The Female Positivity Effect in the Perception of Others

The Female Positivity Effect in the Perception of Others

JOURNAL OF RESEARCH IN PERSONALITY ARTICLE NO. 32, 370–388 (1998) RP982221 The Female Positivity Effect in the Perception of Others Lynn A. Winquis...

96KB Sizes 0 Downloads 25 Views

JOURNAL OF RESEARCH IN PERSONALITY ARTICLE NO.

32, 370–388 (1998)

RP982221

The Female Positivity Effect in the Perception of Others Lynn A. Winquist, Cynthia D. Mohr, and David A. Kenny University of Connecticut We posit a female positivity effect in the perception of others. Using social-role theory, we argue that female perceivers will view others more positively than will male perceivers. In Study 1, a Social Relations Model analysis was undertaken on 10 data sets containing mixed-gender interpersonal perception ratings of the Big Five factors. Results revealed that there is a remarkably consistent, albeit modest, trend indicating that women assign more positive trait ratings to others than men for all of the Big Five factors at all levels of acquaintance. Study 2 replicated the gender difference, controlled for target gender, and explored possible moderators and mediators of this effect. Consistent with social-role theory, there is an indication that sex-role stereotypes mediate the effect.  1998 Academic Press

Researchers interested in gender roles have carefully examined the nature of gender role stereotypes and the extent to which they are manifested in male and female behavior (Eagly, 1987). Indeed, males and females behave in accordance with the social roles that they occupy; males typically occupy agentic roles, whereas females occupy communal roles. Consequently, males are more assertive and controlling, whereas females are more selfless, nurturing, and concerned with others. Given that gender roles impact behavior differentially for males and females, might gender roles differentially inform the manner in which males and females perceive others? Although extensive literature has documented the involvement of genderbased expectancies in social perception (e.g., Bem, 1974; Spence & Helmreich, 1978; Swim, 1994), the research concentrates almost exclusively on the manner in which gender-role stereotype information guides perceptions of others, based upon the gender of the target being perceived. Very little research has been conducted on the way in which gender roles and consequently gender-based expectancies may differentially guide males’ and females’ perceptions of others. Moreover, evidence from person–perception studies reveals that characteristics of the perceiver are at least as important This research was supported in part by grants from the National Science Foundation (DBS9307949) and the National Institute of Mental Health (R01-MH51964). We thank Linda Albright for her helpful comments on an earlier draft. 370 0092-6566/98 $25.00

Copyright  1998 by Academic Press All rights of reproduction in any form reserved.

THE FEMALE POSITIVITY EFFECT

371

as characteristics of the target (Kenny, 1994). It logically follows, then, that if females are more other-oriented and nurturant, whereas males are more dominant and forceful by virtue of the social roles they occupy, then males and females might perceive others differently. The purpose of the present investigation is to document the extent to which women and men perceive others differently. In particular, given the nature of females’ communal role, we argue that females will perceive others more favorably, a phenomenon we dub the female positivity effect. Furthermore, we use the methods and conceptual framework of the Social Relations Model (Kenny & La Voie, 1984), an interpersonal and interactive research paradigm, to obtain unbiased estimates of the relative influence of the perceivers’ characteristics and expectancies upon social judgment. SOCIAL-ROLE THEORY In her program of research, Eagly documents the fundamental differences in the way men and women behave (Eagly & Carli, 1981; Eagly & Crowley, 1986; Eagly & Steffen, 1986). Consistent with gender-role expectations, females were less likely to be aggressive, physically more so than psychologically (Eagly & Steffen, 1986). According to a review by Carli (1982), women are more likely to be friendly and agreeable during a small group task than males, who are primarily task-oriented. Further, women have been found to be more nonverbally expressive, more involved (i.e., more nodding and forward leaning) and better nonverbal decoders (Hall, 1984). Eagly (1987) interpreted Hall’s findings in light of a social-role theory framework, whereby ‘‘the female gender role fosters communal qualities and thereby encourages women to be pleasant, interpersonally oriented, expressive and socially sensitive’’ (p. 106). Eagly maintains that females’ special ability to decode nonverbal cues enhances their social skills, especially their social sensitivity. After documenting the sizable differences in male and female behavior, Eagly (1987) proposed a model to account for such behavior. She argued for the importance of considering the differential distribution of males and females in the division of labor. According to Eagly, it is the differential distribution that leads to stereotypic sex differences in behavior. The specific role occupied by males or females offers a set of gender-role expectancies, which frequently manifest themselves in overt behavior. Eagly consistently demonstrated the strong relationship between gender expectations and behavior exhibited by men and women; men and women typically demonstrate the behavior that is expected of them by virtue of their sex. The social roles men and women occupy also offer a set of sex-typed skills and beliefs that contribute to social role behavior. Thus, men and women behave differently due to the different social roles

372

WINQUIST, MOHR, AND KENNY

they occupy, which dictate different social role-based expectancies and provide relevant skills and beliefs. According to Eagly (1987), the behavior men and women display can be generally described as agentic (dominant, aggressive, and task-oriented) or communal (socially oriented, sensitive, warm, and caring), respectively. Because social roles are involved in the differential behavior of men and women, partly because of gender role expectancies, we argue that social roles, and consequently gender-role expectancies, should also lead to differences in perception. As mentioned earlier, researchers have documented the influence of gender stereotypes on differential perception of men and women (e.g., Bem, 1974; Spence & Helmreich, 1978; Swim, 1994). However, person perception researchers have not typically considered the role of the perceiver’s gender in the perception of others. That is, how might men and women perceive others differently, due to their social role occupation and accompanying gender-role expectancies? To better understand the distinction we make between perceptions of others based upon the target’s gender or, alternatively, perceptions based upon the perceiver’s gender, it is appropriate to turn to the Social Relations Model. We shall see that this model makes this distinction explicit. Social Relations Model The Social Relations Model (SRM) is a general approach for the study of dyadic phenomena. It has been most often applied to answer questions pertaining to interpersonal perception (Kenny, 1994; Kenny & La Voie, 1984). The SRM specifies three components of social perception, which are referred to as target, perceiver, and relationship and are defined as follows: Target: Is the target seen the same way by different perceivers? Perceiver: Does the perceiver see targets in the same way? Relationship: Does the perceiver view the target in a unique fashion?

To illustrate, imagine three perceivers, Sue, Mike, and Bill: Target: Do Mike and Bill see Sue in the same way? Perceiver: Does Sue see Mike and Bill in the same way? Relationship: Does Sue view Mike in a unique fashion?

Most previous interpersonal perception research has focused on the target effect. The reason for this focus is that the target effect provides insight into the most basic interpersonal perception questions regarding consensus, accuracy, and self–other agreement (Kenny, 1994). The focus of the present investigation is the perceiver effect, that is, the degree to which a perceiver generally views targets as strongly or weakly possessing a trait. The perceiver effect represents the typical evaluation that a rater assigns to targets on the trait dimension. The SRM provides estimates

THE FEMALE POSITIVITY EFFECT

373

of the proportion of variance attributable to the perceiver, and research consistently shows evidence of perceiver variance in the ratings of targets (Kenny, 1994). To a large extent, Cronbach’s (1955) critique of accuracy research centered on the fact that social perception researchers were ignoring the perceiver effect, or what he referred to as differential elevation. The presence of perceiver variance implies assimilation (Kenny, 1994). Assimilation concerns the extent to which a single perceiver views two targets as more similar than two different perceivers’ views of the same two targets. For example, Sue sees Mike as intelligent and gives Mike an intelligence rating of 6 on a 7-point scale. The issue of interest is whether or not Sue also assigns Bob, Steve, and others ratings near 6 on intelligence. For illustrative purposes, imagine that Sue gives nearly all targets a rating of 6, and another judge, Bill, rates the same targets, and he gives them all a rating of 4 on the same measurement scale. In this hypothetical situation, assimilation would be exceptionally high. In a systematic review of research on assimilation effects in person perception, Kenny (1994) has shown that perceiver variance is regularly present in personality ratings. He considered three explanations of the differences exhibited in perceivers’ judgments of others: response set, generalized expectations about persons, and expectations about the particular group. The response set explanation approaches assimilation as a methodological artifact. Assimilation is seen as the perceiver’s tendency to use large or small numbers [akin to Cronbach’s (1955) differential elevation]. In Kenny’s (1994) review of the evidence, the response set explanation does not appear to be adequate. Alternatively, Kenny (1994) argued that assimilation should not be viewed as a measurement artifact, but rather as a reflection of the perceiver’s expectations of what others are generally like. Thus, the perceiver effect captures a generalized expectation of the other, rather than simply reflecting how individuals systematically assign numbers. According to Eagly’s theory (1987), gender roles offer one such set of expectancies that influence behavior and potentially perceptions of others. If these expectancies influence the manner in which perceivers view others, one would expect gender to be correlated with the perceiver effect. Using the work of Eagly and colleagues (1987), we propose that female perceivers will view others more positively and will assign to others correspondingly more favorable trait ratings than male perceivers. A review of the few relevant perception studies, in fact, bolsters this hypothesis. Gender of Perceiver There exist a very limited number of papers specifically examining differences in the evaluations that women and men assign to others. Furthermore,

374

WINQUIST, MOHR, AND KENNY

often the dependent measures do not specifically involve personality traits, per se, but rather evaluations of ability, affect, or overall favorability. Jones and colleagues (Jones, Rock, Shaver, Goethals, & Ward, 1968) asked 140 male and female college students to make predictions of a stimulus person’s performance on a problem-solving task. These researchers found that women predicted a higher level of performance by the stimulus person than did men. McDonald (1968) concluded from the data obtained from an interpersonal checklist completed by 528 men and women that women express and experience more love toward others than do men. This study supports the general notion that women feel more positively toward others than do males. Further, in a study examining violations of gender stereotype-based expectancies, Bettencourt, Dill, Greathouse, Charlton, and Mulholland (1997) discovered a main effect of participant gender. Specifically, females rated others more positively on a global favorability measure than did males. Finally, Carli and colleagues have uncovered a similar main effect of participant gender in their studies in which they evaluate the impact of various nonverbal styles and gender on social influence. For example, Carli, LaFleur, and Loeber (1995) reported that, overall, females liked the speakers more than males did. These studies, though consistent, only vaguely suggest that women would rate others more favorably than do men. A review of studies is needed that examines perceiver gender differences across a wide range of personality variables. PRESENT STUDY In this paper we investigate whether differences exist in the way in which men and women view others’ personalities. Based upon the social-role theory, in conjunction with the past findings, we predict that females will view others more positively than males and assign more positive trait ratings. This study serves the secondary purpose of exploring possible explanations for the consistent perceiver effects typically found in person–perception research. To begin to understand better the meaning of the perceiver effect, we need to examine the correlates of the perceiver effect with measures of individual difference. One such potential correlate is gender and will be the focus of the current study. The present investigation also takes into consideration that gender–perceiver relationships may be trait specific. Women may be more lenient judges than men for some traits, but the reverse may be true for other traits. Variables from the Big Five factors (Goldberg, 1990; Extroversion, Agreeableness, Conscientiousness, Emotional Stability, and Culture) were selected for inclusion in the review to classify the results systematically. Moreover, the

THE FEMALE POSITIVITY EFFECT

375

degree of acquaintance between the perceivers and the targets was examined as a potential moderating variable. Study 1 Overview The goal of Study 1 is to explore whether gender differences exist in perceivers’ trait judgments of others. Additionally, if gender differences do exist, we examine whether they vary as a function of varying levels of acquaintance. To address these questions, a search for relevant empirical research studies was conducted. All of the studies selected for review used the same design with participants all engaging in some level of interaction and serving as both perceivers and targets. The data were analyzed using the SRM. Method Selection of Studies. Studies were selected for this review on the basis of several criteria. First, the gender composition had to be mixed; that is, men and women judges had to provide trait ratings for both men and women targets. Second, the data collection design had to permit an SRM analysis in order to obtain unbiased estimates of perceiver variance. All of the studies included in this review had a round-robin design; that is, everyone rated everyone else. Last, the traits measured had to be indicators of the Big Five factors (Goldberg, 1990). Thus, seven studies, two containing three waves of data, resulting in a total of 10 data sets are included in this review. The Appendix presents information about these studies. The possible influence of the varying level of acquaintance of participants is also considered. Zero-acquaintance studies include participants that have never met. Judgments are based upon merely the observation of the target by the perceiver, not interpersonal interaction. Shortterm acquaintance refers to interactions ranging from minutes to hours. Finally, long-term acquaintance studies include participants that have had numerous interactions over an extended period of time (i.e., months, years). Statistical Analyses. An SRM analysis was undertaken for each study using specialized software called SOREMO (Kenny, 1993). The SRM analysis accomplishes the decomposition of the variance in social judgments to represent the three theoretical components: perceiver, target, and relationship. All computed variances are forecasts of what the variance would be if there were an infinite number of participants. To accomplish this, the variances are corrected for the finite number of participants. The correlation of interest is between the perceiver effect and gender. The gender–perceiver correlations for each factor were computed within each study. Gender is coded 0 for men and 1 for women and so a positive correlation means that women evaluate targets more favorably. If a study included two or more variables to represent each of the factors, the correlations for the individual variables were averaged to provide a single measure of the gender–perceiver relationship for the Big Five factors. To provide another frame of reference, the gender–perceiver correlations were transformed into mean differences by the formula: r PGs p /s G where P refers to perceiver effect and G refers to gender, dummy coded (1 ⫽ female, 0 ⫽ male). The mean difference, which is equivalent to a regression coefficient, equals the female mean minus the male mean. Each study used a seven-point scale for perceivers’ responses and responses were always oriented so that larger numbers indicated a more favorable response. So if this measure equaled .5, then that meant

376

WINQUIST, MOHR, AND KENNY

TABLE 1 Gender–Perceiver Correlations of Other-Perception Factors Study Zero acquaintance Albright, Kenny, & Malloy (1988), Study 1, Wave 1 Kenny, Horner, Kashy, & Chu (1992), Year 1, Study 3, Wave 1 Kenny (1990), unpublished data Short-term acquaintance Hallmark & Kenny (1989) Kenny, Horner, Kashy, & Chu (1992), Study 3, Wave 2 Levesque (1990) Malloy & Janowski (1992) Park & Judd (1989) Long-term acquaintance Albright, Kenny, & Malloy (1988), Study 1, Wave 3 Albright, Kenny, & Malloy (1988), Study 3, Wave 3 Means

Extr.

Agr.

Consc.

Em. Stab.

Cult.

Mean

.61

.35

.20

.52

.29

.39

.29

⫺.21

⫺.34

.28

⫺.10

⫺.02

.61

.19

.23

⫺.03

.11

.22

.19 .22

.14

.27 .12

.20 .44

.24

.22 .23

.09 .37 .14

.22 ⫺.12 .14

.14

.04

.13 .21 .12

.15 .23 .12

.11

.32

.24

.35

⫺.04

.20

.25

⫺.27

.24

.03

.15

.08

.29

.08

.14

.23

.12

.18

that females saw others more positively than males by one-half of a point on a seven-point scale. Because studies differ in the number of male and female perceivers, the power in the test of gender differences in the perceiver effect varies considerably across studies. We decided not to present significance test results, because we want to focus on the size and pattern of the results.

Results As noted earlier, the gender–perceiver correlations provide information regarding assimilation, the extent to which either women or men more positively evaluate targets on traits. The gender–perceiver correlations for each factor within the 10 data sets are presented in Table 1. A negative correlation means that men provided more favorable trait ratings, whereas a positive correlation signifies that women were more favorable in their judgments of the targets. Table 1 clearly shows that, in general, women are more lenient perceivers of others than men are. More specifically, although the mean correlations

377

THE FEMALE POSITIVITY EFFECT

TABLE 2 Gender–Perceiver Mean Differences Factors Study Zero acquaintance Albright, Kenny, & Malloy (1988), Study 1, Wave 1 Kenny, Horner, Kashy, & Chu (1992), Year 1, Study 3, Wave 1 Kenny (1990), unpublished data Short-term acquaintance Hallmark & Kenny (1989) Kenny, Horner, Kashy, & Chu (1992), Study 3, Wave 2 Levesque (1990) Malloy & Janowski (1992) Park & Judd (1989) Long-term acquaintance Albright, Kenny, & Malloy (1988), Study 1, Wave 3 Albright, Kenny, & Malloy (1988), Study 3, Wave 3 Means

Extr.

Agr.

Consc.

Em. Stab.

Cult.

Mean

.50

.43

.24

.37

.38

.38

.29

⫺.22

⫺.25

.26

⫺.09

.00

.54

.20

.22

⫺.03

.14

.21

.20 .24

.17

.36 .14

.26 .42

.31

.27 .26

.10 .34 .15

.41 ⫺.11 .19

.14

.05

.14 .21 .12

.22 .15 .13

.02

.30

.16

.37

⫺.03

.16

.33

⫺.50

.34

.06

.25

.10

.27

.10

.17

.22

.16

.19

are small, there is a clear consistent trend demonstrating that women provide more favorable ratings across the factors. Thirty-seven of the forty-four gender–perceiver correlations are positive. The mean correlations for the Big Five factors range from .29 for Extroversion to .08 for Agreeableness. To compare the perceiver–gender correlations across acquaintance levels, we computed the mean of the correlations for each study. In general, the correlations did not vary much across levels of acquaintance and are, therefore, not presented in Table 1. It is worth mentioning, however, that the largest mean correlation was clearly for Extroversion at zero acquaintance, .50, indicating that women rate others much more favorably on this dimension during initial interactions. The mean correlation collapsing across traits and studies was .20 at zero acquaintance, .19 for short-term acquaintance, and .14 for studies in which participants were well acquainted. Acquaintance does seem to lower, but not eliminate, the positivity effect. Table 2 presents the calculated gender–perceiver mean differences for the 10 data sets. As with the correlations, the sign of the mean difference indi-

378

WINQUIST, MOHR, AND KENNY

cates the gender: A positive sign indicates higher ratings by women, and a negative sign denotes higher ratings by men. The average mean difference on the Big Five factors ranged from .10 to .27 points on a 7-point scale. The largest mean difference was .27 for Extroversion, followed closely by Emotional Stability with a reported mean difference of .22. The calculated mean difference for Conscientiousness and Culture were approximately equivalent, .17 and .16, respectively. Finally, the smallest mean difference, .10, was for Agreeableness. In general, the calculated mean differences did not vary considerably by level of acquaintance with the exception of Extroversion at zero acquaintance with the highest average mean difference of .44. Discussion There is a remarkably consistent, although modest, trend revealing that women assign more positive trait ratings to others than do men; it is present for all of the Big Five factors at all levels of acquaintance. The average correlation is .18. The effect, however, does vary somewhat by trait type, with the largest correlations demonstrated for Extroversion and Emotional Stability. One potential explanation for this result, the one that we prefer, comes from Eagly’s (1987) social-role theory. Specifically, she discussed the importance of gender-based expectancies in determining social-role behavior, based upon the different social roles in which each gender occupies. Eagly’s theory implies that gender-based expectancies mediate the relationship between social-role categories and social-role behavior. According to this line of reasoning, it would be the expectancies of gender-appropriate behavior that accompany a particular gender role that largely determine behavior. In the current investigation, the gender-appropriate behavior would consist of female’s positive ratings of others. Perhaps, then, these positive ratings of others can be explained by gender-based expectancies of women. However, there are other interpretations of the result. One such explanation comes from research concerning gender effects on self-perception. Studies have shown that female children (Bledsoe, 1961, 1967) and adults (Sarason & Koenig, 1965) make more positive ratings to themselves than do males. For example, Amatora (1955) found that female children evaluated themselves more favorably than male children on many personality measures. In a sample of adult men and women, Goldrich (1967) concluded that women were more optimistic about their future self-evaluations, interpersonal relations, and professional life. However, men do rate themselves more positively on some dimensions. Maccoby and Jacklin (1974) pointed to a consistent trend in research literature illustrating that men rate themselves higher on self-dimensions relating to strength and potency (McDonald, 1968; Goss, 1968), whereas women

THE FEMALE POSITIVITY EFFECT

379

have more positive social self-concepts (Smart & Smart, 1970; Walker, 1967). Benton, Gelber, Kelley, and Liebling (1969) reported that men expressed a greater feeling of power than women did. They also showed that men rated other men as more powerful than women. Furthermore, women rate themselves as more anxious, extroverted, trusting, and nurturing than men and men rate themselves to be more assertive and dominant than women (Feingold, 1994). Taken together, these research studies demonstrate that women view themselves more positively than do men for socially oriented traits. The results of these studies are consistent with the social-role theory presented earlier, suggesting that women are more communal, whereas males are more agentic. Given the phenomenon of assumed similarity (Marks & Miller, 1988), that is, people see others as being similar to self, it can be argued that if women see themselves positively, they would see others positively. Given the assumed similarity explanation, it is appropriate to examine those Big Five factors that are socially oriented (i.e., Extroversion and Agreeableness), as those should be the dimensions on which women rate themselves higher. If assumed similarity is a viable explanation of the female positivity effect, the gender–perceiver correlations should be highest on the socially oriented dimensions and lowest for dimensions that are not socially oriented (e.g., Culture). In fact, the highest gender–perceiver correlation was for Extroversion, but the lowest gender–perceiver correlation was for Agreeableness. This discrepancy is problematic for the assumed similarity explanation of the female positivity effect. Nonetheless, we still think it is worthwhile to examine assumed similarity as a possible mediator. We have so far ignored gender of target. However, it is pertinent to consider whether females’ positive ratings equally apply to both male and female targets. Perhaps female positivity applies more to male or female targets. Thus, gender of partner should be considered as a potential moderator of the gender effect on perception. The data sets in Study 1 were not structured to address these issues. To expand our understanding of the relationship between gender and partner ratings, we conducted a second study. Study 2 The first goal of Study 2 was to replicate the findings of Study 1. Additionally, this study probed potential explanations of this finding, as well as establishing boundary conditions. We attempt to understand the influence of gender on person perception as an important insight to person–perception processes, as Kenny (1994) asserted, rather than simply a methodological artifact. Participants interacted with one man and one woman and provided trait ratings of each partner following each interaction. Participants interacted ei-

380

WINQUIST, MOHR, AND KENNY

ther face-to-face or through a computer network. The relative influence of communication media is not addressed in the current investigation. Based upon Eagly’s (1987) social-role theory, Study 2 purported to test the possible relevance of gender-based expectancies. For that reason, the extent to which participants endorsed traditional sex roles was also assessed. The relationship between the female social role and positive ratings of others should be explained by females’ gender-based expectancies. Thus, endorsement of traditional sex roles should mediate females’ positive ratings of others. One potential alternative explanation considered was the possible mediating effect of self-perception. As discussed earlier, women have demonstrated more favorable self-perceptions in past research studies. According to the phenomenon of assumed similarity, the positive partner ratings given by women, then, would be explained by their enhanced self-concepts. Thus, perceivers’ self-ratings were assessed to demonstrate whether they would account for the relationship between gender and positive partner ratings. We also considered the possibility that perhaps females rate one gender more positively rather than all interaction partners. To assess this possibility, Study 2 was designed to test whether gender of the partner moderates the female positivity effect in the perception of others. As predicted in Study 1, women were expected to rate their partners more positively. Based on social-role theory, we predict that this relationship should be explained by endorsement of gender-based expectancies. Alternatively, the relationship could be explained by women’s positive self-perceptions. Finally, we considered the possibility that women may only rate one gender more positively. Method Participants. Eighty-four undergraduate introductory psychology students (42 men and 42 women) from a northeastern university received course credit in exchange for their participation. Group members were queried as to the extent of their acquaintance, and in cases where they were acquainted, the group was dismissed. The resulting 84 participants formed 21 groups. Participants interacted one-on-one with one male and one female partner from their group in a block design (Kenny, 1994). Procedure. During testing at the start of the semester and before the onset of the study, most of the potential participants completed the endorsement of traditional sex-role measure. Participants who did not complete the scale prior to participation were administered that questionnaire at the end of the experiment after completing all of the self- and partner ratings. Participants arrived at the psychology building and were individually escorted to separate rooms to avoid contact with other group members before the appropriate time. Individuals within gender were randomly assigned a letter (i.e., person A, B, C, and D) which would represent the order and with whom they would interact. Persons A and B were the male participants, while persons C and D were the female participants. Participants were told that they would be interacting with two college students and at-

THE FEMALE POSITIVITY EFFECT

381

tempting to come to a consensus about a discussion topic. They completed a questionnaire regarding the discussion topic and were then paired for their first interaction. One pair of participants (persons A and B) was taken to the room for interaction depending on the condition to which their group was assigned. The other pair (persons C and D) remained in their individual rooms and completed self-reports including the Self-Monitoring Scale (Snyder & Gangestad, 1986). In the face-to-face condition, the interacting pair was taken to a room where partners were seated across from each other. In the computer-mediated condition, interacting participants were escorted separately to individual rooms in which two computers were connected. The networking program CHAT (Vigor, 1992) was used for the dialogue. Face-to-face participants were introduced to each other. In the computer condition, each participant was told his or her partner’s name. In case the person’s name might be ambiguous with regard to gender, the experimenter referred to the partner as he or she during the instructions. Thus, participants in the computer condition were made aware of their partners’ gender. For each interaction, the pair was asked to discuss and reach a consensus regarding a specific topic. In both conditions, conversations were recorded. After each interaction, they were taken to individual rooms to complete the ratings of their own and interaction partners’ behaviors. During this time, partners C and D now interacted. After all postinteraction ratings were completed, interaction partners were rotated such that persons A and C interacted, followed by persons B and D in the manner set forth for the first interactions. The order of same- and opposite-gender interactions was counterbalanced, and so in half the trials, partners A and C interacted first. After both interactions and ratings were completed, participants were debriefed and dismissed from the study. For more details concerning the procedure, the reader should consult Mohr and Kenny (1997). Measures. The MF Scale of the Personal Attributes Questionnaire (PAQ; Spence & Helmreich, 1978) was adapted to measure endorsement of traditional sex roles. The items of the MF Scale are bipolar, such that the one pole represents masculinity (lower scores), while the other represents femininity (higher scores). The items for the MF Scale were selected such that the attributes could be found in men or women, but would be more socially appropriate for one gender and not the other. Individuals rated the level of masculinity or femininity of both the typical male and female college students on a 7-point Likert scale. The participants completed the eight items for each gender, producing a MF score for women and one for men. After each interaction, participants were asked to rate the same six traits regarding themselves and their partners, each on a 7-point Likert scale. These traits were selected from the Bem Sex-Role Inventory (1974) and included the extent to which they were truthful, assertive, warm, helpful, and conscientious during the interaction, as well as whether they acted as a leader. Ratings across all six attributes were highly correlated and demonstrated high reliability, .83 for partner ratings and .72 for self-ratings. Likely, this strong, positive relationship is a reflection of the social desirability underlying these attributes (Bem, 1974). Thus, ratings were averaged across the six attributes to form a positivity measure for partner and selfperceptions. Higher scores on the positivity measure, then, reflect more socially desirable attributes. Statistical Analyses. Traditional sex-role endorsement was assessed using the MF scale scores of the typical male and typical female. Recall that higher scores indicate more feminine ascribed trait ratings, while lower scores indicate more masculine ascribed trait ratings. For each participant, the measure was derived by subtracting the MF male rating from the MF female rating. Hence, the larger and more positive a participant’s score is on this measure, the more he or she endorses traditional sex roles. The mean level of sex-role endorsement across participants was 10.415, meaning that most participants endorsed traditional sex roles. To center the scale at zero, the mean was subtracted

382

WINQUIST, MOHR, AND KENNY

from each score. Two participants failed to complete the items, rendering a valid score for 82 participants. The analyses to test whether or not sex-role endorsement explained the relationship between gender and trait ratings were conducted using the categorical predictor of gender and continuous predictor of sex-role endorsement. Although we hypothesized that sex-role stereotypes would be a mediator, the measure used in Study 2 to ascertain gender-based expectancies implies moderation, not mediation. (As Baron and Kenny (1986) point out, there are situations in which mediation can be tested by moderation.) To test appropriately for mediation, the gender-based expectancy measure would need to be bipolar, with one pole representing masculinity and the other representing femininity. We would then expect that those who saw themselves as more feminine would show positivity and those who saw themselves as more masculine would not. However, the measure that we used is a difference measure where one pole represents a large difference in the way men and women are rated, with men and women being rated in accordance with traditional sex roles. The other pole of our measure represented no difference between the ratings of the typical man or woman. We would then expect to find the female positivity effect for participants who have large scores on this measure. Thus, we were interested in testing the difference in partner ratings between men and women, expecting that it would be larger for those individuals who endorse traditional sex roles. Because this moderation effect involved an interaction between the categorical and continuous variables, we used the effect-coding procedures recommended by Aiken and West (1991) to estimate and interpret any significant effects. To interpret, but not to estimate, effects, comparison group values were established. Comparison groups for the categorical variables were assigned values of ⫺1 (men) and 1 (women). For the continuous variable of sex-role endorsement, it was necessary to estimate the dependent measures at three levels: strong sex-role endorsement, moderate sex-role endorsement, and low sex-role endorsement. The mean preadjustment level of sex-role endorsement (M ⫽ 10.415) was used as a basis for creating comparison groups. The predicted levels of the dependent measures were estimated for an individual who strongly endorses sex roles (M ⫽ 10.415), an individual who moderately endorses sex roles (M ⫽ 0), and an individual who does not endorse sex roles (M ⫽ ⫺10.415). Regression equations were used to estimate the predicted values for each level (strong, moderate, and low sex-role endorsement). The values presented under Results are based on this procedure and so represent predicted values. For all of the analyses presented, the effects of communication context are controlled for. Order of interaction partner (first versus second) was also tested and demonstrated a consistent and significant effect for both partner and self-ratings and so order of partner was controlled.

Results and Discussion First, analyses were conducted to estimate the influence of gender on partner trait ratings. The main effect of gender was significant [F(1, 73) ⫽ 10.57, p ⫽ .002, d ⫽ .56] as predicted. Consistent with Study 1, women rated their partners significantly more positively (M ⫽ 5.72) than did men (M ⫽ 5.28). In fact, in this more controlled study, we find a larger female positivity effect of about .44 points versus .19 in Study 1. We then considered whether the effect was moderated by partner gender. Following the procedures recommended by Baron and Kenny (1986), an analysis was conducted testing the interaction effect of partner gender and rater gender on partner ratings. The interaction, controlling for the main effects of partner and rater gender, did not achieve significance [F(1, 74) ⫽

THE FEMALE POSITIVITY EFFECT

383

.71, p ⫽ .403]. Thus, it is not the case that women are merely rating one gender more positively, but all of their partners. Next, an analysis was conducted to test the effect of sex-role endorsement on the relationship between gender and partner ratings. The main effects of sex-role endorsement and perceiver gender were also entered into the analysis. The expected interaction between sex-role endorsement and gender was demonstrated at F(1, 73) ⫽ 3.97, p ⫽ .050. The gender difference in partner ratings held only for those who had a traditional sex-role orientation. Among non-sex-role endorsing individuals, no gender difference emerged in partner ratings (M ⫽ 5.52 for women, M ⫽ 5.49 for men, d ⫽ .04). High sex-role endorsing women rated their partners more positively compared to high sexrole endorsing men (M ⫽ 5.91 for women, M ⫽ 5.06 for men, d ⫽ 1.09). Thus, the hypothesis that sex-role endorsement would mediate the gender effect on partner ratings was supported. The subsequent set of analyses was conducted to measure the possible mediating effect of self-ratings on the relationship between gender and partner ratings. Measuring mediation requires a test of three separate analyses in a specific sequence (Baron & Kenny, 1986). The first step is established: Women evaluate partners more favorably than do men. The second step to demonstrate is that gender significantly predicts the mediator or self-ratings. An analysis was conducted measuring the influence of gender on self-ratings. The effect of gender failed to achieve significance [F(1, 73) ⫽ 3.39, p ⫽ .070]. Therefore, the second step of the test for mediation failed, rendering further steps inappropriate. The hypothesis of mediation of self-perception on the relationship between gender and partner ratings was not supported because females did not see themselves any more positively than did males. GENERAL DISCUSSION The current investigation documents what seems to be a prevalent phenomenon: Women rate others more positively than do men. We found this result in a review of 10 data sets and in a carefully controlled study. Generally, our findings suggest that the source of differences between perceivers is not merely a reflection of response sets; rather, we demonstrate one of the potentially many psychologically meaningful processes underlying the variability in the perception of others. By rejecting the notion that this is a methodological artifact, we hope that person–perception researchers are encouraged to investigate further the nature of perceiver effects. Despite nearly a century of trait rating studies, this area has remained virtually unexplored. In fact, many researchers have perceivers rank-order the targets, which eliminates the perceiver effect altogether. We have demonstrated that this approach may be disadvantageous because it eliminates the possible emergence of potentially interesting phenomena. We have chosen to interpret our results in terms of social-role theory.

384

WINQUIST, MOHR, AND KENNY

Eagly’s (1987) theory focuses on what she terms the proximal causes of gender differences in behavior (i.e., social roles), rather than the distal factors of evolution or socialization. Others may wish to interpret the results from a purely biological or socialization perspective. We leave it to evolutionary and social learning theorists to develop such explanations. However, what should be emphasized is that the results we describe fit well with either explanation. Whether due to culture or biology, it is generally agreed upon (e.g., Bem, 1983; Eagly, 1987; Eagly, 1995; Siann, 1994) that women are encouraged to be nice to others (i.e., communal), whereas men are encouraged to compete (i.e., agentic). Given the robustness of the female positivity effect, a systematic investigation of mediating and moderating variables was undertaken in Study 2. One possible mediator considered was self-perception. We hypothesized that if women saw themselves positively and if they assumed similarity, they might see others as similar and so more positively. However, self-perception does not appear to be a mediator. Females’ positive ratings of others were not simply a reflection of their positive self-perceptions. We considered the possibility that the female positivity effect might be moderated by gender of target. We were able to test if the effect occurred exclusively within gender. For instance, is it the case that women rate only other women more favorably? This hypothesis was not confirmed, further strengthening the generalizability of the effect. These results are consistent with results found in the performance appraisal literature. Tsui and Gutek (1984) echo the findings of other performance appraisal researchers in reporting no gender-of-perceiver by gender-of-target interaction. Interestingly, they also report that female perceivers tended to give slightly higher ratings. We further examined whether females’ positive ratings were mediated by sex-role orientation. The more men and women endorse traditional sex roles, the more divergence was expected in their partner ratings. This hypothesis was supported. Seemingly, the female positivity effect is related to sex-role orientation. More specifically, females’ ratings of others may be a manifestation of their relationship orientation and fulfillment of expectations of greater compliance and nurturance. We suspect that further studies of the female positivity effect will show that sex-role orientation is the mediator of the phenomenon. There is one important limitation of the present research: All of the studies included samples taken from college students in the United States. Additionally, perhaps the reason that gender differences were very small for Agreeableness and Conscientiousness is because all targets were viewed as fairly cooperative and responsible because they were volunteering to participate in the study. Also, the results for Culture may have been attenuated because all participants were currently enrolled in college and were therefore seen as having a somewhat similar level of intelligence. More generalizable re-

THE FEMALE POSITIVITY EFFECT

385

sults and a more detailed picture of the gender–perceiver relationship would likely be obtained if subsequent researchers obtained more diverse samples and settings. CONCLUSION The current investigation documents what seems to be a prevalent phenomenon: Women rate others more positively than do men. We found this result in our review of the literature, in a survey of 10 data sets, and in a carefully controlled study. The effect, though small in magnitude, is consistently found for each of the Big Five factors and for all levels of acquaintance. We have labeled this phenomenon the female positivity effect. Further, we have made progress in understanding the source of this phenomenon. First, we demonstrated that the female positivity effect is not due to assumed similarity. However, there is evidence that sex-role orientation mediates the effect, and so the question deserves further probing. Our Study 2 presents the appropriate methodological and statistical model to other interested researchers. Finally, by rejecting the notion that assimilation is merely a methodological artifact, we uncovered significant research evidence concerning expectations perceivers have of others, an area abundant with future research possibilities. APPENDIX Studies Included in Review Studies are described according to the following format: author and date; number of participants; level of acquaintance and type of interaction. 1. Park, B., & Judd, C. M. (1989); N ⫽ 71; previously unacquainted individuals completed a series of standardized interviews, each interview lasting 3 min, one half-hour, four sessions over 4 consecutive days. 2. Levesque, M. J. (1990); N ⫽ 142; following mock jury deliberations, participants rated each other. 3. Albright, L., Kenny, D. A., & Malloy, T. E. (1988); Study 1, Wave 1, N ⫽ 55; Study 1, Wave 3, N ⫽ 52; Study 3, Wave 3, N ⫽ 40; previously unacquainted individuals interacted briefly and rated group members. 4. Hallmark, B., & Kenny, D. A. (1989); N ⫽ 163; previously unacquainted individuals interacted briefly and rated group members. 5. Kenny, D. A. (1990); N ⫽ 84; previously unacquainted individuals rated group members. 6. Malloy, T. E., & Janowski, C. L. (1992); N ⫽ 68; short-term acquaintance; rated group members. 7. Kenny, D. A., Horner, C., Kashy, D. A., & Chu, L. (1992); Study 3, Wave 1, N ⫽ 70; previously unacquainted individuals rated group members.

386

WINQUIST, MOHR, AND KENNY

REFERENCES Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage Publications. Albright, L., Kenny, D. A., & Malloy, T. E. (1988). Consensus in personality judgments at zero acquaintance. Journal of Personality and Social Psychology, 55, 387–395. Amatora, M. (1955). Comparisons in personality self-evaluation. Journal of Social Psychology, 42, 315–321. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Bem, S. L. (1974). The measurement of psychological androgyny. Journal of Consulting and Clinical Psychology, 42, 155–162. Bem, S. L. (1983). Gender schema theory and its implications for child development: Raising gender-aschematic children in a gender-schematic society. Signs, 8, 598–616. Benton, A. A., Gelber, E. R., Kelley, H. H., & Liebling, B. A. (1969). Reactions to various degrees of deceit in a mixed-motive relationship. Journal of Personality and Social Psychology, 12, 170–180. Bettencourt, B. A., Dill, K. E., Greathouse, S. A., Charlton, K., & Mulholland, A. (1997). Evaluations of ingroup and outgroup members: The role of category-based expectancy violation. Journal of Experimental Social Psychology, 33, 244–275. Bledsoe, J. C. (1961). Sex differences in mental health analysis scores of elementary pupils. Journal of Consulting Psychology, 25, 364–365. Bledsoe, J. C. (1967). Self-concepts of children and their intelligence, achievement, interests, and anxiety. Childhood Education, 43, 436–438. Carli, L. L. (1982). Are women more social and men more task oriented? A meta-analytic review of sex differences in group interaction, reward allocation, coalition formation, and cooperation in a Prisoner’s Dilemma game. Unpublished manuscript, University of Massachusetts, Amherst. Carli, L. L., LaFleur, S. J., & Loeber, C. C. (1995). Nonverbal behavior, gender, and influence. Journal of Personality and Social Psychology, 68, 1030–1041. Cronbach, L. J. (1955). Processes affecting scores on ‘‘understanding of others’’ and ‘‘assumed similarity.’’ Psychological Bulletin, 52, 177–193. Eagly, A. H. (1987). Sex differences in social behavior: A social-role interpretation. Hillsdale, NJ: Erlbaum. Eagly, A. H. (1995). The science and politics of comparing women and men. American Psychologist, 50, 145–158. Eagly, A. H., & Carli, L. L. (1981). Sex of researchers and sex-typed communications as determinants of sex differences in influenceability: A meta-analysis of social influence studies. Psychological Bulletin, 90, 1–20. Eagly, A. H., & Crowley, M. (1986). Gender and helping behavior: A meta-analytic review of the social psychological literature. Psychological Bulletin, 100, 283–308. Eagly, A. H., & Steffen, V. J. (1986). Gender and aggressive behavior: A meta-analytic review of the social psychological literature. Psychological Bulletin, 100, 309–330. Feingold, A. (1994). Gender differences in personality: A meta-analysis. Psychological Bulletin, 116, 429–456. Goldberg, L. R. (1990). An alternative ‘‘description of personality’’: The Big-Five factor structure. Journal of Personality and Social Psychology, 59, 1216–1229.

THE FEMALE POSITIVITY EFFECT

387

Goldrich, J. M. (1967). A study in time orientation: The relation between memory for past experience and orientation to the future. Journal of Personality and Social Psychology, 6, 216–221. Goss, A. M. (1968). Estimated versus actual physical strength in three ethnic groups. Child Development, 39, 283–290. Hall, J. A. (1984). Nonverbal sex differences: Communication accuracy and expressive style. Baltimore, MD: Johns Hopkins Univ. Press. Hallmark, B. W., & Kenny, D. A. (1989). The effect of rating a person at a time or a trait at a time on the variance components of the Social Relations Model. Unpublished manuscript, University of Connecticut. Jones, E. E., Rock, L., Shaver, K. G., Goethals, G. R., & Ward, L. M. (1968). Pattern of performance and ability attribution: An unexpected primacy effect. Journal of Personality and Social Psychology, 10, 317–340. Kenny, D. A. (1990). Zero acquaintance replication with self-ratings before other ratings. Unpublished raw data, University of Connecticut. Kenny, D. A. (1993). SOREMO: A FORTRAN program for the analysis of round-robin data structures. Unpublished manuscript, University of Connecticut. Kenny, D. A. (1994). Interpersonal perception: A social relations analysis. New York: Guilford. Kenny, D. A., Horner, C., Kashy, D. A., & Chu, L. (1992). Consensus at zero acquaintance: Replication, behavioral cues, and stability. Journal of Personality and Social Psychology, 62, 88–97. Kenny, D. A., & La Voie, L. (1984). The Social Relations Model. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 18, pp. 142–182). Orlando, FL: Academic Press. Levesque, M. J. (1990). Round-robin ratings following mock jury deliberations: Perceptions of intelligence and influence. Unpublished raw data, University of Connecticut. Maccoby, E. E., & Jacklin, C. N. (1974). The psychology of sex differences. Stanford: Stanford Univ. Press. Malloy, T. E., & Janowski, C. L. (1992). Perceptions and meta-perceptions of leadership: Components, accuracy, and dispositional correlates. Personality and Social Psychology Bulletin, 18, 700–708. Marks, G., & Miller, N. (1988). Perceptions of attitude similarity: Effect of anchored versus unanchored positions. Personality and Social Psychology Bulletin, 14, 92–102. McDonald, R. L. (1968). Effects of sex, race, and class on self, ideal-self and parental ratings in southern adolescents. Perceptual and Motor Skills, 27, 15–25. Mohr, C. D., & Kenny, D. A. (1997). Stereotypes, communication context, and accuracy in impression formation. Unpublished paper, University of Connecticut. Park, B., & Judd, C. M. (1989). Agreement on initial impressions: Differences due to perceivers, trait dimensions, and target behaviors. Journal of Personality and Social Psychology, 56, 493–505. Sarason, I. G., & Koenig, K. P. (1965). Relationships of test anxiety and hostility to description of self and parents. Journal of Personality and Social Psychology, 2, 617–621. Siann, G. (1994). Gender, sex and sexuality: Contemporary psychological perspectives. Bristol, PA: Taylor & Francis. Smart, M. S., & Smart, R. C. (1970). Self-esteem and social-personal orientation of Indian 12- and 18-year-olds. Psychological Reports, 27, 107–115.

388

WINQUIST, MOHR, AND KENNY

Snyder, M., & Gangestad, S. (1986). On the nature of self-monitoring: Matters of assessment, matters of validity. Journal of Personality and Social Psychology, 51, 125–139. Spence, J. T., & Helmreich, R. L. (1978). Masculinity and femininity: Their psychological dimensions, correlates, and antecedents. Austin: Univ. of Texas Press. Swim, J. K. (1994). Perceived versus meta-analytic effect sizes: An assessment of the accuracy of gender stereotypes. Journal of Personality and Social Psychology, 66, 21–36. Tsui, A. S., & Gutek, B. A. (1984). A role set analysis of gender differences in performance, affective relationships, and career success of industrial middle managers. Academy of Management Journal, 27, 619–635. Vigor, K. (1992). CHAT (Version 1.1). Salt Lake City, UT. [Computer Software] Walker, R. N. (1967). Some temperament traits in children as viewed by their peers, their teachers, and themselves. Monographs of the Society for Research in Child Development, 32.