has compared evaluations of attractive and unattractive individuals. Specifically, attractive targets are not only more attractive than unattractive targets; they are also more similar to the self. Given this confound, the conclusion that people are biased in favor of (or against) attractive targets because of the target’s attractiveness, per se, is premature. Similarity and Attractiveness 7 Self-similarity The associa tion between similarity and attraction is among the most robust in the social psychological literature (See Byrne, & Nelson, 1965 for a review). People regard favorably individuals with whom they share attitudes and characteristics (Duck & Craig, 1978; Rosenblatt & Greenberg, 1991; Sprecher & Duck, 1994) and even seem to be drawn to towns and occupations whose names are similar to their own (Pelham, Mirenberg, & Jones, 2002). There is also substantial evidence that similarity of physical attractiveness influences interpersonal evaluation and functioning. Romantic partners and male-friend pairs are more similar on objective and selfrated physical attractiveness than would be expected by chance (Cash & Derlega, 1978; Feingold, 1988), and individuals choose romantic/dating partners whose physical attractiveness level is similar to their own (Berscheid, Dion, Walter, & Walster, 1971; Stroebe, Thompson, Insko, & Layton, 1970). Additionally, physical attractiveness similarity is a significant predictor of relationship progress (White, 1980) with more similar partners, relative to less similar partners, reporting greater increases in love over time. Taken together, these findings (commonly regarded as evidence for a “matching for attractiveness” phenomenon) suggest that physical attractiveness similarity is associated with interpersonal attraction and positive evaluation. Overview The current project assessed the importance of the evaluator’s self-rated physical attractiveness (PA) and evaluator-target PA similarity on perceptions of targets who varied in PA. More specifically, I investigated the possibility that previous research Similarity and Attractiveness 8 regarding the impact of target PA on judgment has confused a bias for beauty with a bias for similarity. The project includes a pilot study and an experiment. In the pilot study, I identified photographs of individuals who the population sampled in the subsequent experiment considered high, medium, or low in PA. In the main experiment participants completed self-assessments and made judgments about a male or female target who was high, medium, or low in PA. More specifically, participants expressed how much they would like the target and rated the target on various personality dimensions (e.g., sociability, integrity, potency, etc.). The empirical foundation reviewed previously drives four hypotheses for the current research. First, I expected participants to regard themselves as more attractive than the average college student. Second, I expected that participants would regard themselves as most similar on attractiveness to relatively attractive individuals. Third, consistently with previous research regarding the impact of target attractiveness on judgment, I expected participants to express more liking for physically attractive, relative to unattractive, targets and to evaluate attractive targets more positively than less attractive targets on items related to sociability (e.g., assertiveness, extraversion, etc.) but not on items related to integrity (Eagly, et. al., 1991). Finally, and most critically, I expected that participants would evaluate most favorably those targets who they viewed as similar to themselves on PA, independent of target PA. More specifically, I expected that (1) participants’ self-rated PA would moderate the influence of target PA on evaluation such that participants high in self-rated PA (high PA participants) would inflate evaluations of high PA targets whereas Similarity and Attractiveness 9 participants low in self-rated PA (low PA participants) would not, and (2) PA similarity would mediate the influence of target PA on target evaluation. Pilot Study Twenty individuals (eight males and 12 females) who were unaware of research hypotheses and procedures rated 115 wallet-size photographs of 18-20 year-old males (59 pictures) and females (56 pictures) for PA. Raters evaluated each photograph on a sevenpoint scale (1 = “not at all attractive” and 7 = “very attractive”). Two male and two female photographs were chosen to represent each of three levels of PA: high, medium, and low. The two males (Ms = 4.91, 4.82) and two females (Ms = 6.13, 6.46) who were rated, on average, as most attractive were chosen as high PA targets. The two males and two females who received average ratings closest to the scale midpoint (= 4) were selected as medium PA targets (male medium target Ms = 4.04, 4.04; female medium target Ms = 4.00, 3.96). Finally, the two males and two females who received the lowest average ratings were selected as low PA targets (male low targets, Ms = 1.50, 1.30; female low targets, Ms = 1.92, 1.89). There were no rater gender effects for the ratings of the 12 target individuals chosen (ps ranging from .11 to .89); male and female raters evaluated each of the targets similarly. Method Participants and Experimental Design One hundred and forty-two introductory psychology students at the University of North Carolina at Chapel Hill participated as partial fulfillment of a course option. The 65 men and 77 women were assigned randomly to one of six experimental conditions Similarity and Attractiveness 10 created by a 2 (target gender: male/female) X 3 (target PA: high/medium/low) factorial design. I also included participant gender as a factor in all statistical analyses. Measures Target assessment. Participants (Ps) completed sixteen items regarding the target person. One item assessed liking for the target (“I think I would like this person...”; endpoints, 1 = not at all and 10 = very much). Twelve items tapped personality dimensions identified by previous research as relevant to the PA stereotype (popularity, honesty, intelligence, etc.). Ps rated the target on each item relative to other college students (endpoints 1 = bottom 5% and 10 = top 5%), and these twelve items were subjected to principal components analysis with direct oblimin rotation. This analysis revealed two factors that accounted for 63% of the variance. The factors displayed adequate simple structure prior to rotation and were interpreted as (1) sociability (adjustment, confidence, assertiveness, popularity, sociability, exciting) and (2) integrity (honesty, morality, concern for others, sincerity, intelligence, kindness). I created two composite indices by averaging items that loaded onto each factor (sociability: 6 items, alpha = .86; integrity: 6 items, alpha = .89). Ps also rated the target’s PA relative to other college students (endpoints 1 = bottom 5% and 10 = top 5%) as a manipulation check. Finally, participants completed an item regarding PA similarity to the target (“In reference to level of physical attractiveness, how similar to you is the pictured person?”; endpoints, 1 = not at all similar and 10 = very similar) and an item assessing overall similarity to the target (“Overall, how similar to you do you think the pictured person is?”; endpoints, 1 = not at all similar and 10 = very similar). Similarity and Attractiveness 11 Selfand Ideal self-assessments. Ps evaluated themselves relative to other college students on ten dimensions including PA (endpoints, 1 = bottom 5% and 10 = top 5%). Ps also evaluated their ideal selves, the kind of people they “would ideally like to be,” on the same 10 dimensions. Procedure The experimenter informed Ps that the experimental session would include multiple pilot studies that were unrelated to one another. Under this pretense, Ps completed self, ideal self, and target assessments in individual rooms. Ps were assigned randomly to evaluate either a male or a female target who was high, medium, or low in PA. The order of the assessments (self, ideal self, and target assessments) was counterbalanced to control for order effects. Ps were under no time restrictions while completing assessments. The experimental session concluded with a full debriefing. Results Manipulation check I investigated Ps’ ratings of target PA with a 2 (participant gender) x 2 (target gender) x 3 (target PA) analysis of variance (ANOVA). The target PA effect was significant, F (2, 130) = 39.75, p < .001. Ps rated high PA targets (M = 7.48, SD = 1.79) as more physically attractive than either medium (M = 6.76, SD = 1.48; Tukey’s HSD p < .08) or low (M = 4.60, SD = 1.59; p < .001) PA targets. Ps regarded medium PA targets as significantly more attractive than low PA targets (p < .001). No other effects reached significance. Ps’ ratings of the targets’ PA confirmed the pilot ratings discussed previously. Similarity and Attractiveness 12 Did participants regard themselves as more attractive than the average college student? In order to investigate this hypothesis, I compared Ps’ self-rated PA to the scale midpoint (5.5). As expected, Ps rated themselves, on average, significantly higher than the scale midpoint (M = 6.81, SD = 1.27), t (141) = 12.33, p < .001. There was no gender difference on selfrated PA, t (140) = .58, p > .50. Men (M = 6.88, SD = 1.23) and women (M = 6.75, SD = 1.30) evaluated themselves similarly on attractiveness. Interestingly, only 18 of the 142 Ps rated themselves in the bottom 50% of college students with regard to PA. Did participants regard themselves as more similar (on attractiveness) to attractive targets? I created a PA similarity index (hereafter referred to as “similarity index”) by (1) subtracting each participant’s rating of target PA from his or her self-rating of PA and (2) taking the absolute value of each difference. The similarity index constituted a unidirectional measure of perceptual distance between self and target PA ratings (see Duck, 1973; Duck & Craig, 1978) and served as the dependent measure in the analyses described here and as the mediator in analyses described below. I subjected the similarity index to the 2 x 2 x 3 ANOVA. The expected target PA main effect was significant, F (2, 130) = 5.37, p < .01. Ps rated high (M = 1.60, SD = 1.09) and medium (M = 1.57, SD = 1.35) PA targets more similarly to the self on attractiveness than they did low PA targets (M = 2.44, SD = 1.77), Tukey’s p’s < .02 and .01, respectively. Index scores for high and medium PA targets were not significantly different, p > .75. Similarity and Attractiveness 13 Did participants evaluate attractive targets more positively on sociability and liking but not on integrity? I subjected sociability and integrity composite indices and liking to the 2 x 2 x 3 (target PA) ANOVA. The target PA main effect was significant for the three outcome measures, and the patterns of effects were generally consistent with expectations. 5 With regards to sociability, Ps rated high PA targets as more sociable than both medium and low PA targets, Tukey’s p’s < .01 and .001, respectively, and rated medium PA targets as more sociable than low PA targets, p < .001, target PA main effect F (2, 128) = 33.66, p < .001 (See Table 1 for Means and Standard Deviations). For integrity, Ps evaluated medium and low PA targets more favorably than high PA targets, Tukey’s post-hoc p’s < .01 and < .05, respectively. Ratings of medium PA and low PA targets did not differ significantly, p >.55, target PA main effect, F (2, 128) = 5.50, p < .01. Finally, for liking, Ps regarded medium PA targets as significantly more likable than low PA targets, p < .02. Additionally, Ps rated high PA targets as descriptively more likable than low PA targets, though the post-hoc comparison did not reach significance, p > .25, target PA main effect F (2, 128) = 14.24, p < .02. Did self-rated PA moderate the influence of target PA on evaluation? In order to test for moderation, I subjected each outcome measure to a four-factor, factorial ANOVA. Participant gender, target gender, and target PA were categorical factors, whereas selfrated PA was continuous. I centered self-rated PA and each outcome measure. As a reminder, I expected that that Ps high in self-rated PA (high PA Ps) would inflate evaluations of high PA targets whereas low PA Ps would not. Where appropriate, I Similarity and Attractiveness 14 computed predicted outcome values for hypothetical participants 1 standard deviation above and below the mean of selfrated PA. In each analysis, the target PA main effect reached significance. These effects were similar to those reported previously, and thus, I do not discuss them further and focus attention here on the crucial higher-order effects. Sociability. When investigating perceptions of sociability, the expected self-rated PA x target PA was marginal, F (2, 116) = 2.95, p < .06. Both high and low PA targets evaluated high PA targets as most sociable and low PA targets as least sociable; however, high PA Ps inflated ratings of high PA targets to a greater degree than did low PA Ps (See Table 2 for predicted values). Within the high PA target condition, the relation between self-rated PA and sociability ratings was marginal, B = .28, t = 1.82, p < .08. The more positive a participant’s selfrated PA, the more sociable he or she regarded the attractive target. The relation between self-rated PA and sociability ratings did not reach significance in either the medium PA, B = .06, t = .48, p > .40, or low PA, B = -.20, t = 1.53, p > .10, target conditions. No other effects involving self-rated PA approached significance. Integrity. For perceptions of integrity, the self-rated PA x target PA interaction was significant, F (2, 116) = 4.97, p < .01. Consistently with expectations, high PA Ps evaluated high PA targets significantly more favorably than did low PA Ps (see Table 2), B = .33, t = 2.19, p < .05. In contrast, low PA Ps evaluated medium PA targets more favorably than did high PA Ps, B = -.22, t = -1.95, p < .06. In the low PA target condition the relation between self-rated PA and integrity ratings did not approach significance, B = .08, t = .72, p > .45. Similarity and Attractiveness 15 Interestingly, the significant selfrated PA X target PA interaction was qualified by a three-way interaction among participant gender, self-rated PA, and target PA, F (2, 116) = 4.27, p < .02. The interaction between self-rated PA and target PA was significant for male Ps, F (2, 52) = 5.10, p < .01, but was not significant for female Ps, F (2, 64) = 1.02, p < .40. Predicted values for the significant interaction indicate that high PA male Ps, relative to low PA male Ps, were particularly favorable towards high PA targets and unfavorable towards medium PA targets. Liking. With regards to liking for the target, the critical self-rated PA x target PA interaction was significant, F (2, 116) = 6.87, p < .01. As expected, high PA Ps expressed the most liking for high PA targets and the least liking for low PA targets (Table 2). Additionally, these Ps evaluated high PA targets more favorably than did low PA Ps, B = .59, t = 3.32, p < .01. In contrast, low PA Ps expressed the most liking for medium PA targets and the least liking for high PA targets; two non-significant trends suggested that these Ps were more favorable towards medium PA, B = -.32, t = -1.62, p = .12 and low PA, B = -.27, t = -1.52, p = .14, targets than were high PA Ps. In summary, participants who regarded themselves as attractive were more favorable towards high PA targets than were participants who did not hold themselves in such high regard. The nature of the personality inferences participants made about attractive and unattractive targets varied as a function of self-perceptions. Further, the results are consistent with the notion that attractiveness similarity contributes to differential perceptions of attractive and unattractive individuals. Similarity and Attractiveness 16 Does attractiveness similarity mediate the influence of target attractiveness on interpersonal evaluation? To investigate the possibility that attractiveness similarity explains the influence of target PA on evaluation, I investigated the potential mediation of the target PAevaluation link by the similarity index. Using a series of regression analyses, I tested the four mediation steps outlined by Baron and Kenney (1986): the IV must predict (a) the DV (Step 1) and (b) the potential mediator (Step 2), the potential mediator must predict the DV (Step 3), and when both the IV and the potential mediator are in the model, the mediator must predict the DV significantly while the influence of the IV must be significantly reduced (Step 4). In order to test for indirect effects of the IV on the DV via the mediator, it was necessary to calculate a single estimate for the influence of target PA (the IV) on the similarity index (the potential mediator) and on each outcome measure (the DVs). Because target PA was a threelevel categorical variable, it could not be represented by a single variable in a regression analysis. Therefore, I created two orthogonal contrasts to represent the threelevel factor and focused my analysis on the contrast comparing evaluations of high and low PA targets. For each analyses described below, the contrast comparing high PA vs. low PA target was the IV and the similarity index was the potential mediator. Importantly, Ps regarded high PA targets as more similar to themselves on PA than they did low PA targets, B = -.41, t = -2.77, p < .01, effectively fulfilling a mediational criterion (Step 2) for each model. Analyses specific to each outcome measure are presented below. Similarity and Attractiveness 17 Sociability. Ps rated high PA targets as significantly more sociable than low PA targets, B = .88, t = 8.15, p < .001 (Step 1), and the similarity index predicted sociability ratings significantly, B = -.28, t = -4.05, p < .001 (Step 3): the more similar to the self a target was rated on PA, the more sociable that target was perceived to be. When including the similarity index in the full model (including target gender, participant gender, and target PA contrasts), both the high vs. low PA contrast, B = .81, t = 7.48, p < .001, and the similarity index, B = -.15, t = -2.43, p < .02, predicted sociability ratings significantly. The indirect effect was marginal, z = 1.77, p < .08 (Step 4). Though attractiveness similarity did not mediate the target PA effect, it did play a unique role in perceptions of sociability. Integrity. Ps rated low PA targets as having significantly more integrity than high PA targets, B = -.25, t = -2.36, p < .02 (Step 1); however, the similarity index did not predict integrity ratings significantly, B = -.09, t = -1.41, p > .15 (Step 3). This latter result effectively derailed the mediation under investigation; however, I proceeded with the complete analysis for exploratory purposes. Not surprisingly, when the similarity index was included in the full model, the high vs. low PA contrast was significant, B = .30, t = -2.64, p < .01, but the similarity index was not, B = -.05, t = -.70, p > .45. The indirect effect was not significant, z = 1.20, p > .20 (Step 4). Attractiveness similarity did not mediate the influence of target PA on integrity ratings. Liking. Finally, Ps liked high PA targets more than low PA targets, B = .36, t = 2.21, p < .03 (Step 1), and the similarity index predicted liking significantly, B = -.41, t = -4.44, p < .001 (Step 3): the more similar to the self a target was rated on PA, the more the participant liked the target. When including the similarity index in the full model, the Similarity and Attractiveness 18 high vs. low PA contrast was rendered non-significant, B = .18, t = 1.19, p > .20, the similarity index was significant, B = -.41, t = -4.70, p < .001, and the indirect effect was significant, z = 2.35, p < .02 (Step 4). Attractiveness similarity completely mediated the influence of target PA on liking. Discussion In an effort to investigate the role of self-perception in the evaluation of targets who varied in physical attractiveness, participants completed self-assessments and expressed liking for and made personality inferences about male or female targets who were high, medium, or low in attractiveness. Consistently with expectations, participants regarded themselves as more attractive than the average college student and as more similar on attractiveness to attractive, rather than unattractive, targets. These results are consistent with the self-promotion literature (e.g., Alicke, 1985) and confirmed that PA similarity is a potential confound in investigations of the link between PA and attraction and/or personality inference. As expected, the impact of target attractiveness on target evaluation was profound; however, the influence of the self was also unmistakable. The physical attractiveness stereotype (i.e., inferences regarding the sociability and integrity of the target) was contingent upon self-perception. Participants who rated themselves highly on attractiveness regarded attractive targets as more sociable (marginally so) and as having more integrity than did those participants who regarded themselves as less attractive. On the other hand, participants who rated themselves as relatively unattractive were more positively disposed to less attractive targets. Similarity and Attractiveness 19 Interestingly, self-perceptions had a particularly profound effect on liking for the targets. Self-rated attractiveness moderated the impact of target attractiveness in the expected manner: attractive participants liked attractive targets the most, whereas less attractive participants favored less attractive targets. Also, attractiveness similarity completely mediated the influence of target attractiveness on liking. That is, there was not a significant effect of target attractiveness on liking once I accounted for attractiveness similarity. Increased affection for attractive people was a function of similarity, and not of beauty. This result is particularly striking given the critical role that physical attractiveness is presumed to play in interpersonal attraction. Interpretation There are at least two perspectives that offer insight into the role of selfperception in judgments of attractive and unattractive individuals. Speaking first about inferences of sociability and integrity, previous discussions of the physical attractiveness stereotype suggest that the stereotype originates in implicit personality theories (Ashmore, 1981) that are formed via (1) direct observation of attractive and unattractive individuals and (2) cultural representations of attractive and unattractive people (Eagly, et. al., 1991). Following from this perspective, the moderation of personality inferences about attractive targets by selfrated attractiveness likely reflects different implicit theories that attractive and unattractive individuals have formed via observation and experience. Though attractive and unattractive people are likely exposed to similar cultural influences, attractive individuals may have had more positive interactions with physically attractive people and may have observed these people in a more positive light than have unattractive participants (possibly as a result of attractiveness similarity itself), Similarity and Attractiveness 20 thus creating more positive implicit personality theories. Indeed, the influence of selfperception on personality inference seems to be embedded in welllearned cognitive structures, and these cognitive structures may develop differently as a function of one’s attractiveness and possibly as a function of egoistic motives that influence one’s perceptions of attractive and unattractive individuals. Such a possibility is interesting fodder for future research. While implicit personality theories do well to explain the role of the self in personality inference, it is rather unlikely that such cognitive structures can account adequately for the role of the self in expressions of liking for others. After all, judgments of liking reflect how one feels about someone else, are, thus, affective in nature, and may be relatively independent of objective assessment of a target’s characteristics (See Byrne, 1971 for a discussion of the affective nature of attraction). One interpretation of the data regarding liking judgments is that implicit, automatic, and positive feelings about the self (implicit egotism, see Pelham, Mirenberg, & Jones, 2002) influence attraction via affective, effortless responses. When evaluating a target who is similar to the self, implicit egotism translates into a positive emotional response to the similar target (i.e., I feel good about myself, and thus, I feel good about those like me). The positive emotional response, in turn, drives affection for the target (see Clore, & Tamir, 2002; Martin, Abend, Sedikides, & Green, 1997 for reviews regarding the impact of affect on judgment). In fact, the transfer of positive feelings regarding the self to a similar target may also combine with the implicit theories discussed previously to contribute to inferences about sociability and integrity. Similarity and Attractiveness 21 It is important to note that the potential influence of implicit egotism depends upon the assumption that most of the participants in the current project had positive automatic associations with the self. Though I present no data to support this contention, previous work suggests that this assumption is valid (see Pelham, Mirenberg, & Jones, 2002; Schwartz, 1986, Taylor, & Brown, 1988). Future research would do well to investigate the impact of attractiveness similarity in those individuals who have less positive, or even negative, automatic self-associations. If implicit egotism accounts for the influence of attractiveness similarity, such an investigation should yield findings that differ markedly, or are even contradictory, to the findings reported here, with such participants not manifesting a bias for similar others. Further empirical inquiry Beyond questions of the mechanism driving the current results, there are a number of other empirical queries to which this project leads. First, is the influence of attractiveness similarity unique? Previous research has confirmed that personality similarity is associated with interpersonal attraction (e.g., Byrne, Griffitt, & Stefaniak, 1967); however, few investigations have considered the role of similarity on individual personality traits. We are left to wonder whether information regarding a target individual’s sociability or integrity might create perceptual similarity that would confer onto the target the same benefits as attractiveness similarity. Second, the moderation of the target attractiveness effect by self-rated attractiveness raises the question of a dispositional characteristic that might differentiate between those who rate themselves relatively favorably on attractiveness and those who do not. Dispositional narcissism (Morf, & Rhodewalt, 2002; Raskin & Hall, 1979) may Similarity and Attractiveness 22 be just such a characteristic. Narcissists, to a greater degree than non-narcissists’, evaluate themselves as more attractive (1) than the average person and (2) than is warranted by objective criteria (Gabriel, Critelli, & Ee, 1994). Narcissists also tend to base interpersonal judgments on physical attractiveness (Campbell, 1999). Accordingly, dispositional narcissism may by a critical factor that underlies self-perceptions (of attractiveness as well as other characteristics), which, in turn, impact target judgment. Conclusions This project revisits the physical attractiveness stereotype and the attractivenessattraction link and provides evidence for the critical role of self-perceptions in affection for and personality inferences about attractive and unattractive individuals. The data indicate that the physical attractiveness stereotype is robust, but that the specific content of the stereotype varies as a function of self-perception. Additionally, relative affection for attractive individuals, a finding that has been documented extensively by previous researchers (Byrne, London, & Reeves, 1968; Snyder, Tanke, & Berscheid, 1977), was a function of similarity rather than target attractiveness per se. While I would not argue that attractiveness plays no role in attraction (possibly a potent role in romantic attraction rather than “liking”), these data do suggest that attractiveness similarity is an important variable to consider when investigating attraction as a function of target physical attractiveness. More generally, the project contributes to our understanding of the way that selfperceptions influence target evaluation. The project leaves open questions regarding dispositional and motivational mechanisms that might drive these results; however, I hope that these data inspire further inquiries into the influence of similarity, the role of Similarity and Attractiveness 23 implicit egotism, and the dynamic interplay of dispositional, self-evaluative, and social factors in interpersonal judgment. As such investigations proceed, researchers interested in the impact of attractiveness on judgment and behavior might practice a new idiom: What is beautiful may be good, but what is beautiful like me is even better. Similarity and Attractiveness 24 ReferencesAlicke, M. D. (1985). Global self-evaluation as determined by the desirability andcontrollability of trait adjectives. Journal of Personality and Social Psychology,49, 1621-1630.Ashmore, R. D. (1981). Sex stereotypes and implicit personality theory. In D. L.Hamilton (Ed.), Cognitive processes in stereotyping and intergroup behavior (pp.37-81). Hillsdale, NJ: Erlbaum.Baron, R. M., & Kenney, D. A. (1986). The moderator-mediator variable distinction insocial psychological research: Conceptual, strategic, and statistical considerations.Journal of Personality and Social Psychology, 51, 110-118.Bassili, J. N. (1981). The attractiveness stereotype: Goodness or glamour? Basic andApplied Social Psychology, 2, 235-252.Berscheid, , E., Dion, K., Walster, E., & Walster, G. W. (1971). Physical attractivenessand dating choice. A test of the matching hypothesis. Journal of ExperimentalSocial Psychology, 7, 173-189.Brown, J. D. (1986). Evaluations of self and others: Self-enhancement biases in socialjudgments. Social Cognition, 4, 353-376.Byrne, D. (1971). The attraction paradigm. New York: Academic Press.Byrne, D., Griffitt, W., & Stefaniak, D. (1967). Attraction and similarity of personalitycharacteristics. Journal of Personality and Social Psychology, 5, 82-90.Byrne, D., & Nelson, D. (1965). Attraction as a linear function of positive reinforcement.Journal of Personality and Social Psychology, 1, 659-663.Byrne, D., London, O., & Reeves, K. (1968). The effects of physical attractiveness, sex, Similarity and Attractiveness 25 and attitude similarity on interpersonal attraction. Journal of Personality, 36, 259-271.Campbell, W. K. (1999). Narcissism and romantic attraction. Journal of Personality andSocial Psychology. 77, 1254-1270.Cash, T. F., & Derlega, V. J. (1978). The matching hypothesis: physical attractivenessamong same-sexed friends. Personality and Social Psychology Bulletin, 4, 240-243.Cash, T. F., Kehr, J., Polyson, J., & Freeman, V. (1977). Role of physical attractivenessin peer attribution of psychological disturbance. Journal of Consulting andClinical Psychology, 45, 987-993.Chaikin, A. L., Gillen, B., Derlega, V. J., Heinen, J. R. K., & Wilson, M. (1978).Students’ reactions to teachers’ physical attractiveness and nonverbal behavior:Two explana tory studies. Psychology in the Schools, 15, 588-595.Clore, G. L., & Tamir, M. (2002). Affect as embodied information. PsychologicalInquiry, 13, 37-45.Dermer, M., & Thiel, D. L. (1975). When beauty may fail. Journal of Personality andSocial Psychology, 31, 1168-1176.Dion, K., Berscheid, E., & Walster, E. (1972). What is beautiful is good. Journal ofPersonality and Social Psychology, 24, 285-290.Duck, S. W. (1973). Personality similarity and friendship choice: Similarity of what,when? Journal of Personality, 41, 543-558.Duck, S. W. & Craig, G. (1978). Personality similarity and the development of Similarity and Attractiveness 26 friendship: A longitudinal study. British Journal of Social and ClinicalPsychology, 17, 237-242.Dushenko, T. W., Perry, R. P., Schilling, J., & Smolarski, S. (1978). Generality of thephysical attractiveness stereotype for age and sex. Journal of Social Psychology,105, 303-304.Eagly, A. H., Makhijani, M. G., Ashmore, R. D., & Longo, L. C. (1991). What isbeautiful is good, but...: A meta-analytic review of research on the physicalattractiveness stereotype. Psychological Bulletin, 110, 109-128.Epley, N., & Dunning, D. (2000). Feeling “holier than thou”: Are self-servingassessments produced by errors in selfor social prediction? Journal ofPersonality and Social Psychology, 79, 861-875.Feingold, A. (1992). Goodlooking people are not what we think. Psychological Bulletin,111, 304-341.Feingold, A. (1988). Matching for attractiveness in romantic partners and same-sexfriends: A meta-analysis and theoretical critique. Psychological Bulletin, 104,226-235.Fong, G. T., & Markus, H. (1982). Self-schemas and judgments about others. SocialCognition, 1, 191-204.Gabriel, M. T., Critelli, J. W., & Ee, J. S. (1994). Narcissistic illusions in self-evaluationsof intelligence and attractiveness. Journal of Personality, 62, 143-155.Green, J. D., & Sedikides, C. (2001). When do self-schemas shape social perception?:The role of descriptive ambiguity. Motivation and Emotion, 25, 67-83.Johnson, D. F., & Pittenger, J. B. (1984). Attribution, the attractiveness stereotype, and Similarity and Attractiveness 27 the elderly. Develomental Psychology, 20, 1168-1172.Kulka, R. A., & Kessler, J. B. (1978). Is justice really blind? The influence of litigantphysical attractiveness or juridical judgment. Journal of Applied SocialPsychology, 8, 366-381.Landy, D., & Sigall, H. (1974). Beauty as talent – Task evaluation as a function of theperformer’s physical attractiveness. Journal of Personality and SocialPsychology, 29, 299-304.Larwood, L., & Whittaker, W. (1978). Managerial myopia: Self-serving biases inorganizational planning. Journal of Applied Psychology, 62, 194-198.Lewicki, P. (1983). Selfimage bias in person perception. Journal of Personality andSocial Psychology, 45, 384-393.Li, N. P., Bailey, M., Kenrick, D. T., & Linsenmeier, J. A. W. (2002). The necessititesand luxuries of mate preferences: Testing the tradeoffs. Journal of Personalityand Social Psychology, 82(6), 947-955.Martin, L. L., Abend, T., Sedikides, C., & Green, J. D. (1997). How would you feel if...?Mood as input to a role fulfillment evaluation process. Journal of Personality andSocial Psychology, 73, 242-253.Miller, D. T., & Ross, M. (1975). Self-serving biases in the attribution of causality: Factor fiction? Psychological Bulletin, 82, 213-225.Morf, C. C., & Rhodewalt, F. (2001). Unraveling the paradoxes of narcissism: A dynamicself-regulatory processing model. Psychological Inquiry, 12(4), 177-196.Nisbett, R. E., & Wilson, T. D. (1977). The halo effect: Evidence for unconscious Similarity and Attractiveness 28 alteration of judgments. Journal of Personality and Social Psychology, 35(4),250-256.Parekh, H., & Kanekar, S. (1994). The physical attractiveness stereotype in a consumer-related situation. Journal of Social Psychology, 134, 297-300.Pelham, B. W., Mirenberg, M. C., & Jones, J. T. (2002). Why Susie sells seashells by theseashore: Implicit egotism and major life decisions. Journal of Personality andSocial Psychology, 82(4), 469-487.Pelham, B. W., Ly, T. M., & Hetts, J. J. (2002). Judging an open book by its cover: Thescope and limits of the physical attractiveness stereotype. Unpublishedmanuscript.Raskin, R., & Hall, C. S. (1979). A narcissistic personality inventory.Psychological Reports, 45, 590.Rosenblatt, A. & Greenberg, J. (1991). Examining the world of thedepressed: Do depressed people prefer others who are depressed? Journal ofPersonality and Social Psychology, 60, 620-629.Ruble, D. N., Eisenberg, R., & Higgins, T. (1994). Developmental changes inachievement evaluation: Motivational implications of self-other differences. ChildDevelopment, 65, 1095-1110.Schwartz, R. M. (1996). The internal dialogue: On the asymmetry between positive andnegative coping thoughts. Cognitive Therapy and Research, 10, 591-605.Sedikides C., & Green, J. D. (2000). On the self-protective nature of Similarity and Attractiveness 29 inconsistency/negativity management: Using the person memory paradigm toexamine self-referent memory. Journal of Personality and Social Psychology, 79,906-922.Sedikides, C. & Skowronski, J. J. (1993). The self in impression formation: Traitcentrality and social perception. Journal of Experimental Social Psychology, 29,347-357.Sigall, H., & Landy, D. (1973). Radiating beauty: Effects of having a physicallyattractive partner on person perception. Journal of Personality and SocialPsychology, 28, 218-224.Snyder, M., Tanke, E. D., & Berscheid, E. (1977). Social perception and interpersonalbehavior: On the selffulfilling nature of stereotypes. Journal of Personality andSocial Psychology, 35, 656-666.Soloman, M. R., & Schopler, J. (1978). The relationship of physical attractiveness andpunitiveness: Is the linearity assumption out of line? Personality and SocialPsychology Bulletin, 4, 483-486.Sprecher, S., & Duck, S. (1994). Sweet talk: The importance of perceived communicationfor romantic and friendship attraction experienced during a get-acquainted date.Personality and Social Psychology Bulletin, 20, 391-400.Stroebe, W., Insko, C. A., Thompson, V. D., & Layton, B. D. (1970). Effects of physicalattractiveness, attitude similarity, and sex on various aspects of interpersonalattraction. Journal of Personality and Social Psychology, 18, 79-91.Svenson, O. (1981). Are we all less risky and more skilled than our fellow drivers? ActaPsychologica, 47, 143-148. Similarity and Attractiveness 30 Taylor, S. E., & Brown, J. D. (1988). Illusion and well-being: A social psychologicalperspective on mental health. Psychological Bulletin, 103, 193-210.Timmerman, K., & Hewitt, J. (1980). Examining the halo effect of physicalattractiveness. Perceptual and Motor Skills, 51(2), 607-612.Walster, E., Aronson, V., Abrahams, D., & Rottman, L. (1966). Importance of physicalattractiveness in dating behavior. Journal of Personality and Social Psychology,4, 508-516.Wetzel, C. G., Wilson, T. D., & Kort, J. (1981). The halo effect revisited: Forewarned isnot forearmed. Journal of Experimental Social Psychology, 17(4), 427-439.Wetzel, C. G., & Insko, C. A. (1982). The similarity-attraction relationship: Is there anideal one? Journal of Experimental Social Psychology, 18, 253-276.White, G. L. (1980). Physical attractiveness and courtship progress. Journal ofPersonality and Social Psychology, 39, 660-668. Similarity and Attractiveness 31 Author NoteRobert S. Horton, Assistant Professor of Psychology, Wabash College. Theresearch described was conducted while the author was a graduate student at theUniversity of North Carolina at Chapel Hill.I would like to thank Dr. Vaida Thompson for advising me during thedevelopment of this project and Dr. Thompson, Dr. Jeff Green, Matt Montoya, Dr. KeithCampbell, Dr. Brett Pelham, and three reviewers for their valuable suggestions on earlydrafts of this manuscript.Correspondence concerning this article should be addressed to Robert S. Horton,P.O. Box 352, Wabash College, Crawfordsville, IN 47933. Electronic mail may be sentto [email protected]. Similarity and Attractiveness 32 Footnotes1 It is important to note that I focused on selfratings of PA rather than objectiveratings of evaluator PA. Objective ratings of evaluator PA tend to exert limited influenceon target evaluations (e.g., Dermer & Thiel, 1975), and more importantly, selfrated PAis relatively more predictive of important psychological factors than are objective ratingsof PA (Feingold, 1992). The unique phenomenological experience of one’s ownattractiveness is an appropriate self-assessment when investigating the role of the self intarget evaluation.2 The astute reader will notice that the high PA male targets were rated asrelatively low on attractiveness, as compared to the high PA female targets. However, formy purposes, it was most critical that high PA targets be regarded as more attractive thanmedium PA targets (and, accordingly, low PA targe ts), within each gender. The chosenphotographs fulfilled this objective. The pictures chosen to represent high PA men wereregarded as more attractive than those chosen to represent medium PA men, (pairwisecomparison p’s ranging from .003 to .028).3 Previous researchers (e.g., Wetzel, & Insko, 1982) have suggested that idealself-similarity, rather than similarity to the actual self, is the most potent predictor ofattraction. I analyzed the current data with and without controlling for ideal self-similarity. Importantly, statistical conclusions regarding the role of target PA and themediating role of evaluator-target PA similarity were the same in both sets of analyses.Thus, ideal self-similarity is not discussed further.4 I used the similarity index, and not the PA similarity item, as the dependentmeasure and potential mediator because I believed the similarity index to be free of an Similarity and Attractiveness 33 important source of online bias. Specifically, previous research suggests that individualswill rate themselves (on single-item measures) as more similar to attractive, rather thanunattractive, targets. This result may be due to inflated self-beliefs, but it may also be dueto momentary fits of the self-enhancement need (i.e., “I bolster myself by rating myselfas similar to this attractive individual.”). For exploratory purposes, I conducted eachanalysis involving attractiveness similarity with the PA similarity item, rather than thesimilarity index. The results of these exploratory analyses were exactly similar to thosereported in the text (results available from the author).5 The three-factor analysis revealed a number of target gender and participantgender effects. These effects are less critical to the goals of this project and thus, are notdiscussed in detail. In general, participants were more favorable towards opposite, ratherthan same, gender targets. Attractiveness mattered more for female, rather than male,targets, and target attractiveness was more influential when participants evaluatedopposite, rather than same, gender targets. These differences are unique in a literature inwhich significant target or participant gender effects have been rare (e.g., Dion,Berscheid, & Walster, 1972). More specific data regarding target gender and participantgender effects are available from the author.6 Given the fact that 1 standard deviation below the self-rated PA mean translatedinto an absolute self-rating of 5.65 (on a 10 point scale), the fact that low PA participantsinflated evaluations of medium PA targets is consistent with expectations.7 Each regression included a main effect contrast for participant gender (coded 1, -1 for male and female, respectively), a main effect contrast for target gender (1, -1), twomain effect contrasts for target PA (high: 1, .5, medium: 0, -1, low: –1, .5), and contrasts Similarity and Attractiveness 34 representing the interactions of the three factors (e.g., the participant gender contrastcrossed with each of the target PA contrasts). Standard errors were computed usingGoodman’s (1960) method (in which the product of the standard errors of the two pathsare added to the denominator term).8 I investigated mediation in two additional ways. First, I used analysis ofcovariance in which target PA was a threelevel categorical predictor. This analysisyielded conclusions similar to those reported in the text and is available upon request.Additionally, I computed indirect effects using participants’ ratings of their target’s PA asa continuous predictor and participant and target gender (both coded 1, -1) as categoricalpredictors. Similarly to the analyses reported in the text, the indirect effect of target PAvia the similarity index on liking was significant, z = 2.42, p < .05; however, the indirecteffects on sociability and integrity were not significant, z = .88 and .93, respectively (ps >.05). Similarity and Attractiveness 35 Table 1.Target Evaluation as a Function of Target Physical Attractiveness TargetAttractiveness Outcome MeasureHighMediumLow Sociability7.34a(1.10)6.64b(.97)5.56c(1.05) Integrity6.36b(1.06)7.10a(1.18)6.89a(1.08) Liking6.68ab(1.61)6.96a(1.86)6.05b(1.46) All main effects reached significance. Means within each row that do not share asubscript differ significantly (p < .05). Standard deviations are given in parentheses. Similarity and Attractiveness 36 Table 2.Predicted Values for Hypothetical Individuals 1 SD Above and Below the Mean for Self-Rated Physical Attractiveness as a Function of Target Attractiveness TargetAttractiveness Outcome MeasureHighMediumLow SociabilityHigh Self-rated PALow Selfrated PA7.616.916.766.625.285.77IntegrityHigh Self-rated PALow Selfrated PA6.715.866.767.317.006.80LikingHigh Self-rated PALow Selfrated PA7.325.836.567.365.616.30 Similarity and Attractiveness 37