Genero Deaux

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Journal of Personality and Social Psychology Copyright 1984 b'y the

1984, Vol. 46, No. 5, 991-1004 American Psychological Association, Inc.

Structure of Gender Stereotypes: Interrelationships


Among Components and Gender Label
Kay Deaux and Laurie L. Lewis
Purdue University
Various components of gender stereotypes were examined in three studies to de-
termine their interrelationship and their influence relative to gender label. In the
first two experiments, male and female college students were given information
about gender and either role behaviors or traits and were asked to assess the prob-
ability that the stimulus person possessed other gender-related characteristics. In
the third experiment, each of four gender stereotype components was presented
in a within-subjects design and subjects made judgments about each other com-
ponent. Results indicate that (a) information about one stereotype component can
implicate other components; (b) specific component information may outweigh
gender identification; and (c) components differ in their ability to implicate other
components of gender stereotypes, with physical appearance playing a dominant
role. The significance of these findings in understanding the structure and operation
of gender stereotypes is discussed.

Stereotypes play an important role in hu- Stitt, & Segal, 1980). In doing so, there is a
man judgment, and investigators have shown tendency to move away from the conception
sustained interest in exploring the content and of stereotypes as negative judgments that de-
consequences of such judgments. From Lipp- viate from some true state, and to move toward
man's (1922) early and insightful analysis of a more neutral view that emphasizes process
the "pictures in our heads," through the work and content rather than "rightness" and
of Katz and Braly (1933), Allport (1954), "wrongness."
Campbell (1967) and others, stereotypes have Such a change in perspective requires a
continued to hold the interest of social sci- change in the kinds of questions that are asked
entists concerned with ways in which we judge as well. Rather than providing simply a de-
and misjudge members of recognizable groups. scriptive account of the contents of gender ste-
In recent years, interest in the topic has ac- reotypes, current investigators have begun to
celerated, with new attention focused both on ask how various aspects of the overall stereo-
the general nature of stereotypes as well as on type are related to one another, how judgments
specific content categories (Hamilton, 1979, may be altered in the face of new and some-
1981; Miller, 1982). At the general level, in- times contradictory information, and the pro-
vestigators are increasingly coming to concep- cess by which such changes take place. To cite
tualize stereotypes as one particular instance only a few examples of this shifting focus, one
of more general cognitive processes (Ashmore can look to the work on prototypes, social
&DelBoca, 1981; Hamilton, 1979;McCauley, schema, and implicit personality theory by in-
vestigators such as Ashmore and Del Boca
(1979), Cantor and Mischei(1977), Rosenberg
and Sedlak (1972), Schneider and Blankmeyer
Support for this research and for the preparation of this (1983), and Howard and Rothbart (1980). Al-
article was provided by grants from Purdue Research though the specific emphases of these studies
Foundation and from National Science Foundation (BNS-
8217313). Additional support was provided by the Center vary, each is concerned with the more dynamic
for Advanced Study in the Behavioral Sciences and the process by which the interpretation of infor-
John D. and Catherine T. MacArthur Foundation. We are mation is affected by preexisting cognitive
grateful to Eliot Smith for his insightful comments on an structures.
earlier version of this article. Gender stereotypes are a particularly timely
Requests for reprints should be sent to Kay,Deaux,
Department of Psychological Sciences, Purdue University, and rich arena for this type of investigation.
West Lafayette, Indiana .47907. Although early investigators of stereotypes
991
992 KAY DEAUX AND LAURIE L. LEWIS

tended to focus on ethnic stereotypes, following asked to make judgments about the probability
the line of Katz and Braly (1933), more recent that the stimulus person possessed other gen-
work has included physical attractiveness, der-related characteristics. Information pre-
mental disorders, aging, and social class (cf. sented was either consistent with the sex of
Miller, 1982). Within the past decade, sex has the person (e.g., a male described in terms of
probably been the most active area of inves- masculine traits) or inconsistent (e.g., a female
tigation, reflecting a rise in all questions related described in terms of male-linked role behav-
to gender. Yet despite the frequent invocation iors). In the third study, using a within-subject
bf gender stereotypes as both cause and con- design, subjects were sequentially given infor-
sequence of various events, relatively little is mation about each component and were asked
known about their structure. to estimate the probability of each other com-
The most widely cited studies of gender ste- ponent. Through this method, it is possible to
reotypes are those of Broverman, Rosenkrantz, assess the possible reciprocal influence of
and their colleagues (Broverman, Vogel, Brov- components. One can ask, for example,
erman, Clarkson, & Rosenkrantz, 1972; Ro- whether traits influence judgments about
senkrantz, Vogel, Bee, Broverman, & Brov- physical characteristics as readily as infor-
erman, 1968). These investigators identified mation about physical characteristics will in-
two clusters of traits that are associated with fluence trait judgments.
women and men: warmth and expressiveness, In each of the three studies, it is also possible
believed by subjects to be more characteristic to compare the influence of gender label with
of women than men; and competence and ra- that of specific component information.
tionality, believed by subjects to be more char- Locksley and her colleagues (Locksley, Bor-
acteristic of men than women. Spence, Helm- gida, Brekke, & Hepburn, 1980; Locksley,
reich, and Stapp (1975) have derived two sim- Hepburn, & Ortiz, 1982) have recently shown
ilar factors. Common to both investigations is that specific case information can reduce the
an emphasis on personality traits as the core effects of stereotypes (or social category in-
of gender stereotypes. Consideration of lay ste- formation) on subsequent judgments. These
reotypes of gender, however, suggests that the authors argue that "as soon as individuating,
common conceptions of male and female are subjectively diagnostic characteristics of a per-
much more diverse, containing references to son are known, stereotypes may have minimal,
appearance, role behaviors, and sexuality. Ac- if any, impact on judgments of that person"
cordingly, Deaux and Lewis (1983) have re- (Locksley et al, 1980, p. 830). Similarly, Eagly
cently presented evidence for a multiple-com- and Wood (1982) have found that stereotypes
ponent construction. This research shows that about social influence rest on assumptions
a number of separate components of gender about differences in status. Consequently,
stereotypes can be identified: specifically, traits, when specific status information is supplied,
role behaviors, occupations, and physical ap- gender no longer is a major determinant of
pearance, each of which has a masculine and subsequent judgments.
feminine version. Although no component is The strength of this argument, however,
seen as the exclusive province of one or the would appear to depend in critical ways on
other sex, masculine and feminine components the types of information presented and the
are significantly more strongly associated with linkage between provided information and the
males and females, respectively. The like-sex subsequent judgment. For example, in the
components (e.g., male role behaviors and Locksley et al. (1980) study, subjects were given
masculine traits) bear some relationship to a simple gender label and three specific ac-
each other, but correlational analysis suggests counts of either assertive or passive behavior.
that they are best viewed as separate factors They were then asked to describe how the per-
that can vary independently. son would behave in situations "similar to
Using this component formulation as a base, those in the original dialogue" (Locksley et
the studies reported here attempt to determine al., 1980, p. 823). Given the near identity of
the degree to which information about one the information and the requested judgment,
component can implicate another. In the first the provided information did indeed have
two studies, subjects were presented with in- more influence than the gender label.
formation about one component and were Although Locksley and her colleagues
STRUCTURE OF GENDER STEREOTYPES 993

(1980,1982) have shown that passive behavior inference between different types of compo-
is more predictive of passive behavior than is nent information, for example, the degree to
gender (assuming passivity is part of the female which people may infer traits from information
stereotype), their approach is not a convincing about role behavior.
repudiation of stereotypes. Implicit in the idea In summary, the intent of the present studies
of stereotypes is the assumption that a large is to learn more about the structure of gender
number of beliefs are linked within the cat- stereotypes and the degree to which various
egory, and that these beliefs may implicate one components are related to one other. Further,
another in the absence of behavioral evidence. these studies assess the extent to which gender
Thus, to use the case of gender as an example, label and various types of component infor-
it is quite probable that a person who is told mation can influence the judgments that in-
that a man fixes his child's lunch on Monday, dividuals make about the existence of other
Tuesday, and Wednesday would subsequently gender-related characteristics, not specifically
predict that the man would fix the lunch on implied by the provided information.
Thursday as well, even though fixing lunches
may normally be believed to be more char- Experiments 1 and 2
acteristic of women. It is not clear, however,
Overview
what this same person would predict if asked
whether the man was a nurse or whether he Experiment 1 used a 2 X 2 X 3 between-
had feminine physical characteristics. Assum- subjects design in which male and female sub-
ing that fixing lunches is not particularly di- jects were presented information that a person
agnostic of being a nurse, one should predict identified as either male or female performed
that gender, rather than specific information, role behaviors that were either masculine,
would influence the occupational judgment. feminine, or mixed. Subjects were then asked
However, if components of stereotypes are to estimate the probability that the stimulus
linked in some type of network, then infor- person had other gender-related traits, physical
mation about one component could influence characteristics, and occupations, and was het-
judgments of other components as well. It is erosexual and homosexual.
just this implicit linkage between various In Experiment 2, a 2 X 3 X 3 between-
components of gender stereotypes that the subject design was used, including both male
present studies address. and female subjects and varying trait infor-
Grant and Holmes (1981,1982) have raised mation (masculine, feminine, or mixed) with
similar questions about the conclusions of gender label (male, female, or not specified).
Locksley and her colleagues (1980, 1982). Subjects estimated the probability that the
Working from the assumption that a number stimulus person had a variety of physical char-
of schemata (or implicit personality theories, acteristics, role behaviors, and occupations,
to use their terms) are bound together in the and was heterosexual and homosexual. When
(ethnic) stereotype, they have shown that when gender was not specified, subjects also esti-
people are given information about one mated the probability that the stimulus person
schema associated with an ethnic stereotype, was male or female.
they will generalize to other schemata also as-
sociated with that group. Accordingly, these Method: Experiment 1
authors argue that stereotypes are important Subjects. A total of 108 students (42 male, 65 female)1
in the social perception process. participated in the experiment as one means of fulfilling
One of the major differences between these an introductory psychology course requirement.
two investigations is the type of judgments Procedure. Subjects were tested in groups of 20-25,
with approximately equal numbers of males and females
that subjects are asked to make. In Locksley in each group. The experiment was described as a study
et al. (1980), the inference is from behavioral of social perception, and the experimenters stressed that
information to a nearly identical behavioral there were no right or wrong answers. Test booklets were
judgment. In Grant and Holmes (1981), sub-
jects infer traits from one schema based on
information about traits in another schema. 1
One additional subject did not indicate his or her sex
The present studies extend this consideration on the forms. That person was included in the correlational
one step further by considering the process of analysis but excluded from all others.
994 KAY DEAUX AND LAURIE L. LEWIS

arranged so that all six conditions were represented in were: (male associated) truck driver, insurance agent, tele-
each testing session. phone installer, chemist, urban planner; (female associated)
The first two pages of the booklet contained eight practice occupational therapist, telephone operator, speech pa-
items, designed to familiarize students with probability thologist, elementary school teacher, nurse's aide.3
scales. (An example of these items is as follows: "How Order of presentation of occupations and traits was
likely or unlikely is it that an average person has red hair?") counterbalanced. Within each section of dependent vari-
Responses to each question were made on a scale that ables, masculine and feminine items were randomly in-
ranged from extremely unlikely (0) to extremely likely termixed. Questions on sexual orientation were placed at
(100). The scale was marked in intervals of 5, with the the end of the booklet for all subjects.
following anchor points specified: extremely unlikely,
moderately unlikely, neither likely nor unlikely, moderately Method: Experiment 2
likely, and extremely likely.
After students had completed the first two pages, the Subjects. A total of 179 students (84 male, 92 female)4
experimenter determined if there were any questions about participated in the experiment as one means of fulfilling
the use of the probability scale. She then went on to explain an introductory psychology course requirement.
that students would be given a limited amount of infor- Procedure. The basic procedures were the same as
mation about a person and would be asked to estimate those in Experiment 1, with the following changes in ex-
the probability that the person had a number of other perimental materials. Stimulus information consisted of
characteristics or engaged in a number of other behaviors. traits rather than roles, using the 16 items from the PAQ
Students were asked to form a clear image of the person scale. The 8 masculine items were used in the masculine
in their minds before answering the questions. condition and the 8 feminine items were used in the fem-
After all students had completed the booklets, the ex- inine condition. Traits used in the mixed condition were
perimenter explained the purpose of the experiment and the following: active, emotional, kind, never gives up easily,
answered any questions that students had. understanding of others, self-confident, stands up well under
Stimulus material. There were six conditions, repre- pressure, helpful to others. Dependent measures were also
senting one of two gender labels (male or female) and three the same in Experiment 2, with the exception that traits
types of role information (masculine, feminine, or mixed). were eliminated and role behaviors were added, the same
Role information presented in each condition was as fol- eight behaviors that were used as stimulus materials in
lows: (masculine) head of household, financial provider, a Experiment I.5
leader, and responsible for household repairs; (feminine)
source of emotional support, manages the house, takes
care of children, responsible for decorating the house; Results: Experiments 1 and 2
(mixed) financial provider, source of emotional support, Relationship among components. To ex-
a leader, takes care of children.
Each type of role information was combined with male plore the relationship among stereotype com-
and female gender information. As an example, subjects ponents, correlations were computed between
in the female/mixed role behavior condition were given summed dependent measures: specifically, be-
the following description: "A woman has been described tween occupations and traits in Experiment 1
in terms of the following characteristics and behaviors:
financial provider, source of emotional support, a leader,
and between occupations and role behaviors
takes care of children. Consider these characteristics care- in Experiment 2. Probability estimates were
fully and think about what type of woman this would be." normalized around each subject's mean judg-
Dependent measures. Subjects were asked to estimate ment to eliminate individual differences in use
the probability on a 0- to 100-point scale that the stimulus of the response scale.
person had each of eight masculine traits and eight feminine
traits, had five male physical characteristics and five female
physical characteristics,2 would be in each of five male-
2
dominated and five female-dominated occupations, and The items assessing physical characteristics did not
the probabilities that the person was heterosexual and was produce a satisfactory factor structure and will not be
homosexual. discussed in the results of Experiments 1 and 2.
3
The personality traits were drawn from the Personal Alpha coefficients for the masculine traits and feminine
Attributes Questionnaire (PAQ; Spence, Helmreich, & traits were .90 and .89, respectively. For the occupational
Stapp, 1974) and consisted of the following items: (mas- factors, alpha coefficients were .79 and .76 for the male
culine) independent, active, competitive, can make deci- occupations in the two studies and .78 and .86 for the
sions easily, never gives up easily, self-confident, stands up female occupations in the two studies.
4
well under pressure, and feels superior; (feminine) emo- Three subjects who did not indicate their sex were
tional, able to devote self (completely to others, gentle, included only in the correlational analysis.
5
kind, aware of the feelings of others, understanding of Factor analysis of the role behaviors showed that all
others, warm in relations with others, helpful to others. four of the presumed masculine behaviors loaded on a
Occupations were selected on the basis of employment single factor (alpha coefficient = .78). However, the second
statistics that showed at least 70% of people in the oc- factor consisted only of the items "emotional supporter"
cupation to be of one sex. In addition, male and female and "takes care of children." Consequently, the other two
occupations were matched for prestige on the basis of items were dropped from the female roles component; the
pretest data. Occupations selected as dependent measures alpha coefficient for the remaining two items was .65.
STRUCTURE OF GENDER STEREOTYPES 995

Table 1
Mean Probability Judgments of Traits, Occupations, and Sexual Orientation: Experiment 1
Mean probability judgments

Masculine Feminine Male Female Hetero- Homo-


Role information and gender label traits traits occ. occ. sexuality sexuality

Female role behaviors


Female label .69 .78 .30 .45 .86 .14
Male label .60 .77 .39 .48 .64 .39
Male role behaviors
Female label .86 .63 .52 .51 .67 .30
Male label .83 .65 .53 .39 .82 .19
Mixed role behaviors
Female label .79 .75 .41 .51 .83 .28
Male label .75 .78 .47 .50 .82 .26

Note. occ. = occupations.

In Experiment 1, ratings of the probability r(179) = -.19, p = .009. The association be-
of a person having masculine traits were not tween judgments of male and female occu-
correlated with the rated probability of being pations was inconsistent across the two studies,
in a male occupation, r(108) = .11, p = .26. r(108) = -.03, ns in Experiment 1; r(179) =
Nor was the correlation between feminine -.53, p = .001, in Experiment 2.
traits and female occupations significant, Effect of information on judgments: Exper-
r(108) = -.1 \,p = .26. In Experiment 2, cor- iment 1. In Experiment 1, subjects were given
relations between probability judgments of information about gender and role behaviors,
occupations and role behaviors were low but and were asked to estimate the probability of
significant for both feminine and masculine particular traits, of occupations, and of sexual
cases: r(179) = .22, p = .003, and r(179) = orientation.7 Mean probability judgments for
.17, p = .02, respectively. These correlations all conditions are shown in Table 1.
support previous evidence (Deaux & Lewis, Two multivariate analyses of variance
1983) that the components of gender stereo- (MANOVAS) were performed on the data, using
types are relatively independent; less than 5% the two male components as dependent vari-
of the variance was explained by the largest ables in one analysis and the two female com-
correlation obtained here.6 ponents as dependent variables in the other.
A second question of interest is the degree Results of the multivariate analysis for the
to which masculine and feminine components masculine measures showed significant effects
within a single domain may be related. Al- of both role information (p < .001) and gender
though self-report measures of traits have label (p = ,004). Neither the three-way inter-
shown independence (cf. Spence et al., 1974; action, the two-way interactions, nor the main
Bern, 1974), it is not necessarily the case that
observers will view these dimensions as or-
thogonal. In fact, Foushee, Helmreich, and 6
In Deaux and Lewis (1983), these correlations are
Spence (1979) have reported that observers somewhat higher than in the present report because they
use a bipolar schema when rating masculinity were
7
not normalized around each subject's mean judgment.
and femininity of target persons. Such a bi- In rating the probability that'a person would have a
particular occupation, subjects appeared to have altered
polar view was evident in the present results their use of the scale. Rather than judging the probability
as well. In Experiment 1, judgments of the that, for example, a man would be a telephone installer
probability of male traits showed a significant (logically a low number), subjects apparently switched to
negative correlation with judgments of female estimating the probability that a telephone installer would
traits, r(108) = -.5l,p = .001. In Experiment be a male, thus explaining why these means are higher
than one would expect. For other components, however,
2, the correlation between judgments of male it can not be assumed that the estimates should add to
and female roles was lower but still significant, 100% because elements are not mutually exclusive.
996 KAY- DEAUX AND LAURIE L. LEWIS

Table 2
Summary of Univariate F-Tests From Multivariate Analyses: Experiment 1
Gender label Role information

Dependent measure p» P I2 Fb P r,2

Masculine traits 4.96 .03 .05 20.85 <001 .31


Feminine traits 0.33 .57 .003 13.0 <.001 .22
Male occupations 2.78 .10 .03 15.39 <.001 .24
Female occupations 1.0 .32 .01 0.48 .62 .01

2,95.

effect of subject sex was significant in the mul- variate analysis, subject sex was significant only
tivariate analysis. Multivariate analysis of the in judgments of feminine characteristics. Uni-
feminine measures showed significant effects variate tests showed that there were borderline
of role information (p < .001) and a weaker effects for both feminine traits (p = .06) and
effect for sex of subject (p = .05). In this case, female occupations (p = .07); in both cases
the effect of gender label was not significant; female subjects made slightly higher proba-
nor were any of the interactions significant. bility estimates than did males.
Subsequent univariate tests on each dependent Although judgments of traits and occupa-
variable are shown in Table 2, together with tions are influenced primarily by information
the variance estimates for each independent about role behavior, estimates of sexual pref-
variable. erence are affected by the interaction between
As shown by the ij2 values in Table 2, prob- role behavior and gender information, F(2,
ability estimates of both occupations and traits 95) = 9.75, p < .001, for judgments of het-
were heavily influenced by jthe provided role erosexuality; F(2, 95) = 7.05, p < .001 for
information and, to a much lesser extent, by homosexuality. The pattern of these interac-
gender information. Only in the case of judg- tions can be seen in Table 1. For both the male
ments of female occupations was the provided and female stimulus persons, role information
information not influential. Statistical com- that is inconsistent with gender leads to a sharp
parison of the effect sizes supports the visual decrease in the estimated probability of being
interpretation.8 Differences between the con- heterosexual and a sharp increase in the es-
tributions of role and gender were significant timated probability of being homosexual. In
for male traits (p < .01), male occupations the case of the male stimulus person, this es-
(p < .01), and female traits (p < .05), but did timate of homosexuality reaches nearly .40
not differ for female occupations. when female role behaviors are identified with
Within this pattern, however, there is one the person.
anomalous finding. Although the effect of gen-
der was significant for judgments of masculine Results: Experiment 2
traits, the direction was the opposite of that In Experiment 2, subjects were given in-
which would be predicted, that is, females were formation about gender and traits and were
judged more likely to have instrumental traits
than were males. One explanation for this di- 8
vergence is that all of the selected role behaviors The formula for comparing variance estimates, gen-
might have implied some instrumental activity. erously provided by B. }. Winer, is as followsiz = ZE\ -
Z£J/(2/W-3)" 2 .
In judging the female stimulus person, this 'Those items suggesting the most instrumentality
implied instrumentality might have led to a (managing the house and being responsible for decoration)
contrast effect in judgments, resulting in higher were also the same two items that did not load on a com-
ratings of instrumental traits for the woman mon factor in Experiment 2, suggesting a methodological
than for the man.9 artifact. At the same time, other investigators have also
reported instances in which women are judged higher in
Sex of subject played only a limited role in instrumentality (e.g., Eagly & Steffen, 1984), suggesting
these judgments. As indicated by the multi- that the finding might not be totally artifactual.
STRUCTURE OF GENDER STEREOTYPES 997

Table, 3
Mean Probability Judgments of Role Behaviors, Occupations, and Sexual Orientation: Experiment 2
Mean probability judgments
Trait information and Male role Female role Male Female Hetero- Homo-
gender label behaviors behaviors occ. occ. sexuality sexuality

Feminine traits
Female label .50 .86 .30 .61 .83 .26
Person label .59 .85 .36 .62 .68 .36
Male label .67 .82 .47 .56 . .66 .40
Masculine traits
Female label .68 .63 .56 .43 .78 .27
Person label .82 .65 .55 .40 .84 .15
Male label .84 .63 .55 .35 .82 .20
Mixed traits
Female label .64 .76 .45 .60 .71 .30
Person label .71 .79 .40 .55 .72 .36
Male label .74 .74 .45 .44 .86 .17
Note. occ. = occupations.

asked to estimate the probability of selected trait information and, to a lesser extent, by
role behaviors, of occupations, and of sexual gender identification. Statistical comparison of
orientation. Mean probability judgments for the variance estimates shows that the contri-
all dependent variables in all conditions are bution of trait information was significantly
shown in Table 3. Again, MANOVAS were per- greater in judgments of male occupations (p <
formed on the data, followed up with uni- .01), female occupations (p < .05), and female
variate tests when appropriate. In judgments role behaviors (p < .01). In the case of male
of masculine characteristics, the multivariate role behaviors, the contribution of both factors
analysis showed significant effects for trait in- was substantial and not significantly different.
formation and gender label (p < .001, in both Males were consistently believed to be more
cases) and no effects of subject sex or any in- likely than females to display male role be-
teractions. Judgments of female characteristics haviors, no matter what kind of trait infor-
were significantly affected by gender label (p = mation was provided. In contrast, estimates
.02), trait information (p < .001) and subject of the probability of female role behaviors were
sex (p = .004). Once again, none of the in- influenced only by the type of trait information
teractions was significant. Summaries of the provided, and were unaffected by gender in-
univariate tests are shown in Table 4, together formation.
with variance estimates. There was a main effect for trait information
Estimates of the probability of occupations on judgments of both heterosexuality, F(2,
and role behaviors were heavily influenced by 159) = 2.97, p = .05, and homosexuality, F(2,

Table 4
Summary of Univariate F-Tests From Multivariate Analyses; Experiment 2
Gender label Trait information
Dependent measure F" P J F" P r,2

Male role behaviors 24.02 <.001 .23 36.69 <.001 .32


Female role behaviors 0.64 .53 .008 29.03 <001 .27
Male occupations 3.08 .05 .04 22.52 <001 .22
Female occupations 5.64 .004 .07 24.21 <.001 .23
i = 2, 158.
998 KAY DEAUX AND LAURIE L. LEWIS

159) = 7.39, p = .001, suggesting that pos- accordingly, and vice versa. Similarly, either
session of feminine traits increases the per- role or trait information has been shown to
ceived probability of homosexuality and re- affect other judgments, specifically occupa-
duces the probability of heterosexuality. How- tional likelihood and sexual orientation. In
ever, in both cases these main effects were fact, the latter estimates are quite striking. A
qualified by the interaction of gender label and male who is described by such traits as emo-
trait information: for heterosexuality, F(4, tional, gentle, understanding of others and
159) = 4.03, p = .004; for homosexuality, .F(4, kind, for example, is given a .40 chance of
159) = 3.88, p = .005. These significant in- being homosexual. Similarly, a mean estimate
teractions reflect two tendencies. First, judg- of .39 for being homosexual was obtained
ments of female sexuality are relatively unaf- when a male was described as being a source
fected by trait information whereas judgments of emotional support, managing the house,
of male sexuality are influenced quite strongly taking care of children, and decorating the
by trait information. Second, sexuality esti- house. Although estimates of heterosexuality
mates for the neutral person tend to parallel were higher in each case, it is still notable how
judgments of the male, suggesting that the in- strongly the estimates shifted in the face of
definite person is more closely associated with quite minimal information. These results ex-
a male than with a female. tend the findings of Storms, Stivers, Lambers,
In areas other than sexuality, however, this and Hill (1981), who found that "masculine"
association between person and male is not women were rated as more homosexual than
terribly strong. (Nor did Deaux & Lewis, 1983, "feminine" women.
find any consistent pattern of association.) In general, the effects of specific trait or role
Further, when subjects were asked to estimate behavior information was greater than the in-
whether the person was male or female, prob- formation provided by gender label. Thus as
ability judgments paralleled provided trait in- suggested by Locksley et al. (1980), specific
formation. Rated probabilities of the person information about a target may account for
being a male were .62, .58, and .46 when sub- more variance than gender in subsequent be-
jects were provided masculine, mixed, and havior and trait predictions. At the same time
feminine trait information, respectively; es- and in a manner consistent with the arguments
timated probability that the person was a fe- of Grant and Holmes (1981, 1982), these ef-
male were .45, .59, and .66 for the same three fects are not restricted to judgments in do-
conditions. mains identical to those provided in the initial
Sex of subject effects were again limited to information. Before this argument is extended,
the judgment of feminine characteristics. Uni- however, it is important to determine its gen-
variate tests indicated that sex differences were erality.
significant in the judgment of female occu- Although we have argued that there are a
pations, F( 1, 158)= 10.12,^= .002, and bor- number of components of stereotypes, only
derline in the judgments of female roles, F(l, two were manipulated in the previous studies.
158) = 3.16, p = .08. In both cases, judgments Role behaviors and traits may have unique
by women were higher than those by men. properties, and/or they may be more closely
linked to other components than would be
Discussion occupations or physical characteristics. Thus
in the third study, information about each of
Gender stereotypes appear to consist of a the four components identified in earlier re-
number of separate components that are rel- search (Deaux & Lewis, 1983) was presented
atively independent of each other. At the same and judgments were made about the proba-
time, when a person is given information about bility of the stimulus person having charac-
one particular component, he or she will use teristics in each of the other components. By
that information to infer the existence of other presenting subjects with each type of infor-
gender-related characteristics that are consis- mation, it is possible to determine the weight
tent with the initial information. Thus, if peo- that each component has on judgments of
ple are given specific information about role other components. Further, one can begin to
behaviors, they shape their estimates of traits assess the degree to which components have
STRUCTURE OF GENDER STEREOTYPES 999

reciprocal effects on one another or, in con- groups of approximately 40. All conditions were repre-
trast, the degree to which two components may sented within a single testing session, and the general pro-
cedures were identical to those of Experiments 1 and 2,
have a more unidirectional influence (for ex- with the additional explanation that the subject would be
ample, physical characteristics may affect in- making judgments about several different people, rather
ferences of traits, but traits may not influence than only one.
judgments of physical characteristics as
strongly). Additional goals of the third study Results
were to replicate the previous findings, using
improved component manipulations devel- As in the previous studies, multivariate tests
oped by Deaux and Lewis (1983). were conducted and followed up with uni-
variate tests. These analyses were conducted
separately for each independent variable. A
Experiment 3 summary of these analyses, together with sig-
Method nificance levels of the differences between
variance estimates, are presented in Table 5.
Subjects and design. A total of 171 students (83 male, Again, it can be seen that the influence of
88 female) participated in the experiment as one means specific component information is far greater
of fulfilling an introductory psychology course requirement.
A 2 X 3 X 3 X 4 between-within design was used, with than gender label. The i?2 values show that
sex of subject, gender label (man, woman, or person), and gender label never accounts for more than 19%
type of stimulus information (masculine, feminine, or of the variance, whereas type of stimulus in-
mixed) as between-subjects factors. Component infor- formation accounts for much larger amounts
mation was a within-subjects factor. Within each condition
for four successive stimulus persons, a subject was given of the variance—up to 79% in the judgment
information about one of four stereotype components of male occupations when male physical char-
(traits, role behaviors, occupations, or physical character- acteristics were the independent variable. In
istics) and was asked to estimate the probability that thegeneral, the results obtained when either roles
stimulus person had each of the other components.10 The or traits were the independent variables are
order of presentation was counterbalanced.
Stimulus material. The four information components comparable to the results of Experiments 1
used as independent variables were traits, role behaviors, and 2, respectively. In Experiment 3, some of
physical appearance, and occupation. Trait manipulations the interactions between gender label and
were the same as those used in Experiment 2. Role be- stimulus information also proved to be sig-
haviors provided in each of the three conditions were as
follows: (masculine) financial provider, head of household, nificant, although typically at much lower lev-
takes initiative with the opposite sex, assumes financial els than either of the main effects. To the extent
obligations; (feminine) takes care of children, tends the that a pattern was detectable, it appears that
house, does the laundry, cooks the meals; (mixed) financial judgments of the sex-unspecified person were
provider, takes care of children, cooks the meals, head of more extremely polarized toward the sex link-
household. Physical descriptors for the masculine condition
were tall, strong, sturdy, and broad-shouldered, and for age of the stimulus information.
the feminine condition, soft voice, dainty, graceful, and Notable in Table 5 is the substantial amount
soft. When occupation was the independent variable, only of variance that information about physical
a single item was provided because of the obvious im- characteristics explains. For every dependent
possibility of a person having four occupations. Four dif-
ferent examples were used in both the masculine and fem- variable, the effect of physical appearance in-
inine conditions, however, counterbalanced across con- formation is considerable, significantly out-
ditions. These eight occupations were telephone installer, weighing the contribution of gender label that
police officer, automobile mechanic, construction worker, never accounts for even 1% of the variance.
hairdresser, secretary, telephone operator, and registered In contrast, trait information is considerably
nurse.
Dependent variables. Each item that was part of a less potent, exerting significant influence but
component in the manipulation of independent variables statistically outweighing the information of
also served as a dependent variable on three occasions. simple gender label only in judgments of male
Thus, for example, subjects were asked to judge the prob- occupations.
ability that a stimulus person had each of the eight physical
characteristics on three separate occasions; when presented
with role information, when presented with trait infor- 10
mation, and when presented with an occupational label. Subjects in the mixed-information condition did not
Alpha coefficients for each of these dependent variable receive either physical characteristics or occupation as an
sets ranged from .75 to .95. independent variable because of the noncredibility of the
Procedure. Subjects participated in the experiment in combined information.
1000 KAY DEAUX AND LAURIE L. LEWIS

The influence of each type of information themselves to be strongly influenced by gender-


on judgments of other stereotype components related information, specifically when either
can be seen in the means of Table 6. It is clear role behaviors or physical characteristics are
that the probability estimates parallel the pro- the independent variables. The interaction be-
vided stimulus information. For example, tween gender label and stimulus information
given masculine traits, the estimate of male is particularly strong in the latter case: F(2,
roles is .79 and the estimate of female roles 103) = 9.85, p = .001, for judgments of ho-
is .54; given feminine traits, the estimate of mosexuality (F = 6.86,p = .001, for judgments
male roles is .58 and the estimate of female of heterosexuality). A male described in terms
roles is .70. This pattern holds true even when of feminine physical characteristics is esti-
information is combined across sex of stimulus mated to have a .58 probability of being ho-
person, as it is in Table 6. mosexual and a .56 probability of being het-
Judgments of sexual orientation again show erosexual. The corresponding mean judgments

Table 5
Summary ofUnivariate F-Tests From Multivariate Analyses: Experiment 3
Type of stimulus
Gender label information Difference
between
Dependent variable ri2 values

Independent variable: Traits"


Male roles 17.11 <.001 .19 40.96 <.001 .35 m
Female roles 7.55 .001 .09 14.81 <.001 .16 ns
Male occ. 10.33 <.001 .12 19.53 <.001 .21 .05
Female occ. 10.83 <.001 .13 14.24 <.001 .16 ns
Male phys. char. 3.79 .02 .05 12.53 <.001 .14 ns
Female phys. char. 12.13 <.001 .14 14.90 <.001 .16 ns

Independent variable: Rolesb


Male traits 7.71 .001 .09 37.22 <.001 .33 .01
Female traits 2.45 .09 .03 14.05 <.001 .16 .05
Male occ. 7.81 .001 .09 56.05 <.001 .43 .01
Female occ. 1.71 .18 .02 1.59 .21 .02 ns
Male phys. char. 0.10 .91 .001 11.06 <.001 .13 .01
Female phys. char. 1.10 .33 .01 6.21 ..003 .08 ns

Independent variable: Occupations0


Male traits 4.20 .02 .08 7.66 .007 .07 ns
Female traits 0.14 .87 .003 13.53 <.001 .12 .05
Male roles 0.37 .69 .007 30.31 <.001 .23 .01
Female roles 9.60 <.001 .16 40.76 <.001 .29 ns
Male phys. char. 0.61 .54 .01 60.77 <.001 .38 .01
Female phys. char. 4.41 .02 .08 47.39 <.001 .32 .05

Independent variable: Physical characteristics'


Male traits 1.31 .28 .03 147.69 <.001 .60 .01
Female traits 1.16 .32 .02 48.41 <.001 .33 .01
Male roles 4.40 .02 .08 117.67 <.001 .54 .01
Female roles 3.98 .02 .08 38.73 <.001 .28 .05
Male occ. 1.51 .22 .03 367.87 <.001 .79 .01
Female occ. 0.28 .76 .006 68.17 <.001 .41 .01
Note. occ. = occupations; phys. char. = physical characteristics.
' dfe = 2, 150. b dfs = 2, 152.' dfa = 2, 99 for gender label; dfe -- 1, 99 for stimulus information.
STRUCTURE OF GENDER STEREOTYPES 1001

for a female described by male physical char- probability estimate of having feminine traits.
acteristics are .40 and .69. In each case, probability judgments of one
Exploration of the relative influence of component (e.g., feminine traits) served as a
components on one another was done by con- repeated measure when three different com-
ducting a series of repeated measures ANOVAS ponents were the independent variables (e.g.,
on the data, thereby assessing, for example, roles, physical characteristics, and occupa-
which type of information led to the highest tions), with sex linkage of the stimulus infor-
mation (i.e., male, female, or mixed) held con-
stant in each analysis. Although this form of
Table 6
analysis does not have the power that the use
Mean Probability Judgments: Experiment 3 of a nonhierarchical model with a complete
factorial design would yield, it can provide
Type of information suggestive evidence regarding the links between
components.
Dependent variable Masculine Feminine Mixed The majority of these analyses yielded sig-
Independent variable: Traits nificant findings, suggesting that the compo-
nents were not equivalent in their influence.
Male roles .79 .58 .69 For example, when the required judgment was
Female roles .54 .70 .64
Male occ. .56 .39 .45
the probability of the person having masculine
Female occ. .48 .62 .53 traits, information about male role behavior
Male phys. char. .65 .55 .63 led to a higher judged probability (.79) than
Female phys. char. .52 .63 .53 did information about masculine physical
Heterosexual .80 .73 .80 characteristics (.75) or a male occupation (.72),
Homosexual .25 .32 .24
with the difference among these means sig-
Independent variable: Roles nificant beyond the .001 level. (These means
can be found in the first column of Table 6.)
Male traits .79 .63 .73 Although the patterns of these differences were
Female traits .62 .74 .71
Male occ. .57 .29 .43 not totally consistent across various compo-
Female occ. .47 .51 .52 nent combinations, inspection of judgments
Male phys. char. .65 .55 .58 made when independent and dependent vari-
Female phys. char. .50 .58 .52 able components were gender-consistent sug-
Heterosexual .81 ,75 .76 gest some tentative regularities. Role behavior
Homosexual .22 .27 .25
information tended to lead to the highest
Independent variable: Occupations
judged probability of traits, and vice versa, for
both male-linked and female-linked compo-
Male traits .72 .65 nents, suggesting that these two components
Female traits .58 .67
Male roles .75 .61
have the strongest implicative links. Within
Female roles .47 .65 the masculine domain, occupation and phys-
Male phys. char. .70 .54 ical characteristics had the strongest influence
Female phys. char. .39 .56 on one another, for example, the probability
Heterosexual .75 .73 of a male occupation was judged highest when
Homosexual .28 .35
masculine physical information (as opposed
Independent variable: Physical characteristics to trait or role behavior information) was pro-
vided, and the probability of masculine phys-
Male traits .75 .48 ical characteristics was highest when the person
Female traits .57 .74
was described as having a male-dominant oc-
Male roles .74 .46
Female roles .48 .66 cupation. For the female domain, feminine
Male occ. .64 .26 traits led to the highest judged probability of
Female occ. .36 .62 a person having feminine physical character-
Heterosexual .76 .68 istics, whereas judgments of the probability of
Homosexual .30 .39
a female occupation were highest when either
Note. occ. = occupations; phys. char, physical char- trait or physical information was provided. As
acteristics. noted earlier, these results must be considered
1002 KAY DEAUX AND LAURIE L. LEWIS

tentative, yet they do suggest that the linkages and Holmes (1981) found that people made
between components vary in their strength. trait inferences on the basis of other trait in-
formation and Locksley et al. (1980) asked for
General Discussion judgments of behavior following information
on very similar behavior, subjects in the present
Gender stereotypes are best described in study showed that they are not averse to cross-
terms of a set of components. Although these ing domains, for example, moving with ap-
components are relatively independent of one parent ease from information about physical
another, information about one component characteristics to inferences about traits, oc-
can implicate other components in systematic cupations, and role behaviors. Such ready
ways. Gender information alone does lead to generalizations underline the importance of
the inference of certain sets of characteristics, studying stereotypes as a collection of com-
with the strength of the inference depending ponents that implicate each other, rather than
on the particular component. In most cases, isolating a single component.
however, the influence of gender can be out- On the basis of these results, we may begin
weighed by other information, such as role to think of the process of stereotyping as a
behaviors, traits, and the like. sequence of inferences. Given only gender la-
From one perspective, these findings support bel, people will infer a variety of gender-related
the contention of Locksley that information characteristics, as Deaux and Lewis (1983)
about specific characteristics of the stimulus have shown. For example, people will estimate
persons should have a greater impact on sub- the probability that a male is strong as ,66 and
jects' judgments than information about their that a woman is strong as .44. Given more
categorical membership. Yet we would not specific, and presumably more diagnostic in-
conclude that the findings demonstrate that formation, the influence of gender label is di-
stereotypes are unimportant. As Grant and minished, and the resultant judgment is based
Holmes (1981, 1982) have shown, clusters of primarily on the identifying information. As
information are gathered together under the Locksley et al. (1982) have suggested, the im-
stereotype umbrella, and information about pact of this information will not be affected
one component may implicate others. To un- by its consistency with the previous gender
derstand stereotypes in general, and gender label, supported in our case by the nearly
stereotypes in particular, one must be con- complete absence (with the exception of judg-
cerned with this linkage. Taking one isolated ments of sexual orientation) of interactions
element and showing that it can outweigh the between gender label and stimulus informa-
influence of gender does not neutralize the tion.
concept of stereotype nor does it provide much On the basis of these results, one may also
insight into the durability and pervasiveness speculate that concepts of male and female
of stereotypes. One may, for example, describe are not totally separate stereotypes. First, there
a male in terms of feminine traits and min- is the finding that implications from provided
imize a person's use of typically male-asso- information are not altered by gender label,
ciated traits in the further description of that suggesting that the concepts do not have
person, thereby demonstrating that specific unique influences on subsequent processing.
trait information can outweigh gender in sub- Second, it appears that people deal with in-
sequent trait judgments. Yet at the same time, consistency between gender label and infor-
it appears that people will take that trait in- mation by inferring consistency among com-
formation and paint another portrait of the ponents and disregarding gender label. Such
man in question that shows evidence of ste- an interpretation would be consistent with a
reotypical thinking—in this case, a male with view of the individual maintaining a bipolar
a wide range of typically feminine character- conception of masculinity and femininity,
istics, including role behaviors, occupations, simply shifting from one end of the dimension
and physical attributes. to the other when faced with specific gender-
That the scope of these stereotypical linkages linked information. (Negative correlations be-
is considerable is further evidenced by the pat- tween masculine and feminine components in
tern of judgments in this study. Whereas Grant the same domain are also consistent with this
STRUCTURE OF GENDER STEREOTYPES 1003

interpretation.) Thus the popular terminology of information about genitalia and chromo-
of the opposite sex may reflect an underlying somes, people rely on physical characteristics
principle in ,the cognitive construction of to make gender attributions—to a much
gender. greater extent than they use information about
In depicting this sequence of inferences, we role behaviors and traits. Ashmore and Del
should note that the present data suggest that Boca (1979) have identified three different
the influence of the various components is not types of attributes associated with stereotypes:
equally strong. Trait information, for example, defining, identifying, and ascribed, exemplified
does not always outweigh gender label whereas by genitalia, physical stature, and traits. If we
physical appearance cues always do. It is, of conceptualize these three types of attributes
course, unwise to make any absolute state- as a hierarchy in which the overlap between
ments about the strength of these components, the stereotypical male and female increases as
as their impact may depend on the particular one moves from defining to ascribed charac-
selection of items used to create the compo- teristics, then the results found in the present
nent. At the same time, however, it should be study may appear more understandable. Males
noted that, in the abstract, each of these com- and females are viewed as more different on
ponent measures is associated with gender to physical dimensions than they are on psycho-
about the same degree. Thus Deaux and Lewis logical dimensions; hence information about
(1983) reported that the probabilities of mas- physical characteristics may be more diag-
culine components being associated with a nostic than is the less differentiated trait in-
male are .67, .80, .71, and .75 for physical formation. ''
characteristics, role behaviors, occupations, Previous research on physical attractiveness
and traits, respectively. Similar patterns were (Berscheid & Walster, 1974) demonstrates how
observed for other combinations of gender and pervasive the effects of physical appearance
component information. Interesting about can be. In the case of gender stereotypes, rec-
these figures is the fact that the association ognizing the importance of physical descrip-
between physical characteristics and gender is tors has some important implications. Al-
somewhat weaker than is true for the other though one may, with greater or lesser degrees
components. Yet once that information about of difficulty, modify one's traits, role behaviors,
physical appearance is provided, people seem or even occupation, it is somewhat more dif-
to rely heavily on it to make inferences about ficult to modify aspects of physical appearance.
other aspects of the person. Thus if that aspect of gender-related behavior
The importance of physical appearance cues that is most difficult to change is also the most
to an understanding of gender stereotypes can potent source of stereotyping, the continued
not be minimized. Although we most.im- strength of these stereotypes seems likely. As
mediately identify women and men by their a further implication, consider the case of ini-
physical appearance, it is surprising that in- tial encounters. On such occasions, one may
vestigators have given so little thought to this learn little about an individual's personality
aspect of gender stereotypes. In the present or role behaviors in the family, but the observer
case, a relatively limited amount of infor- is immediately aware of the physical infor-
mation about physical characteristics proved mation. Self-fulfilling prophecies and expec-
to be far more influential than gender label in tancy confirmation sequences (cf. Darley &
subsequent judgments, and it outweighed gen- Fazio, 1980) may thus begin to operate before
der to a greater extent than did any of the potentially disconfirming information is avail-
other components. able, ensuring the perpetuation of gender ste-
McArthur (1982) has recently suggested that reotypes.
physical appearance is critically involved in In summary, -approaching gender stereo-
many aspects of the stereotyping process. Al- types by simple elicitation of associations with
though her analysis does not deal specifically
with gender, other evidence is consistent with 1
the prominence of physical appearance found ' The items included in the present definition of physical
characteristics carry some trait implications as well. Were
in the present study. Kessler and McKenna "purer" physical items selected, it is possible that the effect
(1978), for example, found that in the absence of physical information might have been even stronger.
1004 KAY DEAUX AND LAURIE L. LEWIS

the label of female or male may suggest greater Hepburn, and Ortiz. Social Psychology Quarterly, 45,
strength than these labels have. At the same 274-276.
Hamilton, D. L. (1979). A cognitive-attributional analysis
time, the present exploration of the linkages of stereotyping. Advances in Experimental Social Psy-
among stereotype components attests to the chology, 12, 53-81.
network of beliefs that are associated with Hamilton, D. L. (Ed.). (1981). Cognitive processes in ste-
gender and underlines the importance of con- reotyping and intergroup^ behavior. Hillsdale, NJ: Erl-
sidering both the structure and process of gen- baum.
Howard, J. W, & Rothbart, M. (1980). Social categori-
der stereotypes rather than simply descriptive zation and memory for in-group and out-group behavior.
content. Journal of Personality and Social Psychology, 38, 301-
310.
Katz, D., & Braly, K. W. (1933). Racial stereotypes of
References one hundred college students. Journal of Abnormal and
Social Psychology, 28, 280-290.
Kessler, S. J., & McKenna, W. (1978). Gender: An eth-
Allport, G. W. (1954). The nature of prejudice. Reading, nomethodological approach. New York: Wiley.
MA: Addison-Wesley. Lippman, W. (1922). Public opinion. New York: Harcourt,
Ashmore, R. D., & Del Boca, F. K. (1979). Sex stereotypes Brace.
and implicit personality theory: Toward a cognitive- Locksley, A., Borgida, E., Brekke, N., & Hepburn, C.
social psychological conceptualization. Sex Roles, 5, (1980). Sex stereotypes and social judgment. Journal
219-248. of Personality and Social Psychology, 39, 821-831.
Ashmore, R, D., & Del Boca, F. K. (1981). Conceptual Locksley, A., Hepburn, C., & Ortiz, V. (1982). On the
approaches to stereotypes and stereotyping. In D. L. effect of social stereotypes on judgments of individuals:
Hamilton (Ed.), Cognitive processes in stereotyping and A comment on Grant and Holmes's "The integration
intergroup behavior (pp. 1-35). Hillsdale, NJ: Erlbaum. of implicit personality theory schemas and stereotypic
Bern, S. L. (1974). The measurement of psychological images." Social Psychology Quarterly, 45, 270-273.
androgyny. Journal of Consulting and Clinical Psy- McArthur, L. Z, (1982). Judging a book by its cover: A
chology, 42, 155-162. cognitive analysis of the relationship between physical
Berscheid, E., & Walster, E. (1974). Physical attractiveness. appearance and stereotyping. In A. H. Hastorf & A. M.
Advances in Experimental Social Psychology, 7, 158- Isen (Eds.), Cognitive social psychology (pp. 149-211).
216. New York: Elsevier.
Broverman, I. K., Vogel, S. R., Broverman, D. M., Clark- McCauley, C., Stitt, C. L., & Segal, M. (1980). Stereotyping:
son, F. E., & Rosenkrantz, P. S. (1972). Sex-role ste- From prejudice to prediction. Psychological Bulletin,
reotypes: A current appraisal. Journal of Social Issues, 87, 195-208.
28(2), 59-78. Miller, A. G. (Ed.). (1982). In the eye of the beholder:
Campbell, D. T. (1967). Stereotypes and the perception Contemporary issues in stereotyping. New York: Praeger.
of group differences. American Psychologist, 22, 817- Rosenberg, S., & Sedlak, A. (1972). Structural represen-
829. tations of implicit personality theory. Advances in Ex-
Cantor, N., & Mischel, W. (1977). Traits as prototypes: perimental Social Psychology, 6, 235-293.
Effects on recognition memory. Journal of Personality Rosenkrantz, P., Vogel, S., Bee, H., Broverman, I., & Brov-
and Social Psychology, 35, 38-48. erman, D. M. (1968). Sex-role stereotypes and self-con-
Darley, J. M., & Fazio, R. H. (1980). Expectancy confir- cepts in college students. Journal of Consulting and
mation processes arising in the social interaction se- Clinical Psychology, 32, 287-295.
quence. American Psychologist, 35, 867-881. Schneider, D. J., & Blankmeyer, B. L. (1983). Prototype
Deaux, K., & Lewis, L. L. (1983). Components of gender salience and implicit personality theories. Journal of
stereotypes. Psychological Documents, 13, 25. (Ms. No. Personality and Social Psychology, 44, 712-722.
2583). Spence, J. T., Helmreich, R., & Stapp, J. (1974). The
Eagly, A. H., & Steffen, V. J. (1984). Gender stereotypes personal attributes questionnaire: A measure of sex-role
stem from the distribution of women and men into stereotypes and masculinity-femininity. JSAS Catalog
social roles. Journal of Personality and Social Psychology, of Selected Documents in Psychology, 4, 43. (Ms. No.
46, 735-754. 617)
Eagly, A. H., & Wood, W. (1982). Inferred sex differences Spence, J. T, Helmreich, R., & Stapp, J. (1975). Ratings
in status as a determinant of gender stereotypes about of self and peers on sex-role attributes and their rela-
social influence. Journal of Personality and Social Psy- tionship to self-esteem and conceptions of masculinity
chology, 43, 915-928. and femininity. Journal of Personality and Social Psy-
Foushee, H. C., Helmreich, R. L., & Spence, J. T. (1979). chology, 32, 29-39.
Implicit theories of masculinity and femininity: Dualistic Storms, M. D., Stivers, M. L., Lambers, S. M., & Hill,
or bipolar? Psychology of Women Quarterly, 3, 259- C. A. (1981). Sexual scripts for women. Sex Roles, 7,
269. 699-707.
Grant, P. R., & Holmes, J. G. (1981). The integration of
implicit personality theory schemas and stereotype im-
ages. Social Psychology Quarterly, 44, 107-115.
Grant, P. R., & Holmes, J. G. (1982). The influence of Received July 22, 1982
stereotypes in impression formation: A reply to Locksley, Revision received November 9, 1983

You might also like