gender differences in idiosyncratic sex-typed self-images and self-esteem

26
Sex Roles, Vol. 29, Nos. 3/4, 1993 Gender Differences in Idiosyncratic Sex-Typed Self-Images and Self-Esteem 1 Emda Orr and Edna Ben-Eiiahu Ben-Gurion University of the Negev The present study focused on the relationships between idiosyncratic sextyping and self-esteem. Idiosyncratic, feminine, masculine, and nonstereotyped sets of self-attributes, termed self-images (SI), rather than conventional masculinity, femininity and their interaction (androgyny), were used. Findings from two studies with two groups of 337 Israeli early adolescents boys and girls, and 280 male and female university students, using two instruments of sex-typed self-attributes (newly devised Self-Related Self-Image Scale and Spence Personal Attributes Questionnaire) and based on idiosyncratically defined sex images, supported our hypotheses. The findings showed that males' self-esteem was predicted significantly by their masculine SI, that females' self-esteem was predicted significantly by their nonstereotyped SL and that the feminine SI predicted significantly the self-esteem of neither. Findings based on conventional sex role orientation scores supported former findings of significant relationships between masculinity and male and female self-esteem. The results showed that the self-esteem of male students, but not that of boys, was significantly predicted by their nonstereotyped SI too. Partial support was found for the hypothesis that more females than males are likely to redefine conventional masculine attributes as nonstereotyped. The whole set of findings were interpreted as indicating that self-esteem is dependent on socially rewarded and self-synchronized sex-typed attributes. The self-system is defined here as an internal, organized, multidimensional system of constructs whose content is the specific representations of our unique identities, special abilities, achievements, preferences, appearances, and temperament, (Markus, 1977; Marsh, 1989; Rosenberg, 1965). These representations are organized as individual naive theory about our own ex- 1Study 1 is based on the master's thesis submitted to Ben-Gurion University by the second author under the advise of the first one. 271 0360-00?5/93/0800-0271507.00/0© 1993 Plenum Publishing Corporation

Upload: emda-orr

Post on 12-Aug-2016

217 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Sex Roles, Vol. 29, Nos. 3/4, 1993

Gender Differences in Idiosyncratic Sex-Typed Self-Images and Self-Esteem 1

Emda Orr and Edna Ben-Eiiahu Ben-Gurion University of the Negev

The present study focused on the relationships between idiosyncratic sextyping and self-esteem. Idiosyncratic, feminine, masculine, and nonstereotyped sets of self-attributes, termed self-images (SI), rather than conventional masculinity, femininity and their interaction (androgyny), were used. Findings from two studies with two groups of 337 Israeli early adolescents boys and girls, and 280 male and female university students, using two instruments of sex-typed self-attributes (newly devised Self-Related Self-Image Scale and Spence Personal Attributes Questionnaire) and based on idiosyncratically defined sex images, supported our hypotheses. The findings showed that males' self-esteem was predicted significantly by their masculine SI, that females' self-esteem was predicted significantly by their nonstereotyped SL and that the feminine SI predicted significantly the self-esteem of neither. Findings based on conventional sex role orientation scores supported former findings of significant relationships between masculinity and male and female self-esteem. The results showed that the self-esteem of male students, but not that of boys, was significantly predicted by their nonstereotyped SI too. Partial support was found for the hypothesis that more females than males are likely to redefine conventional masculine attributes as nonstereotyped. The whole set of findings were interpreted as indicating that self-esteem is dependent on socially rewarded and self-synchronized sex-typed attributes.

The self-system is defined here as an internal, organized, multidimensional system of constructs whose content is the specific representations of our unique identities, special abilities, achievements, preferences, appearances, and temperament, (Markus, 1977; Marsh, 1989; Rosenberg, 1965). These representations are organized as individual naive theory about our own ex-

1Study 1 is based on the master ' s thesis submit ted to Ben-Gurion University by the second author under the advise of the first one.

271

0360-00?5/93/0800-0271507.00/0 © 1993 Plenum Publishing Corporation

Page 2: Gender differences in idiosyncratic sex-typed self-images and self-esteem

272 Orr and Ben-Eliahu

perience and functioning (Epstein, 1973). According to Epstein, the self- theory has two functions: hedonistic--that is, maintaining a sense of well- being by self-enhancement processes; and instrumental---sustaining realistic predictions of one's personal competence in specific domains. Others (e.g., Markus, 1977) highlight the functional aspects of the self as organizer of information. Findings indicate that the self-esteem affects emotional well- being, instrumental functioning, perception, and memory (Bandura, 1978; Dweck & Elliot, 1983; Eccles, Adler, & Meece, 1984; Markus, Crane, Bern- stein, & Siladi, 1982).

We postulate that only socially rewarded domain-specific self-percep- tions have hedonistic value, and that a major inconsistency between phe- nomenological components in the self is disruptive to self-esteem (e.g., Festinger & Carlsmith, 1959; Kelly, 1955; McClelland, 1961; Rogers, 1963), unless something is done to prevent it. One solution would be to give a low weight to self-perceptions which are discordant to self-esteem. By dis- counting the socially discordant self-perception, a person is able to maintain a realistic self-perception without risk of self-esteem (Pelham, 1991). Hence, the implication is that in order to affect hedonistic goals, self-per- ceptions should be inter-coordinated (consistent) and appraised by the in- dividual as socially desirable and important.

Masculine, Feminine, and Nonstereotyped Self-Images

Theory and common sense suggest that gender-related constructs are important components of the self-system (Alpert-Gillis & Connell, 1989). However, a systematic approach to gender related self-constructs embed- ded in self-system theories are scarce. Current research has been investi- gating the relat ionships between gender-re la ted self-constructs and self-esteem, mostly using the concepts of femininity (or expressivity and compassion), masculinity, (or instrumentality), and androgyny (High levels of both, e.g., Alpert-Gillis & Connell, 1989; Bern, 1981; Markus, Crane, Bernstein, & Siladi, 1982; Marsh & Byrnes, 1991). The problem with these gender-related concepts is that they lack internal reliability as components of the self-system because the self-system is defined as an idiosyncratic phe- nomenological construct (Harter, 1982; James, 1892/1963), whereas femi- ninity (or expressivity and compassion) and masculinity (or instrumentality), as measured by Bern (1974) and Spence (Spence & Helmreich, 1978), are not. The generic definition of these latter attributes as feminine or mas- culine is not determined by the individual but by conventional, normative gender stereotypes, and are labeled sex role orientation. Idiosyncratic stereo- types in Western democratic countries are probably different, at least to

Page 3: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem 273

some extent, from conventional stereotypes (Myers & Gonda, 1982). Idi- osyncratic stereotypes may include a broader or a narrower range of at- tributes compared to conventional stereotypes. Hence, the interpretation of the sex role orientation scores as parameters of the self-system is not accurate. For example, with the sex role orientation parameters (Bem Sex Role Inventory, Bern, 1974; Personal Attribute Questionnaire, Spence, Helmreich, & Stapp, 1974)) a finding of significant relationships between masculinity and self-esteem might have been obtained because persons who perceived themselves as masculine tended to have high self-esteem or be- cause persons who perceived themselves as having specific socially valued attributes tended to have high self-esteem (O'Heren & Orlofsky, 1990).

We suggest the concepts of feminine, masculine, and nonstereotyped self-images, which are defined by the individual as related to gender. To op- erationalize feminine, masculine, and nonstereotyped self-images (Sis), we asked respondents to appraise themselves from a list of attributes and then asked them whether each of these attributes is more characteristic of men than of women, of women than of men, or equally characteristic of both. A masculine SI score was obtained by the mean of self-appraisals of at- tributes judged more characteristic of men than of women. A feminine SI score was the mean of self-appraisals of attributes judged more charac- teristic of women than of men. The nonstereotyped SI score was the mean of self-appraisals of attributes judged equally characteristic of both men and women.

We do not ignore the importance of conventional gender stereotypes. We assume that most people in Western societies know, for instance, that crying easily is considered feminine, and that being ambitious is considered masculine. Nevertheless, findings show that many individuals do not accept this convention, and redefine these attributes as equally typical of male and female (Myers & Gonda, 1982). We suggest that this redefinition has a hedonistic value (self-esteem gain) for females, and therefore females are more likely than males to redefine gender-related attributes as non- stereotyped.

Self-Esteem and the Generic Self-System Model

We propose a model of the self in which the concepts of male and female contain two sets of central gender-related bipolar phenomenological constructs (Kelly, 1955)--"male-female" and "masculine SI-feminine SI." The latter construct comprises the stereotyped SI---which is one pole of another bipolar "stereotyped SI-nonstereotyped SI" construct. According to our model, the set of gender self-images is organized under one pole

Page 4: Gender differences in idiosyncratic sex-typed self-images and self-esteem

274 Orr and Ben-Eiiahu

of the male-female construct, according to one's biological gender. Hence, conventionally stereotyped self-attributes are organized by the individual into three sets: "feminine SI," "masculine SI" (the "stereotyped SI"), and the "nonstereotyped SI." We postulate that this organization functions as a booster. People with higher levels of self-esteem organize their self-im- ages in a way that maximizes self-worth.

We assume that self-esteem is maximal under two conditions: (a) an individual considers specific self-perceptions on self-attributes as assets (i.e., important and socially desirable; Pelham, 1991); and (b) self-perceptions and self-attributes are synchronized (i.e., there is consistency between com- ponents of the self. These assumptions concerning the generic self-system model leads to two distinct male and female self-system patterns. Since masculine, instrumental, individualistic attributes are more highly rewarded than feminine, expressive, other-oriented attributes (Deaux, 1976; Sampson, 1988), they are more likely to be considered assets, and are more likely to contribute to self-esteem. A model should predict, then, that a masculine self-image will enhance male and female self-esteem. However, the above- mentioned principle of synchronization prevents us from endorsing this hy- pothesis. The organization of socially rewarded self-assets as masculine SI is synchronized with male but not with female generic self-perception, and therefore is likely to enhance male but not female self-esteem. Female self- esteem can be achieved, however, by the organization of socially rewarded self-attributes as nonstereotyped SI. By such organization, rewarded self- assets of females are connected to self-esteem, without creating an incon- sistency between gender self-perception and reliance on a masculine self-image for self-esteem.

We share with Bern (1981) the notion that gender-related self-con- cepts are organized and serve personal and interpersonal functions. How- ever, nonstereotyped SI is different from Bem's conceptualization of androgyny. According to Bem (a) individuals differ in the extent of which they tend to perceive their social world in gender-stereotyped concepts, and (b) both females and males cognitive and instrumental behavior is af- fected by this tendency. Bern underscores male's and female's common de- nominators, by suggesting cognitive processes shared by both genders, and by substituting her former feminine vs masculine cognitive schemas with the dichotomy of sex-typed vs. androgynous schemas. We underscore gender differences, and assume that males and females live in different social con- texts, where masculinity is more socially rewarded than femininity. There- fore females but not males self-esteem is threatened. The threat is the inconsistency between one's generic self-perception, and the perception that the cross-gender (masculine) attributes are more highly socially rewarded. Our concept of nonstereotyped SI was constructed, then, as a hypothetical

Page 5: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem 275

solution to this problem. Our operationalization of nonstereotyped SI is also different from Bem's (1981) operationalization of androgyny (a) in that we define diosyncratic rather than conventional stereotypes, and (b) non- stereotyped SI is based on self-perception of attributes redefined as neither masculine nor feminine rather than on high (or low) scores of both mas- culinity and femininity. Hence, we suggest three-dimensional gender related self-images (masculine, feminine, and nonstereotyped). One is likely to en- dorse feminine and masculine Sis, along with nonstereotyped SI, while ac- cording to Bern, the dichotomy "sex-typed--androgynous" is a unidimensional continuum.

There is some indirect support for our hypothesis that deviating from conventional stereotypes does not negatively affect women's emotional well-being (O'Heren & Orlofsky, 1990). The details of this finding, how- ever, are not fully consistent with our hypotheses. Previous results do sup- port our hypotheses that males who appraise themselves highly on masculine sex-typed attributes also tend to report high levels of self-esteem, and that the endorsement of feminine sex-typed attributes is not related or is slightly negatively related to self-esteem (e.g., Antill & Cunningham, 1979). Seemingly contrary to our hypothesis in the majority, but not in all, of the studies, females who appraised themselves highly on masculine sex- typed attributes, tended to report high self-esteem (see Whitley, 1983 and 1988, for review). This inconsistency can be explained, as the above studies used conventional constructs of femininity and masculinity. We predict that the use of idiosyncratic measures will yield support to the present hypothe- ses.

Our last hypothesis deals with female motivation to define attributes as nonstereotyped. We contend that the tendency of females to define at- tributes as nonstereotyped is affected by the discordance between being a female and perceiving oneself as having masculine characteristics. If this is the case, one should predict that females are more likely than males to redefine as nonstereotyped conventional masculine attributes, and that no difference should be found in the likelihood of males and females to re- define as nonstereotyped conventional neutral and feminine attributes. The specific hypotheses tested were as follows: (1) Early adolescent boys' mas- culine SI and girls' nonstereotyped SI are associated with self-perceptions of competence and self-esteem. (2) Masculine SI is not related to girls' self-perception of competence or self-esteem. (3) Feminine SI is not related to either boys or girls self-perceptions of competence or self-esteem. (4) Significantly more girls than boys will define conventional masculine attrib- utes as nonstereotyped and no significant difference will be found between boys and girls in the number of conventional feminine and neutral attrib- utes defined as nonstereotyped. These hypotheses were tested on a popu-

Page 6: Gender differences in idiosyncratic sex-typed self-images and self-esteem

276 Orr and Ben-Eliahu

lation of early adolescents, whose gender differences in self-esteem begin to emerge (Bush, Simmons, Hutchinson, & Blyth, 1977), and for whom gender becomes a central issue (Erikson, 1950; Pleck, 1975).

STUDY 1

Method

Participants

The sample consisted of 337 Israeli children, 84 boys, and 73 girls in the seventh grade, and 95 boys and 85 girls in the eighth grade, from three elementary schools in Beersheva. Since the transition to junior high school has been found more detrimental to girls' than to boys' self-esteem (Sim- mons, Blyth, Van Cleave, & Bush, 1979), students from elementary school----Grades 1-8 who did not experience transition from one school sys- tem to another, were selected. The respondents' mean age was 13.2, ranging from 11.8 to 15 years. The children were Jewish, they came from a wide range of social-economic backgrounds, about half of the grandparents im- migrated to Israel from Islamic countries, and the others had European or American backgrounds. Parents' mean education was 12 years, ranging from some years in elementary school to postgraduate degrees in the uni- versity.

Instruments

Self-Esteem. This was measured by Rosenberg's Self Esteem Scale (Rosenberg, 1965) of ten items tapping from 1 (low) to 4 (high) context free self-appraisals, such as self-satisfaction, self-worth, self-pride, self-re- spect, and other positive context-free attitudes toward one's self. The final score was the mean of total items scores after correction for directionality. The scale has been widely used in both Israel (Yogev & Ilan, 1988; Yogev & Ronen, 1982) and in the United States, and satisfactory psychometric properties were reported (Rosenberg, 1965). The internal consistency in the present study was alpha = .73.

Domain-Specific Self-Perceptions (SP). These were measured by Har- ter's Perceived Competence Scale for Children developed for subjects from the third to the ninth grades (Harter, 1982). The scale consists of 38 items tapping academic competence, interaction with peers, competence in sports and outdoor activities, appearance, conventional behavior (moral), and

Page 7: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem 277

global self-esteem. Items consist of statements such as "Some kids feel they are very good at their school work, but other kids worry about whether they can do the school work assigned to them. Which of the two kinds of kids is more like you?" They were then asked whether their choice was really true, or sort of true, for them. The perception of the lowest self- competence was scored 1 and the highest 4. The scale yielded a profile of six scores, the mean score for each subscale. Factor analYSiS (varimax) con- ducted on the items of the scale yielded five factors similar to Harter's domain specific SPs explaining 36% of the variance. Global self-esteem, however, did not appear as a distinct factor and its items were distributed across the five factors. Therefore, we did not use this scale in further analy- ses. The alpha coefficients of the domain-specific subscales ranged from .64 for conventional behavior to .80 for academic competence, with a mean of .73.

Sex-Related Self-Image Scale (SSIS). Prevalent sex-related attributes scales were developed for the American population. In order to diminish cultural bias, we developed a new two-part scale, based on previous ones. The Self-Appraisal Subscale consisted of 22 items and their antonyms. He- brew translation of the attributes used in previous Israeli and American sex role studies (Bern, 1974; Elingas, 1982; Spence & Helmreich, 1978) were given to a sample of 59 females and 38 male undergraduate psychol- ogy students. They were requested to sort them out as masculine, feminine, or neither. The eight attributes judged as masculine by the larger number of students (e.g., aggressive), the eight judged as feminine (e.g., loyal), and the four judged as neutral (e.g., sincere) were chosen. The items included expressive (e.g., easy to cry; loud), instrumental (e.g., ambitious), and per- son-related (friendly) attributes. We added 2 items that dealt with satisfac- tion with gender. Two items, "self-confident" and "inferior," were deleted from the calculations of Sis because of their confounded meaning with "self-esteem," the dependent variable. The seventh- and eighth-grade re- spondents were requested to evaluate, on a 5-point scale from (A) not at all to (E) very much, how accurately each item described them. The Gender Stereotype Subscale consisted of the same attributes without their anto- nyms, and respondents were requested to decide whether each item was masculine (very characteristic or quite characteristic of men), nonstereotyped (makes no difference), or feminine (quite characteristic or very characteristic of women). Each respondent got three SI scores---feminine, masculine, and nonstereotyped----calculated by the following method: Item scores in the Self-Appraisal Scale were sorted into masculine, feminine, and no differ- ence subscales according to the individual's choices on the Gender Stereo- type Scale. Feminine SI score was calculated by the mean of the items of the Self-Appraisal Scale, defined by the individual in the Gender Stereo-

Page 8: Gender differences in idiosyncratic sex-typed self-images and self-esteem

278 Orr and Ben-EUahu

type Scale as more characteristic of women. In a parallel method, the mas- culine and nonstereotyped SI scores were the mean of Self-Appraisal Scale items, defined as more characteristic to men, or equally characteristic to both men and women (in that order). Test-retest reliabilities, with six weeks between tests, were calculated on an independent group of 36 boys and 35 girls from the seventh (n = 38) and eighth (n = 33) grades, and were rs = .79, .77, .71 (masculine, feminine, and nonstereotyped Sis, re- spectively). No significant correlation was found between girls' three SI scores, and only one out of the boys' intercorrelations was significant (see Table I).

Procedure

The tests were administered in the classroom, in six different order sets, by a graduate psychology student during a 45-minute session. Age, grade, gender, and father's and mother's education were obtained. To avoid order effect on sex-related self-image scores, the Self-Appraisal Scale was administered to half of the respondents first, and to the other half following the Gender Stereotype Scale.

Results and Discuss ion

No significant differences were to be found between ages or grades on the variables of sex-related SI, domain-specific SP, and self-esteem, nei- ther did we find significant differences between results obtained by the dif- ferent orders of administration. Therefore we combined the two grade levels into one group. Fathers' and mothers' education was correlated sig- nificantly with girls' self-esteem and feminine SI. Fathers' education cor- related significantly, also, with girls' masculine SI (Table I). Therefore, in subsequent analyses parental education was controlled for.

Self-Esteem, Domain-Specific Self-Perceptions, and Sex-Related SI

A 2 x 9 (sex, self-components) analysis of variance, mixed model, with parental education as covariates, indicate that boys' and girls' self-esteem, cognitive competence, and social acceptance are not significantly different (Ms = 3.26, 3.20; 3.04, 2.99; 2.92, 2.98, respectively); boys perceived them- selves more physically competent than girls and with a better appearance than girls (Ms = 2.83, 2.62 and 2.89, 2.66; p < .01) and girls perceived themselves higher on the moral domain than boys (Ms = 3.09, 2.82, p <

Page 9: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem 279

i:3

H

0

0

¢,) r../3

.o

0

e.l

0 0

"~" I

t"q

I I I I

" ' I ' I ' I . . . . .

I

I I I I I

~. ~ . ,-'I. m ~. ~ . ~ . ~ . ~ . .

~ . ~ . ~ . ~ . ~ . . ~ . m ~ .

I

"q. ~ . ~ ~. ~. . . . . .

,.~ .o . -

~ , I ! ~ ~'J '~' I~,~ ~ ~ ~ ¢.i, 8 . ~ . ~ . ~ ~

.o

v v v ~

Page 10: Gender differences in idiosyncratic sex-typed self-images and self-esteem

280 Orr and Ben-Ellahu

.001). Boys perceived themselves significantly more masculine than girls (Ms = 3.30, 2.90, p < .001). Girls perceived themselves more feminine than boys (Ms = 3.27, 2.82,p < .001), and no difference was found between their nonstereotyped SI (Ms = 3.32, 3.40).

In support of our hypotheses, boys' masculine SI was significantly cor- related with self-esteem (r = 0.26, p < .001), whereas girls' nonstereotyped SI was significantly correlated with self-esteem (r = 0.36, p < .001). Boys' nonstereotyped and girls' feminine SI were significantly but modestly cor- related with their self-esteem (r = 0.16, 0.14, respectively, p < .05; see Table I). The correlations between domain-specific SP and sex-related SI support our hypotheses even further. Three domains of competence---cog- nitive, social, and physical SPs--were correlated significantly with boys' masculine SI, whereas none correlated significantly with girls' masculine SI. All five domains of competence--cognitive, social, physical, appearance and moral SPs were correlated significantly with girls' nonstereotyped SI, whereas only cognitive SP was significantly (and very modestly) correlated with boys' nonstereotyped SI. Feminine SI was not correlated significantly with any domain of girls' SP, and was negatively correlated with boys' SP of appearance.

In two sets of multiple regression analyses (stepwise with parental education entered first as forced choice), boys' and girls' self-esteem were regressed on all five domain-specific SPs and the three sex-related Sis (Ta- ble II). The findings showed that cognitive, social, moral SPs, and masculine SI accounted significantly for 32% of boys' self-esteem variance, and cog- nitive, social, appearance, moral SPs, and nonstereotyped Sis accounted for 51% of girls' self-esteem variance. Boys' feminine and nonstereotyped and girls' feminine and masculine Sis did not account significantly for self-es- teem. The findings supported the model that predicted boys' masculine and girls' nonstereotyped Sis affect self-esteem independently of other domain- specific self-perceptions. The extent of prediction was significant but quite modest. The relationship of the two Sis with the cognitive and physical SPs were stronger, especially for girls (see Table II).

Our hypothesis was that in order to avoid self-inconsistency, girls are likely to define the socially rewarded instrumental-conventionally masculine attributes as nonstereotyped, and thus gaining in self-esteem without risking inconsistency between their generic self-perception as females and their self-image as masculine. In order to support this hypothesis, the finding should show gender differences in the distribution of idiosyncratic defini- tions. That is (1) more boys than girls define conventional masculine attrib- utes as masculine, (2) more girls than boys define them as nonstereotyped, and (3) that conventional feminine and neutral attributes are judged simi- larly by both. As Table III shows, the results tend to support our hypothe-

Page 11: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem

Table II. Multiple Correlation (Stepwise) Between SSSI, Domain-Specific SPs, and Self-Esteem

281

R 2

Criterion Source I]d R 2 c h a n g e F df

Boys Self-esteem

Masculine SI

Girls Self-esteem

Nonstereotyped SI

Cognitive SP e .25 c .18 .18 c 35.8 c Social SP .29 c .28 .10 c 31.33 c

Masculine SIe .16 a .30 .02 a 18.79 c

Moral SP .17 a .32 .02 a 18.79 c

Cognitive SP .20 b .04 .04 b 6.95 b

Cognitive SP .29 c .29 .29 c 58.1 c Social .26. c .40 .11 c 47.13 c Appearance SP .26 b .46 .06 b 41.13 c

Nonstereotyped SI .16 a .49 .03 a 34.73 c Moral SP .16 a .51 .02 a 29.15 c

Physical SP .23 b .09 .09 ~ 15.17/' Cognitive SP .21 b .13 .04 b 11.26 c

4,174

1.177

5,150

2,153

< .05. P < .01.

V .001. alues in the final equation.

eSp: self-perception; SI: self-image.

ses, but not significantly so. Forty-two percent of girls in comparison with 52.5% of boys defined conventional masculine attributes as such, and 43.3% of girls redefined them as nonstereotyped in comparison with 36.9% of boys. The same trend was not found in conventional feminine and neutral attributes.

Although three out of our four hypotheses were supported in Study 1, it was not clear how similar the sex-related self-images are to adults in this early stage of adolescence. Moreover, we claim that our new self-system model is different from former sex role orientation models because it re- places variables based on conventional femininity and masculinity by the idiosyncratic variables of masculine and feminine SI. This claim has not been tested as yet. One could argue, also, that some peculiarities in the Israeli sample affected the findings. Therefore a replication of the study with an instrument with published records has to be performed before ac- cepting our interpretation and a generalization of the results. Study 2 was designed to respond to the above comments.

Page 12: Gender differences in idiosyncratic sex-typed self-images and self-esteem

282 Orr and Ben-El iahu

"=

<

. .

. .

f f =

.~,

0

0

~2 =

t~

II

&

' , O ~ e - i o6

v ~ v v ~ ~ ~ .

~ v

. . . ~ ~

= .,.e ~ ~ = " - ' " .,= ,,., "= "~ = ._

0 0 0 . . . . .~ • = '== .... ~ .o ~ o o o

V V

Page 13: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem 283

The second study was performed with a population of young adults, and used the Personal Attribute Questionnaire (PAQ; Spence & Helm- reich, 1978) in addition to the SSIS. Our hypotheses were (1) former sex role orientation findings would be replicated when conventional strategy of scoring feminine and masculinity was used, and (2) findings from the generic self-system model in Study 1 would be replicated when idiosyncratic calculations of feminine, masculine, and nonstereotyped SI scores were per- formed. These hypotheses were tested first with the SSIS, and subsequently with the PAQ.

STUDY 2

The specific hypotheses tested in Study 2 were that (1) conventional masculinity will be significantly correlated with male and female self-es- teem; (2) idiosyncratic masculine SI will be significantly correlated with male self-esteem, but idiosyncratic nonstereotyped SI will be significantly correlated with female self-esteem; (3) neither conventional femininity nor idiosyncratic feminine SI will be significantly correlated to either male or female self-esteem; and that (4) significantly more females than males will define conventional masculine attributes as nonstereotyped and no signifi- cant difference will be found between males and females in the number of conventional feminine and neutral attributes defined as nonstereotyped.

Method

Participants and Procedure

The data for Study 2 was collected anonymously, in one session during class hours. The respondents were 183 female and 97 male (mostly Jewish) Israeli first- and second-year undergraduate social science students at Ben- Gurion University of the Negev. The SSIS, a translation of the shortened PAQ (Spence & Helmreich, 1978) and the Rosenberg's Self Esteem Scale (Rosenberg, 1965) were administered in that order. The self-appraisal scales were administered prior to the male female comparisons in both the SSIS and the PAQ. The task was a course requirement and took about 20 minutes. The mean age for male students was 24 (SD = 2.0), and for female students 22 (SD = 1.6).

Page 14: Gender differences in idiosyncratic sex-typed self-images and self-esteem

284 Orr and Ben-Eliahu

Instruments

Self-esteem and idiosyncratic masculine, feminine, and nonstereo- typed Sis were tapped and calculated in the same way as in Study 1 by Rosenberg's Self Esteem Scale (Rosenberg, 1965) and by the SSIS. In order to tap conventional gender scores in the SSIS, an independent pilot study was conducted. A sample of 50 males and 50 females was randomly selected from a group of 253 first-year undergraduate psychology students. They were given the list of SSIS items and required to judge whether each item was (A) generally considered feminine, (b) generally considered masculine, or (C) I don't know. The items aggressive, independent, ambitious, active, strong personality, makes decisions easily, and dominant were judged by at least 70% of students as generally considered masculine; cries easily, friendly, childlike, cheerful, loyal, gentle, feels easily hurt, and artistic were judged by at least 70% of students as generally considered feminine. Con- ventional masculinity was calculated by adding together the scores of the Self-Appraisal Subscale on the masculine items. Conventional femininity was calculated by the same procedure on the feminine items.

For Study 2 we used also the two parts of the shortened PAQ (Spence & Helmreich, 1978): the Personal Attribute Scale and scale of Male-Female Comparisons. The short PAQ is a self-report version of the longer PAQ with 24 items divided into three 8-item scales. The masculine-instrumental scale (M) includes socially desirable attributes for both sexes that males are both stereotypically believed to possess, and in self-report perceive themselves as possessing, to a greater extent than females. The feminine- expressive scale (F) consists of attributes determined by a parallel opera- tion. The M-F scale consists of items on which social desirability for the sexes differed, and which differentiate the sexes stereotypically and in self- report. Each item is scored from 0 (A) to -4 (E). Responses to the M and to M-F items are keyed in a masculine direction. The Male-Female Com- parison scale consists of the attributes on the Personal Attributes Scale, but respondents are asked whether each item is much more characteristic of male (A), slightly more male (B), no difference (C), slightly more female (D), or much more characteristic of female (E). The scale has been widely used and its reliability and validity well documented (Spence & Helmreich, 1978). As previously, two different methodological strategies--idiosyncratic and conventional--were used to summarize individual differences in sex- typed self-appraisal in the PAQ. In the first idiosyncratic strategy, sex- re- lated SI scores were obtained in the following way: each item in the Male-Female Comparison scale that got (A) or (B) was identified as mas- culine. Then, scores on the respective items in the Personal Attribute Scale were summed and averaged to yield the masculine SI score. In a parallel

Page 15: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem

Table IV. Gender Differences Between SSSI Scale and PAQ Scores Based on Conventional and Idiosyncratic Calculations

285

Men (n = 97) Women (n = 183)

Mean (sd) Mean (sd) F

Self-esteem 34.0 (4.2) 32.4 (4.6) 8.20 b SSIS

Masculinity 3.9 (0.60) 3.8 (0.51) 4.29 a Femininity 3.9 (0.43) 4.1 (0.40) 13.05 c Masculine SI 3.4 (0.62) 2.9 (0.81) 22.00 c Feminine SI 2.9 (0.70) 3.8 (0.72) 81.13 c Nonstereotyped SI 3.7 (0.44) 3.6 (0.42) 1.80

PAQ Masculinity a 2.7 (0.50) 2.4 (0.44) 15.20 c Femininity 2.8 (0.46) 3.0 (0.46) 17.50 c Masculine SI 2.0 (0.54) 1.4 (0.70) 43.60 c Feminine SIe 2.1 (0.91) 2.0 (0.95) 0.20 Nonstereotyped SI 1.8 (0.52) 1.8 (0.38) 0.07

ap < .05. bp < .01.

pT < .001. he PAQ summary scores: Masculinity = 21.3 (4.0), 19.4 (3.5); Femininity =

22.0 (3.7), 24.0 (3.7), males and females respectively. eBased on 82 females and 47 males who described at least one item as feminine.

fashion, items that got (C) were identified as nonstereotyped, and those that got (D) or (E) were identified as feminine. The nonstereotyped and feminine Sis parallel items were totaled and averaged after corrections for directionality and the items "self-confident" and "superior" were deleted.

In the conventional strategy, scoring followed the original PAQ in- structions (Spence & Helmreich, 1978), and scores on M, F, and M-F scales were obtained by totalling the 8 respective items on the Personal Attributes Scale. Masculinity was measured by the M scale score, femininity by the F scale score.

Results and Discuss ion

Contrary to our younger participants, male students' self-esteem was significantly higher than that of females (see Table IV for this and the following findings). Males got significantly higher masculine scores than fe- males on each scale, and females got significantly higher feminine scores than males on the two SSIS, and the conventional F (PAQ) scales. Both had similar scores on the nonstereotyped SI. Israeli and American univer-

Page 16: Gender differences in idiosyncratic sex-typed self-images and self-esteem

286 Orr and Ben-Eliahu

sity students got similar median scores on the M scale, but contrary to expectations Israelis showed a lower value of 22 compared to the American 24 for the F scale (Spence & Helmreich, 1978). Females' feminine SI did not appear higher than that of males, and 55% of the female and 52% of the male respondents did not consider even one attribute of the PAQ more characteristic of women than of men. The results seem to indicate that for Israeli university students the PAQ includes masculine items but not a rep- resentative sample of feminine attributes.

In order to asses the validity of the generic self-system (GSS) and the sex role orientation (SRO), models, two multiple regression analyses were performed, where (1) male and (2) female self-esteem scores were regressed on idiosyncratic masculine, feminine, and nonstereotyped Sis. (Table V). The findings replicated, with one modification, those obtained in the early adolescent sample. Male self-esteem was predicted significantly by the masculine SI as in Study 1, but also by the nonstereotyped SI, F(2,90) = 20.0, p < .001. Female self-esteem was predicted by the nonstereotyped SI, F(1,176) = 49.0,p < .001. In a parallel fashion, in the following analyses (3) male and (4) female self-esteem was regressed on conventional mas- culinity, femininity, and their interaction. For males, only masculinity con- tributed significantly to the prediction of self-esteem, F(1,95) = 39.3, p < .001, replicating the finding that masculinity is correlated significantly with self-esteem (e.g., Deutsch & Gilbert, 1976; Lamke, 1982; Massad, 1981; Silver & Ryan, 1979). For females, only the interaction of femininity and masculinity contributed significantly to the prediction of self-esteem, F(1,173) = 58.8, p < .001. Comparisons of the means of four groups----(1) those who ranked above the median (high) on both femininity and mas- culinity; (2) those who ranked high on masculinity, and under the median (low) on femininity; (3) those who ranked low on masculinity and high on femininity; and (4) those who ranked low on both, showed that Group 4 was significantly lower than each of the other groups, who were not sig- nificantly different from each other [one-way analysis of variance, F(4,178) = 8.60, p < .001]. The differences between means were tested by the Scheff6 multiple range test with p < .05. Not feminine SI nor femininity were correlated with self-esteem, as predicted.

The multiple regression analyses supported our hypothesis that the findings with the early adolescence sample would be replicated with older participants when using idiosyncratic SSIS scores. With regard to male stu- dents, our results supported the hypothesis that sex role orientation findings would be replicated when conventional sex-typed scores were used. How- ever, two findings were not predicted: (A) male self-esteem was significantly correlated with nonstereotyped SI, and (B) only nondifferentiatedness (low levels of masculinity and femininity) was related to lower self-esteem. The

Page 17: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem 287

T a b l e V. M u l t i p l e C o r r e l a t i o n s B e t w e e n S e l f - E s t e e m a n d C o n v e n t i o n a l a n d I d i o s y n c r a t i c Parameters of the SSIS [Males (n = 97) Under the Diagonal; females (n =

183) Above]

Idiosyncratic scores - -Pearson correlations

1 2 3 4

1. Self-esteem .04 .06 .48 c 2. Feminine SI .01 .10 - .19 a 3. Masculine SI .45 c .03 .01 4. Nonstereotyped SI .48 c .01 .24 b

Conventional scores - -Pearson correlations

1 2 3

1. Self-esteem .35 c .43 c 2. Femininity .28 c .24 c 3. Masculinity .54 c .23 a

Multiple correlations: Dependen t variable self-esteem

Source

R 2 R 2 change F df Males

Nonstereotyped SI .41 c .22 .22 Masculine SI .32 c .30 .08 20.0 c Masculinity .55 c .30 .30 39.3 c

Females Nonstereotyped SI .50 c .25 .25 49.0 c Femininity x Masculinity .50 c .25 .25 58.8 c

2,90 1,95

< .05. bpP< .01.

~p < .001.

self-esteem of other sex role orientations were similar. Although not pre- dicted, Finding A does not contradict the rationale of our model, which implies that self-esteem benefits can be gained from sex-typed, self-syn- chronized, socially valued attributes. The model does not exclude the pos- sibility that male self-esteem can also benefit from nonstereotyped, socially valued attributes.

Finding B is not consistent with the earlier reported significant cor- relation of masculinity with self-esteem (Deutsch, & Gilbert, 1976; Silver & Ryan, 1979), and with significant correlation of androgyny with self-es- teem (Alpert-Gillis & Connell, 1989; Bem, 1974; Spence & Helmreich, 1978). Our findings may indicate that the conventional sex role orientation was not a valid construct for explaining female self-esteem in our sample. A more plausible explanation is that those who ranked low on many self-

Page 18: Gender differences in idiosyncratic sex-typed self-images and self-esteem

288 Orr and Ben-Eliahu

attributes also ranked low on self-esteem. The inconsistency between pre- sent and former findings could be attributed to the different selection of items, different populations, or real changes in sex typing, perhaps influ- enced by feminist ideology.

Our fourth hypothesis was confirmed in the students sample with a possible modification (see Table III). Conventional masculine attributes were defined as masculine by significantly fewer females than males, 23.7% vs. 43.7%, and redefined as nonstereotyped by significantly more females than males. 70.3% vs. 51.6%, zZ(df 2) = 12.06, p < .001. Sex differences within the masculine and nonstereotyped idiosyncratic definitions were also significant: x2(df I) = 8.38, p < .001, 4.50, p < .05, respectively. No sig- nificant difference was found in the percent of nonstereotyped definitions made by females and males in the conventional neutral attributes, 78.7% vs. 76.0%, respectively. However, a similar female trend of redefining con- ventional feminine attributes as nonstereotyped appeared too: 44.8% vs. 31.8%, idiosyncratic feminine definition, and 50.5% vs., 57.9% redefinition as nonstereotyped, males and females, respectively, x2(df 2) = 6.76, p < .05. The specific comparisons were not significant, but a hypothesis, that conventional feminine attributes are also redefined as nonstereotyped, could not be discarded. More will be said on this issue in the general dis- cussion.

The last series of tests were performed on results of the PAQ. As earlier, first two stepwise multiple regression analyses were performed, in which male and female self-esteem scores were regressed on PAQ deriving idiosyncratic masculine, feminine, and nonstereotyped SI (Table VI) [Since 101 females and 51 males did not describe even one attribute of the PAQ as more characteristic of women than of men, the regression analyses were conducted, first, with the respondents (82 females and 46 males) who de- scribed at least one attribute as such. In these analyses, the contributions of the feminine SI to the predictions of male and female self-esteem were not significant. Subsequently, therefore, self-esteem was regressed on the masculine and nonstereotyped Sis, with N consisting of the entire sample. It is clear, however, that the list of attributes provided by the PAQ, are not appropriate for our aim of research, at least in Israel.] The findings indicated that male self-esteem was predicted significantly by masculine Sis, F(1,94) = 24.1, p < .001, and that female self-esteem was, in a parallel way, predicted by their nonstereotyped SI, F(1,179) = 5.3, p < .05, repli- cating the findings in the early adolescent sample. In a parallel fashion, in the following third and fourth analyses, male and female self-esteem was regressed on conventional masculinity, femininity, and their interaction. For males and females, masculinity contributed significantly to the prediction of self-esteem, F(1,95) = 79.4, p < .001, and F(1,181) = 61.0, p < .001,

Page 19: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem 289

Tab le VI. Mul t ip l e C o r r e l a t i o n s B e t w e e n S e l f - E s t e e m and P A Q C o n v e n t i o n a l and Idiosyncrat ic Scores [Males (n = 97) Under the Diagonal; females (n = 183) Above

Idiosyncratic correlations

1 2 3 4

1. Self-esteem .17a .14 a .22 b 2. Feminine SI d - .01 .10 - .23 c 3. Mascul ine SI .38 c .07 .03 4. Nons te reo typed SI .18 a - .23 a .18 a

Conventional scores----Pearson correlations

1 2 3

1. Self-esteem ..16 a .50 c 2. Feminini ty .08 .24 b 3. Masculinity .67 c .-.03

Multiple correlations: Dependent variable self-esteem

Source

[3 R 2 R 2 change F df

Men Masculine SI .38 c .14 .14 24.1 c 1,94 Masculinity .68 c .46 .46 79.4 c 1,95

Women Nonstereotyped SI .22/' .05 .05 5.3 b 1,176 Masculinity .50 c .25 .25 61.0 c 1,181

< .05. ~pP< .01.

~B < .001. ased on 81 females and 47 males who described at least one item as feminine.

respectively. In sum, when conventional gender scores were used, the re- sults replicated former findings of the SRO model (e.g., Deutsch & Gilbert, 1976; Lamke, 1982; Massad, 1981; Silver & Ryan, 1979).

The correlations between masculinity and male and female self-es- teem were higher than those between nonstereotyped SI and females self- esteem, and masculine SI and male self-esteem (See Table VI). These results may cast doubt about the contribution of the idiosyncratic SI scores. However, the PAQ masculinity scores included the self-esteem confounding the items "self-confident" and "Superior," whereas the same items were deleted from the SSIS, and the PAQ derived idiosyncratic scores. We did not delete these items in the present case, because one purpose of the conventional scores was comparison with former findings. After deletion of the two items, the prediction of male self-esteem by conventional mas-

Page 20: Gender differences in idiosyncratic sex-typed self-images and self-esteem

290 Orr and Ben-Eliahu

culinity was reduced to R 2 = .26, in comparison to R 2 = .14---its predict- ability by masculine SI. The conventional masculinity, then, predicted male self-esteem better than the masculine SI. Female self-esteem prediction by masculinity after the deletion was similar to that obtained with nonstereo- typed SI, R 2 = .05. Hence, the findings indicate that the idiosyncratic defi- nition of gender stereotypes is not likely to change the predictability of female self-esteem by the self-appraisals. The contribution is the specifica- tion of a process that is likely to facilitate this predictability in women.

The fourth hypothesis was less strongly supported (see Table III). In similar fashion to the SSIS results, less females than males tended to define conventional masculine attributes as masculine (26.6% vs. 39.9%) and more than males as nonstereotyped (63.6% vs. 51.6%). The difference, though, was significant at the p = .07 level only. No difference between males and females was found for nonstereotyped definitions of conventional feminine attributes (50.2% vs. 52.6% respectively). The PAQ does not include neu- tral items; therefore this part of the hypothesis could not be tested by it.

GENERAL DISCUSSION

The present major finding supports the hypothesis that female self-es- teem is distinctly predicted by idiosyncratic definitions of gender-related self- attributes compared to its prediction by the same conventionally defined attributes. The results showed that although females perceived themselves significantly more feminine than males, their self-esteem was significantly related to their nonstereotyped self-image. Masculinity (SSIS) was posi- tively and nondifferentiatedness was negatively (PAQ) related to female self-esteem when conventional definitions were used. The findings were mixed in regard to males: males perceived themselves significantly more masculine than females, and in agreement with the hypothesis, their self- esteem was significantly related to conventional masculinity and to idiosyn- cratic masculine self-image. Contrary to our hypothesis, however, the self-esteem of university male students was also predicted by the nonstereo- typed self-image when the SSIS was used. Put differently, contrary to fe- males, masculine attributes with either conventional or idiosyncratic definitions remained a cornerstone of male self-esteem, whereas the self- esteem of females was predicted only by the nonstereotyped attributes when they were idiosyncratically defined. Attributes defined as masculine or feminine were included within the self-assessment of both sexes, but female self-esteem was unlikely to depend on any gender-related attribute, whereas only the feminine self-image was not related to the self-esteem of males. On the other hand, male as well as female self-esteem seemed to

Page 21: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem 291

benefit from the nonstereotyped SI. The findings were inconclusive in re- spect to the "motivation" hypothesis that more males than females define conventional masculine attributes as masculine----but more females than males redefine them as nonstereotyped. Although a trend in this direction was to be found in the three analyses, the difference was significant only in the SSIS of university students.

Although the general pattern of the findings seems to support the generic self-system model, the implications of specific findings should be further discussed: (1) The prediction of female self-esteem by the non- stereotyped SI was impressive only for the SSIS of university students. The predictions for schoolgirls, and students on the PAQ were significant but modest. (2) The motivation hypothesis was significant only in the SSIS of university students but not with the other two analyses. (3) The self-esteem of male university students was predicted by the SSIS masculine SI, but also and more strongly by the SSIS nonstereotyped SI. (4) Female univer- sity students tended to redefine as nonstereotyped not only masculine but also conventional feminine SSIS attributes.

In the early adolescent sample, the gender-related SI accounted for only 2% of the self-esteem variance, whereas domain specific SPs predicted 28% of boys' and 49% of girls' self-esteem. These findings imply that sex- related self-images are not likely to be very important at this age. The above-mentioned values, however, seem to underestimate the contribution of the gender-related images to the self-system. Domain-specific SPs were more strongly correlated than the self-esteem with masculine SI in boys and with nonstereotyped Sis in girls. Hence, the gender-related Sis, al- though having modest relationships with self-esteem, are shown to be con- sistently related to the entire self-system. Further investigation is needed for a better understanding of the implications of age for our model. A multilevel cross-sectional or longitudinal research design should follow the development of the sex-related self-images in children and adolescents, and gender-related idiosyncratic definitions of domain-specific self-perceptions should be included.

The SSIS nonstereotyped SI predicted 25% of female self-esteem variance compared to 5% of the equivalent PAQ score. The SSIS has been constructed for the Israeli population, whereas the PAQ was developed in the United States and was translated to Hebrew. The finding that the con- ventional M and F scores did not provide better predictions of self-esteem than the idiosyncratic scores supports the interpretation that it is the spe- cific inadequacy of the PAQ set of attributes that accounts for the modest prediction, rather than whether the attributes are conventional or idiosyn- cratic. To conclude, the replication of the finding of significant correlation between the nonstereotyped SI and female self-esteem in the three analy-

Page 22: Gender differences in idiosyncratic sex-typed self-images and self-esteem

292 Orr and Ben-Eliahu

ses, and especially the strong correlation for adult females, when an instru- ment adapted for culture has been used, adds support for this hypothesis.

A similar explanation is applicable to the findings about the tendency to redefine conventional masculine as nonstereotyped. Age differences and the cultural bias of the PAQ could explain why the motivation hypothesis was significant only in the SSIS of university students but not in the other two analyses. More importantly, the tendency to redefine conventional mas- culine attributes as nonstereotyped was tested for the entire sample without making a distinction between those with high and low levels of self-esteem. The consistent findings of significant correlations between female self-es- teem and the nonstereotyped SI, and the lack of such correlations with masculine SI indicate that this tendency is stronger for confident females than for less confident ones. Hence, this tendency for the entire population of females was, most probably, not strong enough to reach significance.

Of a more theoretically serious concern are the findings that male university students' self-esteem wa predicted more strongly by the SSIS nonstereotyped SI than by the masculine SI. The finding may imply that the gender definition is less important than the social desirability of the specific attributes (Spence & Helmreich, 1981). This argument is not com- patible, however, with the replicated findings that female self-esteem was not predicted by the masculine SI, and that the feminine SI was not related to the self-esteem of any group. We can only speculate that the findings indicate that the nonstereotyped definition has more than one source. It may be determined in part by female motivation to maintain consistency between their sex and their gender-related SI, and it is probably deter- mined, also, by changes in the social reality and values faced by university students of both sexes. Research designed specifically to investigate this hypothesis should compare the sex-related Sis of respondents in cultures that vary on sex roles distinctions.

Finally, female university students tended to redefine as nonstereo- typed not only masculine but also conventional feminine SSIS attributes. Hypothetically, a modification could be introduced into our model, i.e., our argument was that feminine SI is not related to self-esteem because femi- nine attributes are less socially rewarded in comparison with masculine at- tributes. This reasoning could have been expanded by the argument that attributes are likely to loose social value once they are defined as feminine. Femininity, and not only feminine attributes, is degraded in this case. Hence, when an attribute is considered an asset, a female will tend to de- fine it as nonstereotyped in order to gain in self-esteem. Simultaneously, our model does not exclude the possibility that under some specific con- ditions socially desirable attributes will be defined by women as feminine, and that these feminine attributes will be related to self-esteem. Further

Page 23: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem 293

research may focus on the question as to what extent such a process pre- vails, and look for individual differences between females who get self-es- teem gains from their feminine Sis and those who do not.

The present model does assume that at least on one level males and females are tuned to gender social categorization (stereotypes). It was well documented and also confirmed in our data that males and females are likely to differ in their self-appraisals. Males appraised themselves signifi- cantly higher on conventional masculine, and females appraised themselves significantly higher on conventional feminine attributes. This tendency im- plies, at least, implicit tuning to gender differences, and the two sexes did not differ in the extent of their conformity to their gender stereotypes in this respect. However, the sexes tended to differ in their adherence to the explicit categorization of the same attributes. This illustration implies that simple models are not likely to explain or resolve gender-related phenom- ena, and in our model we tried to relate to the phenomenological aspects of both conventional and the idiosyncratic gender concepts.

Our findings seem optimistic, because they reveal a process by which females are able to feel feminine and worthy in spite of their (presumed) fewer social rewards. The findings show, also, that during early adolescence the self-esteem of boys and girls is not significantly different. However, dur- ing young adulthood, significant gender differences in self-esteem does ap- pear, Even at the university, where feminine equality is expected to prevail, and even within the "feminine" faculty of social sciences, men tended to perceive themselves more worthy than women. These findings, together with our model, indicate that women are able to disregard the limits pre- scribed for them by sex-typed social stereotypes. However, in spite the in- trapersonal efforts made by some portion of women, the average female self-esteem was lower than that of males. Personal, nonconformist, internal organization is only one side of the coin in the campaign for women's equality. The other side should be political and social efforts for changing unequal social values that result in unequal rewards for being "masculine" and "feminine."

Implications for Self-Esteem Theory

The sex role orientation literature describes three alternative models of the relationship between femininity, masculinity and self-esteem. As summarized by Alpert-Gillis and Connell (1989, p. 98)), one model implies that nonadherence to gender stereotypes is adaptive: persons who combine socially desirable masculine and feminine traits (androgyny), specifically, instrumentality and expressivity, show higher levels of self-esteem as a part

Page 24: Gender differences in idiosyncratic sex-typed self-images and self-esteem

294 Orr and Ben-Eliahu

of their overall healthier functioning (Bern, 1974; Spence & Helmreich, 1978). A second model implies that masculinity alone is the strongest pre- dictor of self-esteem, because society rewards masculine traits to a greater extent than feminine traits (Jones, Chernovetz, & Hanssen, 1978), and the third model implies that establishing a sex-typed identity is an adaptive milestone in childhood and adolescence (Erikson, 1950; Kohlberg, 1966; Mussen, 1969). Masculine boys and feminine girls should have higher self- esteem than cross-gender sex role orientation. As reported earlier, incon- sistent results were reported with the first model, the findings supported the second model, and did not support the third model, thus rejecting the hypothesis that self-esteem is related to consistency within the self-system.

Our argument was that femininity and masculinity scores derived from PAQ (Spence, Helmreich, & Stapp, 1974), BSRI (Bern, 1974) or other simi- lar instruments (e.g., Hall & Halberstadt, 1980) do not tap sex-typed in- ternal concepts within the self-system.

Our argument was that femininity and masculinity scores derived from PAQ (Spence, Helmreich, & Stapp, 1974), BSRI (Bern, 1974) or other simi- lar instruments (e.g., Hall & Halberstad, 1980) do not tap sex-typed internal concepts within the self-system. Rather they tap internal self-attributes but do not assess whether or not, and in what direction, a person perceives these traits as sex typed. By contrast, idiosyncratic measures include this sex-typed aspect of self-attributes and therefore could be considered a bet- ter operation of sex-typed phenomenological self-constructs. As such, our concepts of masculine, feminine, and nonstereotyped Sis enabled us to sup- port the consistency hypothesis without rejecting the social desirability hy- pothesis underlying the masculine model, and the nonadherence hypothesis underlying the androgyny model.

REFERENCES

Alpert-Gillis, L. J., & ConneU, J. P. (1989). Gender and sex-role influences on children's self-esteem. Journal of Personality, 57, 97-114.

Antill, J. L., & Cunningham, J. D. (1979). Self-esteem as a function of masculinity in both sexes. Journal of Consulting and Clinical Psychology, 47, 783-785.

Bandura, A. (1978). The self-system in reciprocal determinism. American Psychologist, 33, 344-358.

Bem, S. L. (1974). The measurement of psychological androgyny. Journal of Consulting and Clinical Psychology, 42, 155-162.

Bem, S. L. (1981). Gender schema theory: A cognitive account of sex typing. Psychological Review, 88, 354-364.

Busch, D., Simmons, R., Hutchinson, B., & Blyth, D. (1977). Adolescents perception of sex-roles in 1968 and 1975. Public Opinion Quarterly, 41, 459-474.

Deaux, K. (1976). Sex: A perspective on the attribution process. In J. Harvey, W. lckes, & R. Kidd (Eds.), New directions in attribution research (Vol. 1). Hillsdale, N. J.: Erlbaum.

Page 25: Gender differences in idiosyncratic sex-typed self-images and self-esteem

Differences in Self-Images and Self-Esteem 295

Deutsch, C. J., & Gilbert, L. A. (1976). Sex-role stereotypes: Effects of perceptions of self and others and on personal adjustment. Journal of Counselling Psychology, 23, 373-379.

Dweck, C. S., & Elliot, E. S. (1983). Achievement motivation. In M. Hetherington (Ed.), Carmichael's manual of child psychology, social and personality development (Vol. 3). New York: Wiley.

Eccles, J., Adler, T., & Meece, J. (1984). Sex differences in achievement: A test of alternate theories. Journal of Personality & Social Psychology, 46, 26-43.

EUingas, S. (1982). Psychological androgyny: A comparison between city and Kibbutz girls by male and female versions of Bem's Sex-Roles Questionnaire. Unpublished master's thesis, The Hebrew University of Jerusalem (in Hebrew).

Epstein, S. (1973). The self-concept revisited, or the theory of a theory. American Psychologist, 28, 404-416.

Erikson, E. (1950). Childhood and society. New York: Norton. Festinger, L., & Carlsmith, J. M. (1959). Cognitive consequences of forced compliance. Journal

of Abnormal Social Psychology, 58, 203-210. Hall, J. A., & Halberstadt, A. G. (1980). Masculinity and femininity in children: Development

of the Children Self-Attributes Questionnaire. Developmental Psychology, 16, 170-180. Harter, S. (1982). The perceived competence scale for children. ChiM Development, 53, 87-97. James, W. (1892/1963) Psychology. New York: Fawcett Publication. Jones, W., Chernovetz, O. C., & Hanssen, R. R. (1978). The enigma of androgyny: Differential

implications for males and females? Journal of Consulting and Clinical Psychology, 46, 298-313.

Kelly, G. A. (1955). The psychology of personal construct. New York: Norton. Kohlberg, L. (1966). A cognitive development analysis of children's sex-role concepts and

attitudes. In E. Maccoby (Ed.), The development of sex differences. Stanford, CA: Stanford University Press.

Lamke, L. (1982). The impact of sex orientation on self-esteem in early adolescence. Child Development, 53, 1530-1535.

Massad, C. M. (1981). Sex role identity and adjustment during adolescence. Child Development, 52, 1290-1298.

McClelland, D. C., (1961). The achieving society. Princeton, NJ: Van Nostrand. Markus, H. (1977). Self-schemata and processing information about the self. Journal of

Personality and Social Psychology, 35, 63-78. Markus, H., Crane, M., Bernstein, S., & Siladi, M. (1982). Self-schema and gender. Journal

of Educational Psychology, 42, 38-50. Marsh, H. W. (1989). Age and Sex effect in multiple dimensions of self-concept:

Preadolescence to early adulthood. Journal of Educational Psychology, 81, 417-430. Marsh, H. W., & Byrne, B. M. (1991). The differentiated additive androgyny model: relations

between masculinity, femininity and multiple dimensions of self-concept. Journal of Personality and Social Psychology, 61, 811-827.

Myers, A. M., & Gonda, G. (1982). Empirical validation of the Bem Sex-Role Inventory. Journal of Personality and Social Psychology, 43, 304-318.

Mussen, P. H. (1969). Early sex-role development. In D. A. Goslin (Ed.), Handbook of socialization theory and research. Chicago: Rand McNally.

O'Heren, C. A., & Orlofsky, J. L. (1990). Stereotypic and nonstereotypic psychological adjustment. Journal of Personality and Social Psychology, 58, 134-144.

Pelham, B. W. (1991). On confidence and consequence: The certainty and importance of self-knowledge. Journal of Personality and Social Psychology, 60, 518-530.

Pleck, J. H. (1975). Masculinity - Femininity: Current and alternative paradigms. Sex Roles, /, 161-177.

Rogers, C. R. (1963). Actualizing tendency in relation to "motives" and to consciousness. In M. R. Jones (Ed.), Nebraska symposium on motivation. Lincoln: University of Nebraska Press.

Rosenberg, M. (1965). Society and adolescent self-images. Princeton, NJ: Princeton University Press.

Page 26: Gender differences in idiosyncratic sex-typed self-images and self-esteem

296 Orr and Ben-Eliahu

Sampson, E. E. (1988). Indigenous psychologies of the individual and their role in personal and social functioning. American Psychologist, 43, 15-22.

Silver, L. E., & Ryan, V. L. (1979). Self-rated adjustment and sex-typing on the Bern Sex-Role Inventory. Is masculinity the primary predictor of adjustment? Sex Roles, 5, 739-763.

Simmons, R. G., Blyth, D. A., Van Cleave, E. F., & Bush, D. M. (1979). Entry into early adolescence: The impact of school structure, puberty, and early dating on self esteem. American Sociological Review, 44, 948-967.

Spence, J. T., & Helmreich, R. L. (1978). Masculinity and femininity. Austin: University of Texas Press.

Spence, J. T., & Helmreich, R. L. (1981). Androgyny vs. gender schema: A comment on Bem's gender schema theory. Psychological Review, 88, 365-368.

Spence, J. T., Helmreich, R. L., & Stapp, J. (1974). The personal Attributes Questionnaire: a measure of sex-role stereotypes and masculinity-femininity. JSAS Catalogue of Selected Documents in Psychology, 4, 127.

Whitley, B. E., Jr. (1983). Sex-role orientation and self-esteem: A critical meta-analytic review. Journal of Personality and Social Psychology, 44, 765-778.

Whitley, B. E., Jr. (1988). Masculinity, Femininity, and self-esteem: A multitrait-multimethod analysis. Sex-Roles, 18, 419-431.

Yogev, A., & Ilan, Y. (1988). Self-image, and achievement motivation in a developmental town and its rural suburbs. Theory in Education, 48, 83-96, (Hebrew).

Yogev, A., & Ronen, R. (1982). Peer support tutorial in the school. Theory in Education, 33, 47-58 (in Hebrew).