GENDER ROLE INTERNALIZATION, MULTIPLE ROLES, AND CHINESE WOMEN'S MENTAL HEALTH

Psychology of Women Quarterly, 25 (2001), 181–196. Blackwell Publishers. Printed in the USA.
Copyright © 2001 Division 35, American Psychological Association. 0361-6843/00 $9.50
GENDER ROLE INTERNALIZATION, MULTIPLE ROLES,
AND CHINESE WOMEN’S MENTAL HEALTH
Taryn N. Tang and Catherine S. Tang
The Chinese University of Hong Kong
The influence of gender role internalization as a moderator in the relationship between women’s multiple roles and
psychological distress was investigated. Study 1 identified three components of gender role internalization, which were
labeled “Traditional Ideal Person,” “Self-Sacrifice,” and “Competence Without Complaint,” and found that it did not,
overlap with existing gender-typed measures among 128 female Chinese university students. The multidimensionali-
ty of gender role internalization was confirmed in Study 2 among a sample of 225 women in the paid Hong Kong work-
force. As expected, role quality was a better predictor of psychological distress than role quantity. Gender role inter-
nalization accounted for significant portions of explained variance even after taking role quality into account.
Internalization of Traditional Ideal Person and Competence Without Complaint messages exacerbated distress in cer-
tain areas when role quality was low. However, internalization of Self-Sacrifice messages mitigated distress for Chinese
women with low work quality.
Investigations into the relationship between women’s mul-
investigation was the examination of the degree to which
tiple roles and well-being have continued to be of empiri-
changing social roles of women at the societal level were
cal interest to researchers over the past three decades due
reflected at the individual psychological level. Hence, we
to the fact that so many women are enacting roles that
wanted to examine the degree to which internalization of
demand simultaneous responsibility and commitment
gender role messages for women affects mental health, and
(Green & Russo, 1993; McBride, 1988). The evidence indi-
the relationship between role quality and mental health.
cates that role quality is more important than role quantity
in predicting women’s mental health and well-being
Gender Role Internalization
(Barnett & Baruch, 1985; Barnett & Marshall, 1992;
Baruch & Barnett, 1986; Meleis, Norbeck, & Laffrey,
The internalization of gender role messages for women
1989). Although increased formal participation of women
has been linked to various mental health problems that are
in paid employment while maintaining their traditional
more prevalent in women than men such as depression
roles is a global phenomenon, the resultant plethora of
(Ali & Toner, 1996), eating disorders (Streigel-Moore,
research into multiple roles and mental health has been
1995; Surrey, 1991; Pinhas, Toner, Ali, Garfinkel, &
conducted mostly in Western societies (Green & Russo,
Stuckless, 1999), agoraphobia (Fodor, 1974; McHugh,
1993). Furthermore, despite the inherently gendered
1996), chronic fatigue syndrome (Abbey & Garfinkel,
nature of this phenomenon, there has been surprisingly lit-
1991), and functional somatic syndromes (Toner, 1994;
tle research on the influence of gender-related constructs.
Toner & Akman, 2000). Despite numerous theoretical
Therefore, in this study we examined the relationship
postulations, however, there is a notable lack of empirical
between multiple roles and mental health among a group
investigation regarding the influence of the internalization
of employed Chinese women in Hong Kong. We focused
of gender role messages in women’s lives (Toner, Ali,
on the roles of paid worker, spouse, and parent as these are
Stuckless, Weaver, Akman, Tang, Quattrochioci, & Esplen,
the three most frequently reported roles for women that
1999). A major shortcoming of this literature is the lack of
have been identified in other studies (Barnett & Baruch,
validated instruments to empirically assess women’s gen-
1985; Baruch & Barnett, 1986). Of equal importance to this
der role internalization. Thus, Toner and her colleagues
have endeavored to develop a tool to assess the extent to
which women have internalized gender role messages and
Address correspondence and reprint requests to: Taryn N. Tang,
the extent to which internalization affects mental health.
Women’s Mental Health Research, Centre for Addiction and
The premise of gender role internalization is that there
Mental Health, 250 College Street, Toronto, ON, M5S 1R8,
is a contradiction within the modern role for women
Canada. E-mail: taryn_tang@camh.net
(Toner et al., 1999). Women are expected to possess many
181

---

182
TANG AND TANG
diverse traits and behaviors, such as being both competi-
personality traits. This finding suggests that the rise in
tive and nurturing, compliant and assertive, and to appear
women’s endorsement of masculine traits has been added
in control without any signs of vulnerability, while still
to their continued endorsement of feminine traits rather
demonstrating the traditionally feminine trait of emotion-
than replacing it. Twenge associates the rise in women’s
ality (Bepko & Krestan, 1990). This idea of conflict within
masculine scores on the BSRI and the PAQ with cultural
current cultural prescriptions for women is related to
changes in women’s participation in traditionally mascu-
Pleck’s (1981) notion of sex role strain for men. According
line roles, particularly in the labor force. The relatively
to this theory, sex role strain results from the internaliza-
small change in feminine scores since the early 1970s,
tion of societal gender role stereotypes and norms that are
however, reflects the more stagnant nature of traditional-
contradictory and inconsistent. Historical change con-
ly feminine roles such as housework and childcare. That
tributes to the inconsistencies in sex role expectations such
is, while women have taken on more masculine work
that the emergence of modern roles for men often coexists
roles, they still continue to shoulder much of the respon-
with the persistence of traditional male roles. Sex roles are
sibility inside the home (Hochschild, 1989). As suggested
so highly idealized and difficult to maintain, most individ-
by Eagly’s (1987) social role theory, women and men
uals will feel as though they have failed to live up to their
develop different traits because they enact different roles
sex role. Herein lies the strain and “Catch 22” that comes
in society. Thus, given the changes that have emerged in
from internalizing sex roles: Violating sex role norms leads
women’s social roles, an examination of current gender
to negative social and psychological consequences, but
role messages that have been internalized by women
conforming to certain prescribed sex roles also leads to
seems warranted. In the next section we examine the
adverse consequences.
extent to which aspects of Chinese culture parallel the
These propositions of Pleck’s (1981) sex role strain par-
social changes that have occurred for women in Western
adigm can also be applied to female gender roles. Gender
culture.
roles constitute an integral part of identity and define the
social behavior of women and men in society (Cheung,
Women’s Experiences in Hong Kong
Lai, Au, & Ngai, 1997). Socialization of gender roles
begins early in life and role differentiation is maintained
In Hong Kong, a predominantly Chinese society influ-
through prescriptions of cultural heritage, social norms,
enced by both Western and Chinese cultures, the expan-
attitudes, and beliefs. In Western culture, masculinity has
sion of economic opportunity and a more liberal social cli-
been associated with traits such as independence, ration-
mate provided an impetus for women’s relief from house-
ality, competitiveness, and objectivity whereas femininity
hold obligation and facilitated increased participation in
has been associated with traits such as dependence, intu-
the labor force (Ng, 1995; Tang, Au, Chung, & Ngo, 2000).
ition, submissiveness, and emotionality (Lips, 2000).
However, despite social changes towards increasing status
According to Lips (2000), the content of socially
and expanding roles for women, remnants of traditional
accepted gender roles changes over time, and roles that
Chinese values, including beliefs about the place of
may have been acceptable at an earlier point in one’s life
women, still persist among Hong Kong people (Tang, Lee,
may not be socially desirable at a later point. Eagly and
& Cheung, 1999). Evidence indicates that public attitudes
Mladinic (1989) suggested that in the 1970s many women
towards women with successful careers are both positive
felt compelled to develop and value high levels of mascu-
and negative. Such women are described as achievement-
line gender traits to enter and succeed in fields formerly
oriented, rational, decisive, and competent as well as self-
dominated by men, but more recently, women’s and men’s
ish, aggressive, and lacking in nurturance. The contradic-
attitudes towards feminine gender attributes appear to be
tory nature of attitudes towards the status of women in
more positive.
Hong Kong is exemplified further by Chinese university
However, in a recent meta-analysis of changes in mas-
students’ egalitarian beliefs that women and men are enti-
culine and feminine traits among college student samples,
tled to the same level of autonomy in dating, marriage,
Twenge (1997) found that since 1973 women have
divorce, and work, coupled with their traditional beliefs
increasingly reported stereotyped masculine personality
about women’s roles in the family (Tang, Yik, Cheung,
traits for themselves (as measured by the Bem Sex Role
Choi, & Au, 1995).
Inventory (BSRI; Bem, 1974) and the Personal Attributes
Evidence pointing to the emergence and spread of this
Questionnaire (PAQ; Spence, Helmreich, & Stapp, 1974).
modern domestic ideology in Hong Kong can also be seen
Men’s self-reported masculine personality traits have also
in the preference among the younger generation for
increased over this same time period, although to a lesser
nuclear families and in the rise of a child-centered men-
degree. Endorsement of stereotypical feminine personali-
tality (Ng, 1995). In this child-centered world, women are
ty traits, on the other hand, experienced little change over
portrayed in advertisements, films, and on television as
time among women and a small increase among men.
contented moral guardians in the modern home. The
Also, there was a considerable decrease in sex differences
“husband-as-breadwinner and wife-as-dependent” model
of masculine personality traits, but this was not accompa-
of the modern nuclear family is also explicitly codified as
nied by a similar decrease in sex differences of feminine
the normative basis of social policies in Hong Kong.

---

Gender Role Internalization and Chinese Mental Health
183
The experiences of women occupying senior positions
gests that role quality is more important to mental health
in modern industrial Hong Kong demonstrates the
and well-being than role quantity. However, the extent to
ambivalence with which women confront changing gender
which this relationship holds for Chinese women in Hong
roles. In his interviews with 35 Hong Kong female execu-
Kong is not known. In a related study of work and family
tives, Cashmore (1989) found that on the job, these
roles among Chinese living in urban Shanghai, Lai (1995)
women reported that they could not bargain with men on
found that both work and family roles related to women’s
equal terms, which forced them to play and reconcile con-
mental health status, but only the work role was associat-
tradictory roles. On the one hand, they had to prove them-
ed with men’s mental health status. The results of this
selves worthy of the position of authority; on the other, to
study suggest that women’s mental health may be affected
make life easy for themselves, they were obliged to live up
by many sources, but does not answer the question of
to men’s expectations that women are soft, gentle, and vul-
whether it is the quantity or the quality of women’s roles
nerable. To resolve the dilemma, some chose to reject the
that contributes to well-being. Moreover, as the theoreti-
stereotype and insisted on behaving “like a man.” Many,
cal and empirical literature in Hong Kong has shown, soci-
however, chose to rely on feminine charm, even if they rec-
etal opinions towards women’s participation in multiple
ognized that it simply reconfirmed male-held stereotypes.
roles are often ambiguous and conflicting. Therefore, it is
Many of these women executives in Cashmore’s (1989)
plausible that the degree to which women have internal-
study found their careers conflicting with family roles.
ized gender role messages may affect not only their well-
Some asserted that, in the end, their family responsibilities
being, but also the relationship between well-being and
were more of a concern for them than their careers. Many
multiple roles.
had experienced stress and difficulties in fulfilling their
As discussed previously, the discrepancy between theo-
multiple roles. Moreover, these women both recognized
ry and research has resulted in little empirical investiga-
and contributed to the patriarchal dampening of career
tion of gender role internalization. Consequently, a mea-
opportunities for women. Most of Cashmore’s respon-
sure of this construct does not exist. In the absence of a val-
dents expressed a preference for appointing men rather
idated instrument, candidate items designed to measure
than women to positions of power and heavy responsibili-
the construct of female gender role internalization were
ty for fear that marriage and children would change
“borrowed” from researchers in the West to assess its rel-
women’s career ambitions. These findings clearly illustrate
evance to multiple roles and psychological distress. Toner
that the traditional conceptions of women as mother and
et al. (1999) have undertaken to develop an instrument
homemaker have not been obliterated in Hong Kong.
that would address the internalization of prescribed gen-
Furthermore, the ambivalent nature of female experience
der role messages for women that may affect well-being.
and ideology in Hong Kong is also evident. These women’s
Currently, there are a number of mental health concerns
daily lives are full of conflicting currents and possibilities.
that are more prevalent in women than men, including
Therefore, an analysis of women’s internalization of gen-
eating disorders, depression, somatoform disorders, and
der role messages with regard to their participation in
most anxiety disorders (American Psychiatric Association,
multiple roles is necessary.
1994). Therefore, according to Toner and her colleagues,
According to Ng (1995), gender role theory, which pos-
it is imperative that research be undertaken to examine
tulates the tenacity of culture and the inertia of past social-
the various factors that may account for these differences.
ization practices, is inadequate in answering these ques-
One such factor that warrants study is gender role inter-
tions about contradictory trends. What is needed, he sug-
nalization and its impact on the lives of women.
gests, is a systematic explanation that can grasp the signif-
To this end, two studies were conducted. First, gender
icance of these contradictory processes, take note of the
role internalization candidate items borrowed from Toner
advances that women have genuinely made, and still offer
et al. (1999) were administered to a sample of female uni-
an explanation for the persistence and regeneration of
versity students to assess their validity and psychometric
gender inequalities. Moreover, Cheung, Lai, Au, and Ngai
properties for use among Hong Kong Chinese women. In
(1997) contended that reliance on Confucian classics and
the second study, the degree to which gender role inter-
folklore often confuses the ideal with reality. These pre-
nalization affects multiple roles and psychological distress
scriptions may often be stereotypic expectations that
was investigated among a sample of women in the Hong
might not be followed across all social sectors. To what
Kong workforce.
extent these expectations are incorporated into the gender
role internalization of Chinese women’s cultural heritage
STUDY 1
in Hong Kong has not yet been demonstrated.
Purpose
The Present Study
The aim of the study was to determine a preliminary fac-
This study was an investigation of gender role internaliza-
tor structure among Chinese women for the candidate
tion in relation to Chinese women’s multiple roles and psy-
items measuring the construct of gender role internaliza-
chological distress in Hong Kong. Research evidence sug-
tion. Exploratory factor analyses were performed on the

---

184
TANG AND TANG
gender role internalization candidate items in a sample of
identified and these domains were used as guides in con-
Chinese women attending university. After establishing
structing items. Some examples of these themes and cor-
the internal consistency of this scale, we examined the
responding candidate items include: women’s experience
extent to which gender role internalization correlated with
of the conflicting demands of career and family (“I am torn
two gender-typed constructs (gender role stress and gen-
between trying to reach my own goals and taking care of
der role orientation) and whether it added anything new to
others”), self-blame (“I often apologize for things that I
the literature. We expected gender role internalization to
know are not my fault”), and the value placed on physical
be unrelated to gender role orientation, one’s endorse-
attractiveness (“I can’t feel good about myself unless I feel
ment of stereotypical masculine and feminine personality
physically attractive”). This proposed scale is currently
traits. However, moderate correlations were expected
being developed in Canada. An early version, consisting of
between gender role stress and gender role internalization
62 candidate items, was used in this study. Participants
because both constructs delineate negative psychological
were asked to rate the extent of agreement to each state-
outcomes with high endorsement. Our goal in undertaking
ment on a 7-point Likert scale, ranging from 1 (strongly
these steps was to assess the appropriateness of the gender
disagree) to 7 (strongly agree). The items were back-trans-
role internalization construct rather than a measure per se
lated from English to Chinese ensuring that the meaning
among Chinese women.
of items was as close to the original as possible while main-
taining cultural significance.
Method
Gender role stress. Gender role stress refers to the
Participants and Procedures
cognitive appraisal that one is not living up to the stan-
The questionnaires were administered in large group set-
dards of one’s stereotypical gender. In this study, both the
tings to 128 female undergraduate and graduate students
40-item Masculine Gender Role Stress (MGRS; Eisler &
attending The Chinese University of Hong Kong.
Skidmore, 1987) and the 39-item Feminine Gender Role
Undergraduate students received course credit for partic-
Stress (FGRS; Gillespie & Eisler, 1992) scales were
ipating. Ages of the participants ranged from 18 to 28
administered to all participants. Each item was rated on a
years, with a mean of 20.16 (SD = 1.48). Eighty-eight per-
7-point Likert scale, ranging from 0 (not stressful) to 6
cent indicated they were born in Hong Kong, 12% in
(extremely stressful). Examples of items on the MGRS
mainland China, and one participant was born in Taiwan.
scale are: “Being married to someone who makes more
Number of years lived in Hong Kong ranged from 6 to 28,
money than you,” “Admitting that you are afraid of some-
with a mean of 19.25 (SD = 3.12). Four participants were
thing,” and “Staying home during the day with a sick
married and the rest single.
child.” Examples of items on the FGRS scale are: “Having
others believe that you are emotionally cold,” “Negotiating
the price of car repairs,” and “Finding that you have
Measures
gained 10 pounds.” These two scales have been adopted
Gender role internalization. Candidate items for the
for use in Hong Kong Chinese samples of university stu-
Gender Role Internalization (GRI) scale for women were
dents (Tang & Lau, 1995) and human service profession-
generated and refined from a variety of sources by a
als (Tang & Lau, 1996), and have demonstrated good
Canadian research team working in women’s mental
internal consistency and expected relationships with
health (Toner et al., 1999).1 These sources included: (1) a
health adjustment, burnout, and gender role orientation.
book entitled Too good for her own good, which discusses
In this study, the coefficient alpha was .85 for the MGRS
the internalization of gender role messages for women
and .89 for the FGRS scales.
(Bepko & Krestan, 1990); (2) a large community sample of
adult women and men who provided qualitative descrip-
Gender role orientation. Gender role orientation is
tions of what they considered ideal and what society con-
one’s possession of socially desirable, stereotypical mascu-
sidered ideal in women; (3) reflections from women in
line and feminine personality traits. We used the Li Sex
therapy about standards they felt they did not live up to;
Trait Inventory (LSTI; Li, 1981), an indigenous measure-
(4) current women’s magazines and self-help books;
ment for Chinese modeled after the Bem Sex Role
(5) theoretical and empirical literature on gender role
Inventory (Bem, 1974). The inventory consists of a 20-item
messages and women’s mental health issues; and (6) a
Feminine (F) scale of expressive characteristics, a 20-item
team of experts in gender issues who serve on the execu-
Masculine (M) scale of instrumental characteristics, and 20
tive council of The Psychology of Women Division in the
gender-neutral items that can be used as a measure of social
American Psychological Association and therapists who
desirability. Participants rated all items on a 7-point Likert
work with female clients. This extensive process of item
scale ranging from 1 (not appropriate at all) to 7 (very
generation and refinement is consistent with Toner et al.’s
appropriate). Coefficient alpha in this study for the F, M,
(1999) conceptualization of gender role internalization as
and social desirability scales were .88, .80, and .57, respec-
deriving from multiple socializing forces. Several themes
tively. Due to poor internal consistency of the social desir-
within the construct of gender role internalization were
ability scale, only the F and M scales were used in analyses.

---

Gender Role Internalization and Chinese Mental Health
185
Results
gender role stress (M = 3.37, SD = 0.48) than masculine
gender role stress (M = 2.60, SD = 0.47), t (126) = 23.58,
Exploratory Factor Analyses
p < .001. Women also scored higher on feminine gender
Individual item distributions from both the sample of uni-
role orientation (M = 4.45, SD = 0.58) than masculine gen-
versity students (N = 128) and the sample of working
der role orientation (M = 3.92, SD = 0.79), t (126) = 6.09,
women (N = 225) in Study 2 (see descriptions of the sam-
p < .001.
ple in later section) were analyzed for skewness and unbal-
Table 2 shows the correlations of the composite GRI and
anced distributions. Following procedures suggested by
the three subscales with the feminine and masculine gen-
Clark and Watson (1995), items in which responses were
der role stress scores and the feminine and masculine gen-
highly skewed for both samples were eliminated from the
der role orientation scores of the LSTI scales for university
candidate scale, but only after determining that important
women. Feminine gender role stress correlated with the
construct-relevant information would not be lost. On the
Traditional Ideal Person subscale; masculine gender role
basis of these criteria, 21 items were eliminated from the
stress correlated with all the GRI subscales. Feminine and
candidate scale, leaving a remaining pool of 41 items for
masculine gender role orientation were largely unrelated to
structural analysis.
gender role internalization, with two exceptions to this pat-
Next, the correlation matrix of the 41 gender role inter-
tern. Masculine orientation correlated positively with the
nalization items for the sample of university students only
Competence Without Complaint subscale and negatively
(N = 128) was subjected to principal components analysis
with the Traditional Ideal Person subscale.
with both orthogonal and oblique rotations. Examination
of the screen plot suggested three components. All factor
solutions between two and five components were generat-
Discussion
ed. The three-component solution obtained with Oblimin
Exploratory factor analyses based on female university stu-
rotation, accounting for 31% of the variance of gender role
dents suggested a three-factor model of gender role inter-
internalization ratings, was selected as the most inter-
nalization, which we labeled Traditional Ideal Person,
pretable. There were no double loadings and items with a
Self-Sacrifice, and Competence Without Complaint.
rotated loading greater than .40 were incorporated within
Alphas were in an acceptable range of .73 to .80. With few
an index, resulting in 12 items being deleted. Table 1
exceptions, gender role internalization was unrelated to
shows the items comprising each index, the label given to
feminine and masculine gender role orientation. This find-
each index, and the rotated loadings. The first index
ing is not surprising because gender role orientation and
accounted for 11% of the variance and was termed
gender role internalization differ in important ways.
“Traditional Ideal Person” because items reflect tradition-
Gender role orientation refers to the extent to which one
al cultural prescriptions about interpersonal relationships
believes oneself to have socially desirable or positive per-
and giving to others. The second index, “Self-Sacrifice,”
sonality characteristics. Gender role internalization, how-
also accounted for 11% of the variance. Items convey a
ever, is the extent to which one has taken in societal
sense of self-neglect and selflessness towards others. The
messages, the majority of which have negative connota-
third index was labeled “Competence Without Complaint”
tions. For example, the feminine-oriented characteristic
and accounted for 9% of the variance. This component
“sensitive to the needs of others” is similar in content to
asks women about “doing it all” without asking for or
the GRI message “To be a good woman I always have to
needing help from others. The women in this sample
be sensitive to the needs of others,” however, the tone is
scored highest on the Competence Without Complaint
distinctly positive in the former and negative in the latter.
subscale (M = 4.10, SD = 0.83), second on the Traditional
Masculine gender role stress correlated moderately
Ideal Person subscale (M = 3.95, SD = 0.78), and lowest
with all indices of gender role internalization, whereas
on the Self-Sacrifice subscale (M = 3.10, SD = 0.84).
feminine gender role stress correlated with Traditional
Coefficient alpha was .85 for the 29-item GRI composite,
Ideal Person only. In particular, it seemed that the con-
.80 for the 13-item Traditional Ideal Person subscale, .78
cepts of handling many responsibilities while staying in
for the 8-item Self-Sacrifice subscale, and .73 for the 8-
control (Competence Without Complaint subscale) and
item Competence Without Complaint subscale. The
being very giving to others (Self-Sacrifice subscale) were
scores of the GRI composite and the three subscales were
related to male but not female stressors for Chinese
derived by averaging the scores across items identified by
women. Although we expected more significant relation-
the factor-analytic results.
ships between GRI subscales and feminine gender role
stress, the relationships with masculine gender role stress
Relationship to Other Gender-Typed Constructs
can be explained in light of the content of the items on the
Gender role internalization was compared to some gen-
MGRS scale and the items that loaded on to the
der-related constructs, namely feminine and masculine
Competence Without Complaint and Self-Sacrifice sub-
gender role stress, and feminine and masculine gender
scales. Specifically, these two GRI subscales tap into
role orientation. Consistent with previous research (Tang
notions of emotional inexpressiveness, control, and instru-
& Lau, 1995, 1996) women scored higher on feminine
mental success, ideas that are compatible and indeed

---

186
TANG AND TANG
Table 1
Final Items and Factor Loadings, Gender Role Internalization (N = 128)
Item
Loading
Component 1 (Traditional Ideal Person)
When I assert my own needs I feel selfish.
.62
I can’t feel good about myself unless I feel physically attractive.
.59
I feel uncomfortable asserting my own needs.
.59
When I express myself I tend to get overly emotional.
.55
I can’t feel confident unless I feel attractive.
.55
If I am unhappy it is because my expectations are too high.
.54
If I am not always thinking about my family, I am not a good woman.
.54
If I don’t get what I need it is because I ask for too much.
.51
I will never be a complete person if I am not in a relationship.
.48
If I am too ambitious I will never be an adequate mother.
.47
I am torn between trying to reach my own goals and taking care of others.
.47
To be a good woman I always have to be sensitive to the needs of others.
.41
I don’t feel that I can leave a relationship even when I know that it is not satisfying.
.41
Component 2 (Self-Sacrifice)
I find myself nurturing others but not myself.
.70
I find that I focus all of my energy on the needs of others at the cost of satisfying
my own needs.
.70
I often give up my own wishes in order to make other people happy.
.70
I feel that the needs of others are more important than my own needs.
.66
I feel that I must be supportive of others, but that I must succeed in life without
any added help.
.55
While I am expected to be there for the emotional needs of others, I feel that I am
not allowed to ask for my own needs to be met.
.55
When a relationship in my life is unsuccessful, I always feel that it is my fault, even
when I know that it is not.
.47
I often apologize for things that I know are not my fault.
.45
Component 3 (Competence Without Complaint)
In order to feel confident, I must be able to handle many responsibilities without
feeling overwhelmed.
.67
I feel as though I should be able to do it all and never look overwhelmed.
.66
I feel that I must push myself to the limit so as not to let other people down.
.56
In order to feel worthwhile, I must excel at both a career and my personal life.
.53
No matter how I feel I must always try to look my best.
.51
If I am sexually aggressive then I am not a good woman.
.47
I should never look like I am losing control even if everything is falling apart.
.45
If I don’t accomplish everything I should, then I must be a failure.
.43
Note. Variance accounted for: Component 1 = 11%; Component 2 = 11%; Component 3 = 9%.
Table 2
Intercorrelations Between Gender Role Internalization, Gender Role Stress,
and Gender Role Orientation (N = 128)
Measures
1
2
3
4
5
6
7
8
1. GRI Composite
—
.85***
.66***
.72***
.31***
.39***
.15
–.06
2. Traditional Ideal Person
—
.34***
.42***
.42***
.38***
.15
–.20*
3. Self-Sacrifice
—
.27**
.06
.25**
.03
–.02
4. Competence Without Complaint
—
.15
.21*
.13
.19*
5. Feminine Gender Role Stress
—
.70***
.29**
–.10
6. Masculine Gender Role Stress
—
.15
–.03
7. Feminine Gender Role Orientation
—
–.03
8. Masculine Gender Role Orientation
—
*p < .05. **p < .01. ***p < .001.

---

Gender Role Internalization and Chinese Mental Health
187
correlated with Eisler and Skidmore’s (1987) construct of
business firms. A total of 890 questionnaires were distrib-
masculine gender role stress. These results, however, are
uted to women and 245 were completed and returned for
complicated by a high correlation between the masculine
a response rate of 28%. This figure is low relative to
and feminine gender role stress scales. Thus, the results of
response rates derived from similar sampling methods in
this study suggest that gender role internalization is relat-
North America, but a low response rate in the area of 30%
ed to, but does not overlap with, other gender-typed mea-
is comparable to a previous study using a similar method-
sures, and that it is a valid concept for use among Hong
ology in Hong Kong (Tang & Cheung, 1997).
Kong Chinese women.
Participants
Data from 20 women were considered incomplete or
STUDY 2
unusable and excluded from analyses. The final sample for
Study 2 consisted of 225 women: 119 (53%) unmarried
Purpose
workers, 42 (19%) married workers, and 64 (28%) married
The goal of this study was to explore how gender role
and working mothers. Eighty-nine percent (n = 201) of the
internalization influences the relationship between multi-
sample were born in Hong Kong, 8% (n = 17) were born
ple roles and psychological distress among women in the
in mainland China, 1% (n = 3) reported they were born
paid workforce in Hong Kong. The literature has consis-
elsewhere, and 2% (n = 4) did not indicate where they
tently shown that role quality is a better predictor of men-
were born. The women ranged from 21 to 55 years of age,
tal health and well-being than role quantity (e.g., Barnett
with a mean of 32.30 (SD = 7.56). On average, the women
& Baruch, 1985; Baruch & Barnett, 1986; Meleis et al.,
had been in the workforce for a total of 10.14 years (SD =
1989). Moreover, certain attitudes and psychological vari-
6.91) and 8.83 years (SD = 6.96) at the job they held at the
ables seem to moderate and mediate the strength of this
time of the study. The average number of paid work hours
relationship. However, the role of gender role internaliza-
was 47.41 (SD = 10.12). The mean number of years mar-
tion, as defined by Toner et al. (1999), has not been sys-
ried was 8.94 (SD = 6.26). The mothers in the sample had
tematically investigated.
on average 1.59 (SD = 0.79) children currently living with
It is plausible that the degree to which a woman has
them. This was a highly educated, middle-class sample.
internalized certain gender role messages will influence
Approximately 50% of these women had received a uni-
not only her mental health but also the way in which she
versity education and reported their annual household
experiences the key roles—that of paid worker, wife, and
income in excess of $70,000 (U.S.).
mother—in her life. The expression of mental health
problems that are more prevalent in women than men has
Measures
been linked to gender roles and gender role internaliza-
Gender role internalization. This preliminary 29-item
tion (Ali & Toner, 1996; Streigel-Moore, 1995; Toner,
measure, derived through factor analysis from Study 1,
1994; Toner & Akman, 2000). Therefore, we hypothesized
was administered to test its internal consistency and factor
that the greater the extent to which women have internal-
structure as well as its relationship with multiple roles and
ized gender role messages, the greater the psychological
psychological distress (see Study 1).
distress. Consistent with previous research, we also
expected that role quality would be a better indicator of
Psychological distress. The 28-item General Health
psychological distress than role quantity.
Questionnaire (GHQ; Goldberg & Hillier, 1979; Goldberg
& Williams, 1988) was used as the dependent measure of
METHOD
general psychological distress. The GHQ has been adapt-
ed and translated for use in Chinese populations, demon-
Procedures
strating high internal consistencies (Chan, 1985; Shek,
Participants were recruited through the Hong Kong
1989). Respondents rated each item on a 4-point scale,
Social Workers’ Union (SWU), the Hong Kong Nurses’
with a high score representing a high level of distress.
Association (HKNA), and three small- to medium-sized
Four subscales can be computed from the GHQ as differ-
local business firms. Questionnaires were sent to the
ent indicators of psychological distress. In this study, the
female members of the SWU and the HKNA and respon-
internal consistency of the 7-item Anxiety subscale was
dents were asked to return the completed questionnaire
.84; ? = .87 for the 7-item Depression subscale; ? = .59 for
in the postage-paid return envelope that was supplied.
the 7-item Social Dysfunction subscale; and ? = .83 for the
For the three business firms, questionnaires were hand-
7-item Somatic Symptoms subscale. A mean score across
delivered to representatives at each company, and respon-
items of each subscale was calculated for each respondent.
dents were asked to return the questionnaires in the same
manner.
Role quality. This was assessed with Baruch and
Five hundred questionnaires were mailed to female
Barnett’s (1986) measure of rewards and concerns for the
members of the SWU, 300 to the female members of the
paid worker, wife, and mother roles. Additional items,
HKNA, and 90 were delivered to women from three local
reflecting different aspects of women’s experiences in

---

188
TANG AND TANG
their roles as paid workers, wives, and mothers were added
the EQS program (Bentler, 1989) to perform a maximum
to the original measure for comprehensiveness. These
likelihood CFA, allowing the correlation between factors
scales have been shown to be valid among Hong Kong
to be freely estimated in the three-factor model. The
Chinese in the past (Tang, Chan, Cheung, & Ma, 1999).
results showed a better fit for the three-factor model, ?2
Respondents were asked to indicate on a 4-point scale to
(24) = 62.57, AGFI = .88, NNFI = .94, than the one-fac-
what extent, if at all, each of the items was rewarding or
tor model, ?2 (2) = 44.55, AGFI = .63, NNFI = .86. For
distressing (1 = not at all and 4 = very much). Coefficient
the three-factor model in this study the ?2 / df value was
alpha was .89 for the 21-item work rewards, .90 for the 21-
2.61, which falls within a range of acceptable values
item work concerns, .94 for the 19-item wife rewards, .94
(Wheaton, Muthen, Alwin, & Summers, 1977).
for the 19-item wife concerns, .89 for the 17-item mother
Coefficient alpha was .91 for the GRI composite, .83 for
rewards, and .91 for the 17-item mother concerns scales.
the Traditional Ideal Person, .79 for the Self-Sacrifice, and
Role reward and role concern scores, representing the
.78 for the Competence Without Complaint subscales.
average amount of rewards and concerns experienced in
Sample means for the gender role internalization sub-
each role, were calculated by summing the ratings across
scales were similar to those found for the university stu-
items and dividing by the number of items. A balance
dents in Study 1. The working women in this sample
score was calculated as the difference between the mean
scored highest on the Competence Without Complaint
reward score and the mean concern score. Higher balance
subscale (M = 3.60, SD = 1.02), second on the Traditional
scores indicate higher role quality.
Ideal Person subscale (M = 3.51, SD = 0.94), and lowest
on the Self-Sacrifice subscale (M = 3.02, SD = 0.94).
Results
Regression Analyses
Confirmatory Factor Analyses
We used hierarchical multiple regression analysis to test
Confirmatory factor analyses (CFA) tested the structure of
the main research question about the influence of gender
a three-factor model of gender role internalization against
role internalization on the relationship between multiple
a one-factor model to determine dimensionality. Before
roles and psychological distress, results of which are shown
conducting CFA, three parcels were formed for each sub-
in Tables 3 through 6. The same variables were used to
scale of the GRI scale (Cattell & Burdsal, 1975). We used
predict the four outcome measures of anxiety, depression,
Table 3
Hierarchical Regression Analysis Predicting Anxiety (N = 225)
Predictors
?R2
R2
B
?
Step 1: Age
–.01
–.11
Education
.02
.05
Household income
–.05
–.11
Work hours
.03
.03
.03
.08
Step 2: Wife role
.15
.13
Mother role
.01
.05
.01
.01
Step 3: Work role quality
–.22
–.28***
Wife role quality
–.06
–.09
Mother role quality
.15***
.20
–.27
–.30*
Step 4: Traditional
.13
.21*
Self-Sacrifice
.10
.15
Competence
.06**
.25
–.07
–.12
Step 5: Traditional X Work role quality
–.14
–.18
Self-Sacrifice X Work role quality
.26
.32**
Competence X Work role quality
–.09
–.15
Traditional X Wife role quality
.07
.11
Self-Sacrifice X Wife role quality
–.01
–.01
Competence X Wife role quality
–.01
–.01
Traditional X Mother role quality
–.16
–.16
Self-Sacrifice X Mother role quality
.05
.05
Competence X Mother role quality
.05
.31
–.00
–.00
Note. Traditional = Traditional Ideal Person; Competence = Competence Without Complaint.
*p < .05. **p < .01. ***p < .001.

---

Gender Role Internalization and Chinese Mental Health
189
Table 4
Hierarchical Regression Analysis Predicting Depression (N = 225)
Predictors
?R2
R2
B
?
Step 1: Age
–.01
–.13
Education
–.03
–.07
Household income
–.03
–.07
Work hours
.04
.04
.03
.08
Step 2: Wife role
.10
.09
Mother role
.02
.06
–.21
–.17
Step 3: Work role quality
–.32
–.44***
Wife role quality
–.03
–.05
Mother role quality
.22***
.28
–.17
–.20
Step 4: Traditional
.17
.29**
Self-Sacrifice
–.04
–.07
Competence
.06**
.34
.03
.05
Step 5: Traditional X Work role quality
–.13
–.18
Self-Sacrifice X Work role quality
.19
.26*
Competence X Work role quality
–.17
–.27**
Traditional X Wife role quality
.02
.04
Self-Sacrifice X Wife role quality
–.03
–.05
Competence X Wife role quality
.05
.10
Traditional X Mother role quality
–.16
–.17
Self-Sacrifice X Mother role quality
.02
.02
Competence X Mother role quality
.09**
.43
.03
.04
Note. Traditional = Traditional Ideal Person; Competence = Competence Without Complaint.
*p < .05. **p < .01. ***p < .001.
Table 5
Hierarchical Regression Analysis Predicting Social Dysfunction (N = 225)
Predictors
?R2
R2
B
?
Step 1: Age
–.00
–.02
Education
.01
.02
Household income
–.05
–.17*
Work hours
.03
.03
.00
.01
Step 2: Wife role
.14
.19a
Mother role
.02
.05
–.12
–.14
Step 3: Work role quality
–.17
–.34***
Wife role quality
–.04
–.11
Mother role quality
.13***
.18
–.04
–.08
Step 4: Traditional
.18
.45***
Self-Sacrifice
–.02
–.06
Competence .10***
.28
–.05
–.13
Step 5: Traditional X Work role quality
–.06
–.13
Self-Sacrifice X Work role quality
.10
.20a
Competence X Work role quality
–.04
–.10
Traditional X Wife role quality
–.02
–.04
Self-Sacrifice X Wife role quality
–.05
–.12
Competence X Wife role quality
.05
.15
Traditional X Mother role quality
–.07
–.12
Self-Sacrifice X Mother role quality
.08
.13
Competence X Mother role quality
.04
.32
–.03
–.05
Note. Traditional = Traditional Ideal Person; Competence = Competence Without Complaint.
ap < .06. *p < .05. **p < .01. *** p < .001.

---

190
TANG AND TANG
Table 6
Hierarchical Regression Analysis Predicting Somatic Symptoms (N = 225)
Predictors
?R2
R2
B
?
Step 1: Age
–.02
–.19*
Education
.03
.08
Household income
–.04
–.10
Work hours
.06*
.06
.04
.10
Step 2: Wife role
.24
.20*
Mother role
.03
.09
–.01
–.01
Step 3: Work role quality
–.18
–.24**
Wife role quality
–.06
–.10
Mother role quality
.15***
.23
–.31
–.34**
Step 4: Traditional
.14
.22*
Self-Sacrifice .05
.07
Competence
.03*
.26
–.02
–.04
Step 5: Traditional X Work role quality
–.11
–.15
Self-Sacrifice X Work role quality
.11
.21*
Competence X Work role quality
.01
.01
Traditional X Wife role quality
.13
.19
Self-Sacrifice X Wife role quality
–.03
.05
Competence X Wife role quality
.05
.04
Traditional X Mother role quality
–.34
–.33**
Self-Sacrifice X Mother role quality
.06
.06
Competence X Mother role quality
.09**
.35
.06
.07
Note. Traditional = Traditional Ideal Person; Competence = Competence Without Complaint.
*p < .05. **p < .01. *** p < .001.
social dysfunction, and somatic symptoms. The demo-
The fourth step of the regression equations tested
graphic indicators of age, education, household income,
whether components of gender role internalization could
and working hours entered the equations in the first step
account for additional variance in psychological distress
as control variables. Low family income predicted
beyond the effects of role quality. The R2 increment was
women’s social dysfunction, ? = –.17, p < .05, and higher
significant across all the measures, ranging from 3% for
age predicted somatic symptoms, ? = –.19, p < .05. At this
somatic symptoms to 10% for social dysfunction. The only
stage, no other variables reached statistical significance in
variable that reached statistical significance, however, was
predicting distress.
the Traditional Ideal Person subscale. High endorsement
In the second step, the role occupancy variables of
of messages to attain the traditional ideal person predict-
wife and mother were entered. Being married was a sig-
ed anxiety, ? = .21, p < .05; depression, ? = .29, p < .01;
nificant predictor of somatic symptoms, ? = .20, p < .05,
social dysfunction, ? = .45, p < .001; and somatic symp-
and a near-significant predictor of social dysfunction, ? =
toms, ? = .22, p < .05.
.19, p < .06. Occupancy in the wife role did not predict
In the final step of the regression analyses, the GRI
anxiety or depression. Occupancy in the mother role
subscales by Role quality interaction terms were tested.
(women with three roles) did not predict any indices of
This final stage examined the impact of gender role inter-
psychological distress. The increment in proportion of
nalization as a moderator on the relationship between role
variance explained by these two role occupancy variables
quality and psychological distress. To reduce the potential
was not statistically significant across the four distress
for multicollinearity between the interaction term and the
measures.
component parts of this term, centered scores for the
Role quality scores entered the equations in the third
three GRI subscales and the three role quality variables
step and produced a significant R2 increment in each
were used in these analyses (Cohen & Cohen, 1983). At
regression equation, ranging from 13% for social dysfunc-
this stage, nine interaction terms were tested. To examine
tion to 22% for depression. However, this increase was
the full range of relationships between psychological dis-
mainly due to the work role; in that low work role quality
tress, multiple roles, and gender role internalization, all
predicted anxiety, (? = –.28), depression (? = –.44), social
interaction terms were entered, regardless of whether the
dysfunction (? = –.34), and somatic symptoms (? = –.24),
main effects were significant.
p < .001 for all. Low quality in the mother role also pre-
The interaction term Self-Sacrifice by Work role quali-
dicted anxiety, ? = –.30, p < .05, and somatic symptoms,
ty was a consistent predictor of psychological distress. It
? = –.34, p < .01. Wife role quality did not predict any
was significant in predicting anxiety, ? = .32, p < .01;
indices of distress.
depression, ? = .26, p < .05; somatic symptoms, ? = .21, p <

---