Sex role conflict, social desirability, and eating-disorder attitudes and behaviors

Sex role conflict, social desirability, and eating-disorder attitudes and behaviors

Mark E. Johnson

Anorexia and Bulimia are often referred to as modern disorders that are the byproducts of modern societal values. The disorders, however, have been evident for centuries. The first reports of anorexia date back to 1689, and bulimia was identified as such in 1873 (Nagel & Jones, 1992). However, reported cases of these disorders remained relatively rare until the 20th century. Since the 1960s, however, the prevalence of eating disorders has been increasing rapidly in the United States (e.g., Pyle, Halvorson, Newman, & Mitchell, 1986). Currently the national chapter of Anorexia and Related Eating Disorders (ANAD) estimates that a minimum of 20% of all American women between the ages of 16 and 30 can be diagnosed with a major eating disorder (Nagel & Jones, 1992).

Perhaps the most important questions regarding eating disorders involve identifying the potential contributing or predisposing factors in their development. Many explanatory theories exist, ranging from family-of-origin dynamics to genetic predispositions or biological factors, to societal values about weight and women, to sex role orientation and attitudes. Much evidence exists to suggest that societal values affect how women perceive themselves and how they may respond behaviorally. In the same time span during which reported cases of anorexia and bulimia began to increase dramatically, there was an increase in diet-related articles and advertisements in women’s magazines (Wiseman, Gray, Mosiman, & Ahrens, 1992). In fact, women’s magazines promote weight control 10 times more often than men’s magazines do, both in terms of number of articles devoted to the topic and of advertisements displayed in the magazine (Anderson & DiDomenico, 1992).

Similar developments can be noted in the body shape of women who are societally identified as physical role models for all women. Specifically, contestants for the Miss America pageant and centerfolds in Playboy magazine have reported weights that are 10 to 13% below the ideal weight for their height and body build. There has been a significant decrease in the weight of Miss America contestants since 1979 (Wiseman et al., 1992). Much related to physical social pressures are more indirect, yet equivalent, social values about women’s roles: Sex role orientation and social values are likely to be related to how a women perceives herself and how she responds behaviorally.

Findings from studies conducted in the United States exploring the relationship between sex role orientation and sex role attitudes and eating behavior have been less than conclusive with regard to how sex role orientation may relate to eating disorders. However, some trends are beginning to emerge. The primary sex role orientations identified in the literature (largely based on Bem’s work in the 1970s) are femininity, masculinity, androgyny, and undifferentiatedness.


Femininity has been found to receive higher endorsement among women with bulimia than among women with no eating disorders (Holleran, Pascale, & Fraley, 1988; Scott, 1987; Szymanski & Chrisler, 1990). Some researchers have therefore concluded that hyperfemininity is a trait among women with eating disorders (Steiger, Fraenkel, & Leichner, 1989). Those findings appear to hold true particularly if the endorsed feminine traits were identified as negative (Paxton & Sculthorpe, 1991). The findings, nevertheless, do not always hold. Some investigators have found that although women with eating disorders rate femininity as a higher ideal than those without eating disorders do, they did not differ with regard to actual self-descriptions of their own level of femininity (Pettinati, Franks, Wade, & Kogan, 1987). In fact, some studies have reported either no relationship between femininity and eating behavior (Timko, Striegel-Moore, Silberstein, & Rodin, 1987) or equal levels of endorsed femininity for women with and without eating disorders (Sitnick & Katz, 1984). One study even identified a sample of women in which the normal controls endorsed higher levels of femininity than women with bulimia did, clearly challenging the hyperfemininity hypothesis for bulimia (Lewis & Johnson, 1985).

It has also been found that more feminine role ideals held by fathers for their daughters are related to higher levels of eating-disorder symptoms in their daughters (Silverstein, Perdue, Wolf, & Pizzolo, 1988). Similarly, beliefs in traditional sex role ideology and higher valuation of feminine traits among fathers was related to increased numbers of symptoms of eating disorders among their daughters (Silverstein et al., 1988).


Research regarding the relationship between masculinity and eating behavior has yielded similarly conflicting results. Several studies revealed not only a relationship between low levels of masculinity and eating disorders in general (Sitnick & Katz, 1984), but also low masculinity among persons with anorexia and persons with bulimia as separate groups (Holleran et al., 1988). Those findings have been explained by the theory that masculinity has generally been associated with greater self-confidence, self-esteem, and independence and thus may have a protective role against emotional difficulties (Bem, 1972, as cited in Sitnick & Katz, 1984; Brems, in press).

Conversely, however, masculinity has also been identified as being highly positively related to the drive for thinness (Cantrell & Ellis, 1991), and high masculinity was also found to be related to symptoms of eating disorders for women with high needs for social desirability (Barnett, 1986). Similarly, emphasis on socially desirable masculine traits has been found to be related to symptoms of eating disorders (Timko et al., 1987). Researchers have argued that there is a vulnerability for eating-disorder symptomatology where there is a lack of development of masculine traits, traits that appear essential for women’s survival in today’s society. Additionally, Paxton and Sculthore (1991) found that greater discrepancy between ideal masculinity and actual self-rated masculinity was related to symptoms of eating disorders.

Although Pettinati et al. (1987) found no relationship between masculinity and eating disorders, they did find that lower masculinity was associated with increased depression. Possibly, eating behavior is affected by masculinity through that indirect relationship.


Like masculinity, androgyny has been found to be both positively and negatively related to eating disorders. It was found to be positively correlated with a drive for thinness (Cantrell & Ellis, 1991) and perfectionism as measured by the Eating Disorders Inventory (EDI; Szymanski & Chrisler, 1990). However, in another study it was identified to be negatively correlated with bulimia (Thornton, Leo, & Alberg, 1991).


Undifferentiatedness was not assessed directly in all studies that examined the relationship between sex role orientation and eating disorders, but it has been mentioned as a possible explanation for the inconsistent patterns across different studies. One study that did specifically examine undifferentiatedness found that persons with bulimia were significantly more likely to be classified as undifferentiated than controls without eating disorders were (Lewis & Johnson, 1985). It has been argued that the absence of sex role orientation (i.e., undifferentiatedness) is related to eating disorders because such absence of self-definition is likely related to low self-esteem, making undifferentiated women more vulnerable to developing eating-disorder symptoms (Lewis & Johnson, 1985).


In addition to the findings discussed, some researchers have been unable to reveal any sex role orientation or sex role ideology differences between women with and without eating disorders (Beren & Chrisler, 1990; Scott, 1987; Xinaris & Boland, 1989). Others have argued that any type of gender-role typing (whether masculine or feminine) is related to eating disorder (Thornton et al., 1991). Some of those researchers are adherents of the hyperfemininity hypothesis, which suggests that eating disorders are particularly prevalent among feminine-typed women (Steiger et al., 1989). They also have cited research that reveals that bulimics are less liberated, or more traditional, with regard to sex role attitudes and behaviors than normal controls are (Rost, Neuhaus, & Florin, 1982). Other researchers support the theory that nontraditional sex role orientation, that is, high masculinity in women, is related to eating problems (Silverstein, Carpman, Perlick, & Perdue, 1990). The hypothesis of high sex role typing of any sort (traditional or non-traditional) stands in contrast to findings that have suggested that the lack of a sex role identification for oneself (i.e., undifferentiatedness) is the most likely predisposing factor.

Clearly, research has not been successful in clarifying the relationship between sex role orientation and eating disorders. At best, it can be concluded that sex role orientation does appear to have some sort of relationship with eating disorders, but that the relationship may differ from woman to woman. Some of the most likely underlying factors that may serve to explain the research discrepancies revealed thus far may be social desirability, absence versus presence of sex role orientation, and conflict between sex role ideology (i.e., attitudes about sex role) and sex role orientation (i.e., perception of own sex role).

For some women, femininity may be a mediator of eating disturbance if they perceive themselves as feminine but view femininity as undesirable from a sex role ideology perspective. For other women, masculinity may be the mediator if that is the trait that is perceived as desirable but not descriptive of the self. Alternatively, the absence of a defined self-perceived sex role may be most disturbing and thus related to eating disorders. It also may be that social desirability, rather than sex role-related variables per se, is the mediating variable. Finally, it must be noted that many studies investigated only bulimia symptoms, only anorexia symptoms, or eating disorder symptoms without specifying whether a woman displayed more bulimic or more anorexic symptoms. It is entirely possible that sex role orientation has a different relationship with bulimia than it does with anorexia.

The current study was designed to further investigate the relationship between eating disorders and sex role orientation by taking into consideration previously unstudied factors that may affect the relationship. Specifically, we examined the possibility that it is conflict between a woman’s expressed sex role identity and her actual sex role orientation that is related to eating-disorder attitudes and behaviors, rather than sex role orientation per se. Further, we considered whether a woman’s level of expressed social desirability is related to symptoms of eating disorders. In addition, we also explored the relationship between sex role orientation and eating-disorder symptoms as it has been studied traditionally, in an effort to compare our results with previous findings.



Participants were 321 female college students from two major U.S. universities, one in the Northwest, the other in the Midwest. The participants ranged in age from 18 to 52 years (M= 24.66, SD = 7.58, Mdn = 21). Ethnicity was as follows: 2.9%, African American; 2.4%, Asian American; 2.6%; Hispanic; 6.6%, Native American; 82.9%, White; and 2.6%, other. The breakdown of student’s year in college was: 27.6%, freshman; 15.5%, sophomore; 23.5%, junior; 26.9, senior; and 6.5%, graduate.


Sex Role Ideology Scale (SRI; Kalin & Tilby, 1978). The SRI was designed to measure sex role ideology on a continuum ranging from traditional to feminist beliefs about sex role. The scale requires respondents to evaluate 30 items on a 7-point scale ranging from strongly disagree (1) to strongly agree (7). A total score is calculated by summing all responses, with higher scores indicating increasingly feminist ideology. Two studies have demonstrated that the scale has construct validity, as well as internal consistency (coefficient alpha = .79) and stability over time (test-retest = .87; Kalin & Tilby, 1978).

Bem Sex Role Inventory (BSRI; Bem, 1974). The BSRI was developed to measure sex role orientation. It consists of 60 adjectives and requires respondents to indicate how well each item describes them on a 7-point scale ranging from never or almost never true (1) to always or almost always true (7). Three subscale scores – masculinity, femininity, and social desirability – are obtained by averaging the ratings on the 20 items contained in each scale.

Bem (1974) originally reported high internal consistency scores for the scales on the BSRI (femininity, .80; masculinity, .86; androgyny, .86). Similarly acceptable figures have been reported since, ranging from .75 to .78 for femininity and .87 for masculinity (Lippa, 1986). Test-retest reliabilities over a four-week interval were equally high (femininity, r = .90; masculinity, r = .90; androgyny, r = .93). Construct validity has been reported as adequate in a literature review by Brems and Johnson (1990).

Eating Disorder Inventory (EDI; Garner, Olmsted, & Polivy, 1983; Garner & Olmsted, 1984). The EDI was designed to assess psychological and behavioral traits common in anorexia nervosa and bulimia nervosa. Respondents are asked to indicate to what degree the attitudes, feelings, or behaviors expressed by each of the 64 items apply to them, by rating each item on a 6-point Likert-type scale that ranges from always (1) to never (6).

Eight subscales were derived from the EDI by adding the weighted ratings (weights range from 1 to 3) of those items contained within each scale. Higher scores point toward higher degrees of pathology. The scales, labeled in a self-explanatory fashion, are Bulimia (bingeing and purging; 7 items), Body Dissatisfaction (disturbed body image; 9 items), Drive for Thinness (overconcern with weight control; 7 items), Ineffectiveness (inadequacy, poor self-concept; 10 items), Perfectionism (high expectations for achievement; 6 items), Interpersonal Distrust (difficulty forming close emotional relationships; 7 items), Interoceptive Awareness (inability to identify and label emotions and sensations related to hunger or satiation; 10 items), and Maturity Fears (avoidance of adult responsibility; 8 items). Reliability coefficients for the scales range from .58 to .86; validity is reportedly adequate (Garner & Olmsted, 1984).


After giving informed consent, respondents completed a biographical data sheet inquiring about educational level, ethnicity, marital status, and age. Then they completed the EDI, SRI, and BSRI. The order of the three instruments was varied randomly across participants to control for testing effects.

Design and Statistical Analyses

Preliminary analyses. Participants were classified according to sex role ideology (feminist or traditional), masculinity (high or low), and social desirability (high or low). For that purpose, medians were calculated for the SRI and the BSRI masculinity subscale, and participants were then categorized according to median-split procedures. Once categorized along ideology and masculinity, participants were classified into three groups, according to whether they experienced conflict between those two dimensions. One group was classified as consistent (low in conflict); they were feminist in ideology and high in masculinity, or traditional in ideology and low in masculinity. The second group was classified as discrepant (i.e., high in conflict) and consisted of those who were traditional in ideology but high in masculinity. The third group consisted of participants who were feminist in ideology but low in masculinity; it was classified as discrepant.

Primary analyses. Two separate multivariate analyses of variance (MANOVAs) were calculated. The first was a 2 (social desirability) x 3 (sex role conflict) MANOVA, with the dependent variables being the eight subscales of the EDI. The second MANOVA was calculated to provide comparability of findings from the present study with previous studies. That analysis was a 2 (social desirability) x 4 (sex role orientation) MANOVA, using the eight EDI subscales as dependent variables.


Preliminary Analyses

The median for sex role ideology was 106. Classification resulted in 158 (49.2%) feminist and 151 (47%) traditional participants. The median for masculinity was 4.85, and classification resulted in 140 (51.5%) with high masculinity and 132 (48.5%) with low masculinity. The median for social desirability was 4.7, and classification resulted in 152 (47.4%) participants with low expressed levels of social desirability and 145 (45.2%) with high levels. Conflict classification resulted in 159 (58.5%) participants who were consistent in their expressed sex role ideology and sex role orientation, 59 (21.7%) with a traditional ideology/high masculinity discrepancy, and 54 (19.9%) with a feminist ideology/low masculinity discrepancy. For classification of participants according to sex role orientation, the medians for the BSRI scales were 5.15 for femininity and 4.85 for masculinity. Classification resulted in 70 masculine and 70 feminine participants, as well as 68 undifferentiated and 78 androgynous participants. For each median split, participants who fell exactly at the median were unable to be classified and were therefore dropped. As a result, the final pool of participants was 272 for the primary analyses and 268 for the secondary analyses.

Primary Analyses

Results of the first MANOVA revealed main effects for social desirability and conflict, F(16, 518) = 2.38, p [less than] .005, and F(8, 247) = 2.48, p [less than] .05, respectively; but no significant interaction between the two variables, F(24, 755) = 0.83, p [less than] .05. For the social desirability main effect, univariate analyses revealed significance for five of the eight subscales, namely, for Drive for Thinness, F(1, 260) = 6.29, p [less than] .01; for Bulimia, F(1,260) = 7.94, p [less than] .005; for Body Dissatisfaction, F(1, 260) = 6.05, p [less than] .01; for Perfectionism, F(1, 260) = 4.17, p [less than] .05; and for Interoceptive Awareness, F(1, 254) = 7.68, p [less than] .005. For all five scales, participants endorsing higher levels of social desirability also endorsed higher levels of eating disorder symptoms. Relevant means and standard deviations are displayed in Table 1.

For the conflict main effect, univariate analyses revealed significant differences on two variables, ineffectiveness and distrust, F(2, 266) = 10.63, p [less than] .001, and F(2, 266) = 3.04, p [less than] .05, respectively. Analysis of the means revealed that on both variables, the low masculinity/feminist group scored the highest (M = 27.14, M = 18.07), followed by the nondiscrepant group (M = 24.47, M = 19.77), and the high masculinity/traditional group (M = 21.62, M = 17.27). That pattern suggests that it may be more the level of masculinity than the presence of discrepancy between masculinity levels and ideology that is related to EDI scores. To test that possibility, correlations were calculated between masculinity and EDI scores; they confirmed a pattern of significant negative relationships between four EDI variables – ineffectiveness, perfectionism, distrust, and maturity – and masculinity, r(320) = -.317, p [less than] .001, r(320) = -.256, p [less than] .001, r(320) = -.183, p [less than] .001, and r(320) = -.148, p [less than] .05, respectively.


In considering these findings, it is important to consider three limitations. First, this study, as well as all of the studies cited in the literature review, included only women in the United States. Thus, results are not necessarily generalizable to women from other countries without further investigation. Second, because of the preponderance of White participants (82.9%), it was not possible to examine data for different ethnic groups separately. Because it is possible that ethnicity may influence the variables under investigation, future research could be designed to allow for such between-group comparisons. Next, the study relied strictly upon self-report inventories to assess social desirability, sex role orientation, and eating-disorder attitudes and behaviors. Finally, the study included the entire range of women found on a college campus rather than just women identified as having eating disorders. Although this allowed for an investigation of the relationships in question across the entire college population, it is not necessarily conclusive regarding women diagnosed as having an eating disorder.


Anderson, A. E., & DiDomenico, L. (1992). Diet versus shape content of popular male and female magazines: A dose-response relationship to the incidence of eating disorders? International Journal of Eating Disorders, 11, 283-287.

Barnett, L. R. (1986). Bulimarexia as a symptom of sex role strain in professional women. Psychotherapy, 23, 311-315.

Bassoff, E. S., & Glass, G. V. (1982). The relationship between sex roles and mental health: A meta-analysis of twenty-six studies. The Counseling Psychologist, 10, 105-112.

Bem, S. (1974). The measurement of psychological androgyny. Journal of Consulting and Clinical Psychology, 12, 155-162.

Beren, S. E., & Chrislet, J. C. (1990). Gender role, need for approval, childishness, and self-esteem: Markers of disordered eating? Research and Communications in Psychology, Psychiatry, and Behavior, 15, 183-191.

Brems, C. (1995). Women and depression: A comprehensive analysis. In E. E. Beckham & W. Leber (Eds.), Handbook of depression (2nd ed. pp. 539-566). New York: Guilford.

Brems, C., & Johnson, M. E. (1990). Reexamination of the Bem Sex Role Inventory: The Interpersonal BSRI. Journal of Personality Assessment, 55, 484-498.

Cantrell, P. J., & Ellis, J. B. (1991). Gender role and risk patterns for eating disorders in men and women. Journal of Clinical Psychology, 47, 53-57.

Garner, D. M., Olmsted, M.P., & Polivy, J. (1983). Development and validation of multidimensional eating disorder inventory for anorexia nervosa and bulimia. International Journal of Eating Disorders, 2, 15-34.

Garner, D. M., & Olmsted, M. P. (1984). Eating Disorder Inventory. Odessa, FL: Psychological Assessment Resources.

Holleran, P. R., Pascale, J., & Fraley, J. (1988). Personality correlates of college age bulimics. Journal of Counseling and Development, 66, 378-381.

Kalin, R., & Tilby, P. J. (1978). Sex role Ideology Scale. Psychological Reports, 42, 732-737.

Lewis, L. D., & Johnson, C. (1985). A comparison of sex role orientation between women with bulimia and normal controls. International Journal of Eating Disorders, 4, 247-257.

Nagel, K. L., & Jones, K. H. (1992). Sociological factors in the development of eating disorders. Adolescence, 27, 107-113.

Paxton, S. J., & Sculthorpe, A. (1991). Disordered eating and sex role characteristics in young women: Implications for sociocultural theories of disturbed eating. Sex Roles, 24, 587-598.

Pettinati, H. M., Franks, V., Wade, J. H, & Kogan, L. G. (1987). Distinguishing the role of eating disturbance from depression in the sex role self-perceptions of anorexic and bulimic inpatients. Journal of Abnormal Psychology, 96, 280-282.

Pyle, R. L., Halvorson, P. A., Newman, P. A., & Mitchell, J. E. (1986). The increasing prevalence of bulimia in freshman college students. International Journal of Eating Disorders, 5, 631-647.

Reeves, P. C., & Johnson, M. E. (1992). Relationship between family-of-origin functioning and self-perceived correlates of eating disorders among female college students. Journal of College Student Development, 33, 44-49.

Rost, W., Neuhaus, M., & Florin, I. (1982). Bulimia nervosa: Sex role attitude, sex role behavior, and sex role related locus of control in bulimarexic women. Journal of Psychosomatic Research, 26, 403-408.

Scott, D. W. (1987). The involvement of psychosexual factors in the causation of eating disorders: Time for a reappraisal. International Journal of Eating Disorders, 6, 199-213.

Silverstein, B., Carpman, S., Perlick, D., & Perdue, L. (1990). Nontraditional sex role aspirations, gender identity conflict, and disordered eating among college women. Sex Roles, 23, 687-695.

Silverstein, B., Perdue, L., Wolf, C., & Pizzolo, C., (1988). Bingeing, purging, and estimates of parental attitudes regarding female achievement. Sex Roles, 19, 723-733.

Sitnick, T., & Katz, J. L. (1984). Sex role identity and anorexia nervosa. International Journal of Eating Disorders, 3, 81-87.

Steiger, H., Fraenkel, L., & Leichner, P. P. (1989). Relationship of body image distortion to sex role identifications, irrational cognitions, and body weight in eating disordered females. Journal of Clinical Psychology, 45, 61-65.

Szymanski, L. A., & Chrisler, J. C. (1990). Eating disorders, gender-role, and athletic activity. Psychology: A Journal of Human Behavior, 27, 20-29.

Thornton, B., Leo, R., & Alberg, K. (1991). Gender role typing, the superwoman ideal, and the potential for eating disorders. Sex Roles, 25, 469-484.

Timko, C., Striegel-Moore, R. H., Silberstein, L. R., & Rodin, J. (1987). Femininity/masculinity and disordered eating in women: How are they related. International Journal of Eating Disorders, 6, 701-712.

Wiseman, C. V., Gray, J. J., Mosiman, J. E., & Ahrens, A. H. (1992). Cultural expectations of thinness in women: An update. International Journal of Eating Disorders, 11, 85-89.

Xinaris, S., & Boland, F. J. (1989). Disordered eating in relation to tobacco use, alcohol consumptions, self-control, and sex role ideology. International Journal of Eating Disorders, 9, 425-433.

COPYRIGHT 1996 Heldref Publications

COPYRIGHT 2004 Gale Group