GENDER-STEREOTYPIC IMAGES OF OCCUPATIONS
CORRESPOND TO THE SEX SEGREGATION OF EMPLOYMENT
Mary Ann Cejka & Alice H. Eagly
Personality & Social Psychology Bulletin, Apr99

This study examines the role of gender stereotypes in justifying the social system by maintaining the division of labor between the sexes. The distribution of the sexes in 80 occupations was predicted from participants' beliefs that six dimensions of gender-stereotypic attributes contribute to occupational success: masculine physical, feminine physical, masculine personality, feminine personality, masculine cognitive, and feminine cognitive. Findings showed that, to the extent that occupations were female dominated, feminine personality or physical attributes were thought more essential for success; to the extent that occupations were male dominated, masculine personality or physical attributes were thought more essential. Demonstrating the role of gender stereotypes in justifying gender hierarchy, occupations had higher prestige in that participants believed that they required masculine personality or cognitive attributes for success, and they had higher earnings to the extent that they were thought to require masculine personality attributes.



This study addresses stereotypes' justification of inequality in the social system as a whole, which is a function that has been termed system justification by Jost and Banaji (1994). Although stereotypes have long been thought to serve important motivational functions in terms of preserving one's personal status and justifying one's in-group, their function in justifying the social system more generally has received little empirical attention. If gender stereotypes justify existing social arrangements, they must support the division of labor between the sexes, which is a profoundly important feature of social structure that scholars have long argued is the root cause of the subordination of women (e.g., Lerner, 1986). In modern nations, this division of labor manifests itself not merely in the assignment of the majority of domestic work to women but also in the sex segregation of the employed workforce. The present research addresses the extent to which gender stereotypes reflect and promote sex segregation in employment.

To understand how gender stereotypes may justify the division of labor in employment, this study examines the extent to which people believe that success in occupations dominated by one sex requires personal characteristics typical of that sex. Such beliefs would foster the segregation of employment, just as beliefs that the domestic role requires feminine qualities and that employment roles require masculine qualities foster the assignment of domestic work to women and paid employment to men (Eagly & Steffen, 1984). Even though employment roles may be thought in general to require characteristics that are more masculine and less feminine than the characteristics required by the domestic role, employment roles are likely to be quite heterogeneous in the qualities that they are believed to require. Female-dominated occupations, such as the domestic role, may be thought to require attributes that are stereotypically feminine, and male-dominated occupations may be thought to require attributes that are stereotypically masculine. Moreover, if gender stereotypes justify women's subordination, prestige and income should be associated with occupations being thought to require masculine characteristics.

To appreciate the importance of the issues addressed by this research, it is necessary to understand how effectively occupational sex segregation has been maintained. The dramatic increase in women's paid employment in the 20th century may falsely give the impression that the sexual division of labor is fast disappearing. However, despite some decline in the segregation of employment in recent decades in industrialized countries, 53% of women would have to change occupations in the United States if women were to be distributed into occupations as men are distributed (Reskin & Padavic, 1994). Many occupations are almost totally dominated by one sex: For example, dental hygienists and secretaries are at least 98% women, whereas automobile mechanics and carpenters are at least 98% men (U.S. Bureau of Labor Statistics, 1998). Moreover, individual jobs within occupations (e.g., the job of dining hall manager rather than the general occupation of manager) are more highly segregated than the occupations themselves, with contemporary estimates of the proportion of women who would have to change jobs to be distributed as men running as high as 77% (Tomaskovic-Devey, 1995). Suggesting that this sex segregation is not about to disappear is Lippa and Connelly's (1990) (see also Lippa, 1995) successful prediction of the sex of 90% of their student research participants from their occupational preferences.

The idea that the sex ratios of occupations should be strongly related to gender stereotypic images of occupations follows from a central claim of Eagly's (1987) social role theory of sex-correlated differences in social behavior. This theory maintains that ideas about gender are shaped by observations of women and men in the roles that they commonly play in daily life. Consistent with the general principle of correspondence bias (Gilbert & Malone, 1995), perceivers thus reason that people have the psychological characteristics that are demanded by their family and occupational roles. Considered as a social group, women are believed to possess attributes suited for the roles they generally occupy, and men are likewise believed to possess role-appropriate attributes. Because perceivers thus infer people's characteristics from observing what they do in their daily lives, gender stereotypes can be regarded as emergents from perceivers' observations of the work that each sex commonly does (see also Yount, 1986).

Although the assignment of domestic work primarily to women is no doubt an important contributor to gender stereotyping (Eagly & Steffen, 1984), the increase in women's labor force participation in industrialized countries means that perceivers may increasingly derive their ideas about the sexes from the different types of paid work that women and men perform. Despite some authors' claims that women's increased labor force participation should bring a sharp decline in the extent to which men and women are perceived to be different (e.g., Lueptow, Garovich, & Lueptow, 1995), this prediction is not reasonable to the extent that women's paid occupations are perceived as similar to the domestic role.

Prior research has not provided a strong test of the hypothesis that the qualities thought to be required for occupational success correspond to the gender stereotype of the female or male group that is numerically dominant in the occupation, even though other studies have shown a relation between occupations' sex ratios and the ascription of some gender-stereotypic attributes to job holders in these occupations (e.g., Kalin, 1986; McLean & Kalin, 1994; Shinar, 1978). For example, there is evidence of a positive relationship between the rated masculinity versus femininity of occupations and their sex ratio (Beggs & Doolittle, 1993; Shinar, 1975). Other research has demonstrated that occupations dominated by women, compared with those dominated by men, were perceived to require a higher level of feminine personality traits for job success and a lower level of masculine personality traits (Glick, 1991). However, given that gender stereotypes include physical and cognitive characteristics as well as personality attributes (Deaux & Lewis, 1983,1984; J. E. Williams & Best, 1982), the gender-stereotypic characteristics that are relevant to employment should encompass a wide range of attributes. Therefore, in this research, we predict occupations' sex ratios from three classes of gender-stereotypic attributes that can be relevant to occupational success: personality traits, physical attributes, and qualifies of cognition and cognitive style. Although student respondents judged occupations' stereotypic qualities in this study, their judgments should differ little from those of other citizens in view of evidence that stereotypes of social groups are widely shared in society (see Deaux & Kite, 1993; Hamilton & Sherman, 1994).

In this research, the distributions of the sexes into occupations are assessed subjectively by participants' estimates of occupations' sex distributions and objectively by census data. A closely related variable is sex differences in attraction to occupations, which is also assessed subjectively and objectively. For the subjective assessment, participants rated whether men or women are more attracted to the occupations. For the objective assessment, participants indicated their own attraction to the occupations, and men's and women's ratings were differenced to yield estimates of sex differences in actual attraction. According to social role theory (Eagly, 1987), people should tend to prefer situations, including occupations, that favor qualifies thought to typify their own gender, assuming that to some extent gender roles become internalized in self-concepts (Cross & Madson, 1997). Therefore, both women and men should be attracted to gender-stereotypic occupations. Indeed, theories of occupational choice (e.g., Gottfredson, 1981) maintain that individuals' occupational aspirations tend to become limited to alternatives regarded as appropriate for their gender.

Integral to the study's focus on the system-justifying implications of gender stereotypes is its examination of the extent to which beliefs about the gender-stereotypic qualities required for success in occupations predict occupational prestige and earnings. The study thus predicts both prestige and earnings from measures of the gender-stereotypic personality, physical, and cognitive characteristics believed to be essential for success in occupations. If social structure is patriarchal, rewards such as prestige and high wages should be associated with occupations that favor masculine characteristics. Consistent with this prediction, Glick (1991) found that prestige and earnings are associated with the extent to which occupations are thought to require masculine personality traits (see also Glick, Wilk, & Perreault, 1995). The weak negative relation that Glick (1991) reported between earnings and the belief that occupations require feminine personality traits is consistent with England, Herbert, Kilbourne, Reid, and Megdal's (1994) finding that earnings relate negatively to the perception that occupations involve nurturance. It is thus possible that system justification is reflected in both a positive relation of masculine qualities to earnings and a weaker negative relation of feminine qualities to earnings.

METHOD

Participants and Procedure
Participants were 189 introductory psychology students (81 men, 108 women) who volunteered for a study on perceptions of occupations to fulfill a course requirement. These students completed questionnaires and then were thanked, debriefed, and dismissed.

To minimize context effects among the measures, these participants were divided into two groups who completed different measures. One group, consisting of 144 of the 189 participants, rated occupations on gender-stereotypic attributes and then rated either the average woman or the average man on these same attributes. These participants were divided into eight subgroups (n >/= 17), each of which rated 10 of the 80 occupations included in the study. The second group, consisting of the remaining 45 of the 189 participants, rated other aspects of all 80 occupations (attraction to them, prestige, earnings, sex distribution). Because the unit of analysis in the research was occupations rather than participants, all ratings were aggregated across the participants who rated each occupation (see the section on measuring instruments below).

In addition, 191 other undergraduates (123 men, 68 women) participated in the pretesting required to select the occupations (42 of these participants) and to construct measures of occupations' gender-stereotypic attributes (149 of these participants).

Selection of Occupations
A preliminary group of 335 occupations consisted of all occupations that the U.S. census indicated had more than 25,000 employees (Bergmann, 1986, Appendix A). To ensure that students had at least some knowledge of these occupations, they were pretested for their familiarity. Using a 5-point Likert-type scale ranging from 1 (nothing at all) to 5 (a lot), the 42 pretest participants rated how much knowledge they had about the people employed in each occupation. From among occupations receiving relatively high familiarity ratings (>/= 2.80), 80 were selected to be as evenly spaced as possible along a continuum that ranged from extremely male dominated to extremely female dominated. Examples of occupations are airline pilot, architect, barber, bus driver, computer programmer, hotel clerk, psychologist, mail carrier, speech therapist, telephone operator, and elementary school teacher. These 80 occupations were distributed among the eight versions of the questionnaire so that each version included 10 occupations ranging from male dominated to female dominated. For each version, there were two different orders of the 10 occupations.

Measurement of Belief in Gender-Stereotypic Attributes' Importance to Occupational Success
Using a 5-point Likert-type scale (1 = not at all important, 5 = essential), each participant rated the 10 occupations included in his or her version of the questionnaire on the extent to which each of 56 attributes would be necessary for success in the occupation. As shown in Table 1, these 56 attributes constituted six gender-stereotypic dimensions, which were derived on the basis of a factor analysis of pretest data. Each dimension was represented by eight attributes, with the exception of the feminine personality dimension, which was represented by 16 attributes. This larger number of attributes for the feminine personality dimension stemmed from exploratory factor analyses performed on the 149 pretest participants' ratings of nine occupations on 83 gender-stereotypic attributes chosen from research by J. E. Williams and Best (1982) and Deaux and Lewis (1984). These pretest data yielded a seven-factor solution (masculine and feminine versions of personality, physical, and cognitive attributes, with feminine personality attributes loading on two separate factors). However, in the subsequent study, the two feminine personality dimensions that had appeared in the pretest results were highly correlated, r(78) = .89, p < .0001, and thus were combined.

Because the unit of analysis in the research was occupations, means were computed for participants' ratings of each occupation on each attribute. These means were then averaged across the attributes included in each dimension to produce an overall value representing the extent to which each occupation was perceived to require masculine or feminine personality, physical, or cognitive attributes for success. The ratings were thus aggregated across the participants and across the individual attributes that were constituents of each dimension.

To establish that aggregation across the individual attributes making up each dimension was appropriate, alpha coefficients were calculated (in the Attributes x Occupations matrix for each gender-stereotypic dimension) with the data aggregated across the participants. These alpha coefficients, which assessed the homogeneity of each of these six dimensions across the attributes included in it, were appropriately high. To establish that aggregation across the participants who rated each occupation was appropriate, alpha coefficients were also calculated (in the Participants x Occupations matrix for each dimension) with the data aggregated across the attributes. This second set of calculations was performed separately for each of the eight subgroups of participants, each of which rated 10 occupations, and the resulting alphas were then averaged over the subgroups. These alpha coefficients, which assessed the homogeneity of each of the six dimensions across the participants who performed the ratings, were also appropriately high.

The gender stereotypicality of each dimension was established by comparing the participants' ratings of the average man and woman on each dimension. On a 5-point likelihood scale, these participants thus rated the likelihood that the average man or the average woman would possess each of the 56 traits. These ratings were averaged across the attributes in each dimension. Men were perceived as significantly more likely than women to have masculine cognitive, physical, and personality attributes, and women were perceived as significantly more likely than men to have feminine cognitive, physical, and personality attributes. The difference on the masculine cognitive dimension was relatively small although significant.

Finally, the discriminant validity of the gender-stereotypic dimensions was established by showing that they were not highly intercorrelated when the mean values for the occupations were correlated across the 80 occupations.

Other Measuring Instruments
Attraction to occupations. Using 5-point Likert-type scales (1 = not at all attracted, 5 = extremely attracted), participants rated each occupation according to how much they themselves were attracted to it. These ratings were aggregated separately across the male and female participants to produce measures of the extent to which men were attracted to each occupation and the extent to which women were attracted. The across-participant alpha coefficients were .92 for the male participants and .88 for the female participants. The sex-related difference in attraction to each occupation was estimated by subtracting the mean score for female participants' attraction to each occupation from the mean score for males' attraction. Participants also rated on a 5-point scale (1 = women much more attracted, 5 = men much more attracted) the degree to which they believed that women or men would be attracted to each occupation. The across-participant alpha coefficient was .99 for this measure.

Prestige and earnings. Participants rated each occupation's prestige on a 5-point Likert-type scale (1 = extremely low, 5 = extremely high). They also estimated yearly earnings for each occupation. The across-participant alpha coefficients were .99 for prestige and .92 for earnings. Actual earnings were obtained from census data (U.S. Department of Labor, 1993).

Sex distribution into occupations. Participants' perceptions of the occupations' sex distribution were assessed by their estimates of each occupation's percentage of employees who are women. The across-participant alpha coefficient was .99 for this measure. The actual sex distributions of the occupations were obtained from census data (U.S. Bureau of the Census, 1993).

RESULTS

Prediction of the Distribution of the Sexes Into Occupations From Belief in Gender-Stereotypic Attributes' Importance to Occupational Success
The main dependent variables of this study directly or indirectly index the sexual division of labor in the paid workforce: participants' estimates of occupations' sex distribution, the actual sex distribution, participants' perceptions of sex differences in attraction to the occupations, and the actual sex difference in attraction (constructed by differencing the male and female participants' attraction). These four variables were strongly intercorrelated: Perceptions of the distribution of the sexes into occupations were highly related to the actual distribution, and these variables were in turn highly related to perceived sex differences in attraction to occupations and actual sex differences in attraction. The relationships of the actual sex difference in attraction to the other three distributional variables were, however, somewhat weaker than the other relationships.

To examine the extent to which gender-stereotypic representations of occupations mirror the distribution of women and men into occupations, each of these four distributional variables was regressed onto the six dimensions that represent occupations' gender-stereotypic attributes. As indexed by the very substantial multiple Rs, prediction was quite successful in all cases, and the results were similar across the four variables. The personality and physical feminine dimensions related positively to female dominance of occupations, and the personality and physical masculine dimensions related negatively to female dominance (i.e., positively to male dominance). All of these relationships were significant except for the two physical dimensions as predictors of the actual sex difference in attraction. In contrast, the masculine and feminine cognitive dimensions failed to relate significantly to any of the distributional variables, with the exception of a negative relationship of the feminine cognitive dimension to actual sex distribution. Because this relationship was opposite to prediction and did not replicate across the other three distributional variables, we do not attempt to interpret it.

Importance of Gender-Stereotypic Attributes to Occupational Success, Compared Across the Dimensions
Belief in the contribution that gender-stereotypic qualifies make to occupational success also varied substantially across the six dimensions of gender stereotypes: Some types of characteristics were judged to be generally more important than others. To display this variation overall and within male- and female-dominated occupations, we classified the occupations as male dominated if their incumbents were 75% or more male (e.g., construction laborer, truck driver, civil engineer), female dominated if their incumbents were 75% or more female (e.g., bank teller, dental hygienist, typist), and integrated if their incumbents were less than 75% but more than 25% female (manager/administrator, bartender, pharmacist). A mixed analysis of variance (ANOVA) was then calculated, treating the three levels of sex distribution as a between-occupations variable and the six gender-stereotypic dimensions as a within-occupations (i.e., repeated measures) variable. Although the sex distribution did not produce a significant main effect, both the main effect of dimension, F(5, 385) = 141.49, p< .001, and the Distribution x Dimension interaction, F(10, 385) = 13.56, p < .001, were highly significant.

Consistent with the contrasts across the dimensions, occupations were not thought to be generally very demanding of physical characteristics. Feminine physical characteristics were perceived as least important to occupational success in general, with masculine physical traits somewhat more important. The most important dimension overall was masculine cognitive abilities, with feminine cognitive abilities and personality characteristics being next most important, followed by masculine personality characteristics. Yet, the ordering of these requisites changed substantially when viewed within occupations that were relatively dominated by one sex. For male-dominated occupations, masculine cognitive characteristics were rated as most important, followed by feminine cognitive, masculine personality, masculine physical, feminine personality, and feminine physical characteristics. For female-dominated occupations, feminine personality attributes were rated as most important, followed by masculine cognitive, feminine cognitive, masculine personality, feminine physical, and masculine physical.

Prediction of Occupations' Prestige and Earnings From Belief in Gender-Stereotypic Attributes' Importance to Occupational Success
Occupations gained prestige to the extent that they were thought to require masculine personality or cognitive qualities. Feminine personality and cognitive qualities made only smaller, nonsignificant contributions to prestige. Not surprisingly, occupations lost prestige to the extent that masculine or feminine physical qualities were thought essential for success. Only masculine qualities of personality were significantly related to estimated and actual earnings. Occupations were thus well paid to the extent that they were believed to require masculine qualities of personality for success.

Occupations' sex distributions related to earnings but not to prestige. Consistent with prior research showing that occupations' prestige bears an uncertain relation to their sex ratios (see Jacobs, 1989), the four variables representing the distribution of the sexes into occupations were unrelated to prestige. However, occupations' estimated and actual earnings showed consistently negative relationships to the four sex-distributional variables, indicating that male dominance of occupations was associated with higher wages.

DISCUSSION

This research demonstrates that gendered mental images of occupations correspond to the sex segregation of the occupations and, moreover, that high prestige and wages are associated with masculine images. Our research participants thus thought that, to the extent that occupations are female dominated, success in them requires feminine qualities of personality and/or feminine physical qualities; they similarly thought that, to the extent that occupations are male dominated, success in them requires masculine qualities of personality and/or masculine physical qualities. That prediction of occupations' sex distributions on the basis of gender-stereotypic qualities was very successful, as indexed by relatively high multiple correlation coefficients, and shows how profoundly the occupational structure has been shaped by gender.

It is highly informative to examine which stereotypic dimensions best predicted the distribution of the sexes into occupations. The strongest predictor in all four of the analyses on the variables assessing occupations' sex ratios was feminine qualities of personality-the niceness-nurturance cluster of traits that appear as stereotypic of women in virtually all studies of gender stereotypes (e.g., De Lisi & Soundranayagam, 1990; J. E. Williams & Best, 1982; see review by Ashmore, Del Boca, & Wohlers, 1986). Thus, success in female-dominated (vs. male-dominated) occupations was associated considerably more with qualities such as being gentle, nurturing, helpful to others, sociable, kind, cooperative, and supportive. Masculine personality attributes were important as well: Success in male-dominated occupations was associated with masculine qualities of personality (e.g., being competitive, dominant, aggressive) to a greater extent than was success in female-dominated occupations.

The prediction from stereotypic physical attributes to the distribution of the sexes into occupations is also notable. Success in male-dominated (vs. female-dominated) occupations was more strongly associated with masculine physical attributes (e.g., muscular, physically vigorous), whereas success in female-dominated (vs. male-dominated) occupations was more strongly associated with feminine physical attributes (e.g., pretty, cute, petite). These two classes of physical characteristics were reliably associated with occupations' sex distributions despite the lesser overall importance ascribed to physical attributes as determinants of occupational success. Specifically, physical attributes were thought to be generally rather unimportant to occupational success in comparison with personality attributes and cognitive attributes. Nonetheless, masculine and feminine physical attributes related in the predicted manner to the distribution of women and men into occupations, with the exception of the actual sex difference in attraction, which was computed as a difference between the male and the female participants' self-reported attraction to the occupations. This failure of physical attributes to predict the actual sex difference in attraction is not surprising, because our participants were university students-most of whom aspire to professional and managerial occupations-for which physical characteristics are thought to be quite unimportant.

By identifying the types of gender-stereotypic requisites of occupational success that are correlated with sex differences in occupational preferences, the present study goes beyond the simple generalization that each sex's occupational preferences are related to the sex ratio of occupations. The most important predictor of the male and female participants' differential attraction to occupations was belief that success requires feminine qualities of personality, and the belief that success requires masculine personality attributes was the next most important prediction With respect to attraction to occupations on other gender-stereotypic bases, sex similarity prevailed. These findings are generally consistent with Pratto, Stallworth, Sidanius, and Siers's (1997) demonstration that men tend to have occupations that enhance group-based inequality and that women tend to have occupations that attenuate inequality. They showed that interest in hierarchy-enhancing careers was associated with valuing outcomes, such as leadership opportunities and being famous, that are generally compatible with the present study's stereotypically masculine qualities of personality; interest in hierarchy-attenuating careers was associated with valuing outcomes, such as working with people and helping others, that are compatible with the present study's stereotypically feminine qualities of personality. Partially consistent with the present research is Lippa's (1998) argument that women are relatively more attracted to occupations that are oriented toward people and men are relatively more attracted to occupations that are oriented toward things. Although in the present research women's caring and nurturing orientation toward people was reflected in their attraction to occupations believed to require feminine personality attributes, men's more controlling and dominating orientation toward people may have been reflected in their attraction to occupations believed to require masculine personality attributes. Although male-dominated occupations may be in some sense oriented toward things, they may also favor a type of orientation toward people that is different from what typifies female-dominated occupations.

Another noteworthy feature of our findings is the general weakness of prediction of occupations' sex distributions from the cognitive dimensions of gender stereotypes. Judgments of the gender-stereotypic cognitive attributes required for success in occupations failed to show gender-congruent relations to occupations' sex ratios. The absence of such prediction may in part reflect the lesser stereotypicality of the masculine cognitive dimension. Nonetheless, these findings suggest that the sex segregation of employment may not be importantly related to requirements of mathematical and analytical competence in male-dominated occupations, although much discussion of employment segregation has focused on the cognitive abilities that may be necessary for success in male-dominated occupations (see Betz & Fitzgerald, 1987).

Although this study was not designed to test causal relations, the findings are consistent with reasonable assumptions about reciprocal causal links between the sex segregation of occupations and gender stereotyping. As social role theory (Eagly, 1987) argues, people derive their images of women and men from observing their sex-typical work. Both direct and indirect observation (e.g., through the media) provide exposure to women's and men's roles in the private and public spheres. From this perspective, occupational distributions are causes of gender stereotypes. However, the reciprocal causal link is suggested by our findings on actual sex differences in attraction to occupations: Gender-stereotypic images of occupations would foster sex segregation of employment by producing gendered expectations about the occupations appropriate for each sex and influencing women's and men's tendencies to aspire to particular occupations. Relevant to these processes is research showing that people in gender-incongruent occupations tend to experience role conflict (e.g., Luhaorg & Zivian, 1995; C. L. Williams, 1989) and that personal difficulties are thought to plague such people, especially women in male-dominated fields (C. L. Williams, 1989; Yoder & Schleicher, 1996). Other research suggests that gendered occupational stereotypes mediate sex discrimination in hiring decisions (Glick, Zion, & Nelson, 1988). Thus, the causal relationships between occupational sex ratios and gender stereotypes are no doubt bidirectional, and research in various paradigms has demonstrated relevant mediational processes.

Accuracy of Perceptions of Occupations
Whether beliefs about occupations are accurate is important to our research because of the possibility that perceivers might use the sex ratio of occupations as a heuristic to suggest the extent to which job holders need to display masculine or feminine qualities to be successful. Consistent with this view, Krefting, Berger, and Wallace (1978) argued that the sex ratio of occupations and not actual job content is the most important predictor of beliefs that occupations require feminine or masculine qualities. However, less congenial to this interpretation is the considerable variability in the attributes ascribed to male-dominated and female-dominated occupations and the overall perceived importance of masculine cognitive characteristics to occupational success, even in female-dominated occupations. Thus, participants did not in simple fashion assume that, regardless of job content, female-dominated occupations require a full range of feminine qualities and that male-dominated occupations require a full range of masculine qualities. Had participants used sex ratios as a simple heuristic, the strength of the prediction of the occupational division of labor from our six gender-stereotypic dimensions should have mirrored the relative stereotypicality of the dimensions. Although it is more likely that perceivers have at least a somewhat accurate understanding of the qualities required for success in differing occupations, our study does not yield a formal test of the accuracy of perceivers' beliefs about the qualities required for success in occupations.

One aspect of the accuracy of perceiving occupations can be examined empirically by this study-namely, perceptions of occupations' sex ratio and earnings. Participants' estimates of the distribution of men and women into each occupation were thus related to the actual distributions reported in census data, producing a correlation consistent with overall accuracy, r(78) = .91, p < .001. Nonetheless, confirming research by McCauley and his colleagues (McCauley & Thangavelu, 1991; McCanley, Thangavelu, & Rozin, 1988), participants' errors reflected systematic underestimates of the extent to which male-dominated and female-dominated occupations were segregated. The percentage of women in the female-dominated occupations (i.e., 75% or more female) was underestimated by 9.30%, and the percentage of women in the male-dominated occupations (i.e., 75% or more male) was overestimated by 17.14%. Although this finding fits other research showing that stereotypes, including gender stereotypes, often underestimate group differences (see Eagly & Diekman, 1997; McCauley, 1995; Swim, 1994), in the present study the underestimation of segregation may reflect a more general contraction bias against extreme judgments of magnitude (Poulton, 1989). This interpretation becomes especially plausible in view of the additional tendency we obtained for participants to underestimate high wages and overestimate low wages, despite overall accuracy of participants' estimates of occupations' earnings as shown by the correlation of the estimated and the actual wage variables, r(78) = .94.

Improving Women's Status by Reducing the Sex Segregation of Employment
This research followed from our assumption that the division of labor between men's work and women's work is the key to understanding the causes of women's subordination (e.g., Lerner, 1986). Women are positioned in the social structure to have greater domestic responsibility than men regardless of their employment status (e.g., Blair & Lichter, 1991; Shelton, 1992). Also very important, as illustrated by this research, is the tendency of women and men to be concentrated in different types of occupations. Regardless of the accuracy of perceivers' beliefs about occupations' demands, success in female-dominated occupations (compared with male-dominated occupations) is thought to follow relatively more from feminine qualities of personality and from feminine physical attributes; success in male-dominated occupations (compared with female-dominated occupations) is thought to follow relatively more from masculine qualifies of personality and masculine physical attributes. Perceivers thus assume a degree of fit between the sex of job holders and the requirements of occupations, and this gender congruence would tend to foster sex segregation. Nonetheless, they also perceive some commonality in the qualities occupations require regardless of their sex ratios. Cognitive skills, both masculine and feminine, were generally important, although feminine qualities of personality were also very important for female-dominated occupations.

Our findings on occupations' prestige and earnings suggest that movement by women into stereotypically masculine roles is a possible route to raising women's status in society. The critical findings are that earnings related positively to the belief that occupations require masculine personality characteristics, and prestige related positively to the beliefs that occupations require masculine personality characteristics and cognitive characteristics. However, the male-stereotypic qualities associated with highly paid, prestigious occupations can discourage women's entry into such occupations, because competence and comfort in such occupations might require that women view themselves as possessing a greater measure of masculine qualities. Yet, if large enough numbers of women enter an occupation that it becomes female dominated rather than male dominated, its wages might lag because of the wage penalty associated with feminized occupations (Anker, 1997), and its image might be redefined along more feminine lines. However, any such redefinitions probably occur very slowly, as suggested by the small magnitude of the shifts that have occurred in the qualities thought to be required for success in the managerial role (e.g., Brenner, Tomkiewicz, & Schein, 1989; Frank, 1988; Russell, Rush, & Herd, 1988), which has become far less male dominated. Women thus face the daunting reality of a social structure in which high wages and prestige are associated with occupations that are thought to require masculine personal characteristics.