Abstract
The effects of sex composition of dyads and surrounding others in groups on the accuracy in women’s expected evaluations (metaperceptions) were investigated using the Group Actor-Partner Interdependence Model (GAPIM). The dataset comprised 26 groups of four to six previously unacquainted participants who completed an unstructured social interaction followed by round-robin evaluations and metaperceptions on evaluative adjectives. Women expected less positive evaluations from opposite-sex partners, and this did not vary by the sex composition of surrounding others. However, women were less accurate in predictions of how they were evaluated by male partners (compared with female partners) if women were the majority in the surrounding group. Also, women formed less accurate metaperceptions of female partners when women were the minority, rather than the majority in the surrounding context. Implications for theory development in intergroup relations and contributions of the GAPIM to diversity research are discussed.
Although many might assume that an increased representation of women in groups and other social contexts would yield positive social consequences for women, empirical research has yielded mixed conclusions. Research is available to support the conclusion that women who work in groups with other women report more positive experiences (Chatman & O’Reilly, 2004), more negative experiences (Ely, 1994), or no difference in their experiences based on gender similarity to others (Riordan & Shore, 1997). Some of those conclusions contradict decades of research showing that in-group favoritism occurs in judgments of similar others in diverse environments (Brewer, 1979, 2007). If demonstrating in-group favoritism, women should form more positive perceptions of other women in mixed-sex groups and organizations. Moreover, this long-standing conclusion might lead women to expect to be viewed positively by other women in their groups.
However, findings from recent research showed that women might form more negative perceptions of each other when fewer women are represented in an organization (Ely, 1994; Ryan et al., 2012). The irony of this finding is that when women are least represented in an organization is precisely when support from fellow female colleagues might be especially important. Based on the results, however, if women presume that the few other women in the group or organization will be supportive of them because they share a social identity, they might make the socially and psychologically costly error of expecting social support that is not available.
This study examined the impact of gender representation on women’s expected evaluations from male and female group members by measuring women’s metaperceptions—perceptions of what group members think of them (Laing, Phillipson, & Lee, 1966)—in mixed-sex groups. Relying on past research on intergroup metaperceptions, this study investigated the role of sex representation in groups on the accuracy with which women form metaperceptions of male and female group members, also known as meta-accuracy. The specific ways in which the representation of same-sex and opposite-sex others in a group affected the accuracy of women’s metaperceptions were analyzed using a formal analytic model—Group Actor-Partner Interdependence Model (GAPIM; Kenny & Garcia, 2012; Kenny & Livi, 2009).
Intergroup Metaperceptions and Meta-Accuracy
Knowing how others evaluate the self affects the decisions one makes to approach or avoid opportunities, relationships, and sources of social support. Importantly, forming metaperceptions of group members that are inaccurate might impair interpersonal processes and socially emergent states (e.g., shared mental models; Thompson & Fine, 1999) that are important for successful communications in groups and teams (King, Kaplan, & Zaccaro, 2008).
Fortunately, some research suggests that people do a good job at figuring out how others on average might view them (Carlson & Furr, 2009; DePaulo, Kenny, Hoover, Webb, & Oliver, 1987; Kenny & DePaulo, 1993), even if they are based on first impressions (Carlson, Furr, & Vazire, 2010). Carlson, Vazire, and Furr (2011) found perceivers generally can form accurate metaperceptions of what others think of their personality traits even if the metaperceptions are different from how those perceivers view themselves on those traits. However, intergroup interactions, such as those that would occur in mixed-sex groups, might provide an exception to the rule. Intergroup interactions often involve social status differences between the represented group identities, which may lead to some asymmetries in how group members form perceptions of each other. For instance, lower status group members might be especially motivated to attend to what higher status group members think of them (Neuberg & Fiske, 1987; Stevens & Fiske, 2000). Although research suggests individuals in groups form accurate self-perceptions of their relative status in the group (Anderson, Srivastava, Beer, Spataro, & Chatman, 2006), their metaperceptions of other group members might not show the same degree of accuracy. Social interactions involving members of different social categories or identities have been shown to yield not only atypical (Vorauer, 2006) but also inaccurate metaperceptions (Miller & Malloy, 2003; Santuzzi, 2007).
The metaperceptions of women in mixed-sex interactions are interesting to examine because women are often stereotyped as being particularly socially sensitive and are believed to be better at reading how others view them, despite research conclusions to the contrary (Ickes, Stinson, Bissonnette, & Garcia, 1990; Snodgrass, 1985, 1992). Earlier research suggests that women’s meta-accuracy might depend on status and power differences among social partners. In mixed-sex interactions, women might be more sensitive to male partners’ perceptions of them as compared with those of female partners if there also is a difference in status (Snodgrass, 1985). Research suggests that lower status individuals might be especially concerned about how higher status partners view them in group contexts (Ames, 2004; Snodgrass, 1992) and may put more effort into forming those metaperceptions (Kaplan, Santuzzi, & Ruscher, 2009). Status differences could be implied by the relative representation of men and women in mixed-sex groups, such that the less represented group holds lower status. Women in male-dominated groups, all else being equal, could be in a position to form more accurate metaperceptions of male partners than female partners, as they might have a greater interest in knowing what higher status (male) partners think of them. However, other factors such as concerns about stereotypes also might arise and interfere with meta-accuracy in mixed-sex groups, particularly when there is an unequal sex representation.
Sex Composition and Meta-Accuracy
Frey and Tropp (2006) suggested that the key to understanding how women might form metaperceptions in mixed-sex contexts is to first understand when women expect to be seen as members of a social group (women) rather than as individuals. Revisiting the now classic notions outlined in social identity theory (Tajfel & Turner, 1986), self-categorization theory (Turner, Hogg, Oakes, Reicher, & Wetherell, 1987), and distinctiveness theory (McGuire, 1984), Frey and Tropp argued that the presence and relatively larger number of out-group members in a mixed-group context might activate the group identity. With the group identity salient, people might expect to be viewed by others in light of the characteristics associated with their group labels instead of their own individual qualities. By this logic, women in mixed-sex groups might expect to be targets of sex-based stereotypes while interacting with men, especially when women are the minority in the group. Seeing themselves as part of women as a group, they also might expect to benefit from in-group favoritism from female group members (Brewer, 1979), especially when women are the minority in the group.
Furthermore, women are expected to be more sensitive to group identity salience than men in mixed-sex situations. Pichevin and Hurtig (1996) found that the sex of group members was more salient to female observers than male observers in a mixed-sex context. Corroborating this logic, Chatman and O’Reilly (2004) showed that being part of groups with less representation of the same sex yielded less positive affect among women (but not men) in work teams. This suggests that women might be especially aware of sex composition of their groups and, as a consequence, report experiences that depend on their relative representation in the group. This provides a plausible explanation for why women might report more discomfort than men when they are the minority in mixed-sex group situations. The goal of this study is to determine whether being the minority, and thus having the social identity salient, affects women’s accuracy in forming metaperceptions of group members.
Sex Composition at the Dyad and Group Level
The way in which relative representation of women and men might affect women’s ability to accurately judge what partners think about them might occur at different levels of analysis. Sex composition may function at both the dyad level (meta-accuracy with individual partners) and in the context of surrounding others (sex composition of others in the group) to affect interpersonal perceptions.
This also means that meta-accuracy can be driven by different forces at different levels. Metaperceptions might be inaccurate because either they are adjusted in an inaccurate direction or they fail to be adjusted in the correct direction. Importantly, inaccuracy in metaperceptions does not require that there be a mean-level difference between a perceiver’s metaperception and partners’ actual evaluations of the perceiver. Thus, the mean levels could look similar even if the process by which the metaperceptions were derived was biased (West & Kenny, 2011). For example, women in mixed-sex groups might form metaperceptions of men that are biased by the negative discomfort of being the minority in a mixed-sex situation. However, the presence of bias may not be reflected in mean differences between the metaperceptions and the actual evaluations from those group members. An example of this pattern was shown in research examining smokers’ metaperceptions of other smokers and nonsmokers (Santuzzi, 2007). Although no significant differences in mean-level metaperceptions emerged depending on whether the partner was a smoker or nonsmoker, a social relations analysis revealed different sources of influence on metaperceptions of smokers compared with nonsmokers. Moreover, smokers’ metaperceptions of nonsmokers were less accurate than smokers’ metaperceptions of other smokers. Thus, metaperceptions can show different levels of accuracy across partners even if the mean levels of those metaperceptions are not significantly different.
Sex Composition of Dyads
The social psychology literature has identified several sources of bias in the way individuals form metaperceptions of specific partners. For instance, some research has shown that individuals might form inaccurate metaperceptions if they rely too heavily upon private information that others would not know or would not be able to see (Chambers, Epley, Savitsky, & Windschitl, 2008; Vorauer & Ross, 1999). Assuming that both you and your interaction partner have access to the same personal information might yield inaccurate metaperceptions if the partners form judgments of you based on group stereotypes instead of the individuating characteristics evident to you in your self-perception. As such, women could lose accuracy in their metaperceptions of male partners if they fail to take into account stereotypes and other factors that might affect male partners’ evaluations of them in mixed-sex situations.
However, incorrectly forming metaperceptions based on expected stereotypes—or meta-stereotypes—that are not apparent in partners’ actual evaluations might also reduce meta-accuracy (Vorauer, 2006; Vorauer, Main, & O’Connell, 1998). As noted earlier, a mixed-group context can highlight group identities and corresponding stereotypes. When individuals try to form predictions of what previously unacquainted others think of them, they might assume that stereotypes about their groups will be the basis of the judgments made by out-group members (see Frey & Tropp, 2006). Given that stereotypic judgments are likely to be at a higher level of abstraction relative to individuated judgments (Klein & Snyder, 2003; Sherman, 1996), meta-stereotypes will lead to less meta-accuracy among women forming metaperceptions of men if the actual evaluations from men are more individuated.
Despite past evidence that women might be more attuned to men’s perceptions as compared with women’s perceptions of them (e.g., Snodgrass, 1985), the fact that stereotypes become salient in such intergroup interactions introduces more sources of error in metaperception judgments. Even if they are attuned, women who are forming metaperceptions of male partners might vary in their beliefs about the extent to which group-based stereotypes factor into men’s judgments of them. Thus, meta-accuracy might be lower among women when forming metaperceptions of male as compared with female group members. The nature of the meta-accuracy will be examined within mixed-sex dyads as well as mixed-sex dyads that are surrounded by a mixed-sex group.
Sex Composition of Surrounding Others
Within a group, any two members can be defined as a dyad. For each dyad, the remaining group members become the surrounding others. Given that the composition of groups might affect metaperceptions in the intergroup context (Frey & Tropp, 2006), the sex composition of the surrounding group should affect the metaperceptions that women form of specific male and female partners.
Speaking to this point, Vorauer (2003) found evidence that intergroup metaperceptions might vary if the judgment occurs in the presence of additional racial in-group or out-group members, rather than in isolation from others. Examining only White perceivers, she found that low-prejudice individuals paired with an Aboriginal observer formed more negative metaperceptions of that observer while in the presence of another White person (who was exhibiting prejudice). The pattern was reversed for high-prejudice individuals in the same situation. They expected more positive evaluations from the out-group observer than an in-group observer in the presence of the in-group partner. A plausible explanation for this pattern is that the low-prejudice individuals felt they had more control over how the out-group observer viewed them when the other in-group member was not present, and thus, formed positive metaperceptions. When another in-group member is present during the intergroup interaction, the control over identity-based impressions is reduced (Vorauer, 2003). Although the research did not consider the accuracy of the metaperceptions, it did find that the group identity of surrounding others significantly affected intergroup metaperceptions. If these findings can be generalized to mixed-sex groups, then women might form different metaperceptions of men depending on whether other women are in the surrounding group.
Also, women might form different (and less accurate) metaperceptions of other women depending on the sex composition of the surrounding others. For example, women might not form accurate metaperceptions of other women if they rely on an assumption that other women will see them as they see themselves (Chambers et al., 2008; Vorauer & Ross, 1999). If female partners’ evaluations are being affected by the sex composition of the surrounding group, then women holding on to an assumed transparency with those female partners will be forming inaccurate metaperceptions of them. They will inaccurately assume that female partners are evaluating their individuating features and fail to fully account for how the social environment affects female partners’ evaluations.
The above logic could inform recent research showing that women might form more negative evaluations of other women in male-dominated groups. As noted by Ryan et al. (2012), the extent to which women can accurately expect positive social experiences with other women may depend on the overall representation of women in groups (e.g., organizations). Moreover, women who fail to adjust their metaperceptions in situations where in-group positivity is less likely to occur will be more inaccurate in their metaperceptions formed of other women. The present study examined how the number of in-group and out-group members in the surrounding social context affects women’s meta-accuracy of male and female partners. In light of past research on metaperception biases and the findings of Ryan et al. (2012), the presence of fewer other women than men in a mixed-sex context, thus making the social identity salient, might lead women to incorrectly expect positive evaluations from other women. Thus, meta-accuracy of female partners should be lower in mixed-sex groups when women are the minority.
Method
Participants
The data 1 comprised round-robin ratings from 26 groups ranging in size from 4 to 6 undergraduate students from a large Midwestern university (N = 122). Participants in each group were unacquainted prior to participating in the study and were informed that the study will examine multiple levels of influence on how individuals form perceptions of each other. Both women (n = 79) and men (n = 43) participated in unsystematic group compositions. Fourteen groups had women represented as the numerical majority, 6 groups had men in the majority, 2 groups had equal representation, and 5 groups had only women represented. The majority of the sample self-reported being White (75%), with 8% Black, 8% Hispanic, and 6% Asian. Three percent selected “Other” or did not report a race designation. Participant ages ranged from 18 to 24 years with a median age of 19 years.
Procedure
Participants completed a demographics inventory that asked them to report their sex, race, major area of study, initial self-perceptions on the traits to be rated (see below), and anxiety levels. Anxiety and initial self-perceptions were tested as potential control variables when examining preliminary mean differences in metaperceptions (Christensen, Stein, & Means-Christensen, 2003). To measure anxiety, participants were asked to what extent they were comfortable in meeting new people, talking on the phone, performing on stage, and public speaking. They responded to each item on a 1 (not at all) to 4 (very much) rating scale. The mean score on the four items (Cronbach’s α = .71) and initial self-perceptions (Cronbach’s α = .87) did not change the results of the preliminary metaperception means differences tests; thus, results were presented for statistical models that do not include these additional variables.
Each group of participants then engaged in an unstructured get-to-know-you discussion for 10 minutes. Participants were instructed to learn as much as they could about each of their group members and that they would be completing personality judgments of their group members at the end of the discussion. Following the interaction, group members were asked to rate each group member on a list of 12 positive adjectives and 12 negative adjectives (Saucier, 1994), on a 5-point rating response scale (1 = not at all to 5 = very much). The positive adjectives were self-controlled, mature, broad-minded, optimistic, wise, understanding, purposeful, alert, generous, clear-headed, considerate, and reasonable; the negative adjectives were greedy, deceitful, closed-minded, juvenile, immature, narrow-minded, inconsiderate, hypocritical, half-hearted, short-sighted, unstable, and bigoted. The items have been shown to load on a single general evaluation factor yielding strong internal consistency (α = .84) and strong convergent correlations with social desirability (r = .88; Saucier, 1994).
In addition, participants formed metaperceptions of each group member by guessing how each group member rated them on each evaluative adjective on the list. Finally, each participant completed a second measure of self-perceptions (although the data were not used in the present analysis).
Analysis Using the GAPIM
The GAPIM (Kenny & Garcia, 2012) is an extension of the Actor-Partner Interdependence Model (APIM; Kenny, 1996) for use with group data. Similar to a social relations analysis (Kenny & LaVoie, 1984), GAPIM can account for the random sources of variation at the individual, dyad, and group levels in metaperceptions and evaluations. In addition, the GAPIM allows for fixed effects such as group membership and group composition to be tested in mixed-sex groups.
In dyadic social interactions involving only one actor and one partner, the sex of the dyad can affect social perceptions in one of three effects: the actor’s sex, the partner’s sex, and/or the similarity of the actor to the partner (Actor sex × Partner sex interaction; see West, Popp, & Kenny, 2008). If focusing only on women’s metaperceptions (i.e., holding actor as constant), then the extent to which women’s metaperceptions differ due to the sex of the partner would simply be tested as the effect of partner sex on metaperceptions. However, when examining dyads in the context of larger groups, it is necessary to include several more interactions and higher-order interactions to study the effects of surrounding others in the group (Kenny & Garcia, 2012). For example, the extent to which women’s metaperceptions of women differ by the number of women in the surrounding group could be examined by testing the interaction between sex of partner and sex of surrounding group members (excluding partner).
GAPIM also can flexibly accommodate tests of moderating variables, allowing for formal tests of meta-accuracy differences across group compositions. Thus, the extent to which women’s metaperceptions predict actual evaluations and whether the surrounding group composition moderates that relationship can be built into the model in much the same way as other multilevel models. For example, a model to test whether women’s meta-accuracy differs for male as compared with female partners would include partner sex, metaperceptions (grand mean-centered), and the interaction between partner sex and metaperceptions predicting partners’ actual evaluations.
The effects of sex composition in this study were tested using the dyadic GAPIM model (GAPIM-D) adapted for use in SAS (v. 9.3) PROC MIXED with maximum likelihood estimation. Group size was entered as a control variable in the first step of all analyses. The sex for each group member was effect coded (male = 1; female = −1).
Results
Preliminary Analysis: Women’s Metaperceptions in Mixed-Sex Groups
Women showed a significant difference in metaperceptions depending on the sex of the specific partner, b = −.02, SE = .01, t(58.8) = −2.34, p = .02. Female metaperceivers formed less positive metaperceptions of male partners compared with female partners. For the purposes of comparison, a similar analysis predicting men’s metaperceptions showed no difference between male and female partners, b = .005, SE = .02, t(26.2) = .30, p = .77. However, the mean-level metaperceptions for women were not directly related to the surrounding group composition, b = .04, SE = .03, t(69.2) = 1.56, p = .12. Moreover, a nonsignificant interaction between sex of partner and sex of surrounding others suggested that the less positive metaperceptions that women formed of male compared with female partners did not appear to be affected by the sex composition of surrounding others, b = −.03, SE = .02, t(75.1) = −1.36, p = .18.
Note that the hypotheses in this study focused on the accuracy of women’s metaperceptions, regardless of mean differences. For example, were women correct or incorrect by adjusting their metaperceptions to be more negative when interacting with male partners? Moreover, were women less accurate when they did not adjust their metaperceptions of male and female partners according to the surrounding context?
The actual evaluations from partners in groups significantly differed by the partner’s sex, b = −.15, SE = .03, t(114) = −4.82, p < .001. Male partners generally gave women significantly less positive evaluations than female partners gave. Furthermore, those evaluations did not differ by the sex composition of surrounding others (i.e., no partner sex by surrounding others interaction), b = .001, SE = .05, t(96.1) = 0.04, p = .97. To be accurate, women would thus have to adjust their metaperceptions to the male partner’s, accurately and consistently across group composition contexts.
Women’s Meta-Accuracy of Partners in Mixed-Sex Groups
In general, women’s metaperceptions predicted their partners’ actual evaluations of them, b = .16, SE = .07, t(116) = 2.39, p = .02. Although a modest slope, this result may be interpreted as some degree of generalized meta-accuracy among women across male and female partners. It is also interesting to note that, in general, being in mixed-sex groups (coded as 1) compared with same-sex groups (coded as 0) yielded less meta-accuracy among female metaperceivers. Specifically, the interaction between group type (0 or 1) and metaperceptions (grand mean-centered) predicting evaluations was negative and significant, b = −.44, SE = .15, t(99.9) = −2.88, p = .005. If women can be assumed to have relatively lower status than men in mixed groups, this result is consistent with past research showing less meta-accuracy among other relatively lower status group members in mixed-group contexts (viz., smokers; Santuzzi, 2007).
To test the meta-accuracy differences based on the sex of the specific partner predicted in Hypothesis 1, the analysis included women’s metaperceptions (grand mean-centered), partner sex, and the interaction between metaperceptions (grand mean-centered) and the partner sex as predictors of the actual evaluations women received. Accuracy of women’s metaperceptions were not significantly different with a male partner as compared with a female partner, b = −.09, SE = .06, t(249) = −1.50, p = .14. Thus, women formed metaperceptions that were significantly more negative for male than female partners, but these did not seem to affect accuracy in those metaperceptions. Hypothesis 1 was not supported.
Importantly, the above analysis collapsed across the varying sex compositions of surrounding others in the groups. Meta-accuracy in dyads was examined to see if women’s accuracy of male and female partners was affected by the sex composition of the others surrounding the dyad. The model predicting actual evaluations by partners included partner sex, sex composition of surrounding others, metaperceptions, all two-way interactions, and the three-way interaction among sex of partner, sex composition of surrounding others, and metaperceptions.
Women’s meta-accuracy was influenced by the combination of partner sex and sex composition of surrounding others, demonstrated in a significant three-way interaction (Metaperceptions × Partner sex × Surrounding others), b = .30, SE = .11, t(248) = 2.63, p = .01. The three-way interaction was decomposed using dummy codes for sex of partner and re-centering the proportions of surrounding others to interpret the two-way interactions at each partner sex (male or female) and at two different levels of surrounding others (male vs. female majority). When women were surrounded by more women than men in the group, their metaperceptions of male partners were significantly less accurate than for female partners. The interaction between metaperceptions and partner sex (holding female-dominated surrounding others at 0) was significant, b = −.38, SE = .14, t(242) = −2.80, p = .01. Simple slopes analyses revealed that meta-accuracy for female partners in female-dominated groups was positive and significant, b = .37, SE = .09, t(102) = 4.35, p < .001. Women showed nearly no meta-accuracy with male partners in the female-dominated groups, b = −.01, SE = .11, t(213) = −0.11, p = .91.
When surrounded by a male-dominated group, women were not significantly more accurate with male partners than female partners. The two-way interaction for metaperceptions and partner sex (holding male-dominated surrounding others at 0) was not significant, b = .32, SE = .25, t(263) =1.31, p = .19.
The three-way interaction also could be analyzed by isolating the two-way interactions of surrounding others on meta-accuracy for each partner sex. When women formed metaperceptions of female partners, they were significantly less accurate when more men than women were in the surrounding group context. The interaction between metaperceptions and surrounding others (holding female partner sex at 0) was significant, b = −.36, SE = .13, t(120) = −2.72, p = .01. This suggests women’s meta-accuracy for female partners in female-dominated (b = .37) and male-dominated groups (b = −0.04) are statistically different. Confirming Hypothesis 2, when women were the minority in the group, their meta-accuracy of female partners was significantly lower. In contrast, when forming metaperceptions of male partners’ evaluations, more men in the surrounding group did not yield significantly different meta-accuracy. The two-way interaction between metaperceptions and surrounding others (holding male partner sex at 0) was not significant, b = .25, SE = .19, t(215) = 1.33, p = .18.
In summary, having a male partner yielded more negative metaperceptions but did not seem to directly relate to meta-accuracy for female metaperceivers. Instead, the sex representation in the surrounding group seemed to be the strongest factor in determining whether women formed accurate metaperceptions of how partners evaluated them. Women formed less accurate metaperceptions of male partners than of female partners when there were more women in the surrounding others. Women showed significantly less meta-accuracy of their female interaction partners in male-dominated groups compared with female partners in female-dominated groups.
Discussion
The results from this study suggested that sex composition of dyads and surrounding others might affect women’s meta-accuracy for evaluation. When women formed metaperceptions of male partners, their metaperceptions of those partners were more negative. However, sex of partner alone did not have a direct significant influence on dyadic accuracy in this study. Instead, women’s meta-accuracy of male (and female) partners seemed to depend on the sex composition of surrounding others. Women did show less meta-accuracy of male partners compared to female partners, but only when other women were the majority in the surrounding context.
Confirming the second hypothesis, women showed less meta-accuracy for evaluation with female partners in groups where women were the minority compared with when women were the majority. Combined with the evidence that women’s mean-level metaperceptions were not different under those conditions, one plausible inference is that the less accuracy was due to a failure to adjust their metaperceptions of female partners based on the sex composition of the surrounding group. Another possibility is that when women were in the minority, they paid more attention to the men in the majority when forming metaperceptions rather than their in-group members (Ames, 2004). The reduced attention to in-group members when the out-group was the majority might have introduced error into the metaperceptions that women formed of other women in these situations.
Similarly, the relatively lower meta-accuracy for male partners compared to female partners was observed when women were the in the numerical majority in the group. Note that male partners generally gave more negative evaluations of group members. One explanation is that the higher representation of women in the group might have kept group membership salience to a minimum, thus leading women to underestimate the extent to which their group identity would be the basis for how male partners evaluated them (Frey & Tropp, 2006). An alternative explanation is that, similar to the logic in the previous paragraph, group members might pay more attention to the numerical majority members, at the cost of accuracy in impressions of the minority members. Although women are in-group rather than out-group members and typically not inferred to be the power holders in mixed-sex groups, the fact that they were the majority group represented might have shifted the power in the group (and the attention of perceivers) toward them. Future research should directly test perceptions of power in intergroup contexts due to numerical representation and how such representation might affect the impressions formed of numerical minority and majority in-group and out-group members.
This finding that women seemed to lose meta-accuracy when forming metaperceptions of other women in male-dominated groups might inform conclusions from notable past work on interpersonal sensitivity in mixed-sex social interactions. Snodgrass (1985) examined same-sex and mixed-sex dyads and found that the least accurate sex combination was women with other women. By comparison, the results from that study seem to match the results for female–female dyads in the current study when the surrounding group was male-dominated. If experimenters or other participants in the Snodgrass study were male, the mixed-sex environment might have disrupted the expected meta-accuracy between two women. The description of the procedures in Snodgrass (1985) did not allow for answers to these questions. However, future studies on mixed-sex dyads should at a minimum note the social context in the study and ideally examine social context as a key variable in the analysis.
The conclusions from the present study highlight the important role of social context not only in intergroup relations but also in the research process of studying intergroup relations. More often than not, the social context surrounding a dyadic interaction is either assumed to be irrelevant or treated as an error (see Vorauer, 2003, for an exception). However, in light of the present results, the social context even in laboratory research on intergroup perceptions should be carefully considered. For example, the sex and other visible characteristics of laboratory experimenters should be treated as variables in laboratory research involving intergroup relations topics. The results from this study demonstrated through a direct empirical examination that the sex of surrounding others is a systematic factor—a primary systematic factor—in metaperceptions and meta-accuracy of evaluation in mixed-sex contexts. Similarly, Vorauer (2003) found that the presence of a racial in-group member might affect expected evaluations from an out-group member. Researchers might inadvertently affect the results of their intergroup perceptions studies if they neglect to consider the demographic features of the laboratory personnel as part of the social context.
Limitations
The conclusions from this study are limited due to the fact that data were collected from previously unacquainted participants in a single, simultaneous group interaction. Relying on a simultaneous group interaction preserved a constant social context and controlled the content, such that every group member was exposed to the same social information from group members. However, the sex composition effects might have been different if partners interacted one-on-one rather than in a common group context. Also, the data were collected from previously unacquainted participants after a single laboratory interaction. The role of sex (and other demographic characteristics) might play a stronger role in initial group interactions but then dissipate over time as groups learn to manage their differences (see Moreland, Levine, & Wingert, 2013). Indeed, research has suggested demographic diversity shows less impact on work team processes over time (Chatman & Flynn, 2001; Harrison, Price, Gavin, & Florey, 2002; see King, Hebl, & Beal, 2009, for a review). Finally, the group compositions represented in this study were adequate for examining the hypotheses about women’s metaperceptions and meta-accuracy in mixed-group contexts. However, the data did not allow for a full exploration of men’s metaperceptions and meta-accuracy across all contexts. None of the groups in this study were homogeneous male groups, thus precluding a fair test of whether metaperceptions and meta-accuracy differ by the metaperceiver’s sex in mixed-sex and same-sex contexts.
Future Research
Although beyond the scope of the present study, future research should consider replication of these results within other intergroup contexts, such as mixed-race or mixed-age groups. Research has shown that reactions to group heterogeneity might differ depending on the type of characteristic that varies in the group (e.g., Riordan & Shore, 1997). Continued research specifically on the process of metaperception formation in intergroup contexts is especially important. Past research has shown that intergroup metaperceptions are formed differently and less accurately than those formed in homogeneous contexts, particularly for the relatively lower status metaperceivers (Miller & Malloy, 2003; Santuzzi, 2007). This might suggest a qualification to existing literature showing that individuals are generally able to infer what specific partners think of them (Carlson et al., 2011). Unfortunately, the present study was limited in that each participant was exposed to only a single mixed-sex context, thus precluding a direct test of sensitivity across contexts. In addition, the present study only examined metaperceptions of evaluation, whereas past research has focused on personality trait information (e.g., Carlson & Furr, 2009). Future research should consider the extent to which metaperceivers accurately detect the impact of surrounding social information on how they are rated by others, who form evaluations and other trait judgments using a within-subjects design. In such a study, the same metaperceivers could report metaperceptions of specific others in male-dominated and female-dominated social contexts to directly examine whether metaperceptions and meta-accuracy change across situations.
Footnotes
Acknowledgements
The author wishes to thank Kristy Garza, Mallery Weber, and Elizabeth Yan for their assistance with the data collection for this project.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The data collection for this project was supported by a National Institutes of Health (NIMH) National Research Service Award, No. MN14257, to the University of Illinois in support of the author as a postdoctoral trainee in the Quantitative Methods Program of the Department of Psychology, University of Illinois at Urbana-Champaign.
