Factor Structure of The Bem Sex Role Inventory in Samples of Ethnically Diverse Young Adults in The U.S.

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 23

Regular Article

Journal of Asia Pacific Counseling


Factor Structure of the Bem Sex Role ⓒ 2015 The Korean Counseling Association

Inventory in Samples of Ethnically www.japc.asia


2015, Vol.5, No.1, 1-22
Diverse Young Adults in the U.S.

Donghyuck Lee1
Susan Kashubeck-West2

Abstract
This study examined the fit of Bem’s proposed two-factor structure for the Bem Sex Role Inventory (1974/1981),
using a nationally representative sample (N = 3,709) collected from the U.S. in 2002. The sample included four ethnic
groups of young adults: African American, Asian American, European American, and Hispanic American.
Confirmatory factor analyses (CFA) indicated that the two-factor structure did not fit the data for men or women from
any of the four ethnic groups. Exploratory factor analyses indicated that three-factor solutions fit the data best for all
groups except Asian American men, for whom a four-factor structure was more suitable. The factor structures differed
across the four ethnic groups. Follow-up CFA indicated that although fit was much improved from that of Bem’s
(1981) two-factor structure, the data were fit adequately only for men and women from the European American group.
Implications of the findings are discussed.

Keywords: Gender, Gender Roles, Masculinity, Femininity, Bem Sex Role Inventory

For several decades, scholars have been interested in the measurement of masculinity and femininity
(Constantinople, 1973; Lewin, 1984; Spence, 1993/1999). During this time, many instruments purporting to
measure these constructs have been developed, and some of them have been used widely by researchers,
counselors, educators, and human resource personnel (Hoffman, 2001). For example, the Minnesota
Multiphasic Personality Inventory Masculinity-Femininity Scale (Hathaway & McKinley, 1943), the
Femininity Scale of the California Psychological Inventory (Gough, 1952), the Bem Sex Role Inventory (BSRI;
Bem, 1974/1981), and the Personal Attributes Questionnaire (Spence, Helmreich, & Strapp, 1974) have been
promoted as exemplary measures of masculinity and femininity. Among these inventories, the BSRI has been
the most commonly used measure in research on gender and gender roles (Hoffman, 2001).

1
Konkuk University
2
University of Missouri St. Louis

Corresponding Author
Donghyuck Lee, Department of Education, Konkuk University,
120 Neungdong-ro, Gwangjin-gu, Department of Education, Seoul, South Korea.
Email: dhlee@konkuk.ac.kr
2 LEE AND KASHUBECK-WEST

Although the BSRI has been used frequently to measure femininity and masculinity, its psychometric
adequacy has been questioned in a number of empirical studies. Studies have indicated that the BSRI may not
assess the traits of femininity and masculinity as Bem claimed (e.g., Agbayani & Min, 2006; Blanchard-Fields,
Suhrer-Roussel, & Hertzog, 1994; Choi, Fuqua, & Newman, 2007, 2009; Helgeson, 1994; Marsh, Antill, &
Cunningham, 1989; Peng, 2006; Spence & Sawin, 1985). Criticism has been particularly evident in studies of
minority groups, given that studies involving minorities, married couples, and senior citizens have generated
different factor structures than those found with typical European American college student samples (Choi &
Fuqua, 2003). According to the social constructionist view, gender roles are socially constructed, rather than
biologically manufactured, and are affected by one’s cultural background (Lorber, 1994; Lucal, 1999; West &
Zimmerman, 1987). Therefore, minority groups may view masculinity and femininity differently and may
develop different gender roles than majority groups.
This study examines Bem’s (1974/1981) proposed factor structure of the BSRI in four different ethnic group
samples selected from the U.S. In particular, we used a nationally representative sample of young adults to
examine the generalizability of Bem’s two factors, Masculinity and Femininity. Use of a nationally
representative sample adds to the existing literature base that consists primarily of local convenience samples.
If the purported factor structure of the BSRI is found not to hold with this sample in this study— as it has not
with some convenience samples— then a stronger case can be made for reconsideration of the use of the BSRI
in samples of ethnic minority participants. Given the BSRI has been used in numerous countries outside the
U.S., our results have implications for international use as well.

Development of the BSRI

Initially, Bem (1974) constructed a 60-item version of the Bem Sex Role Inventory (BSRI) that included
masculinity and femininity subscales, as well as filler items related to a social desirability subscale. In her
selection of items for each subscale, Bem surveyed 50 male and 50 female Stanford undergraduates with regard
to 200 personality characteristics to determine the personality characteristics that were considered as feminine
or masculine by both females and males. Based on the survey results, the personality characteristics that were
viewed to be more desirable for a woman than for a man were assigned to the Femininity subscale. Similarly,
those personality characteristics viewed as more desirable for a man than a woman were assigned to the
Masculinity subscale. Using this process, Bem included 20 items each for the Masculinity and Femininity
subscales. For the third subscale, 20 items reflecting both positive and negative characteristics, regardless of
gender, were also selected from the same 200-item pool. This construction process resulted in items being
selected based on the judgment of a small group of university students who were not representative of the U.S.
population.
In 1981, Bem developed a short form of the BSRI in response to studies on the factor structure of the
original BSRI that had shown that some items on the original form had not loaded on their expected factor
(Pedhazur & Tetenbaum, 1979; Whetton & Swindells, 1977). Bem (1981) deleted some items from the original
BSRI based on factor loading values and reduced the total number of items to 30. Because the problematic
FACTOR STRUCTURE OF THE BSRI 3

items were screened out, this short form was thought to be a clearer measure of the constructs than the original
form (Campbell, Gillaspy, & Thompson, 1997; Martin & Ramanaiah, 1988).
With respect to reliability, Bem (1981) reported two types of reliability coefficients: internal consistency and
test-retest reliability. First, the internal consistency coefficients obtained from two Stanford University student
samples were .80 and .82 for the femininity subscale and .86 and .86 for the masculinity subscale. The test-
retest reliability coefficients using a 4-week interval were .82 (female raters) and .89 (male raters) for the
femininity subscale and .94 (female raters) and .76 (male raters) for the masculinity subscale. The reliability
coefficients of the BSRI reported in other studies have been similar to the results reported by Bem (e.g.,
Agbayani & Min, 2006; Ballard-Reisch & Elton, 1992).

Construct Validity of the BSRI

Since the BSRI’s development, many empirical studies have examined the factor structure of the BSRI with
various samples and have consistently revealed that the two-factor solution of the original BSRI is not
supported (Blanchard-Fields et al., 1994; Martin & Ramanaiah, 1988). Although some studies confirmed two
distinct factors (e.g., Ballard-Reisch & Elton, 1992; Thompson & Melancon, 1986), many researchers have
reported that the BSRI items represent more than two factors (e.g., Choi et al., 2009; Pedhazur & Tetenbaum,
1979; Windle & Sinnott, 1985; Zhang, Norvilitis, & Jin, 2001). For instance, Pedhazur and Tetenbaum
reported that a factor analysis of BSRI scores from a sample of graduate students attending one of three
universities in New York City resulted in four factors that they labeled as interpersonal sensitivity,
assertiveness, self-sufficiency, and bipolar masculinity/feminity. However, like the two-factor model, this four-
factor model also has not been found consistently across other studies (e.g., Martin & Ramanaiah, 1988).
Much of the research on the factor structure of the BSRI has shown that feminine items tended to be grouped
as one factor whereas masculine items were not (e.g., Bledsoe, 1983; Choi et al., 2009; Gaudreau, 1977;
Maznah & Choo, 1986; Thompson & Melancon, 1986; Windle & Sinnott, 1985). However, Choi and Fuqua
(2003) showed that half of the femininity items did not load significantly in most studies of the BSRI factor
structure and that typically only 10% of the total variance was accounted for by the femininity factor. Choi and
Fuqua suggested that the femininity items of the BSRI under-represented the construct of femininity. With
regard to the masculinity scale, most studies on the factor structure of the BSRI obtained two to three factors
from the Masculinity items. Those two or three factors accounted for only 5% to 20% of the total variance
(Choi & Fuqua, 2003).
The confusing factor structure of the BSRI also has been evident in studies with samples from ethnically
diverse groups. For example, A. C. Harris (1994) showed that the BSRI did not perform as well with ethnic
minority groups, such as African Americans and Hispanic Americans, as it did with European American
participants. More specifically, the results obtained from 3,000 African Americans, Hispanic Americans, and
European Americans who patronized shopping malls in Chicago showed that African American respondents
tended to consider both masculine and feminine items as desirable for both genders. In addition, although
Hispanic American respondents differentiated the BSRI items into two different factors, these factors were not
4 LEE AND KASHUBECK-WEST

congruent with the original BSRI factors. Mirande (1997) found that Hispanic American men tended to
endorse feminine items such as Warmth, Tenderness, and Sensitivity as desirable for men, suggesting that
masculinity was conceptualized differently by Hispanic males as compared to European American males.
Similarly, in samples of Asian Americans, results have not supported a two-factor solution of femininity and
masculinity. Maznah and Choo (1986) found a five-factor solution in a U.S. sample of Malaysian, Chinese, and
Indian college students. Agbayani and Min (2006) also showed that the two-factor solution of the BSRI did not
fit the data from Filipino American male and female college students; using exploratory factor analysis (EFA),
they found a four-factor structure. Studies conducted in Asian countries also have not supported a two-factor
model. For example, a study performed in China using a sample of Chinese college students yielded six
components for the masculinity items and six components for the femininity items. Peng (2006) examined a
sample of Taiwanese college students, nurses, police officers, and managers and found that both the
masculinity and femininity items loaded on four factors. Taken together, these results suggest that there are
issues with the construct validity of the BSRI when used with ethnically diverse samples. These issues may be
a result of gender roles being constructed differently across cultural groups.

Cultural Differences with respect to Gender Roles

Theories on gender have emphasized that gender roles are socially constructed, rather than a function of
innate, or essential, differences between males and females (Lorber, 1994; Lucal, 1999; West & Zimmerman,
1987). In addition, gender roles have been considered to be “culturally established sets of behaviors,
appearances, mannerisms, and other cues that we have learned to associate with members of a particular gender”
(Lucal, 1999, p. 784). Therefore, individuals develop different gender roles in part due to their different cultural
backgrounds, and thus, their perceptions of masculinity and femininity need to be understood within the
context of their own cultural backgrounds. A brief discussion of gender role conceptualization in the three non-
dominant cultural groups in this study is presented below.

Asian American Gender Roles. Asian American men have demonstrated broader perceptions of
masculinity compared to European American men (Chua & Fujino, 1999). For example, Chua and Fujino
found that Asian American male college students viewed performing domestic duties as a part of masculinity
whereas European American male college students did not. In addition, whereas the European American men
saw masculinity and femininity as separate constructs, Asian American men did not make clear distinctions
between them.
Other research has demonstrated that Asian American women agree that women should have a family-
oriented life focused on raising children and looking after their husbands, a hegemonic view of femininity (Kim,
1994). However, this same group has also experienced a shift in gender roles during the process of immigration
due to the loss of men’s economic power and the family’s increased dependency on wives’ salaries (Kibria,
1990; Palley, 1994). Given the increase in Asian American women’s employment, they currently have more
decision-making power and autonomy in the family. Thus, it is likely that gender roles are shifting among
FACTOR STRUCTURE OF THE BSRI 5

Asian American men and women, given changes in labor market participation, exposure to dominant cultural
beliefs in the U.S., acculturation, generational shifts, and so on.

Hispanic American Gender Roles. Due to the construct of “machismo,” which has long been seen as
synonymous with Latino male identity, Hispanic males have been seen as having exaggerated forms of
masculine gender role orientations (Boulding, 1990; Mosher, 1991). However, recent studies (Arciniega,
Anderson, Tovar-Blank, & Tracey, 2008; Torres, Solberg & Carlstrom, 2002) have proposed multidimensional
views of machismo, which have included “traditional machismo” and “caballerismo.” Traditional machismo
has been characterized as aggressive, sexist, chauvinistic, and hypermasculine, whereas caballerismo has been
referred to as nurturance, family-centeredness, and chivalrousness. Discussions about Hispanic femininity have
suggested that Hispanic women tend to have more traditional gender roles than other ethnic minority women.
However, some empirical studies have indicated that Hispanic women are achievement-motivated (Christensen,
1979), and those who are highly acculturated to the U.S. have more liberal attitudes toward the roles of women
(Kranau, Green, & Valencia-Weber, 1982).

African American Gender Roles. African American men historically have been pressured to meet
hegemonic conceptions of masculinity (Blackwell, 1975). However, those African American men who were
not able to meet such expectations have developed an alternative masculinity to compensate for their feelings
of powerlessness and shame generated from the inability to reach mainstream standards (Franklin, 1984; S. M.
Harris, 1992; Kochman, 1981). This alternative masculinity has emphasized toughness, thrill-seeking, and the
use of violence (Majors, 1989; Majors & Billson, 1992). However, this alternative masculinity cannot be
considered as a true African American masculinity because it is incompatible with Afrocentric values, which
emphasize collectivism, spirituality, and oneness with nature (S. M. Harris, 1992).
African American femininity can be understood as including expressiveness, instrumentality, and
resilience (Hill, 2002). Because of racial discrimination, African American mothers have encouraged their
daughters to develop certain traits, such as strength, assertiveness, and community leadership, which are
incongruent with traditional hegemonic femininity (Gilkes, 2001). Collins (2000) noted that, historically,
African American women have struggled with being treated as existing outside the traditional boundaries for
womanhood in the U.S., given their lack of legal protections, the harsh economic exploitation they endured,
and virulent stereotypes of Black women. Cole and Zucker (2007) used a nationally representative sample of
Black and White women to explore perspectives of femininity. Their results suggested that Black women
reported more interest, compared to White women, in behaviors considered traditionally feminine, such as
wearing attractive clothing and attending to the decoration of one’s home. At the same time, Black women
were more likely to describe themselves as feminists, even while endorsing traditional gender role ideology.
The findings of Cole and Zucker emphasize the importance of further exploration of gender roles in African
American women.
6 LEE AND KASHUBECK-WEST

Purpose of Study

The current study used a nationally representative data set to examine the psychometric properties (internal
consistency and factor structure) of the BSRI in four ethnic samples: African American, Asian American,
Hispanic American, and European American males and females. Given that the two-factor structure of the
BSRI has been challenged by a number of empirical studies, this study assessed whether the original factor
structure of the BSRI fit the nationally representative data, especially in the samples of ethnic minorities. Our
first hypothesis was that the data from these samples would not fit the original two-factor structure proposed by
Bem (1981). This hypothesis was tested using confirmatory factor analysis (CFA). Assuming support for
hypothesis 1, then we planned to examine the factor structure of the BSRI within each ethnic group, using
exploratory factor analysis (EFA) to investigate cultural differences with respect to gender roles. We expected
that the factor structures of the BSRI generated from the four ethnic groups would be different from each other
(hypothesis 2), given the influence of culture on gender roles. By using a nationally representative data set, we
build upon other studies that have used local convenience samples.

Method

Procedure

The data came from the National Longitudinal Study of Adolescent Health (Add Health), a widely used
nationally representative data set (Harker, 2001). Add Health data was collected in 1994 (Wave 1), 1996
(Wave 2), and 2002 (Wave 3), using systematic sampling methods. More specifically, 80 high schools that
were representative of U.S. schools with respect to region of the country, urbanicity, size, type, and ethnic
diversity were selected for data collection. Each selected school included an 11th grade and enrolled more than
30 students. Of the 80 schools, more than 70% participated in the data collection. Schools that chose not to
participate were replaced by other schools within the stratum. Then, feeder schools were identified with the
help of the participating schools. From among the feeder schools, one was selected with probability
proportional to the number of students it contributed to the high school. As a result, 132 schools participated in
the study. Data collection was performed mostly in the home. Interviewers visited participants’ homes and
interviewed participants using questionnaires. Respondents to the survey were adolescents in grades 7 through
12, their parents, and school administrators. The adolescents became young adults in Wave 3. The current data
were obtained from Wave 3. For the data analysis, SPSS 18.0 and MPlus 4.21 were used.

Participants

The sample (N = 3,709) was composed of 712 African Americans (259 males, 453 females), 266 Asian
Americans (109 males, 157 female), 580 Hispanic Americans (263 males, 317 females), and 2,151 European
FACTOR STRUCTURE OF THE BSRI 7

Americans (892 males, 1,259 females). The average age was 22.05 years (SD = 1.73, range 18-27). Regarding
education level, 2,079 (56.1%) participants had attended college or graduate school, and 1,629 (43.9%)
participants had not attended college. At the time of the Wave 3 survey, 1,358 (36.6%) participants responded
that they were currently attending school. Among the participants who were attending school, 23 (0.6%)
participants attended high school, 425 (11.5%) attended a two-year college, 785 (21.2%) attended a four-year
college, and 110 (3.0%) were in graduate school. Regarding participants’ sexual orientation, 3,394 (91.5%)
participants responded that they were 100% heterosexual, and 252 (6.8%) participants answered that they were
mostly heterosexual but somewhat attracted to people of their own sex. Of the remaining, 46 (1.2%) said that
they were bisexual, and 6 (0.2%) identified their sexual orientation as mostly gay or lesbian or 100% gay or
lesbian. These results are presented in Table 1.

Table 1
Education Level, School Attendance, and Sexual Orientation Frequencies by Ethnicity and Gender
Hispanic African Asian European Males Females
13 9 0 55 32 45
Junior-High School
(2.2%) (1.3%) (0.0%) (2.6%) (2.1%) (2.1%)

289 319 75 869 702 850


High School
Education (49.8) (44.8) (28.2) (40.4) (46.1) (38.9)
Level 271 372 185 1178 768 1238
College
(46.7) (52.2) (69.5) (54.8) (50.5) (56.6)

7 12 6 48 20 53
Graduate School
(1.2) (1.7) (2.3) (2.2) (1.3) (2.4)

193 285 128 752 496 862


Yes
Attending (33.3) (40.0) (48.3) (35.0) (32.6) (39.5)
School 387 427 137 1399 1027 1323
No
(66.7) (60.0) (51.7) (65.0) (67.4) (60.5)

526 672 249 1947 1478 1916


100% Heterosexual
(90.8) (94.6) (93.6) (90.7) (97.4) (87.8)

37 32 16 167 36 216
Mostly Heterosexual
(6.4) (4.5) (6.0) (7.8) (2.4) (9.9)

12 5 0 29 2 44
Bisexual
Sexual (2.1) (0.7) (0.0) (1.4) (0.1) (2.0)
Orientation 2 1 0 1 2 2
Mostly Homosexual
(0.3) (0.1) (0.0) (0.0) (0.1) (0.1)

1 0 0 1 0 2
100% Homosexual
(0.2) (0.0) (0.0) (0.0) (0.0) (0.1)

Not attracted to either 1 0 1 1 0 3


females or males (0.2) (0.0) (0.4) (0.0) (0.0) (0.1)
8 LEE AND KASHUBECK-WEST

The characteristics of the sample were somewhat different by ethnicity and gender (see Table 1). First, the
percentage of participants who had attended college differed by ethnicity (χ2 = 54.55, df = 9, p < .001). Second,
the percentage (48%) of Asian Americans who were attending school at the time of the survey was
significantly higher than that of the other ethnic groups (χ2 = 24.49, df = 3, p < .001). More females than males
attended college (χ2 = 23.19, df = 3, p < .001), and more females than males were attending college at the time
of the survey (χ2 = 18.32, df = 1, p < .001). Finally, a higher percentage of females identified themselves as
bisexual or lesbian/gay than males (χ2 = 112.59, df = 5, p < .001). The analyses reported in the Results section
examining overall gender by ethnic group differences in Masculinity and Femininity subscale scores were
conducted both controlling and not controlling for education. The results were almost identical. Therefore, we
present the findings for the analyses not controlling for education.

Measures

The BSRI (Bem, 1974/1981) is a self-report measure of gender role orientation that requires participants to
rate themselves on 30 characteristics, using a 7-point scale ranging from 1 (never or almost never true) to 7
(always or almost always true). The 10 Femininity items are traits commonly viewed as desirable for women,
such as Affectionate, Compassionate, and Sensitive to the Needs of Others, while the 10 Masculinity items are
traits commonly viewed as desirable for men, such as Competitive, Independent, and Forceful. The remaining
10 items are filler items. The 10 filler items were not included in this study because those items are not related
to gender roles. Mean scores for each subscale are calculated and can range from 1 to 7. Higher scores
represent greater endorsement of the characteristics in the subscale. In the current study, coefficient alphas for
each ethnic group (by gender) were as follows: .85/.83 (males/females for Masculinity) and .91/.92
(males/females for Femininity) in African Americans, .74/.77 (males/females Masculinity) and .89/.89
(males/females for Femininity) in Asian Americans, .80/.80 (males/females for Masculinity) and .91/.91
(males/females for Femininity) in European Americans, and .81/.81 (males/females for Masculinity)
and .92/.92 (males/females for Femininity) in Hispanic Americans.

Results

Preliminary Analysis

Preliminary analyses were conducted to check for possible violations of assumptions that might invalidate
the primary analyses. Inspection of skewness, kurtosis, and histograms for all measured variables revealed that
there was no evidence of deviation from the normal distribution for the variables. Both skewness and kutosis
for all variables were smaller than the absolute value of 2.0, the generally accepted criterion for normal
distributions (Tate, 1998). In addition, violation of the assumption of linearity was not evident in the
examination of bivariate scatter plots.
FACTOR STRUCTURE OF THE BSRI 9

Descriptive Statistics

Masculinity and Femininity subscale scores were calculated by gender for each ethnic group (see Table 2).
Potential differences in these scores were tested using a four (ethnic groups) by two (genders) Multivariate
Analysis of Variance (MANOVA). The results showed significant effects for gender, F(2, 3700) = 77.38, p
< .001; ethnicity, F(6, 7400) = 7.73, p < .001; and the interaction between gender and ethnicity, F(6, 7400) =
4.00, p < .01. Follow-up ANOVAs were conducted. Significant main effects and one interaction for gender and
ethnicity were found. More specifically, examination of the means indicated that females scored higher than
males on Femininity, F(1, 3701) = 116.96, p < .001, but that there was no significant mean difference by
gender on Masculinity, F(1, 3701) = .03, p > .05. Regarding ethnic group, the results showed that there were
significant mean differences by ethnic groups on Femininity, F(3, 3701) = 13.81, p < .001, but not on
Masculinity, F(3, 3701) = 2.37, p > .05. Post hoc tests (Tukey HSD) on Femininity scores showed that
European Americans scored significantly higher than African and Hispanic Americans. Moreover, Asian
Americans scored higher than Hispanic Americans. There was no mean score difference on Femininity
between Asian Americans and European Americans. Meanwhile, the interaction effect for gender and ethnic
groups was significant for Masculinity scores, F(3, 3701) = 5.62, p < .01, but not for Femininity scores, F(3,
3701) = .66, p > .05. Examination of the means revealed that males scored higher on Masculinity than females
in the Hispanic American, Asian American, and European American samples, but African American females
scored higher on Masculinity than African American males.
Finally, bivariate correlation coefficients between Masculinity and Femininity subscale scores were
calculated for all ethnic groups. The results showed that Masculinity subscale scores were significantly
associated (p < .01) with Femininity subscale scores for each group. The coefficients were .66 for African
Americans, .41 for Asian Americans, .36 for European Americans, and .63 for Hispanic Americans.

Confirmatory Factor Analysis with the Original Factor Structure

Confirmatory factor analysis (CFA) was performed to examine whether the two-factor solution suggested by
Bem (1981) fit the data for each of the eight groups (four ethnic groups each split by gender). As presented in
Table 3, the results showed that the two-factor structure did not fit the data for any of the groups. In addition to
statistically significant chi-square values (likely due in part to the large sample sizes), the Comparative Fit
Index (CFI) and the Tucker-Lewis Index (TLI) values were smaller than the criterion of .90 for adequate fit
(Kline, 2005), and the Root Mean Square Error of Approximation (RMSEA) as well as the Standardized Root
Mean Square Residual (SRMR) scores were larger than the criterion of .08 (Kline, 2005). Moreover, most
standardized covariance residuals were larger than 3.0, the generally accepted criterion for close fit (Tate, 1998).
Thus, Hypothesis 1, that Bem’s (1974, 1981) two-factor solution would not be an adequate fit for the data from
each ethnic group, was supported.
10 LEE AND KASHUBECK-WEST

Table 2
Mean Scores and Standard Deviations of Masculinity and Femininity Scales by Gender and Ethnic Groups
Masculinity Femininity
M SD M SD
Male 4.88 .03 5.40 .03
Gender
Female 4.89 .03 5.86 .03
African Americans 4.89 .04 5.53 .04
Asian Americans 4.86 .06 5.71 .06
Ethnic Groups
European Americans 4.94 .02 5.76 .02
Hispanic Americans 4.84 .04 5.53 .04
African Americans 4.78 .06 5.27 .06
Asian Americans 4.88 .09 5.51 .09
Male
European Americans 5.00 .03 5.50 .03
Gender x Hispanic Americans 4.85 .06 5.33 .06
Ethnic African Americans 5.01 .04 5.79 .05
Asian Americans 4.83 .07 5.92 .08
Female
European Americans 4.88 .03 6.01 .03
Hispanic Americans 4.82 .05 5.73 .06
Note. Masculinity and Femininity subscales scores could range from 1-7, with higher scores reflecting higher levels of the
construct measured.

Table 3
Fit Indices for Individual Confirmatory Factor Analysis based on Two-Factor Solution in the Eight Groups
χ2 df p value CFI TLI RMSEA SRMR
African Male 777.00 169 < .001 .799 .774 .118 .084
American Female 984.94 169 < .001 .832 .811 .103 .084
Male 349.46 169 < .001 .792 .766 .099 .097
Asian
Female 512.27 169 < .001 .743 .711 .114 .110
Male 633.06 169 < .001 .830 .809 .102 .074
Hispanic
Female 571.42 169 < .001 .873 .857 .087 .075
European Male 1194.05 169 < .001 .866 .850 .082 .075
American Female 1444.72 169 < .001 .881 .866 .077 .077
2
Note. χ = Chi-square; df = Degrees of freedom; CFI = Comparative Fit Index; TLI = Tucker-Lewis Index; RMSEA =
Root Mean Square Error of Approximation; SRMR = Standardized Root Mean Square Residual

Exploratory Factor Analysis

Because the two-factor solution suggested by Bem (1981) did not fit the data, we conducted exploratory
factor analyses (EFA) to identify factor structures for the eight groups that would best summarize the current
FACTOR STRUCTURE OF THE BSRI 11

data. Although Bem (1974, 1981) proposed that Masculinity and Femininity were independent factors, the
Masculinity and Femininity subscale scores were correlated significantly in all four ethnic groups in our study.
Therefore, the maximum likelihood estimation method with oblique rotation (Direct Oblimin, delta = 0) was
used. Before exploring factor patterns for each group, the number of factors was determined by examining the
scree plot (Cattell, 1966), the number of eigenvalues larger than one (Guttman, 1954; Kaiser, 1960), and the
95th percentile of the eigenvalues obtained by parallel analysis (Glorfeld, 1995; Horn, 1965). Finally, the
parsimony and interpretability of the retained factor structures were considered. The EFA results are presented
for each group below, with more information in Table 4.

African Americans. First, a three-factor model was retained to represent the relationships among the 20
BSRI items in the African American male sample. The results indicated that the three retained factors
explained approximately 55% of the total variance in BSRI scores. Communalities associated with most of the
variables ranged from .50 to .70, which indicated that the individual variables were explained reasonably well.
However, there were three items (Forceful, Willing to Take Risks, and Loves Children) with lower
communalities of .20, .38, and .38, respectively. Finally, the absolute value of the factor loadings associated
with each item ranged from .37 to .91. The resulting pattern matrix is displayed in Table 4. The rotated pattern
matrix for the three factors indicated that some masculine items (i.e., Forceful, Willing to Take Risks, and
Dominant) were grouped as one factor. In addition, some feminine items (i.e., Affectionate, Sympathetic, and
Sensitive to Others) were mixed with other masculine items as a factor. Finally, the third factor was composed
of feminine items (i.e., Compassionate, Eager to Soothe, and Gentle).
Three factors also best represented the relationships between the items of the BSRI in the African American
female group. This three-factor solution accounted for approximately 53% of the total variance of the variables.
The communalities associated with most items ranged from .40 to .76, but those of two items (i.e., Loves
Children and Willing to Take Risks) were .36 and .22. This showed that most items were explained reasonably
well by the model. The absolute values of factor loading from the pattern matrix shown in Table 4 ranged
from .38 to .93, which implied that many items had a strong relationship with an associated factor. The rotated
pattern matrix for the three factors indicated that masculine items, such as Dominant, Aggressive, and Forceful
were grouped as one factor. Most of the feminine items, including Sensitive to Others, Understanding,
Compassionate, and Eager to Soothe, were grouped as a second factor. Finally, two feminine items,
Affectionate and Sympathetic, loaded on the third factor with other masculine items (i.e., Defends Beliefs,
Independent, and Assertive). Thus, three-factor solutions were best for both African American women and men,
with one factor of feminine items, one of masculine items, and a third factor that was a mixture of feminine and
masculine items.

Asian Americans. The results showed that a four-factor solution was the most interpretable in the Asian
American male group and accounted for approximately 49% of the variance of the variables. The
communalities associated with most items ranged from .46 to .79, but those of several items (i.e., Defends
Beliefs, Independent, Assertive, and Willing to Take Risks) were smaller than .40. Absolute values of the
12 LEE AND KASHUBECK-WEST

Table 4
Estimated Factor Structure and Factor Loadings of the BSRI in the Eight Groups
Factor 1 Factor 2 Factor 3 Factor 4
Defends Beliefs (.80) Tender (.91) Dominant (.78)
Affectionate (.76) Gentle (.76) Aggressive (.72)
Independent (.71) Compassionate (.61) Forceful (.46)
Strong Personality (.70) Warm (.57) Take risks (.45)
Understanding (.68) Eager to soothe (.43)
Male
Sympathetic (.65) Loves Children (.37)
Sensitive to others (.63)
Assertive (.55)
Leadership Ability (.52)
Take a Stand (.43)

Defends Beliefs (.70) Gentle (.93) Dominant (.77)


Affectionate (.62) Tender (.90) Aggressive (.74)
Independent (.56) Warm (.81) Forceful (.65)
Strong Personality (.56) Compassionate (.61) Take Risks (.41)
Female
Leadership Ability (.54) Understanding (.60) Take a Stand (.38)
Assertive (.53) Eager to Soothe (.54)
Sympathetic (.51) Sensitive to Others (.43)
African Loves Children (.43)
American
Sympathetic (.72) Warm (.80) Forceful (.72) Strong Personality (.55)
Independent (.55) Gentle (.76) Aggressive (.65) Leadership Ability (.49)
Assertive (.54) Compassionate (.74) Dominant (.58) Understanding (.47)
Defends Beliefs (.28) Tender (.73) Loves Children (.47)
Male Eager to Soothe (.65)
Affectionate (.60)
Take a Stand (.45)
Take Risks (.38)
Sensitive to Others (.36)

Understanding (.76) Tender (.83) Aggressive (.63)


Strong Personality (.70) Gentle (.78) Take a Stand (.62)
Sympathetic (.70) Warm (.67) Dominant (.59)
Independent (.66) Eager to Soothe (.57) Leadership Ability (.59)
Female
Sensitive to Others (.65) Compassionate (.48) Loves Take risks (.46)
Affectionate (.53) Children (.42) Forceful (.43)
Defends Beliefs (.50)
Assertive (.47)

Defends Beliefs (.64) Tender (.94) Dominant (.65)


Sensitive to Others (.58) Warm (.90) Leadership Ability (.51)
Strong Personality (.51) Gentle (.70) Aggressive (.49)
Assertive (.47) Compassionate (.61) Take a Stand (.42)
Male Independent (.42) Eager to Soothe (.49) Take risks (.40)
Sympathetic (.48) Forceful (.33)
Affectionate (.47)
Understanding (.43)
Hispanic Loves Children (.42)
American Understanding (.86) Tender (.90) Dominant (.61)
Sensitive to Others (.71) Warm (.73) Aggressive (.61)
Sympathetic (.70) Gentle (.65) Forceful (.53)
Strong Personality (.65) Loves Children (.59) Leadership Ability (.47)
Female Compassionate (.61) Affectionate (.41) Take Risks (.41)
Independent (.58) Take a Stand (.38)
Assertive (.58)
Defends Beliefs (.55)
Eager to Soothe (.47)
Strong Personality (.61) Tender (.89) Dominant (.65)
Independent (.60) Warm (.82) Aggressive (.56)
Defends Beliefs (.54) Gentle (.78) Forceful (.49)
Assertive (.49) Sensitive to Others (.73) Take Risks (.35)
Leadership Ability (.45) Compassionate (.72)
Male
Take a Stand (.44) Sympathetic (.69)
Eager to Soothe (.67)
Understanding (.58)
Affectionate (.55)
European Loves Children (.34)
American Strong Personality (.64) Tender (.89) Aggressive (.69)
Independent (.62) Gentle (.83) Dominant (.68)
Assertive (.59) Warm (.77) Forceful (.62)
Leadership Ability (.55) Compassionate (.75) Take Risks (.39)
Defends Beliefs (.49) Eager to Soothe (.71)
Female
Take a Stand (.45) Sensitive to Others (.67)
Sympathetic (.65)
Understanding (.62)
Affectionate (.53)
Loves Children (.46)
FACTOR STRUCTURE OF THE BSRI 13

factor loadings associated with each item ranged from .26 to .80. As presented in Table 4, the rotated pattern
matrix indicated that masculine items, such as Forceful, Dominant, and Aggressive were grouped as one factor,
and several feminine items, including Affectionate, Sensitive to Others, and Compassionate, were grouped as
another factor along with two masculine items, Willing to Take Risks and Willing to Take a Stand. Masculine
items, such as Defends Beliefs, Independent, and Assertive, loaded on a third factor with a feminine item,
Sympathetic. Finally, masculine items, such as Strong Personality and Leadership Ability, were grouped as the
fourth factor together with feminine items, such as Understanding and Loves Children. This factor structure
was quite different from Bem’s original factor structure and from the factor structures of the African American
male and female samples.
In contrast to the male group, a three-factor structure best explained the relationships between the items of
the BSRI for the Asian American women. This three-factor model explained approximately 48% of the
variance of the variables, and communalities associated with most items ranged from .41 to 73. However, the
items Defends Beliefs, Assertive, Forceful, Willing to Take Risks, Dominant, and Aggressive were associated
with communality values less than .40. The factor loadings of each item ranged from .42 to .83. The rotated
pattern matrix for the three factors indicated that masculine items, such as Aggressive, Leadership Ability,
Willing to Take a Stand, and Dominant, were grouped as one factor. Feminine items, including Tender, Gentle,
Warm, and Eager to Soothe, were grouped as another factor. Finally, some feminine items (i.e., Understanding,
Sympathetic, Sensitive to Others, and Affectionate) and some masculine items (i.e., Strong Personality,
Defends Beliefs, and Assertive) were grouped together as the third factor. In sum, the factor structures for
Asian American men and women differed from each other.

Hispanic Americans. The EFA results showed that a three-factor solution was best in terms of
interpretability and parsimony of the model in the Hispanic American male group. This model accounted for
approximately 50% of the variance in BSRI scores. Communalities associated with the most items ranged
from .40 to .75, and factor loadings associated with the items ranged from .33 to .94. The rotated pattern matrix
indicated that some masculine items, such as Forceful, Dominant, Aggressive, and Leadership Ability, were
grouped as one factor. Other masculine items (i.e., Defends Beliefs, Independent, Strong Personality, and
Assertive) loaded with the feminine item of Sensitive to Others as the second factor. Finally, the rest of the
feminine items were grouped as the third factor.
A three-factor structure was the most interpretable and parsimonious model for the Hispanic American
women. This solution accounted for approximately 51% of the variance in BSRI scores, and the communalities
associated with most items ranged from .42 to .76. However, the communality values for Defends Beliefs,
Independent, Forceful, Willing to Take Risks, and Aggressive, were less than .40. Absolute values of factor
loadings related to the items ranged from .38 to .90. As presented in Table 4, the rotated pattern matrix
indicated that masculine items, including Leadership Ability, Forceful, Willing to Take Risks, and Dominant,
were grouped as one factor. Feminine items, such as Affectionate, Warm, Tender, and Loves Children, were
grouped as another factor, together with the masculine item Willing to Take a Stand. Finally, other feminine
items (i.e., Sympathetic, Sensitive to Others, Understanding, and Compassionate) were grouped as the third
14 LEE AND KASHUBECK-WEST

factor with other masculine items (i.e., Defends Beliefs, Independent, Assertive, and Strong Personality).
Looking at both Hispanic American women and men, the data suggest some similarities in the factor structures
for each group.

European Americans. The results demonstrated that a three-factor solution represented the relationships
between the items of the BSRI in the most interpretable way in the European American male group. The three
factors explained approximately 47% of the variance of the variable. The associated communalities with most
items ranged from .41 to .66, but those of some items, such as Defends Beliefs, Independent, Forceful, Willing
to Take Risks, and Loves Children, were less than .40. Absolute values of factor loadings related to each item
ranged from .34 to .89, and the correlation coefficients between factors ranged from .10 to .51. The rotated
pattern matrix showed that all of the feminine items were grouped as one factor, and the masculine items were
split into two factors: the second factor included the items Defends Beliefs, Independent, Assertive, Strong
Personality, Leadership Ability, and Willing to Take a Stand, and the third factor included the items Forceful,
Dominant, Aggressive, and Willing to Take Risks.
A three-factor solution also was the most interpretable for the European American women. This model
accounted for approximately 48% of the variance in the BSRI items. Communalities associated with most
items ranged from .44 to .68, but those of certain items, such as Defends Beliefs, Independent, Forceful,
Willing to Take Risks, and Loves Children, were less than .40. Factor loadings related to each item ranged
from .39 to .89, and the correlation coefficients between the factors ranged from .06 to .51. The rotated pattern
matrix showed that the items were grouped in the same pattern as the European American men – all of the
feminine items were grouped as one factor and the masculine items were split into two different factors.

Analysis of the New Factors

Thus, the results of the EFAs provide support for hypothesis 2. More specifically, the factor structures of the
BSRI data differed across ethnic groups. While most ethnic groups had three-factor structures, not all did, and
there were differences by gender within some ethnic groups. Mean scores, standard deviations, and Cronbach’s
alphas were calculated for the new factors/subscales that resulted from the EFAs and are presented in Table 5.
Overall, the reliability coefficients were mostly adequate to excellent, with the exception of lower alpha levels
for two of the four factors found for Asian American men and one factor for European American men.
A set of CFAs was performed to examine whether the models retained from the EFA fit the data for the eight
groups. As presented in Table 6, the results showed that the factor structures estimated from the EFA did not fit
the data well in most groups, although the retained factor structures were significantly improved from the
original two-factor solution for all groups except the Hispanic American men. The extracted factor solution for
the Hispanic American male group was significantly poorer than the original two-factor structure (∆χ2 [2] = -
14.97, ∆df = 2, p < .001). The retained models fit the data well only for the European American men and
women. More specifically, the CFI and TLI values typically were above the criterion of .90, and the RMSEA
and SRMR scores were less than the criterion of .08 for adequate fit. In addition, the test of improvement of
FACTOR STRUCTURE OF THE BSRI 15

model fit (∆χ2) in both European American groups showed that the model fit was significantly improved (male:
∆χ2 [2] = 273.33, ∆df = 2, p < .001, female: ∆χ2 [2] = 564.38, ∆df = 2, p < .001). Meanwhile, the estimated model
from the EFA marginally fit the data in the Asian American male group. Although the CFI and TLI values
were not above .90, the RMSEA and SRMR scores were slightly less than the criterion of .08 for adequate fit.

Table 5
Reliability Estimates and Mean Scores (Standard Deviation) of New Factors by Ethnic Group
Cronbach’s Alpha for
Ethnic Group Subscale Male Female
Males/Females
F1 .92/.84 5.32 (1.25) 5.65 (1.11)
African
F2 .88/.91 5.27 (1.25) 5.83 (1.07)
American
F3 .72/.76 3.90 (1.27) 4.35 (1.25)
F1 .61/.86 5.27 (0.96) 5.71 (9.93)
Asian F2 .88/.84 5.38 (0.95) 5.89 (0.86)
American F3 .69/.73 3.73 (1.19) 4.42 (0.99)
F4 .76/NA 5.81 (0.88)
F1 .80/.80 5.26 (1.20) 5.62 (1.04)
European
F2 .91/.91 5.35 (1.17) 5.66 (1.02)
American
F3 .64/.71 4.56 (1.07) 4.20 (1.13)
F1 .77/.90 5.64 (0.98) 5.61 (0.95)
Hispanic
F2 .91/.86 5.50 (0.99) 6.01 (0.84)
American
F3 .74/.70 4.04 (1.11) 3.80 (1.20)

Table 6
Fit Indices for Individual Confirmatory Factor Analysis based on the Retained Models from EFA in the Eight Groups
P P RMSE
χ2 df ∆χ2 ∆df CFI TLI SRMR
value value A
African Male 626.51 167 < .001 150.49 2 < .001 .848 .827 .103 .082
American Female 850.41 167 < .001 134.53 2 < .001 .859 .840 .095 .095
Male 272.27 164 < .001 77.19 5 < .001 .875 .856 .078 .079
Asian
Female 433.73 167 < .001 78.54 2 < .001 .800 .772 .101 .106
Male 648.03 167 < .001 -14.97 2 < .001 .824 .799 .105 .074
Hispanic
Female 548.52 167 < .001 22.90 2 < .001 .879 .863 .085 .076
European Male 920.72 167 < .001 273.33 2 < .001 .902 .888 .071 .062
American Female 880.34 167 < .001 564.38 2 < .001 .933 .924 .058 .051
Note. ∆χ2 = improvement of chi-square of the retained model from the one of the original model; ∆df = improvement of degrees
of freedom; CFI = Comparative Fit Index; TLI = Tucker-Lewis Index; RMSEA = Root Mean Square Error of Approximation;
SRMR = Standardized Root Mean Square Residual
16 LEE AND KASHUBECK-WEST

Discussion

This study examined the factor structure of the BSRI to assess whether the two-factor solution of the BSRI
proposed by Bem (1981) would fit a nationally representative data set collected in the United States with
ethnically diverse young adults. As hypothesized, CFA analyses showed that the two-factor model did not fit
the data well for men and women from the following four ethnic groups: African American, Asian American,
European American, and Hispanic American. These results are congruent with a number of previous empirical
studies that examined the factor structure of the BSRI (e.g., Agbayani & Min, 2006; Choi et al., 2007; Windle
& Sinnott, 1985). Importantly, this study extended previous findings to a nationally representative sample of
young adults. Therefore, it can be argued that the BSRI does not perform as intended for the majority of young
adults in the United States.
Exploratory factor analyses were conducted to explore the BSRI factor structure for men and women in each
ethnic group. The results showed that three factors were retained for all groups except for Asian American men,
for whom a four-factor structure was best. As hypothesized, the item compositions of each factor were different
across the ethnic groups and across males and females, except for European American males and females.
There were some interesting findings in this EFA data. First, looking at the data across the ethnic groups, it
appeared that both men and women identify masculine items that could be seen as harsher or as more extreme
characteristics of the male gender role (such as Dominant, Aggressive, Forceful, and Willing to Take Risks) as
separate from the rest of the traditional masculine characteristics. All eight groups had a factor that could be
considered to represent this more extreme version of masculinity. This result was consistent with findings from
some other studies. For example, Choi et al. (2007) found that the items of Forceful, Dominant, and Aggressive
were categorized as a different factor from other masculine items in their sample of graduate and undergraduate
students. Moreover, Blanchard-Fields et al. (1994) also reported that Forceful, Dominant, and Aggressive items
loaded on a separate factor from other masculine items in their sample of adults and college students. Taken
together, the findings from this study and the previous studies suggest that the composition of masculinity as a
gender role has shifted over time, such that the more powerful or domineering aspects of the role have been
identified as something separate from other aspects of the masculine role.
With respect to African American participants, the data from both men and women suggested that at least
two characteristics identified by Bem (1974/1981) as feminine, Sympathetic and Affectionate, were
characteristics more associated with masculine traits such as Independent, Assertive, Defends Beliefs, and
Leadership Ability. It appears that, at least for African American participants, the traditional male role has been
broadened to incorporate some expressive characteristics. At the same time, however, the traditional feminine
role did not change, as no masculine items loaded on the factor composed of feminine items. Thus, one might
conclude that African American masculinity is different from the hegemonic view of the masculine gender role.
Indeed, the data from European American males and females did not show a similar broadening of the
masculine gender role. According to Gilkes (2001) and S. M. Harris (1995), African American gender roles
have been influenced by racial discrimination. For example, African American women have been encouraged
to possess instrumentality with expressiveness (Hill, 2002), and African American males have developed a
FACTOR STRUCTURE OF THE BSRI 17

masculine gender role that emphasizes toughness and strength (Franklin, 1984; S. M. Harris, 1992; Kochman,
1981; Majors, 1989; Majors & Billson, 1992; Oliver, 1989). Interestingly, the African American participants in
this study did not associate instrumental characteristics with other feminine traits, and the masculine
characteristics reflective of toughness and strength loaded on a separate factor. These findings suggest that the
data collected from this nationally representative sample in 2002 were not consistent with such
conceptualizations of African American gender roles. Research centered on understanding African American
gender roles is needed to understand how such roles are conceptualized, supported, and perhaps, prescribed.
A different factor structure was generated with the sample of Asian American males, as feminine and
masculine items were mixed across three factors (the fourth factor was composed of the Forceful, Aggressive,
and Dominant items, similar to the other groups). For example, the masculine items Strong Personality and
Leadership Ability were grouped with the feminine items Understanding, and Loves Children. Moreover, the
masculine items Willing to Take a Stand and Willing to Take Risks loaded on a factor mostly composed of
feminine items. Similar results were found in the sample of Asian American females, as the masculine items
Strong Personality, Independent, Defends Beliefs, and Assertive were grouped with the feminine items,
Understanding, Sympathetic, Sensitive to Others, and Affectionate. Overall, these findings suggested that
Asian American perspectives of masculinity and femininity are very different from the hegemonic perspective.
Chua and Fujino (1999) reported that Asian American males do not make clear distinctions between
masculinity and femininity, and Cheng (1996) indicated that Asian American men value humility and
communalism, characteristics that are often considered feminine in the dominant culture in the U.S. Our
findings lend support to the notion that as a result of immigration and increasing leadership roles in the family
and the workplace (Kibria, 1990; Palley, 1994), Asian American women have experienced concomitant
changes in their expected gender roles.
With regard to the Hispanic American results, there were some similarities and differences between the men
and women in their responses. For Hispanic American men, the masculine items loaded on two factors. The
first factor included four masculine characteristics (Defends Beliefs, Strong Personality, Assertive, and
Independent) and one feminine characteristic (Sensitive to Others). The second factor was composed of the rest
of the feminine items, and the third factor was similar to one found across the ethnic groups and reflected the
more extreme masculine characteristics (Dominant, Aggressive, Willing to Take Risks, Forceful, Leadership
Ability, and Willing to Take a Stand). Thus, for the Hispanic American men, the gender roles were relatively
separate in terms of masculinity and femininity. Interestingly, for Hispanic American women, some masculine
items (e.g., Strong Personality, Independent, Assertive, and Defends Beliefs) were grouped with feminine
items (e.g., Understanding, Sensitive to Others, Sympathetic, Compassionate, and Eager to Soothe) to form a
factor. A second factor was composed of five feminine items (Tender, Warm, Gentle, Loves Children, and
Affectionate) and one masculine item (Willing to Take a Stand), and the third factor was the extreme masculine
factor that was found across the groups. These results suggest that Hispanic American women conceptualize
gender roles in some hegemonic ways (a strong masculine role and a nurturing feminine role) and in a different
way (the third factor that combined half feminine and half masculine characteristics). Further exploration of
gender roles from the perspective of Hispanic American women would increase our understanding of how
18 LEE AND KASHUBECK-WEST

Latinas experience gender role expectations.


The final ethnic group to be examined was European American men and women. Unlike the other three
ethnic groups, men and women in this group had identical factor structures. Thus, European American women
and men conceptualized gender roles in the same fashion. All of the feminine items loaded on one factor, the
more extreme masculine items loaded on another, and the rest of the masculine items were the third factor.
These results suggested that conceptualizations of femininity in the dominant culture have not changed
significantly since Bem’s work in the 1970’s and early 1980’s. At the same time, conceptualizations of
masculinity appeared to have split into a more extreme factor and a factor that might be considered desirable
masculine characteristics by most individuals in the dominant culture (Independent, Assertive, Leadership
Ability, Strong Personality, Defends Beliefs, and Willing to Take a Stand). Perhaps European American young
adults perceive two types of men – those who are hypermasculine and those who are “regular” men.
Our results suggest that the BSRI is reasonable to use with European American young adults, in that the
factor structures: a) were identical across gender; and b) show acceptable psychometric properties in this
sample. Rather than using the Masculinity and Femininity subscales as proposed by Bem (1974/1981), we
suggest users consider calculating scores based on the factors found in our study. At the same time, our results
indicate that the BSRI should not be used with ethnic minority participants. Our data revealed that neither
Bem’s two-factor solution nor our three- and four-factor solutions were adequate with regard to goodness-of-fit
statistics. Indeed, the MANOVA results indicated different Masculinity and Femininity scores (using Bem’s
original factors) across both gender and ethnic group. These findings from a nationally representative sample,
combined with other studies that indicated differing factor solutions for ethnic minority individuals (e.g.,
Agbayani & Min, 2006; A. C. Harris, 1994; Maznah & Choo, 1986), highlight the lack of construct validity of
the BSRI when used with individuals who do not identify as European American. Our results also caution
against the use of the BSRI in participants outside of the United States. Overall, the data suggest that there is no
universality of gender roles across cultural groups.
The findings of this study need to be considered in light of its limitations. First, the data were all self-report
and thus suffered from mono-method bias. Second, the confirmatory factor analyses performed on the retained
models from the EFAs indicated that the retained models only fit the data from the European American groups.
Although the models from the four ethnic groups were significantly improved over the original two-factor
structure (except for the Hispanic male group), the fit indices did not indicate adequate fit for any of the groups
except for European Americans. Third, although the data set used for the study was nationally representative, it
still does not represent the contemporary population because it was collected in 2002. Thus, caution needs to be
exercised in interpreting the study results.
Future research on gender role conceptualizations in different ethnic groups in the U.S. is strongly warranted.
The results of this study suggest that there is some unity across ethnic groups in identifying a more extreme
masculine gender role. Little is known about how desirable these more extreme characteristics are perceived to
be for men across the four ethnic groups. At the same time, there was little similarity across the ethnic groups
of men and women with regard to how the remaining masculine characteristics were grouped. It may be that
conceptualizations of masculinity are both different across ethnic groups and changing over time, such that
FACTOR STRUCTURE OF THE BSRI 19

Bem’s (1981) definition of masculinity is no longer viable. Interestingly, many of the ethnic and gender groups
in this study showed consistency in having one factor comprised only or primarily of feminine items. Perhaps
these characteristics are still seen as part of a typical feminine gender role. To more clearly understand gender
role conceptualizations across these gender and ethnic groups, we recommend that researchers employ both
qualitative (such as focus groups) and quantitative (such as new surveys composed of both positive and
negative personality characteristics) methods in future research.

References

Agbayani, P., & Min, J. (2006). Examining the validity of the Bem Sex Role Inventory for use with Filipino
Americans using confirmatory factor analysis. Journal of Ethnic & Cultural Diversity in Social Work, 15,
55-80. doi: 10.1300/J051v15n01_03
Arciniega, G. M., Anderson, T. C., Tovar-Blank, Z. G., & Tracey, T. J. G. (2008). Toward a fuller conception
of machismo: Development of a traditional machismo and caballerismo scale. Journal of Counseling
Psychology, 55, 19-33. http://dx.doi.org/10.1037/0022-0167.55.1.19
Ballard-Reisch, D., & Elton, M. (1992). Gender orientation and the Bem Sex Role Inventory: A psychological
construct revisited. Sex Roles, 27, 291-306. doi: 10.1007/BF00289930
Bem, S. L. (1974). The measurement of psychological androgyny. Journal of Consulting and Clinical
Psychology, 42, 155-162.
Bem, S. L. (1981). A manual for the Bem Sex Role Inventory. Palo Alto, CA: Consulting Psychologist Press.
Blackwell, J. E. (1975). The Black community: Diversity and unity. New York: University of Massachusetts.
Blanchard-Fields, F., Suhrer-Roussel, L., & Hertzog, C. (1994). A confirmatory factor analysis of the Bem Sex
Role Inventory: Old questions, new answers. Sex Roles, 30, 423-457. doi: 10.1007/BF01420602
Bledsoe, J. C. (1983). Factorial validity of the Bem Sex Role Inventory. Perceptual and Motor Skills, 56, 55-58.
Boulding, K. E. (1990). Three faces of power. Newbury Park, CA: Sage.
Campbell, T., Gillaspy, J. A., Jr., & Thompson, B. (1997). The factor structure of the Bem Sex Role Inventory
(BSRI): Confirmatory analysis of long and short forms. Educational and Psychological Measurement, 57,
118-124. doi: 10.1177/0013164497057001008
Cattell, R. B. (1966). The scree test for the number of factors. Multivariate Behavioral Research, 1, 245-276.
Cheng, C. (1996). “We choose not to compete”: The “merit” discourse in the selection process, and Asian and
Asian American men and their masculinity. In C. Cheng (Eds), Masculinities in organizations (pp.177-
200). Thousand Oaks, CA: Sage.
Choi, N., & Fuqua, D. R. (2003). The structure of the Bem Sex Role Inventory: A summary report of 23
validation studies. Educational and Psychological Measurement, 63, 872-887. doi: 10.1177/00131644032
58235
Choi, N., Fuqua, D. R. & Newman, J. L. (2007). Hierarchical confirmatory factor analysis of the Bem Sex Role
Inventory. Educational and Psychological Measurement, 67, 818-832. doi: 10.1177/0013164406299106
20 LEE AND KASHUBECK-WEST

Choi, N. Fuqua, D. R., & Newman, J. L. (2009). Exploratory and confirmatory studies of the structure of the
Bem Sex Role Inventory Short Form with two divergent samples. Educational and Psychological
Measurement, 69, 696-705. doi: 10.1177/0013164409332218
Christensen, E. W. (1979). The Puerto Rican woman: A profile. In E. Acosta-Belen (Ed.), The Puerto Rican
woman. New York: Praeger.
Chua, P. & Fujino, D. C. (1999). Negotiating new Asian-American masculinities: Attitudes and gender
expectations. The Journal of Men’s Studies, 7, 391-413.
Cole, E. R., & Zucker, A. N. (2007). Black and White women’s perspectives on femininity. Cultural Diversity
and Ethnic Minority Psychology, 13, 1-9. http://dx.doi.org/10.1037/1099-9809.13.1.1
Collins, P. H. (2000). Black feminist thought (2nd ed.). New York: Routledge.
Constantinople, A. (1973). Masculinity-femininity: An exception to the famous dictum? Psychological Bulletin,
80, 389-407. http://dx.doi.org/10.1037/h0035334
Franklin, C. (1984). Black male-black female conflict: Individually caused and culturally nurtured. Journal of
Black Studies, 15, 139-154.
Gaudreau, P. (1977). Factor analysis of the Bem Sex Role Inventory. Journal of Consulting and Clinical
Psychology, 45, 299-302. http://dx.doi.org/10.1037/0022-006X.45.2.299
Gilkes, C. T. (2001). If it wasn’t for the women. Maryknoll. NY. Orbis Books.
Glorfeld, L. W. (1995). An improvement on Horn’s parallel analysis methodology for selecting the correct
number of factors to retain. Educational and Psychological Measurement, 55, 377-393.
Gough, H. G. (1952). California Psychological Inventory: Manual. Palo Alto, CA: Consulting Psychologists
Press.
Guttman, L. (1954). Some necessary conditions for common factor analysis. Psychometrika, 19, 149-162.
Harker, K. (2001). Immigrant generation, assimilation, and adolescent psychological well-being. Social Forces,
79, 969-1004. doi: 10.1353/sof.2001.0010
Harris, A. C. (1994). Ethnicity as a determinant of sex role identity: A replication study of item selection for the
Bem Sex Role Inventory. Sex Roles, 31, 241-273. doi: 10.1007/BF01547717
Harris, S. M. (1992). Black male masculinity and same sex friendship. The Western Journal of Black Studies,
16, 744-781.
Harris, S. M. (1995). Psychosocial development and black male masculinity: Implications for counseling
economically disadvantaged African American male adolescents. Journal of Counseling & Development,
73, 279-287. doi: 10.1002/j.1556-6676.1995.tb01749.x
Hathaway, S. R., & McKinley, J. C. (1943). The Minnesota Multiphasic Personality Inventory. New York:
Psychological Corporation.
Helgeson, V. S. (1994). Prototypes and dimensions of masculinity and femininity. Sex Roles, 31, 653-682. doi:
10.1007/BF01544286
Hill, S. A. (2002). Teaching and doing gender in African American families. Sex Roles, 47, 493-506. doi:
10.1023/A:1022026303937
Hoffman, R. M. (2001). The measurement of masculinity and femininity: Historical perspective and
FACTOR STRUCTURE OF THE BSRI 21

implications for counseling. Journal of Counseling & Development, 79, 472-485. doi: 10.1002/j.1556-
6676.2001.tb01995.x
Horn, J. L. (1965). A rationale and test for the number of factors in factor analysis. Psychometrika, 30, 179-185.
Kaiser, H. F. (1960). The application of electronic computers to factor analysis. Educational and Psychological
Measurement, 20, 141-151. http://dx.doi.org/10.1177/001316446002000116
Kibria, N. (1990). Power, patriarchy, and gender conflict in the Vietnamese immigrant community. Gender
and Society, 4, 9-24. doi: 10.1177/089124390004001002
Kim, B. (1994). Value orientations and sex-gender role attitudes on the comparability of Koreans and
Americans. In H. Cho & O. Chang (Eds.), Gender division of labor in Korea. Seoul, Korea: Ewha
Women’s University Press.
Kline, R. B. (2005). Principles and practice of structural education modeling (2nd ed.). New York: Guilford
Press.
Kochman, T. (1981). Black and white styles in conflict. Chicago: University of Chicago.
Kranau, E. J., Green, V., & Valencia-Weber, G. (1982). Acculturation and the Hispanic woman: Attitudes
toward women, sex-role attribution, sex-role behavior, and demographics. Hispanic Journal of Behavioral
Sciences, 4, 21-40. doi: 10.1177/07399863820041002
Lewin, M. (1984). Psychology measures femininity and masculinity, 2: From “13 gay men” to the
instrumental-expressive distinction. In M. Lewin (Ed.), In the shadow of the past: Psychology portrays the
sexes (pp.179-204). New York: Columbia University Press.
Lorber, J. (1994). Paradoxes of gender. New Haven, CT: Crossing Press.
Lucal, B. (1999). What it means to be gendered me: Life on the boundaries of a dichotomous gender system.
Gender and Society, 13, 781-797. doi: 10.1177/089124399013006006
Majors, R. (1989). Cool pose: The proud signature of black survival. In M. S. Kimmel & M. A. Messner (Eds.),
Men’s lives (pp.83-87). New York: Macmillan.
Majors, R., & Billson, J. M. (1992). Empowerment in counseling. Journal of Counseling & Development, 69,
222-227. doi: 10.1002/j.1556-6676.1991.tb01491.x
Marsh, H. W., Antill, J. K., & Cunningham, J. D. (1989). Masculinity and femininity: A bipolar construct and
independent constructs. Journal of Personality, 57, 625-663. doi: 10.1111/j.1467-6494.1989.tb00566.x
Martin, H. J., & Ramanaiah, N. V. (1988). Confirmatory factor analysis of the Bem Sex-Role Inventory.
Psychological Reports, 62, 343-350. doi: 10.2466/pr0.1988.62.2.343
Maznah, I., & Choo, P. F. (1986). The factor structure of the Bem Sex Role Inventory (BSRI). International
Journal of Psychology, 21, 31-41.
Mirande, A. (1997). Hombres y machos: Masculinity and Latino culture. Boulder, CO: Westview Press.
Mosher, D. L. (1991). Macho men, machismo, and sexuality. Annual Review of Sex Research, 2, 199-247.
Oliver, W. (1989). Sexual conquest and patterns of black-on-black violence: A structural-cultural perspective.
Violence and Victims, 4, 257-273.
Palley, M. L. (1994). Feminism in a Confucian society: The women’s movement in Korea. In J. Gelb & M.
Lieff (Eds.), Women of Japan and Korea. Philadelphia: Temple University Press.
22 LEE AND KASHUBECK-WEST

Pedhazur, E. J., & Tetenbaum, T. J. (1979). Bem Sex Role Inventory: A theoretical and methodological
critique. Journal of Personality and Social Psychology, 37, 996-1016. http://dx.doi.org/10.1037/0022-
3514.37.6.996
Peng, T. K. (2006). Construct validation of the Bem Sex Role Inventory in Taiwan. Sex Roles, 55, 843-851. doi:
10.1007/s11199-006-9136-6
Spence, J. T. (1993). Gender-related traits and gender ideology: Evidence for a multifactorial theory. Journal of
Personality and Social Psychology, 64, 624-635. http://dx.doi.org/10.1037/0022-3514.64.4.624
Spence, J. T. (1999). Thirty years of gender research: A personal chronicle. In W. B. Swann, J. H. Langlois, &
I. A. Gilbert (Eds.), Sexism and stereotypes in modern society: The gender science of Janet Taylor Spence
(pp.255-289). Washington, DC: American Psychological Association.
Spence, J. T., Helmreich, R. L., & Strapp, J. (1974). The Personal Attributes Questionnaire: A measure of sex-
role stereotypes and masculinity-femininity. JSAS Catalog of Selected Documents in Psychology, 4, 43-44.
(Ms. No. 617)
Spence, J. T., & Sawin, L. L. (1985). Images of masculinity and femininity: A reconceptualization. In V. E.
O’Leary, R. K. Unger, & B. S. Wallston (Eds,), Women, gender, and social psychology (pp.35-66).
Hillscale, NJ: Erlbaum.
Tate, R. (1998). An introduction to modeling outcomes in the behavioral and social sciences. MN: Bergess
International Group Inc.
Thompson, B., & Melancon, J. G. (1986). Factor structure of the Bem Sex Role Inventory. Measurement and
Evaluation in Counseling and Development, 19, 77-83.
Torres, J. B., Solberg, V. S. H., & Carlstrom, A. M. (2002). The myth of sameness among Latino men and their
machismo. American Journal of Orthopsychiatry, 72, 163-181. http://dx.doi.org/10.1037/0002-943
2.72.2.163
West, C., & Zimmerman, D. H. (1987). Doing gender. Gender & Society, 1, 125-151.
Whetton, C., & Swindells, T. (1977). A factor analysis of the Bem Sex Role Inventory. Journal of Clinical
Psychology, 33, 150-153. doi: 10.1002/1097-4679(197701)33:1+<150::AID-JCLP2270330131>
3.0.CO;2-F
Windle, M., & Sinnott, J. D. (1985). A psychometric study of the Bem Sex Role Inventory with an old sample.
Journal of Gerontology, 40, 336-343. doi: 10.1093/geronj/40.3.336
Zhang, J., Norvilitis, J. M., & Jin, S. (2001). Measuring gender orientation with the Bem Sex Role Inventory in
Chinese culture. Sex Roles, 44, 237-251. doi: 10.1023/A:1010911305338

Received October 3, 2014


Revision received February 6, 2015
Accepted February 13, 2015
Copyright of Journal of Asia Pacific Counseling is the property of Korean Counseling
Association and its content may not be copied or emailed to multiple sites or posted to a
listserv without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.

You might also like