Thoits, Peggy Multiple Identities Examining Gender and Marital Status Differences in Distress

Multiple Identities: Examining Gender and Marital Status Differences in Distress
Author(s): Peggy A. Thoits

Source: American Sociological Review, Vol. 51, No. 2 (Apr., 1986), pp. 259-272
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2095520
Accessed: 01-09-2016 16:10 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact [email protected].
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
American Sociological Association, Sage Publications, Inc. are collaborating with JSTOR to digitize,
preserve and extend access to American Sociological Review
This content downloaded from 92.241.95.11 on Thu, 01 Sep 2016 16:10:00 UTC
All use subject to http://about.jstor.org/terms
MULTIPLE IDENTITIES: EXAMINING GENDER AND
MARITAL STATUS DIFFERENCES IN DISTRESS *
PEGGY A. THOITS
Princeton University
Based upon assumptions that the social self is comprised of major role-identities and
that role-identities reduce psychological distress, this paper examines the mental
health advantage of married and unmarried men relative to comparable women as a
function of multiple-role occupancy. Panel data from surveys of!, 106 adult heads of
household in Chicago and 720 adults in New Haven are utilized. Possession of
multiple role-identities (up to 6 in Chicago, 8 in New Haven) does significantly
reduce distress in both samples. But identity summation does not consistently reduce
gender or gender by marital status differences in distress. Further exploration
revealed that men and women appear to experience equivalent levels of distress
when they hold the same numbers and types of roles. When sex differences do occur,
they appear to be a function of employment rather than of marriage, contrary to
popular belief. Although structural inequalities in role occupancy appear to produce
status differences in distress, future research will require deliberately stratified
samples to adequately test this hypothesis.
From a structural symbolic interactionist and role requirements should give purpose,
perspective (Stryker, 1980; Thoits, 1983), the meaning, and guidance to one's life. In short,
self is partially composed of the social posi- from this perspective, one's sense of self as a
tions that an individual holds and enacts. That meaningful, purposeful entity is derived in part
is, the self can be conceptualized as a set of from the social roles one accepts and enacts.'
social identities, where identities refer to posi- A sense of meaningful and purposeful exis-
tional designations assigned by others and ac- tence derived from multiple identity involve-
cepted by the individual him- or herself (e.g., ments should therefore reduce feelings of anxi-
spouse, parent, employee, student, church ety and despair (Thoits, 1983). Guidance ob-
member). tained from role expectations should prevent
Attached to positions are sets of normative disordered conduct. Thus, the possession or
behavioral expectations, or roles (Linton, acquisition of role-identities should enhance
1936; Merton, 1957). Roles are "scripts" for psychological well-being and functional be-
the enactment of particular positions. Thus, havior; role-identity lack or loss should be psy-
role-identities not only can help define who onechologically damaging and dysfunctional.
is, but suggest how one ought to behave. Roles This hypothesis conflicts with more
traditional views of multiple-role occupancy.
*Address all correspondence to Peggy Thoits, De-

partment of Sociology, Green Hall, Princeton Uni-
versity, Princeton, NJ, 08544. ' It is important to specify that the positions with
An earlier version of this paper was presented at which I am concerned here are those that are salient
the Self and Identity Conference, Cardiff, Wales, and important, not only socially, but to the individual
July 9-13, 1984. I am grateful to Leonard I. Pearlin him/herself. Otherwise, one would have to mention
and Morton A. Lieberman for making their Chicago such transitory identities as customer, automobile
data available for this paper. The original data driver, and dental patient, as these too represent
collection was supported by grants from the Admin- locations in the social structure to which are attached
istration on Aging, #90-A-644, and the National In- patterned behavioral expectations. Further, it is nec-
stitute on Aging, NIA 5-POl-AG00123. I am also essary to state that certain statuses, such as "di-
grateful to Jerome K. Myers for permission to usevorced"his or "unemployed," do not represent iden-
New Haven data, originally collected and analyzed tities in themselves. The divorced or unemployed
under Public Health Service Contract 43-67-743 and individual is not in a patterned relationship with an-
Research Grant MH 15522 from the National Insti- other person or set of persons. Role-identities are
tute of Mental Health. The data analysis was partially reciprocal relationships, dependent upon recurrent
supported by an NIH Biomedical Research Support interaction between role partners. The status of "di-
Grant through Princeton University. Chuck Call vorced" or "unemployed" reflects the absence of
provided many hours of valuable, careful research reciprocity, rather than its presence. Further, these
assistance on this project. I appreciate the helpful "roles" lack sets of behavioral expectations, one
comments of John M. Darley and Ronald Kessler on reason why they often give the individual so much
this paper. trouble-one does not know how to behave.
American Sociological Review, 1986, Vol. 51 (April:259-272) 259
260 AMERICAN SOCIOLOGICAL REVIEW
For several decades, sociologists have associ- GENDER AND MARITAL STATUS
ated multiple roles with harmful consequences DIFFERENCES IN DISTRESS:
for the person. The greater the number of A BRIEF REVIEW
roles, the greater the potential for role conflict
or overload, and therefore the greater the risk Two of the most consistent findings in the
of physical and psychological disorder (Mer- epidemiological literature are that women ex-
ton, 1957; Goode, 1960; Coser, 1974). In more hibit higher rates of distress and disorder than
recent years, this view has been questioned. men (Gove, 1972, 1978; Dohrenwend et al.,
Several theorists have argued that the benefits 1980) and that the unmarried-i.e., the never
of multiple-role occupancy may far outweigh married, separated, divorced, and widowed-
tensions due to overload and conflict (Marks, are at greater risk of distress and disorder than
1977; Sieber, 1974; Verbrugge, 1983; Thoits, the married (Bachrach, 1975). Less agreement
1983). Marks and Sieber, for example, have exists regarding the distribution of distress by
suggested that multiple role involvements can gender and marital status jointly. Gove has
expand rather than constrict an individual's re- argued that married women are at greater risk
sources, rewards, energy, commitment, sense of disturbance than married men, while unmar-
of ego gratification, and security, resulting in ried men are at greater risk than unmarried
enhanced physical and psychological well- women (Gove, 1972, 1978; Gove and Tudor,
being. 1973). Although it is clear that wives' rates of
Given the plausibility of these two contrast- distress and disorder do exceed husbands'
ing views, it is quite likely that the relationship (Gove, 1978), recent evidence suggests that
between multiple identities and well-being is Gove's assertion about rates among the unmar-
not additive but curvilinear. One might posit a ried may be incorrect, at least with respect to
U-shaped curvilinear relationship between the prevalence of distress (which is the focus of
number of role-identities and psychological this paper). Prevalence studies indicate that the
outcomes, where in general, the more iden- distress of unmarried women exceeds that of
tities possessed, the fewer symptoms of psy- unmarried men in all unmarried statuses (Fox,
chological distrubance. But beyond some op- 1980; Warheit et al., 1976), although this dif-
timal number of identities, numerous and/or ference is not as pronounced as that among
conflicting demands may begin to undermine a marrieds. The risk of distress for women may
sense of orderly, purposeful existence and be higher than that for men, regardless of mar-
thereby increase psychological symptoms. ital status.
In a direct test of these ideas, Thoits (1983) These patterns of distress distributions are
showed that symptoms of distress vary inconsistent with the distributions of major
versely with the number of role-identities pos- role-identities in American society, at least
sessed, and that changes in identities over time among the married. Traditionally, married
are psychologically beneficial or harmful, de- women have held fewer identities (e.g.,
pending upon the direction of the change. spouse, parent) than married men (e.g.,
Other studies report similar results (Kandel et spouse, parent, employed worker). However,
al., 1985; Verbrugge, 1983; Spreitzer et al., over the past forty years more and more
1979; Gore and Mangione, 1983). Only one women have entered the labor force, increas-
study (Thoits, 1983) examined the issue of cur- ing the number of role-identities held by
vilinearity. In that study, little evidence of cur- women, particularly by married women. If
vilinearity was found, but this does not rule out identity accumulation is psychologically pro-
its possibility. tective, as previously argued, then the distress
Taken together, these studies suggest that rate for employed wives ought to be lower than
role acquisition is psychologically protective, that of homemakers and equal to the (low) dis-
perhaps enhancing a sense of purpose and tress rate of employed husbands.
meaning in life, despite potential costs in role A review of the empirical literature yields
strains. These findings seem directly appli- mixed support for this reasoning (see Thoits, in
cable to a problem that has preoccupied press, for a detailed summary of studies). Con-
epidemiologists and sociologists for many sistent support for the identity accumulation
years: why do women, especially married prediction is obtained when homemakers are
women, have higher rates of psychological compared to employed husbands. Regardless
distress and disorder than comparable men? It of parental status, the anxiety and/or depres-
seems reasonable to propose that these dif- sion scores of homemakers are significantly
ferential rates of disturbance might be due in higher than those of husbands. But when em-
part to differential possession of role-identities ployed wives are compared to employed hus-
by men and women. This paper examines this bands and to homemakers, ambiguous findings
hypothesis in some detail. are obtained. A majority of studies indicate
MULTIPLE IDENTITIES 261
that, independent of parental status, employed the same number and the same configurations
wives' distress exceeds that of husbands, con- of role-identities to ascertain the source of
trary to the identity accumulation prediction. men's mental health advantage, and, indi-
Holding parental status constant, just as many rectly, to assess the relative meaning of dif-
studies report equal distress among homemak- ferent roles for the well-being of men and
ers and employed wives (contrary to predic- women.
tion) as report lower distress among employed For these reasons, this paper re-examines
wives (consistent with prediction). gender and gender by marital status differences
These inconsistent findings suggest that in psychological distress using a wider range of
further examination of the role-identities held role-identities as explanatory factors for the
by women and men across marital statuses is relative mental health advantage of married
required. Typically, studies restrict analysis to and unmarried men compared to similar
the psychological effects of a maximum of women. The paper then explores the sources of
three roles-spouse, parent, and employed psychological benefit more closely by com-
worker, ignoring other major role-in- paring the distress levels of men and women
volvements potentially held by men and who share the same number and combinations
women differentially-e.g., friend, neighbor, of role-identities.
relative, student, church member, organi-
zational member. Homemakers, for example,
METHODS
may be active in community affairs or do vol-
unteer work, making their number of role- The data for this paper come from a two-wave
involvements equal to or in excess of employed panel study of life strains conducted by Pearlin
wives' involvements. Employed husbands, and Lieberman in Chicago (see Pearlin and
too, may hold greater numbers of role- Lieberman, 1979; Pearlin et al., 1981, for de-
identities than employed wives by virtue of tails) and a two-wave panel study of life events
participation in union or professional associ- conducted by Myers and his associates in New
ations. The potential explanatory power of a Haven (see Myers et al., 1971, 1974).
fuller range of role-identities has not been ex- In Chicago, a sample of 2,300 individuals
amined. representative of the adult population was
Moreover, researcher attention has focused drawn with multistage cluster sampling tech-
more upon gender differences among the mar- niques. Male and female family heads were
ried than the unmarried. Because separated interviewed on an alternate basis. (Because
and divorced women are more likely to con- women are more often single heads of house-
tinue living with their children and perhaps hold, the final sample was 60.2% female.) The
more likely to retain close involvements with first wave of interviews was conducted in
relatives than comparable men, one might ex- 1972-73. Four years later, among the 88 per-
pect unmarried women to exhibit a mental cent of the respondents who agreed to be rein-
health advantage over their male counterparts, terviewed, 82 percent of those who could be
contrary to recent evidence. The structure of located were interviewed again (N= 1349 or 67
role-involvements among the unmarried has percent of the original sample). Thus, the
not been explored in a manner parallel to that Chicago panel consists of 1,106 respondents.
among the married. Lower proportions of young, nonwhite males
Finally, Bernard (1972) and Gove (1972, of lower income are present in the panel than in
1984; Gove and Tudor, 1973) have long argued the original sample (see Pearlin and Lieber-
that marriage is more beneficial for men than man, 1979 for details). The analysis of this
for women. As evidence, for example, Gove paper is based upon the panel sample only.
points to the lower incidence and prevalence of In New Haven, 1095 adult men and women
mental illness among husbands compared to were selected at random from the catchment
wives and (perhaps erroneously) to the higher area of a community mental health center. A
rates of mental illness among unmarried men total of 938 were interviewed successfully in
compared to unmarried women. However, 1967. Two years later, 720 of the original
since the effects of specific roles on mental cohort were reinterviewed. (The panel was
health may be confounded with the number of 55.8 percent female.) The panel sample does
roles held by each gender differentially, it is not differ significantly from the original sample
not clear whether marriage itself is responsible in sociodemographic characteristics, with one
for the mental health advantage of men. Em- exception: the under-30 age group dropped
ployment, parenthood, or other roles may in- from 25 percent in 1967 to 19 percent in 1969
stead be the source of differential benefit. Con- (Myers et al., 1974). The analysis reported here
sequently, it is necessary to compare the psy- is based on the panel sample of 720.
chological states of men and women who hold The Chicago and New Haven samples are
highly similar in sex, age, and marital status Likert-type items are reverse-coded when ap-
composition (mean ages are 45.6 and 44.9, re- propriate and summed; scores range from 20
spectively; 75 percent and 75.3 percent are (little or no distress) to 80 (high distress).
married, respectively). The Chicago sample in- Cronbach's alpha is satisfactorily high for the
cludes more nonwhite respondents (18.8 per- index, approximately .84 or above at each time
cent) than New Haven (12.9 percent). The two point. Validity studies have shown that this
samples differ most in socioeconomic status. instrument and its variants (e.g., Langner,
In Chicago, the average respondent has had 1962) discriminate significantly between com-
some specialized training after high school munity residents and psychiatric patients with
and has an income in the range of $12,000 to neurotic symptoms. Individuals scoring high
$13,999. In New Haven, the average respon- on the index are likely to be those whose psy-
dent has had 10 to 11 years of formal schoolingchological state impairs their everyday func-
and has an income in the $7,000 to $8,999 tioning to some degree.
range. Thus, the Chicago sample is somewhat Any differences in effects found between the
better educated and has a higher mean family two cities might be attributable to instrument
income than the New Haven sample. differences. The depression subscale of the
From both studies measures of psychologi- Johns Hopkins Symptom Checklist (used in
cal distress, role-involvements, and socio- Chicago) differs most from the Gurin index
demographic characteristics are drawn. These (used in New Haven). The depression scale
are described below. consists primarily of psychological items (e.g.,
Psychological Distress. Two measures of feelings of loneliness, hopelessness), while the
psychological distress are available in the Gurin items are heavily psychosomatic. The
Chicago study. Respondents reported how anxiety scale is more similar to the Gurin
often in the preceding week they had experi- index; it contains a number of psychophysio-
enced each of 23 symptoms, identified from the logical symptoms-7 of which are the same on
presenting symptoms of patients receiving both scales (e.g., headaches, upset stomach,
psychiatric treatment (Derogatis et al., 1971). trembling hands). Thus, the anxiety scores of
When factor analyzed, these symptoms form respondents in Chicago should most closely
two factors, termed anxiety (12 symptoms) and parallel the Gurin scores of respondents in
depression ( 11 symptoms) (Pearlin and New Haven.
Lieberman, 1979). A simple sum of the Link and Dohrenwend (1980) have examined
Likert-type responses to each set of itemsthe
wascorrelations between scores on various in-
computed, as measures of anxiety and depres- struments. They calculate a correlation of .80
sion at Time 1 and Time 2, respectively.2 These between the SCL-90 (which is most closely
measures of distress have satisfactory levels of related to the anxiety and depression subscales
internal consistency at both time points; Cron- used in Chicago) and the Langner index (which
bach's alpha exceeds .80 for each measure. is most similar to the Gurin index used in New
Although the scores on the anxiety and depres- Haven). Although the instruments do differ in
sion subscales cannot be considered measures items, it is likely that all three tap an underlying
of psychiatric disorder per se, these measures factor of generalized psychological distress and
reasonably may be interpreted as indicators of should behave in similar ways.
experienced psychological distress. Anxiety Role-Identities. Because data are limited
and depression correlate strongly (r = .65). But with respect to specific numbers of role re-
because the symptoms making up these scales lationships enacted by individuals in each
form meaningful and separate factors, and be- study, identities are measured here as the sum
cause previous analyses indicate that a variety of social positions held by the individual at
of factors affect these dependent measures each time point. The following six social posi-
differently, analyses are performed for anxiety tions could be held in the Chicago panel:
and depression separately. spouse, parent, employee, relative, friend, and
In New Haven, generalized distress is mea- group member. Specifically, the "number of
sured with an instrument developed by Mac- identities" variable is incremented by " 1" if the
millan (1957) and further modified by Gurin et respondent is married, has children under the
al. (1960). The index consists of 20 psychologi- age of 18 at home, is employed full or part
cal and psychosomatic symptoms. The
I Active parenting is assumed when children under
the age of 18 are at home. This indicator of par-
2 A large number of respondents refused to answer enthood status is used to increase variability in the
one depressive symptom question ("In the past identity-accumulation factor. Roughly 90% of all
week, how often did you lose sexual interest or plea- married respondents are parents in each panel study
sure?"). To reduce the number of missing cases due when any child is used as an indicator of parent
to this symptom, the mean for the sample was sub- status. Sixty to 70%o are parents when children under
stituted when an answer was refused. 18 are specified.
time, has relatives in the area with whom the Sociodemographic Characteristics. In each
respondent gets together at least occasionally, set of data, gender and marital status are indi-
has good friends in the area, and belongs to at cated by dummy variables. "Female" is coded
least one club or organization (including "1" if female, "0" if male. "Married" is coded
church). The identity accumulation variable "1" if married, "O" otherwise, at each time
thus ranges in value from 0 to 6, and is mea- point. Different categories of the unmarried
sured at each time point, four years apart. In (never married, separated or divorced, and
some analyses, each potential identity is widowed) are also coded as dummies (e.g., " 1"
treated as a dummy variable, coded " F" if held, if widowed, "O" otherwise) for unmarried sub-
and "0" otherwise. group analyses. Comparisons between the
In the New Haven panel, eight social posi- genders within marital status categories are
tions could be held: spouse, parent, employee, made in the analyses that follow.
student, organizational member, church To partially control for variations specific to
member, neighbor, and friend. Specifically, the different unmarried statuses and to hold con-
identity-accumulation variable is successively stant certain extraneous factors, four demo-
incremented by "1" if the respondent is mar- graphic characteristics are controlled in each
ried, has children under the age of 18 at home, analysis: age measured in years, race measured
is employed full or part time, is in school, at- as a dummy variable (" 1" if nonwhite, "O" if
tends organizational meetings at least occa- white), education measured ordinally at Time 1
sionally, attends church services at least occa- (7 categories in each panel), and family income
sionally, visits neighbors, and has two or more measured ordinally at Time 1 (14 categories in
close friends. Thus, in New Haven, the Chicago, 9 in New Haven).5
identity-accumulation variable ranges in value
from 0 to 8 and is measured at two time points,
RESULTS
two years apart. Again, in some analyses, each
identity held is treated as a dummy variable. Table 1 indicates the mean number of role-
Admittedly, these identity indicators are identities held, by gender and by gender within
crude, in three ways. First, individuals were various categories of marital status in the two
not asked directly whether they define them- panel studies. As expected, the mean number
selves as spouses, parents, workers, and so on. of identities differs by gender; in general, men
Self-definition in terms of these positions is possess significantly more role-identities than
assumed. Second, to each position may be at- women, and married men's identities out-
tached numerous roles, what Merton (1957) number those of married women in both sam-
has called the "role set." To take a familiar ples. Contrary to expectation, unmarried men
example, professors are often both teachers and women do not differ in their number of
and researchers. Other positions usually entail role-involvements (as measured here), and this
a variety of distinct role relationships, too. holds within specific unmarried statuses as
However, the data do not permit this determi- well.
nation. Finally, more specific positions can- Table 2 summarizes the correlations of gen-
not be assessed for their presence or ab- der with psychological distress and with
sence-e.g., holding an organizational office number of identities, each measured at Time 1
or particular kin relationships. In short, the and Time 2. Within-marital status correlations
identity variables tap only gross variations in are also shown. (In each case, a positive cor-
the presence or absence of major social re- relation coefficient indicates that females have
lationships in different life domains. However,
in previous research (unpublished) with an-
pie, the relationship between identities and distress
other adult sample the author asked respon-
might be quite different in form from the U-shaped
dents to rate the importance of the positional
relationship reported in the Results section.
identities examined here. Without exception, Moreover, if specific roles in the role sets attached to
respondents' mean ratings indicated that these each position could be counted, more complex ef-
identities were either "important" or "very im- fects of each identity on distress might be revealed.
portant" to them. It seems safe to assume that For example, the greater the size of the role set, the
although the identity measures are not direct, more important or salient an identity might be, and
detailed, or complete indicators of social self- thus, the greater the impact of that identity on psy-
chological well-being (Stryker and Serpe, 1982).
conception, they do tap a range of important
Without extensive or detailed measures of identity
role-identities an individual can hold.4
and associated role sets, the findings here must be
viewed with caution as exploratory and suggestive in
4 The limited number and types of role-identities nature.
examined here must be kept in mind when interpret- 5 The mean income category was substituted for
ing the findings of the paper. If the number of iden- any refusals, to reduce the number of missing cases
tities possessed could range from 0 to 12, for exam- in the analyses.
Table 1. Mean Number of Identities at Time I by Gender and Marital Statusa
Chicago Sample New Haven Sample

Total Sample N Mean St.D. Diff. N Mean St.D. Diff.
Men 440 4.5 1.2 315 4.7 1.5
Women 666 4.0 1.1 p<.001 397 4.1 1.5 p<.001
Married
Men 358 4.8 .9 258 5.1 1.3
Women 471 4.4 .9 p<.001 277 4.6 1.3 p<.001
Unmarried (All)
Men 82 3.0 .9 57 3.1 1.3
Women 195 3.0 .9 p=NS 120 2.9 1.2 p=NS
Divorced/Separated
Men 16 3.1 .9 7 3.1 1.3
Women 72 3.2 .9 p=NS 32 3.1 1.2 p=NS
Widowed
Men 9 3.1 1.4 15 2.1 1.1
Women 77 2.9 1.0 p=NS 53 2.6 1.3 p=NS
Never Married
Men 57 2.9 .8 35 3.5 1.1
Women 46 3.0 .8 p=NS 35 3.3 1.1 p=NS
a Marital status is measured at Time 1.
more distress or more identities

unmarried subgroups are examined, with than
the m
negative correlation indicates that females exception of widowed persons in Chicago and
have less, or, equivalently, that males have single persons in New Haven. However, data
more of the factor in question.) Two patterns not shown here indicate that women's mean
are evident in Table 2. distress scores are higher than men's within
First, the data replicate recent prevalence each specific unmarried status in both panel
findings with respect to the distribution of dis-
samples and at both time points.
tress by sex within marital status. Women Second, generally speaking, individuals who
score significantly higher on distress scales exhibit higher psychological distress also pos-
than men, regardless of marital status, when sess fewer identities. This holds for both total
marital status is dichotomized (married versus samples as well as for married individuals. The
unmarried). Due to small numbers of men in exception to the rule, as seen earlier, concerns
specific unmarried statuses, significant sex unmarried females and males; although unmar-
differences in distress are not obtained when ried women are more anxious, depressed, and
Table 2. Correlations of Gender (Male = 0, Female = 1) with Distress Scores and Number of Ident
Time 1 and Time 2, by Marital Statusa
Time I Time 2 Time I Time 2 Time 1 Time 2

Chicago Sample Anxiety Anxiety Depression Depression Identities Identities N
Total Sample .18*** .13*** .22*** .13*** -.20*** -.18*** 1106
Married .20*** .10** .24*** .11*** -.22*** - .18*** 829
Unmarried (All) .13** .14** .13* .13* .03 -.02 277
Divorced/Separated .09 .07 .11 -.04 .06 .07 88
Widowed .19* .140 .140 .160 -.07 .07 86
Never Married -.04 .12 .01 .11 .04 .002 103
Time 1 Time 2 Time 2 Time 2
New Haven Sample Gurin Gurin Identities Identities N
Total Sample .14*** .11** -.20*** - .4*** 720
Married .13*** .09* -.18*** -.09* 542
Unmarried (All) .14* .14* -.07 -.05 178
Divorced/Separated .04 .02 -.03 .07 39
Widowed .06 .09 .16? .15 69
Never Married .19? .17? - .11 -.02 70
? p l 10. * p S .05. ** p .0l. *** p S .001.
a Marital status is measured at Time 1.
generally distressed than unmarried men, they cients, a positive beta coefficient for "female"
do not differ from the men in number of iden- in Table 3 indicates that women are more dis-
tities held. Thus, Tables 1 and 2 only partially tressed than men.)
and indirectly support the initial hypotheses of Two general patterns can be discerned in
this paper. Those who are more psychologi- Table 3 with respect to the effects of identity
cally distressed (women in general and married accumulation in general. First, as expected,
women) also possess fewer identities. How- role-identities reduce anxiety, depression, and
ever, unmarried women represent an exception generalized distress; the identity summation
to expectations. variable is consistently negative in sign and
Do these differential distributions of roles usually significant under each dependent vari-
account for gender differences in psychological able in all panels of Table 3, except for
distress? To examine this question, a series of widowed persons in New Haven (panel E) and
hierarchical regression analyses were per- for the never married (panel F). Second, there
formed first for each total sample, then sepa- is curvilinearity in the relationship between
rately for the married and unmarried, and fi- number of identities and distress. In combina-
nally for each unmarried status. Specifically, tion with the negative identity summation
distress at Time 2 (measured as anxiety, de- coefficient, the positive sign of the squared-
pression, or generalized distress) was re- identities coefficient in each column 2 of Table
gressed on sets of variables entered in the fol- 3 indicates a U-shaped parabolic function; in-
lowing order: (a) sociodemographic factors dividuals with very few and very many role-
(gender, age, race, education, family income), identities are more distressed. (Note that this
(b) number of role-identities at Time 1, (c) pattern is reversed for widowed persons in
number of identities at Time 1 squared (to New Haven [panel E] and never married indi-
check and control for curvilinearity), (d) viduals in panel F, although the coefficients are
changes in identities over time (measured as not significant in these panels. The multicol-
the number of identity gains and losses, re- linearity of the two variables and small num-
spectively), and (e) each individual's distress bers of cases in these comparisons likely ren-
score at Time 1 (anxiety, depression, or gener- der these coefficients unstable.)7
alized distress, depending on the dependent By substituting successive values of the
variable of the equation). In essence, the identity summation variable into the column 2
analyses take advantage of the longitudinal equations, it can be shown that the bend in the
nature of the data, first examining the effectscuvilinear
of relationship occurs between five and
Time 1 identities on Time 2 distress (with ex- six identities in both the Chicago and New
traneous factors held constant), then intro- Haven samples. That is, five role-identities ap-
ducing identity changes over time, and finally, pear to be optimal for psychological well-
adding Time 1 distress scores to control for the being; individuals who hold six or more iden-
level of pre-existing disturbance for each indi- tities increasingly become subject to difficul-
vidual.6 In Table 3, the relevant results of ties that partially offset the benefits of role
equations a, c, and e are summarized in col- summation. Because the average respondent in
umns 1, 2, and 3 respectively, for each depen- each sample possesses fewer than six role-
dent variable. (As with the correlation coeffi- identities (see Table 1), and because it is inap-
propriate to extrapolate a relationship beyond
6 A reader might question the logic of these hierar- the range of actual values obtained for a vari-
chical analyses. Basically, a Lazarsfeldian approach able (see fotonote 4), interpretive emphasis will
is taken here (see Kendall and Lazarsfeld, 1950). An
initial or target relationship (that between gender and
distress) is first established, net of other background 7 In previous analyses with the New Haven data
factors. Then potential explanatory factors are held (Thoits, 1983) curvilinearity in the relationship be-
constant successively, in order to examine changes tween identities and distress was not found. Two
that occur in the target relationship as a result. De- changes from the previous analyses may account for
pending upon the type of variable held constant (e.g., the appearance of curvilinearity in the New Haven
intervening variable, antecedent variable) and data here. First, in the previous analyses, race was
changes that occur in the target relationship (e.g., not controlled and it is here. Second, and more im-
reduction in size, reversal of sign, the emergence of a portantly, in the original paper, parenthood was as-
significant association), different inferences can be sessed simply by whether an individual had had chil-
drawn about causality (e.g., the variable held con- dren or not. In this paper, active parenting is as-
stant explains the relationship, makes it spurious, sessed by the presence of children under 18 in the
suppressed it). In the analyses described here, inter- home (see footnote 3). Because nearly all respon-
vening (explanatory) variables are introduced suc- dents were parents when any child was used as an
cessively, then Time 1 distress is introduced to check identity indicator, this restricted the variation in the
for spuriousness in the target relationship and in the original identity summation factor. Restricted varia-
relationships of the intervening variables with psy- tion very probably accounts for the failure of cur-
chological distress. vilinearity to appear in the previous analyses.
Table 3. The Effects of Number of Identities at Time 1 on Time 2 Distressa
CHICAGO SAMPLE NEW HAVEN SAMPLE
Time 2 Time 2 Time 2

ANXIETY DEPRESSION DISTRESS
A. Total Sample (1) (2) (3) (1) (2) (3) (1) (2) (3)
Female (0,1) .11*** .11*** .04 13*** .12*** .05$ .09* .06 .02
No. of Identities - -.69*** -.49** - -.94*** -.90*** - -.38* -.23
Identities Squared - .58*** .38* 277*** .72*-* - .23 .12
Gains- - -.02 - - 06$ - - .0*
Losses - - .10 - - .14*** - - .08
Time I Distress - - .35*** - - .27*** - - .49*5
R2 (adjusted) .04 .06 .18 .03 .07 .16 .04 .05 .29
N 993 993 693
B. Married'
Female (0,1) 10** .10** .02 .13*** .12** .04 .09* .06 .01
No. of Identities - -.47$ -.10 - -.37 -.32 - -.61** -.43*
Identities Squared - .43 .09 - .32 .24 - .48* .29
Gains - - .03 - - -.02 - - -.13
Losses - - .09** - - .17*** - - .11*
Time I Distress - - .39*** - - .25*** - - .47
R2 (adjusted) .03 .03 .18 .03 .03 .12 .03 .05 .29
N 754 754 521
C. Unmarried (All)b
Female (0,1) .10 .10 .07 .09 .11 .08 .12 .12 .07
No. of Identities - -.64* -.69* - -.85** -1.00*** - -.15 -.11
Identities Squared - .58* .59* - .69* .78** - -.09 -.04
Gains - - -.10 - - -.12$ - - -.05
Losses - - .07 - - .08 - - .02
Time I Distress - - .28*** - - .29*** - - .51
R2 (adjusted) .03 .04 .12 .02 .06 .15 .03 .07 .30
N 239 239 172
D. Divorced or Separatedh
Female (0,1) .01 -.003 .003 -.01 .01 .01 .07 .06 .08
No. of Identities - -.88 -1.28$ - -1.22$ -1.38* - -.88 -.24
Identities Squared - .66 .93 - .92 1.01 - .53 .19
Gains - - -.22$ - - -.06 - - -.23
Losses - - .19 - - .11 - - .05
Time I Distress - - .15 - - .03 - - .71
R2 (adjusted) .02 .02 .06 -.03 .07 .04 -.08 .01 .50
N 72 72 37
E. Widowed
Female (0,1) .11 .15 .10 .10 .14 .08 .11 .12 .07
No. of Identities - -.98* -.87$ - -1.11* -1.03** - .05 .05
Identities Squared - .94* .84$ - .89$ .64$ - -.31 -.25
Gains - - -.05 - - -.14 - - .13
Losses - - .04 - - .17 - - .01
Time I Distress - - .37** - - .55*** - - *43**
R2 (adjusted) -.01 .02 .13 -.01 .08 .44 -.03 -.01 .13
N 78 78 67
F. Never Married'
Female (0,1) .14 .14 .12 .11 .12 .08 .17 .15 .06
No. of Identities - .05 .05 - .43 .20 - .36 .52
Identities Squared - -.06 .01 - -.47 -.19 - -.53 -.63
Gains - - .03 - - -.02 - - -.06
Losses - - -.05 - - -.06 - - -.02
Time I Distress - - .31** - - .26* - - .44
R2 (adjusted) .07 .05 .12 .03 .01 .05 .07 .08 .24
N 89 89 68
: p .10. * p .05. ** pz.01. *** p .001.

a Each equation controls for age, race, education, and family income. Standardized coefficients are
reported.
I Marital status is measured at Time 1.
be placed here on the linear effects of iden- see that in Chicago, women in general and mar-
tities. Subsequent analyses will simply control ried women (panels A and B) are still
for the curvilinear trend. significantly more anxious and depressed
Turning to the issue of explaining gender than men, despite controls for the differential
differences in distress, mixed findings are ob- possession of role-identities. However, in New
tained in Table 3. The associations between Haven, women's greater distress disappears
gender and distress should become smaller or when the number of identities possessed is
nonsignificant when the number of identities controlled (panels A and B). The gender coeffi-
possessed by each individual is controlled in cients for unmarried individuals (in panels C
column 2. Examining the gender coefficients in through F) are not significant, but this is due to
columns 1 and 2 of each panel in Table 3, we the effects of sociodemographic variables (age,
race, education, income) controlled in each the identity accumulation coefficient for anxi-
equation 1, rather than due to the effects of ety. Possession of numberous roles may de-
identities. Similar results are obtained when pend on previous superior functioning-
the six or eight dummy variables for identities functioning reflected by selection into marriage
are substituted for identity summation (data as well. But on the basis of this reasoning, one
not shown). In short, the guiding hypothesis of would expect similar findings among the un-
the paper is only partially confirmed by these married, where previous inferior functioning
results. Specifically, the possession of fewer should have selected these individuals out of
identities appears to explain the higher distress various roles. But this does not occur. Conse-
of women in general and of married women in quently, it seems safe to conclude that social
New Haven, but not in Chicago. Among the selection effects are not a serious threat to the
unmarried, women's higher distress in both analyses of this paper. It is likely, as is shown
samples is a function of background variables below, that these apparent selection effects for
(e.g., lower family income). Thus, after con- the married in Chicago can be attributed to the
trols are introduced for these differences, unstable impacts of specific identity configura-
married men and women exhibit equivalent tions over time.
levels of distress and possess equivalent num- Because gender differences in distress are
bers of identities (Table 1). This might be not consistently reduced by identity accumu-
viewed as support for the guiding hypothesis of lation, it becomes necessary to explore why.
the paper (equal numbers of identities should One underlying assumption of the analysis to
produce equal degrees of distress) but is not this point has been that males and females at
compelling support. Additional exploration is each level of identity summation possess simi-
needed. lar roles. But perhaps this is not the case.
It is relevant to note that in Chicago, when Perhaps the number of identities held masks
Time 1 anxiety or depression is controlled in quite different configurations of roles for men
each column 3, the associations between gen- and women, which in turn might help explain
der and distress become nonsignificant and differences in distress. A second implicit as-
near-zero in size in the total sample (for anxi- sumption has been that each role has equiva-
ety) and in the married subsample as well (for lent impacts on the psychological states of men
both anxiety and depression). Hierarchical and women. But the same roles may not carry
analyses (not shown) verify that gains and similar meanings for men's and women's self-
losses of identities over time are not responsi- conceptions, and thus, for their psychological
ble for this result-the gender coefficients be- well-being. To assess these alternative pos-
come nonsignificant only when prior level of sibilities, it seems advisable to re-examine the
distress is controlled in the last step of the existence of gender differences in distress for
analysis. These findings suggest that women those men and women who hold the same
are more distressed at Time 2 because they number and the same configurations of role-
were already more distressed at Time 1. That identities.
is, some stable, systematic difference between For a preliminary exploratory analysis along
the genders in Chicago appears to account for these lines, only the Chicago panel is used
their higher levels of anxiety and depression. since the sample is larger. Only individuals
As is shown below, this stability is likely due to who possessed friend and kin relations were
the role configurations held differentially by selected for analysis (N=787), since over 80
men and women. percent of all respondents, regardless of mari-
It is important to add that a social selection tal status, held each role (data not shown). This
explanation of the relationship between multiple reduced the number of role combinations it
identities and psychological well-being is gen- would be necessary to examine. At each level
erally ruled out by the findings in each column of identity summation (levels 3 through 6), for
3 of Table 3. If an individual's number of iden- individuals holding each possible combination
tities were due to his or her prior level of psy- of spouse, parent, employed worker, and
chological functioning, a control for prior dis- group member roles, hierarchical equations
tress in each column 3 equation should reduce were estimated. Time 2 anxiety or depression
previously significant identity summation was regressed first on background variables
coefficients to nonsignificance. This occurs (gender, age, race, education, and family in-
only twice in the 18 analyses (in panel A for come), then on the number of identities gained
New Haven and in panel B for Chicago). In the and lost over time, and finally on anxiety or
New Haven total sample, the size of the iden- depression at Time 1.8 For simplicity in pre-
tity accumulation coefficient is not reduced
substantially. However, in panel B, among the 8 No control for curvilinearity is necessary in this
married in Chicago, a selection effect is sug- analysis, as the number of identities is a constant for
gested by the drastic reduction in the size of each equation.
sentation, only the gender coefficients from ble and thus difficult to interpret with confi-
these three steps are presented in columns 1, 2, dence. Second, many role-combinations are
and 3 of Table 4. highly gender-specific-70 percent or more of
It should be noted that these and the next those holding particular numbers and combi-
analyses focus explicitly on interaction effects nations of roles are female (or, in three in-
(the interaction of gender with specific role stances, male [panels 5a, 5c, and 6]). Conse-
combinations). Subsets of individuals are quently, it is very difficult to disentangle gen-
selected first because they possess the same der differences in distress from particular
number and combinations of roles. Then sex role-combination effects. In fact, this is
differences in distress for each subset of indi- perhaps the most telling finding of the paper
viduals are examined. This analysis is equiva- thus far. The role constellations of men and
lent to adding interaction terms to the Table 3 women at each level of identity accumulation
equations-terms for gender multiplied by are so distinct that an attempt to explain gender
each role and gender multiplied by each role differences in distress in their terms may be
combination. This was not done because doomed.
numerous interaction terms drastically in- Third, the most striking and consistent dif-
crease the problem of multicollinearity. Sub- ferences in the role-identities held by men and
dividing the sample into relevant groups and women lie in parenthood and employment.
examining sex differences within those groups Men predominate in almost all configurations
yields the same information about interactions that include employment. Women are over-
without the inefficiencies introduced by mul- whelmingly predominant in configurations that
ticollinearity. Each gender coefficient in Table include parenthood but exclude employment.
4, then, represents an interaction effect (i.e., These patterns again suggest that disentangling
the differential impact of certain role combina- sex differences in distress from differential
tions on the psychological distress of women role-involvements may be extremely difficult,
compared to men). with the data available here. There may be too
Three observations can immediately be few cases of males compared to females (or
made on the basis of the results in Table 4. vice versa) in each comparison to trust the
First, some combinations of role-identities are findings. Nevertheless, overall, there are sur-
surprisingly rare, even in a large community prisingly few significant differences between
sample such as this. Small numbers of cases in men and women who share the same number
many equations may render coefficients unsta- and combinations of roles, which in turn sug-
Table 4. Gender Differences in Distress for Individuals Holding the Same Number and Combinations of
Identities at Time la (Chicago Sample Only)
ANXIETY ~~DEPRESSION
Number of Identity Percent ANXIETY
Identities Combination N Female (1) (2) (3) (1) (2) (3)
3 Friend, Relative, and:

a) Spouse 38 76.3% .46* .46* .44* .55** .55** .47**
b) Parent 27 100.0o - - - - - -
c) Employed Worker 46 45.7% -.04 -.02 -.03 .09 .10 .10
d) Group Member 22 68.2% .005 -.13 -.09 .36 .29 .32

a) Spouse, Parent 96 95.8% -.09 -.09 -.07 -.09 -.10 -.09
b) Spouse, Worker 66 39.4% .23# .14 .11 .27* .16 .13
c) Spouse, Group Member 37 73.0%o .04 .02 -.01 .01 -.02 -.004
d) Parent, Worker 20 95.0o .21 .19 .10 -.02 -.04 -.004
e) Parent, Group Member 6 100.0%o - - - - - -
f) Worker, Group Member 34 58.8% .09 .16 .16 .06 .12 -.08

a) Spouse, Parent, Worker 107 28.0o .23* .21* .21* .07 .05 .07
b) Spouse, Parent,
Group Member 111 99.1% .03 .02 .02 .07 .07 .06
c) Spouse, Worker,
Group Member 47 29.8% .01 .01 -.03 .20 .20 .14
d) Parent, Worker,
Group Member 10 80.0% .38 -.11 -.002 -.01 -.47 -.39

Spouse, Parent, Worker,
Group Member 120 30.8% .20* .16$ .12 .160 .14 .08
# ps.10. * ps.O5. ** p .Ol.

NOTE: Standardized coefficients are reported.
a Each equation controls for age, race, education, and family income. Only gender coefficients are
reported. Column 1 equations control for background variables only. Column 2 equations additionally control
for identity gains and losses over time. Column 3 equations have a control added for Time 1 distress.
gests that these roles are equally meaningful

for male and female self-conception.
However, there are some hints in Table 4
that at some levels of identity summation, the
same roles are not equally meaningful, and
thus not equally protective, for men and
women. In particular, the employed worker
identity appears to distress women more than
men, especially when combined with marriage
(panel 4b) or with marriage plus parenthood eD Z S - aa s = CD
(panels 5a and 6), although these effects tend to
disappear when changes in identities and prior
distress levels are controlled in equations 2 and ca' o) . ll0
3 (see panels 4b and 6). (For men and women O ~ ~ ~~. 20 W BO

who are spouses only [panel 3a], women ap-
pear more distressed; however, this compari-
son is based on only 9 men and 29 women.)
More statistical power can be obtained by *DE -> .o N r2
simplifying these comparisons-i.e., by com-
paring men and women who hold the same
E ; o. I oes st < e Z
numbers and configurations of three basic
role-identities (spouse, parent, and employed
worker), while holding the effects of other
E | 1 < | | | l | | Y X '~~~~~~.(
identities constant. Table 5 summarizes these
simplified comparisons for both Chicago and
New Haven.
Table 5 again presents gender coefficients
only. Each column I under the dependent vari-
able indicates the effect of gender on distress at
Time 2, controlling for background variables
and the effects of other role-identities, indi- Q ^ t o x vo t o x . = ~~~~~~~~~ct )
cated by dummy variables (e.g., friend, rela-

tive, church member). In each column 2, losses
and gains of identities over time are added as
controls. In each column 3, distresses at Time
I has been added as a control. Analyses were
performed separately for individuals holding 0,
1, 2, and 3 basic identities in the combinations
specified in Table 5.9 It should be recalled that
these analyses focus explicitly on the interac- co I z I> X ...I x (
tions of gender with various roles and role
combinations.
Once again we see highly skewed distribu-
tions of major role configurations by sex, in a J -> S z ' = >
both samples. So again, coefficients must be
interpreted with substantial caution. Never-
theless, it is possible to see certain patterns in
the findings. In New Haven, women who are red > >>t : )z
spouses only or parents only are significantly
less distressed than comparable men (panels I a
and I b-although note that in panel lb only 2
men are compared to 29 women), while women
who are employed workers only are signifi-
cantly more distressed than comparable men
(panel Ic). In Chicago, women who hold
spouse and parent roles simultaneously are
significantly less anxious than comparable men
9 The findings in Table 5 are not altered in any way

when the identities-squared term is controlled in
each equation.
(panel 2a). It is difficult to account for dis- from identity accumulation. This paper repre-
crepancies between New Haven and Chicago sents an attempt to apply this general finding to
with respect to the effects of holding single and the phenomenon of sex differences in psycho-
dual family roles for men and women. It is logical disturbance. As a straightforward ex-
possible that the two cities differ in ethnic tension, it seemed reasonable to propose that
composition (by religion, nationality) and that the higher prevalence of anxiety and depres-
ethnic differences in the importance attached sion among women, especially among married
to traditional family roles account for these women, might be explained at least in part by a
patterns. It is also possible that the lower edu- relative lack of role-identities.
cation and income of New Haven respondents This research documented clear sex dif-
might explain these patterns. But small and ferences in psychological distress and in iden-
highly skewed numbers of cases in these com- tity accumulation (in general, and among mar-
parisons probably make these city differences ried individuals, but not among unmarried
unreliable. The patterns of effects in Table 5 as persons), reconfirmed the anxiety- and de-
a whole simply suggest that family and work pression-reducing effects of identities, and
identities are differently meaningful and thus substantiated a curvilinear (U-shaped) re-
differentially protective for men and women. lationship between multiple identities and
Specifically, marriage and family roles appear symptoms of distress. But the number of iden-
to protect women psychologically while em- tities held by men and women did not consis-
ployment appears to protect men. It is also tently account for their differences in distress
clear that women who simultaneously hold in the two community samples. This inconsis-
spouse, parent, and employed worker roles are tent finding turned attention to the questions of
more anxious, depressed, and distressed in whether men and women hold equivalent roles
general than comparable men, although only at each level of identity accumulation and
the anxiety coefficients remain significant afterwhether the same roles carry equivalent
subsequent controls for identity-change and meanings for self-conception.
distress are introduced (panel 3). In short, one Subsequent analysis compared the anxiety,
might say that employment complicates life for depression, and distress scores of individuals
those women who possess both marriage and holding the same number and the same config-
family roles, while lack of employment dam- urations of role-identities. The analysis re-
ages comparable men. These findings contrast vealed that men and women hold quite dif-
with the common contention that the spouse ferent roles at each level of identity accumu-
role benefits men more than women (Bernard, lation. In most cases, role configurations were
1972; Gove, 1972; Gove and Tudor, 1973). either too rare or too skewed by sex to make
Rather, it appears that the employed worker comparisons between males and females reli-
role is the basis of men's mental health advan- able. Overall, significant differences in distress
tage, or alternatively put, the basis of women's were uncommon between men and women
disadvantage, at least at certain levels of iden- who held the same number and types of role-
tity accumulation.10 identities, suggesting that identities carry
equivalent meaning for the sexes. However,
some indications were found that paid work
DISCUSSION AND CONCLUSIONS
benefitted men more than women. Among in-
Previous theoretical and empirical work has dividuals with one or both marriage and family
suggested that psychological benefit accrues roles, unemployed men were more anxious and
distressed than unemployed women. Among
10 To rule out the possibility that the same role employed married parents, on the other hand,
configurations held by men and women represent women were more anxious than men. The
different life-cycle stages (e.g., employed workers
most striking finding of the paper, however,
only who are young single women and older
appears to be the fundamental noncomparabil-
widowed men), the Table 5 analyses were repeated
ity of men's and women's role situations. This
for each sample when stratified by age (under 35
years, 36 to 50 years, and 51 years and over). The noncomparability has three implications.
results in Table 5 are strengthened by these analyses. First, systematic gender differences in dis-
Few significant differences between men's and tress may reflect systematic differences in
women's distress levels are found in either sample men's and women's role structures. Alterna-
among the under 35 or the over 51 age groups at any tive factors, such as differential expressivity,
level or configuration of role-involvements. Among personality traits, or constitutional factors
the 36 to 50 year old married parents, women are
certainly are not ruled out definitively in this
significantly less anxious and less depressed than
research. But the highly skewed distributions
comparable men (in Chicago). Among the employed
of role constellations are very suggestive. Un-
married parents in this age group, women are sig-
nificantly more anxious, more distressed, and more fortunately, the very noncomparability of role
depressed than comparable men. structures makes the confirmation or discon-
formation of that of unmarried women. Instead,

distress unmarried
differen
It is true that few significant differences be- men and women exhibit equivalent levels of
tween men and women were found when com- anxiety, depression, and distress, once age,
pared at the same number and configurations education, and income differences are con-
of role-identities, particularly with respect to trolled, and unmarried men exhibit equal or
depression levels (which may better reflect the lower distress than unmarried women at all
purpose and meaning derived from role in- levels and combinations of role accumulation.
volvements). But the relative scarcity of men Although employment may interact with
for many comparisons and small numbers of traditional family roles in complicated ways
total cases for each set of comparisons render that still must be clarified for both sexes, future
the implications of this finding very tentative. work might concentrate more usefully on gen-
Clearly, a different methodological strategy is der differences in psychological disturbance as
necessary to put a structural explanation of a function of employment, rather than of mar-
gender differences to a test. Systematic sam- ital status per se.
ples of men and women stratified by number
and types of role involvements would seem to REFERENCES
be necessary for confident conclusions (and Bachrach, Leona L. 1975. Marital Status and Men-
this strategy also would be appropriate for tal Disorder: An Analytical Review. U.S. Depart-
epidemiological studies of the prevalence of ment of Health, Education, and Welfare
(ADAMHA), Publication No. 75-217. Washing-
distress by sex more generally).
ton, D. C.: Government Printing Office.
Second, the highly skewed distributions of
Bernard, Jessie. 1972. The Future of Marriage. New
role configurations suggest that social and York: Bantam Books.
cultural forces may still limit women's access Coser, Lewis (with Rose Laub Coser). 1974. Greedy
to particular arenas, especially the public Institutions. New York: Free Press.
arenas. The upshot is that women possess Derogatis, L.R., R.S. Lipman, L. Covi, and K.
fewer roles overall, which puts them at an ini- Rickles. 1971. "Neurotic Symptom Dimensions."
tial mental health disadvantage. Further, the Archives of General Psychiatry 24:454-64.
prevalence of certain role patterns by sex may Dohrenwend, Bruce P., Barbara S. Dohrenwend,
Madelyn Schwartz Gould, Bruce Link, Richard
structure the meaning of various roles for
Neugebauer, and Robin Wunsch-Hitzig. 1980.
self-conception. Less common role configura-
Mental Illness in the United States: Epidemiologi-
tions may weaken the centrality of certain roles cal Estimates. New York: Praeger.
for self-conception, and thus render those roles Fox, John W. 1980. "Gove's Specific Sex-Role
less protective. For example, because em- Theory of Mental Illness: A Research Note."
ployment is less common among women, they Journal of Health and Social Behavior 21:260-67.
may derive from it less purpose and meaning. Gurin, Gerald, Joseph Veroff, and Sheila Feld. 1960.
However, it is certainly no secret that women's Americans View Their Mental Health. New York:
jobs are typically of lower status, offer lower Basic.
Goode, William J. 1960. "A Theory of Role Strain."
pay, and rarely result in the redistribution of
American Sociological Review 25:483-96.
house and childcare duties with a working hus-
Gore, Susan and Thomas W. Mangione. 1983. "So-
band. Consequently, women's higher distress cial Roles, Sex Roles and Psychological Distress:
when employment is combined with marriage Additive and Interactive Models of Sex Dif-
and parenthood may reflect role strain rather ferences." Journal of Health and Social Behavior
than less meaning and purpose derived from 24:300-12.
paid work. Gove, Walter R. 1972. "The Relationship Between
Finally, it has long been argued that marriage Sex Roles, Mental Illness, and Marital Status."
Social Forces 51:34-44.
is more beneficial for men than for women
. 1978. "Sex Differences in Mental Illness
(Bernard, 1972; Gove, 1972; Gove and Tudor,
Among Adult Men and Women: An Evaluation of
1973). On the contrary, this research indicates Four Questions Raised Regarding the Evidence on
that marriage is equally beneficial for men and the Higher Rates of Women." Social Science and
for women, if not more beneficial for women Medicine 12B: 187-98.
(see also Reskin and Coverman, 1985). If any . 1984. "Gender Differences in Mental and
role is a source of disadvantage for women Physical Illness: The Effects of Fixed Roles and
relative to men, employment is. Little evidence Nurturant Roles." Social Science and Medicine
19, 2:77-91.
in this analysis indicates that marriage in itself
Gove, Walter R. and Jeannette F. Tudor. 1973.
represents or carries a mental health disad-
"Adult Sex Role and Mental Illness." American
vantage for females. If it did, married women
Journal of Sociology 78:50-73.
at all levels and combinations of identity ac-
Kandel, Denise, Mark Davies, and Victoria H.
cumulation would exhibit higher distress than Raveis. 1985. "The Stressfulness of Daily Social
comparable men, and this was not found. Roles for Women: Marital, Occupational and
Moreover, if marriage were more beneficial forHousehold Roles." Journal of Health and Social
men, unmarried men's distress would exceed Behavior 26:64-78.
Kendall, Patricia L., and Paul F. Lazarsfeld. 1950. Research in Community and Mental Health,
"Problems of Survey Analysis." Pp. 133-96 in edited by Roberta Simmons. Greenwich, CT: JAI
Continuities in Social Research, edited by Robert Press.
Merton and Paul F. Lazarsfeld. Glencoe, IL: Free Pearlin, Leonard I., Morton A. Lieberman,
Press. Elizabeth G. Menaghan, and Joseph T. Mullan.
Langner, Thomas S. 1962. "A Twenty-two Item 1981. "The Stress Process." Journal of Health and
Screening Score of Psychiatric Symptoms Indi- Social Behavior 22:337-56.
cating Impairment." Journal of Health and Reskin, Barbara F. and Shelley Coverman. 1985.
Human Behavior 3:269-76. "Sex and Race in the Determinants of
Link, Bruce, and Bruce P. Dohrenwend. 1980. Psychophysical Distress: A Reappraisal of the
"Formulation of Hypotheses about the True Prev- Sex-role Hypothesis." Social Forces 63:1038-59.
alence of Demoralization in the United States." Sieber, Sam D. 1974. "Toward a Theory of Role
Pp. 114-32 in Mental Illness in the United States: Accumulation." American Sociological Review
Epidemiological Estimates, edited by Bruce P. 39:567-78.
Dohrenwend, Barbara Snell Dohrenwend, Made- Spreitzer, Elmer, Eldon E. Snyder, and David L.
lyn Schwartz Gould, et al. New York: Praeger. Larson. 1979. "Multiple Roles and Psychological
Linton, Ralph. 1936. The Study of Man. New York: Well-being." Sociological Focus 12:141-48.
Appleton-Century. Stryker, Sheldon. 1980. Symbolic Interactionism: A
Macmillan, Allister M. 1957. "The Health Opinion Social Structural Version. Palo Alto, CA:
Survey: Technique for Estimating Prevalence of Benjamin/Cummings.
Psychoneurotic and Related Types of Disorder in Stryker, Sheldon, and Richard T. Serpe. 1982.
Communities." Psychological Reports 3:325-39. "Commitment, Identity Salience, and Role Be-
Marks, Stephen. 1977. "Multiple Roles and Role havior." Pp. 199-218 in Personality, Roles, and
Strain: Some Notes on Human Energy, Time, and Social Behavior, edited by William Ickes and Eric
Commitment." American Sociological Review S. Knowles. New York: Springer-Verlag.
42:921-36. Thoits, Peggy A. 1983. "Multiple Identities and Psy-
Merton, Robert K. 1957. Social Theory and Social chological Well-being: A Reformulation and Test
Structure. Revised Edition. New York: Free of the Social Isolation Hypothesis." American
Press. Sociological Review 48:174-87.
Myers, Jerome, Jacob J. Lindenthal and Max Pep- . in press. "Multiple Roles: Social and Inter-
per. 1971. "Life Events and Psychiatric Impair- personal Implications." In Modern Women: Man-
ment." Journal of Nervous and Mental Disease aging the Dual Roles, edited by Faye Crosby. New
152:149-57. Haven, CT: Yale University Press.
. 1974. "Social Class, Life Events, and Psy- Verbrugge, Lois. 1983. "Multiple Roles and Physical
chiatric Symptoms: A Longitudinal Study." Pp. Health of Women and Men." Journal of Health
191-205 in Stressful Life Events: Their Nature and and Social Behavior 24:16-30.
Effects, edited by Barbara S. Dohrenwend and Warheit, George J., Charles E. Holzer, Robert A.
Bruce P. Dohrenwend. New York: Wiley. Bell, and Susan A. Arey. 1976. "Sex, Marital
Pearlin, Leonard I., and Morton A. Lieberman. Status, and Mental Health: A Reappraisal." Social
1979. "Social Sources of Emotional Distress." In Forces 55:459-70.

Thoits, Peggy Multiple Identities Examining Gender and Marital Status Differences in Distress

Uploaded by

Copyright:

Available Formats

Thoits, Peggy Multiple Identities Examining Gender and Marital Status Differences in Distress

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Thoits, Peggy Multiple Identities Examining Gender and Marital Status Differences in Distress

Uploaded by

Copyright:

Available Formats

Multiple Identities: Examining Gender and Marital Status Differences in Distress

Author(s): Peggy A. Thoits

*Address all correspondence to Peggy Thoits, De-

American Sociological Review, 1986, Vol. 51 (April:259-272) 259

Table 1. Mean Number of Identities at Time I by Gender and Marital Statusa

Chicago Sample New Haven Sample

more distress or more identities

Time I Time 2 Time I Time 2 Time 1 Time 2

Table 3. The Effects of Number of Identities at Time 1 on Time 2 Distressa

CHICAGO SAMPLE NEW HAVEN SAMPLE

Time 2 Time 2 Time 2

: p .10. * p .05. ** pz.01. *** p .001.

3 Friend, Relative, and:

4 Friend, Relative, and:

5 Friend, Relative, and:

6 Friend, Relative, and:

# ps.10. * ps.O5. ** p .Ol.

gests that these roles are equally meaningful

3 (see panels 4b and 6). (For men and women O ~ ~ ~~. 20 W BO

cated by dummy variables (e.g., friend, rela-

9 The findings in Table 5 are not altered in any way

formation of that of unmarried women. Instead,

You might also like

: p .10. * p .05. pz.01. * p .001.