self-schemas and gender hazel markus, marie crane, stan

Journal of Personality and Social Psychology1982, Vol. 42, No. 1, 38-50

Copyright 1982 by the American Psychological Association, Inc.0022-3514/82/4201-0038$00.75

Self-Schemas and Gender

Hazel Markus, Marie Crane, Stan Bernstein, and Michael SiladiInstitute for Social Research

University of Michigan

Self-schemas are knowledge structures developed to understand, integrate, orexplain one's behavior in a particular domain. Two, studies examined the infor-mation-processing consequences of self-schemas about gender. Systematic dif-ferences in cognitive performance were observed among groups of individualsidentified as masculine schematics, feminine schematics, low androgynous, andhigh androgynous. Those individuals identified as feminine schematics remem-bered more feminine than masculine attributes, endorsed more feminine qualities,required shorter processing times for "me" judgments to these attributes thanto other types of attributes, and were more confident of their judgments. Theseindividuals were able to supply relatively more examples of past feminine be-havior than masculine behavior. A parallel pattern of results was found formasculine stimuli for those individuals identified as masculine schematics. Incontrast, those subjects identified as androgynous recalled as many masculineattributes as feminine attributes and did not differentiate between masculine andfeminine attributes with respect to latency or confidence. A careful comparisonof the two groups of androgynous subjects shows that only the low androgynousshould be considered aschematic with respect to gender. Overall the findingssuggest that individuals differ markedly in the nature of their knowledge struc-tures about gender and in how gender is integrated into the self-concept.

The self-concept contains representationsof our special abilities, achievements, andpreferences, the unique aspects of our ap-pearance, and the characteristic expressionsof our temperament. With social experiencewe gain a diversity of self-relevant infor-mation that becomes organized into cogni-tive structures. It is by means of these struc-tures that we categorize, explain, andevaluate our behavior in various focal do-mains. For the purposes of this investigation,we will refer to these cognitive structures asself-schemas and to the union of these par-ticular schemas in the various domains asthe self-concept.

This research was supported by National Institute forMental Health Grant MH29753. Portions of these datawere presented at the Midwestern Psychological Asso-ciation meetings, Chicago, 1978.

The authors thank Robert Zajonc and Steven Martinfor helpful comments on an earlier version of this articleand Mary Cullen and Kymberly Skinner for preparationof the manuscript.

Requests for reprints should be sent to Hazel Markus,Research Center for Group Dynamics, Institute for So-cial Research, University of Michigan, Ann Arbor,Michigan, 48106.

Self-schemas are assumed to be summa-ries and constructions of past behavior thatenable individuals to understand their ownsocial experience and to organize a widerange of information about themselves. Pre-vious studies (Markus, 1977; Rogers, Kui-per, & Kirker, 1977; Markus, Crane, & Si-ladi, Note 1; Markus, Hamill, & Sentis,Note 2) have gathered converging evidencefor the self-schema construct, showing, forexample, that differential processing of in-formation about the self in various behav-ioral domains is associated with differencesin self-schemas.

The term schema embodies many of theproperties and functions that are assumedto characterize cognitive structures in gen-eral. As such the concept implies a structureof knowledge or a systematic framework thatparticipates in ongoing interpretive activityduring information processing. Little can besaid at this time about the structural natureof schemas, and most of their assumed prop-erties and functions remain to be demon-strated. The same is true of other cognitiveconceptions, such as frames or scripts. Yetthese concepts have been found to have solid

38

SELF-SCHEMAS AND GENDER 39

heuristic value in organizing empirical find-ings and guiding research. The studies to bediscussed focus on the information-process-ing consequences of self-schemas relevant togender. Our approach provides a conceptualframework for exploring whether there aresystematic differences among individuals inhow gender-relevant knowledge is organizedin memory and how it is used in thinkingabout oneself.

We assume that individuals vary enor-mously in the content and organization oftheir self-schemas. For example, some in-dividuals may be intensely concerned withtheir honesty, their masculinity, or their cre-ativity and may develop highly articulatedschemas about themselves in these particulardomains. In contrast, others may be rela-tively undifferentiated, or aschematic (with-out self-schemas), in these domains.

There are some aspects of behavior thatare so prominent and central that virtuallyeveryone develops at least rudimentary sche-mas for them. These universal schemas,which everyone has to one degree or another,develop with respect to those aspects of theself that are particularly salient and avail-able for social evaluation and comment.One's sex, for example, becomes a point offocus and a salient distinguishing trait earlyin life, and it remains a significant parameterof the person's social interactions. The lit-erature on sex role development (cf. Green,1974; Sears, Rau, & Alpert, 1965) indicatesthat by the time children-are 3 years old,they are keenly aware of their sex and be-have as if they have differentiated their self-systems according to this factor. Even thoughvirtually all individuals develop some basicappreciation and understanding of their birological sex, only some individuals seem toconstruct an elaborated self-schema abouttheir gender. In many cultures one couldreasonably expect nearly everyone to havesome understanding and representation ofattributes and meanings that could be char-acterized by the general terms masculinityor femininity, even if these particular termsare not included in verbal repertoire. Forsome individuals, however, these networksof meanings are used in thinking about, de-scribing, and evaluating the self; it is thenthat a self-schema may be thought to exist.

When this occurs we can imagine a mergingof the self-concept with the network ofknowledge relevant to masculinity or femi-ninity. For example, the elements in the fem-inine structure, perhaps some of the stereo-typic elements such as warmth, nurturance,gentleness, or understanding, may now beseen as having relevance and importance forthe self. Some aspects of one's representa-tions of the self may now become associatedwith some of these trait elements, and in turnthe representations of these features mayengage representations of some aspects ofthe self. A self-schema in a particular be-havioral or stimulus domain is, then, an in-tersection in memory between the represen-tations of these behaviors or of these stimuliand the representations of the self.

As an important component of the self-concept, a gender schema is likely to behighly available and centrally implicated ininformation processing about gender in gen-eral and about the gender aspects of the selfin particular. The concept of self-schemasallows us to make some distinctions and in-ferences regarding the recent theorizingabout gender and sex roles. In particular itis significant, in the light of the self-schemaconcept, to inquire whether androgynous in-dividuals are aschematic, that is, whetherthey have no especially articulated cognitivestructures about either masculinity or fem-ininity, or whether they are equally sche-matic with respect to both masculinity andfemininity. It is possible, in fact, that thereexist both kinds of androgynous persons andthat these individuals would be expected tobehave differently and process gender-rele-vant information in different ways.

Despite the fact that most theoretical dis-cussions of masculinity, femininity, and an-drogyny and the relationships among them(e.g., Bern, 1977; Spence, Helmreich, &Stapp, 1975) seem to imply quite clearly thatindividuals who evaluate themselves in dif-ferent ways with respect to gender have dif-ferent kinds and amounts of knowledgeabout gender, there has been almost no at-tention given to the specific cognitive cor-relates of gender identity (cf. Bern, 1981, fora recent exception). The present studies ex-tend our earlier research on differential pro-cessing of gender-related information (Mar-

40 MARKUS, CRANE, BERNSTEIN, AND SILADI

kus et al., Note 1) and explore individualvariation in how gender is represented in theself-concept.

We examined in detail some cognitiveconsequences of the systematic differencesin knowledge structures that are assumed toexist among individuals who are character-ized as masculine, feminine, high androgy-nous (individuals who think that both mas-culine and feminine attributes characterizethem), and low androgynous (individualswho think that neither masculine nor femi-nine attributes characterize them).1 Thesegroups were examined for their tendenciesto differentiate some forms of social infor-mation according to masculinity or feminin-ity. The focus was on differences amongthese groups with respect to the recall ofgender-relevant material, the speed and con-fidence of processing gender-relevant mate-rial, and the ability to describe their owngender-relevant behavior.

The schema concept suggests a variety ofpossible differences in the organization ofgender-relevant knowledge among these fourgroups. Figure 1 outlines (in the extreme)these possible differences. The gender-sche-matic subjects (Panels A and B), who thinkof themselves as distinctly masculine or fem-inine, are assumed to have a large networkof schema-relevant cognitions that are re-trieved as a unit when the schema is acti-vated. This is illustrated by the intersectionsamong these associated cognitions. For theindividual with a femininity schema, all ofthese cognitions are related to the conceptof femininity and thus are available to work-ing memory when the feminine schema isactivated. This schema is assumed to be partof the self-concept. These individuals prob-ably have some structure relevant to mas-culinity, but this structure is not likely to beself-relevant. This is illustrated by the factthat the masculine schema does not intersectwith the self.

For the high androgynous subjects (PanelC), we can assume that both masculine andfeminine attributes are associated with theself-concept. These individuals have notsharply differentiated themselves with re-spect to gender and relate some attributesof both masculinity and femininity to theirself-concept. They should be able to respondin some instances like individuals with a fem-

ininity schema and in other instances likethose with a masculinity schema.

For the low androgynous subjects (PanelD), we can assume that neither masculinenor feminine attributes are associated withthe self-concept, and if these subjects havestructures of masculine or feminine knowl-edge, they are not well developed or elabo-rated. These individuals, for whatever rea-sons, may be seen as aschematic, or trulywithout self-schemas reflecting conventionalaspects of masculinity or femininity. Forthese subjects features of masculinity orfemininity have not been incorporated intothe self-concept.

This self-schema model implies, for ex-ample, that those with feminine self-schemas(the feminine schematics and high androg-ynous individuals) will privilege femininestimuli in information processing. Relativeto masculine schematic or low androgynousindividuals, these individuals can be viewedas experts about femininity and should beable to use knowledge relevant to femininityefficiently. An alternative view of how gen-der-relevant knowledge is organized by thefour groups is one that assumes that themasculine and feminine schematics are alikein that they both have gender schemas, andthat the androgynous groups are alike in thatthey both are aschematic with respect togender. This view holds that masculine orfeminine subjects are not schematic with re-spect to either masculinity or femininity, butwith respect to the domain of gender as awhole. It assumes that masculine and fem-inine subjects, relative to the androgynousgroups, should privilege all gender-relevantstimuli, masculine and feminine alike. In thestudies that follow, we attempted to deter-mine which of these views of how genderinformation is organized is most appropriate.

Study I

Overview

The first study investigates whether de-scribing oneself as masculine, feminine, or

' Subjects who endorse both masculine and femininewords at low levels are discussed by Spence, Helmreich,and Stapp (1975), who call such people undifferen-tiated, and Bern (1977), who refers to these individualsas low-lows.

SELF-SCHEMAS AND GENDER

(C)

41

MASCULINE SCHEMATIC HIGH ANDROGYNOUS

(B)

Self-Concept

(D) <ErllnUvttuqllnk H d.l.rxi. • ]

\.b33i*'

Mcncullne Schema

oC)o

Self-incept

FEMININE SCHEMATIC LOW ANDROGYNOUS

Figure 1. Relationship between gender-relevant knowledge structures and the self-structure for masculineschematics, feminine schematics, high androgynous and low androgynous subjects.

androgynous is related to differential mem-ory for gender-related, information. The ma-jority of studies of free recall indicate thatfactors that facilitate associations among theelements in the stimulus set to be remem-bered also enhance later recall (Anderson& Bower, 1973). Thus an individual who hasa self-schema about gender should demon-strate enhanced recall of gender-relevantstimuli. In this study subjects were asked tocomplete the Bern Sex-Role Inventory(BSRI). The BSRI consists of 20 masculinecharacteristics (e.g., assertive, independent,ambitious), 20 feminine characteristics (e.g.,affectionate, understanding, sensitive to theneeds of others), and 20 neutral character-istics (e.g., sincere, friendly, cooperative).We assumed that the process of describingoneself as required by the BSRI should make

Bone's self-schemas of masculinity and fem-ininity highly available, and that these self-schemas should help impose organization ofthe gender-relevant information and accen-tuate the associations among them.2 Theavailability of self-schemas of masculinity

and femininity should allow for more effi-cient encoding of the gender-relevant adjec-tives and produce enhanced recall for them,for as suggested by Figure 1, not only dothese elements share the gender predicate,they share the self predicate as well.

Specifically we expected that individualsclassified as either masculine or femininewould have enhanced recall for words thatare consistent with their particular genderidentity and that inconsistent trait adjectiveswould be less likely to be recalled. The an-drogynous subjects were those who have notcategorized themselves exclusively in termsof one gender or another, andithus they werenot expected to exhibit differential recall.

For the purpose of this initial study, weassumed that individuals classified as femi-nine on the basis of the BSRI were feminine

2 This organization would go beyond that provided bythe semantic similarity of the adjectives themselves. Thesemantic similarity of the adjectives is a basis for group-ing that is presumably available to all subjects, regard-less of their self-definition with respect to gender.


schematic and that those classified as mas-culine were masculine schematic. We as-sumed that those classified as androgynousdid not differentiate their behavior accord-ing to masculinity or femininity and thuswere aschematic in these domains.

Method

Subjects

Two hundred sixty-seven undergraduate students inintroductory biology classes volunteered to participatein this study.

Materials and Procedures

In small classroom groups (approximately 18 perclass), students were asked to rate themselves on theBern Sex-Role Inventory. They were asked to indicateon a 7-point scale the extent to which each of the 60trait adjectives described them. The scale is anchoredwith 1 (never or almost never true) and 7 (always oralmost always true).

After completing the BSRI students were given ablank sheet of paper and were asked to write down asmany adjectives from the scales as they could remember.Three minutes were allotted for recall.

Results

Using Bern's (1974) procedure for iden-tifying masculine, feminine, and andrpgy-nous subjects, four groups of subjects wereidentified. Subjects classified as masculine(« = 50; 42 males and 7 females) were thosewho rated themselves significantly (p < .05)higher on the 20 masculine words than onthe 20 feminine words. Subjects classified asfeminine (« = 31; 5 males and 25 females)were those who rated themselves signifi-cantly higher on the feminine words than onthe masculine words. Androgynous individ-uals were those whose self-ratings on themasculine words were not significantly dif-ferent from their self-ratings on the femininewords. Two types of androgynous individualswere distinguished: high androgynous sub-jects (n = 100; 51 females and 47 males;those who felt both masculine and feminineterms described them) and low androgynoussubjects (n = 86; 45 females and 41 males;those who did not feel either masculine orfeminine words described them), accordingto the procedure followed by Bern (1977).3

These four groups of subjects were com-

pared for recall of items from the BSRI. Therecall score for each subject consisted of thenumber of correct responses out of the first10 responses.4 To adjust for primacy andrecency effects, recall of the first and lastthree items of the scale (two masculine, twofeminine, and two neutral) were not countedas correct recall (cf. Kuiper & Rogers,1979). Table 1 shows the number of mas-culine, feminine, and neutral words recalledby the four groups of subjects. The meannumber of correct words recalled did notdiffer across groups. Masculine subjects re-called significantly more masculine wordsthan feminine words, f(49) = 3.12, p < .01.They also recalled significantly more mas-culine words than did students classified asfeminine, f(79) = 2.79, p < .01. A parallelpattern of results was found for femininesubjects. These subjects recalled signifi-cantly more feminine words than masculinewords, f(30) = 3.83, p < .001, and they re-called significantly more feminine wordsthan did masculine subjects, t(79) = 4.82,p < .001. Subjects designated as high an-drogynous also recalled more feminine wordsthan masculine words, t(99) = 2.62, p < .01,but the difference in the number of mas-culine and feminine words recalled wasclearly smaller than that exhibited by eitherof the other two groups of subjects. A testfor the proportion of variance accounted forshowed clear differences among these groups(masculine subjects, w = .149; feminine sub-jects, co = .306; high androgynous subjects,<o = .003. Low androgynous subjects showedno significant differences in recall of mas-culine and feminine attributes (t < 1). Withrespect to neutral words, there were no dif-ferences among the four groups in the num-ber of adjectives recalled.5

3 One masculine subject, one feminine subject, andtwo androgynous subjects failed to indicate their sex.

*As there may have been some variation in the exacttime allowed for free recall across the 13 administrationsof the scale, the analysis of only the first 10 responsesis presented here. When all responses are included inthe analysis, however, the pattern of findings is the same.

5 When the results within the masculine, feminine,and androgynous groups are divided according to sexof the subject, the pattern of results for each sex is thesame as the overall pattern. Male (n = 88) and female(n = 96) androgynous subjects show nearly identicalpatterns. The females (n = 96) among the masculine


Table 1Mean Number of Masculine, Feminine, and Neutral Words Recalled

Adjectivetype

MasculineFeminineNeutral

Total

Masculine"

3.422.602.108.12

Feminine1"

2.683.902.098.67

Highandrogynous'

3.003.392.058.44

Lowandrogynous"1

2.982.942.047.96

' n = 50; 42 males, 7 females; one subject failed to indicate his or her sex.b n = 31; 5 males, 25 females; one subject failed to indicate his or her sex.°n = 100; 47 males, 51 females; two subjects failed to indicate their sex.d n = 86; 41 males, 45 females.

The four cells in the upper left-hand cor-ner of Table 1 provide strong support for aself-schema model. Masculine schematicsrecalled substantially more masculine wordsthan feminine words, and feminine sche-matics recalled substantially more femininewords than masculine words. If neutralwords are used as a baseline, however, itappears that the schematics show enhancedmemory for both masculine and femininetraits. Such a result could be interpreted assupport for the idea that the schematics haveschemes about gender as a whole, ratherthan self-schemas either about masculinityor femininity. Given this interpretation thelow androgynous subjects should not showan enhanced memory for gender words, sincepresumably gender words are no more sig-nificant for them than are the neutral attri-butes. Yet this is not the case; the low an-drogynous subjects also recalled more gender-related words than neutral words. The factthat all subjects remembered neutral itemsas well, regardless of schematicity, suggeststhat the neutral words as a set may be lessdistinctive or meaningful words. These dataalone, then, do not provide unambiguoussupport for either a self-schema or a gender-

subjects show an even more pronounced difference inrecall (in favor of masculine words) than the males. Themales (n = 15) among the feminine schematics, how-ever, show somewhat less difference than the group asa whole. Overall male subjects do not differ in recall ofmasculine and feminine trait adjectives (M recall = 3.04and 3.09, respectively). Female subjects show greaterrecall of feminine words (3.52) than masculine words(3.03), but this difference is smaller than that exhibitedby the feminine subjects for feminine and masculinewords (3.90 vs. 2.68).

schema model. An analysis of clustering inrecall produces some additional data rele-vant to this concern.

Clustering in Recall

The way that gender-relevant informationis organized can also be explored by analyz-ing the type of clustering in free recall. Usingan index suggested by Shuell (1969), clus-tering of the masculine and feminine adjec-tives was compared to the clustering of neu-tral adjectives.6 The results indicate thatmasculine, feminine, and high androgynoussubjects all cluster masculine and feminineadjectives more than neutral words. Mas-culine subjects clustered masculine wordssignificantly more than feminine words, andthe other two groups did not differentiate intheir clustering of masculine and femininewords. The low androgynous subjects alsodid not cluster masculine and femininewords differentially, but they showed signif-icantly more clustering of neutral words.

Study 2

Study 1 showed differences in the contextand pattern of recall that depend on one'sview of oneself with respect to gender. Study

6 This index uses Pagan's (1968) normalization of theBousfield and Bousfield (1966) deviation measure. It isderived from the occurrence of runs of items of the sameword type in the recall protocols and is normalized withrespect to the maximum number of repetitions possiblefor lists of the same item compositions. The indices cal-culated do not assume equal availability of items ofdifferent types and are independent of the number ofitems recalled.


2 employed other criteria for identifying self-schemas for gender and investigated the in-fluence of self-schemas for gender on thespeed and confidence of information pro-cessing. It was hypothesized that individualscategorized as masculine and feminine sche-matics would differentially process mascu-line and feminine self-related information,demonstrating greater ease and confidencein processing schema-consistent stimuli. Toexplore the specific cognitive consequencesof androgyny further, the responses of sche-matics were used as a baseline against whichto compare the responses of the androgynoussubjects.

Method

Subjects

Two hundred students in introductory psychologyclasses were administered questionnaires that contained10 self-rating scales. Three of the scales were relevantto masculine behavior (aggressive, dominant, and actsas a leader), and 3 scales were relevant to femininebehavior (gentle, emotional, and sensitive).' Four otherscales not relevant to masculinity or femininity werealso included in the measure. Students were asked torate themselves on each of these traits, using a 10-pointscale with 10 labeled describes me and 1 labeled doesnot describe me. On the basis of their responses to thequestionnaire, 61 students (32 males and 29 females)were selected for participation. From these individualsthree groups of subjects were identified.

Masculine schematic. Individuals who rated them-selves extremely high (8-10) on at least two of the threemasculine characteristics and moderate or low (1-7) onat least two of the three feminine characteristics werelabeled masculine schematic (9 males and 1 female).

Feminine schematic. Individuals who rated them-selves extremely high (8-10) on at least two of the threefeminine characteristics and moderate or low (1-7) onat least two of the three masculine characteristics werelabeled feminine schematic (7 males and 14 females).

Androgynous. Individuals who rated themselves sim-ilarly on both masculine and feminine adjectives werelabeled androgynous (16 males and 14 females). Again,two types of androgynous subjects were identified, highandrogynous individuals and low androgynous individ-uals. Those who rated themselves extremely high (8-10) on at least five of the six relevant characteristicswere labeled high androgynous (n = 14; 7 females and7 males), and those who rated themselves moderate tolow (1-7) on at least five of the six characteristics werelabeled low androgynous (n = 16; 9 females and 7males). The results of the study are analyzed, first, con-sidering the androgynous subjects as a whole and, sub-sequently, looking at each subgroup separately.

Eight weeks after the questionnaire was administered,the 61 subjects were called individually to the laboratoryand received identical treatment. They were not in-

formed of the connection between the laboratory sessionand the questionnaire, and it is unlikely that they couldhave inferred such a connection, as different ostensiblepurposes were presented to them and different experi-menters were used for the studies.

Materials and Procedures

The laboratory session consisted of two separate cog-nitive tasks designed to assess the possible influence ofgender self-schemas on the processing of informationabout the self in these domains.

Task 1: Content, latency, and confidence of self-de-scriptions. This part of the experiment was under pro-gram control using a PDP-11 computer. Subjects werepresented with 60 adjectives from the Bern Sex-RoleInventory (Bern, 1974), one at a time, on cathode-raytube (CRT) displays. They were instructed to respondby pushing a me button if the characteristic was self-descriptive and a not me button if the characteristic wasnot self-descriptive. Each adjective appeared on thescreen for 2 sec or until the subject responded, whichevercame first. Both the latency and the responses were re-corded (although the subjects did not realize that la-tencies were being measured). After the subject re-sponded "me" or "not me" to each word, the question,"How confident are you?" appeared on the CRT dis-play. Subjects responded by pushing buttons on a 6-point scale with endpoints labeled not at all confidentand very confident. Four different randomly determinedorders of presentation were used. For each subject theme button was placed by the dominant hand. To insurethat individuals were associating similar types of be-havior to the trait adjectives, a particular context wasspecified for the self-judgments.8

Task 2: Supplying behavioral evidence for self-de-scription. Following the me/not me task, each subjectwas given a booklet of 20 words (1 on each page) takenfrom the characteristics in the BRSI. Eight were mas-culine words, 8 were feminine words, and 4 were neutralwords. The order of the adjectives in each booklet wasrandomly determined. Subjects were given written in-

7 In a pretest a separate group of 160 undergraduatestudents were presented with the words from the BernSex-Role Inventory (BSRI, Bern, 1974) and the Per-sonal Attributes Questionnaire (Spence, Helmreich,& Stapp, 1975). Students were instructed to indicatewhether they thought a word was masculine, feminine,or neither. The three words most frequently designatedas masculine and the three words most frequently des-ignated as feminine were the words used in this self-rating task to select subjects. The BSRI was deliberatelynot used to identify subjects so that the words that com-prise it could be used in the laboratory study.

8 Subjects were told: "When you are making thesedecisions about yourself, try to imagine yourself in atypical group situation, one that might occur, for ex-ample, in a classroom, in the dorm lounge, or at a meet-ing in a friend's home. You are together to discuss animportant and controversial issue and to make somedecisions about it. Many of the people in the group youknow or are familiar to you, while others are not."


structions to circle each adjective that they consideredto be self-descriptive and were asked to give supportingevidence from past behavior.9

Results

The results are first presented comparingmasculine schematic, feminine schematic,and androgynous subjects.

Content of Self-Description

All subjects, regardless of their classifi-cation, endorsed a substantial proportion ofmasculine, feminine, and neutral adjectives.There were important differences, however,in the endorsement patterns of the threegroups. Masculine schematics endorsed sig-nificantly more of the masculine words (92%)than feminine words (63%), t(9) = 4.53,p< .001. Scheffe comparisons indicated thatthese masculine schematics endorsed signif-icantly more masculine adjectives than didthe feminine schematics, t(29) = 8.97, p <.001. Feminine schematics endorsed signif-icantly more feminine adjectives (86%) thanmasculine adjectives (52%), /(20) = 8.84,p < .001. These subjects, in comparison withthe masculine schematics, endorsed reliablymore feminine words, t(29) = 4.33, p < .001,than masculine words. The third group, theandrogynous students, did not exhibit suchmarked differences in endorsement of mas-culine and feminine words. Although theyendorsed reliably more masculine words(85%) than feminine words (76%), the dif-ference was much smaller than that observedfor the other two groups, t(29) = 2.86, p <.01. The first task can be seen as a reliabilitycheck on the subject selection procedure andindicates that this particular method ofchoosing schematics succeeds in identifyingindividuals who respond in a consistent fash-ion across a large set of gender-relevantstimuli.

Latency of Self-Description

If we assume that the three groups of sub-jects have different self-schemas for gender,then we can hypothesize that these struc-tures should allow them to process gender-relevant stimuli with varying degrees of easeand efficiency. Figure 2 illustrates the re-

sponse latencies for the masculine, feminine,and neutral words judged as self-descriptive.The percentages beneath the bars refer tothe number of words endorsed for the firsttask. Masculine schematics responded "me"significantly faster to the masculine wordsthan to the feminine words, t(9) = 4.25,/><.01. Masculine schematics also re-sponded significantly faster to the masculinewords than did the feminine schematics,t(29) = 2.76, p < .01, The feminine sche-matics responded "me" significantly fasterto the feminine words than to the masculinewords, f(20) = 3.63, p<.0l. And whencompared to the masculine schematics, theseindividuals responded significantly faster tothe feminine words, t(29) = 4.54, p < .001.In contrast, the androgynous subjects, whoit will be recalled, were more likely to en-dorse masculine than feminine adjectives,showed no difference at all in response timefor endorsement of masculine and femininewords. They were as fast at responding "me"to masculine words as were the masculineschematics, and they were as fast at respond-ing "me" to the feminine words as were thefeminine schematics.

The results for the not me responses, al-though slower than the me responses overall,are the mirror images of those for the meresponses. Thus masculine schematics wererelatively quick to respond "not me" to fem-inine words (x = 1.78) and much slower torespond "not me" to masculine words (x =2.35). Feminine schematics showed a similarpattern, responding "not me" quickly tomasculine words (x = 1.90) and significantlyslower to feminine words (x = 2.18). Theandrogynous subjects did not exhibit thisdifference in their rejection latencies to mas-culine and feminine words (x = 1.90 and2.00, respectively). They did not differ sig-nificantly from the masculine subjects in re-

9 The instructions were: "Immediately after you circlean adjective, list the reasons you feel this adjective isself-descriptive. Give specific evidence from your ownbehavior to indicate why you feel a particular trait isself-descriptive. List the first kinds of behavior that cometo your mind. Do not worry about howt other peoplemight interpret a particular behavior; use your ownframe of reference." Several examples were given, andsubjects were given as much time as they needed tocomplete the task.


SUBJECTS

Figure 2. Mean response latency for masculine, femi-nine, and neutral trait adjectives judged as self-descrip-tive.

sponse time to feminine words (t < 1) anddid not differ significantly from the femininesubjects in their response time to masculinewords (t < 1).

Confidence of Self-Description

An analysis of the confidence judgmentsyields very similar results. There were nodifferences in overall confidence among thegroups.10 But masculine schematics were sig-nificantly more confident about their me re-sponses to the masculine words than theywere about their me responses to femininewords, <(9) = 6.22, p< .001. Similarly fem-inine schematics were significantly moreconfident about their me judgments for fem-inine words than they were about their mejudgments for masculine words, ?(20) =5.32, /x.OOl. The androgynous subjects,however, did not exhibit these striking dif-ferences in confidence.11

In contrast to the schematic subjects, theandrogynous subjects did not show muchdifferentiation in endorsement, latency, orconfidence in their responses to masculineand feminine stimuli. These subjects appearto be distinct from the two schematic groups.But what is the nature of this difference?We attempted to explore this result furtherby examining high androgynous and low an-drogynous subjects separately.

There were no differences between highandrogynous and low androgynous subjectswith respect to endorsement or latency. Infact the pattern of responses for the twogroups is virtually identical. Both androgy-nous groups endorsed slightly more mascu-line words, and neither group distinguishedamong masculine, feminine, and neutralwords with respect to latency. All of thesesubjects responded equally quickly to mas-culine, feminine, and neutral adjectives forboth me and not me judgments. Neithergroup of androgynous subjects differed sig-nificantly from masculine subjects in theirrejection latencies to feminine words, nor didthey differ significantly from feminine sub-jects in their rejection latencies to masculinewords.

The only difference that emerged amongthe two groups was in confidence of me en-dorsements. The high androgynous subjectswere significantly more confident in theirendorsements to both the masculine, t(28) =3.92, p < .001, and the feminine trait adjec-tives, f(28) = 4.97, p < .001, than were thelow androgynous subjects. The two groups

10 For each of the three groups of subjects, masculineschematics, feminine schematics, and androgynous, wecompared the scores of the male subjects with those ofthe females. Among the masculine schematics there was1 woman. Among the feminine schematics there were7 men. And among the androgynous there were 16 menand 14 women. This uneven distribution of men andwomen across the three groups precluded a formal over-all analysis of sex differences. It is intriguing to notethat in all three groups, however, both males and femalesconformed to the overall pattern. With respect to theandrogynous subjects, there was only one significantdifference between males and females across content,latency, and confidence of judgments for the masculine,feminine, and neutral words. The male androgynoussubjects endorsed more masculine words than the femaleandrogynous subjects (18.31 vs. 15.36). There were nodifferences, however, between males and females in theirendorsement of the feminine words or in their latencyor confidence for masculine and feminine words. Maleand female androgynous subjects responded in very sim-ilar ways to this task and quite differently from eitherthe masculine or feminine schematics.

11 The androgynous subjects were significantly moreconfident about the endorsement of feminine words,t(29) = 2.20, p < .05, but the difference was muchsmaller than that exhibited by the sex-typed schematics.A test for the proportion of variance accounted forshowed clear differences among these groups (masculinesubjects, 01 = .790; feminine subjects, o> = .565, androg-ynous subjects, <o = .113).


do not differ in their confidence for neutraladjectives, however. With respect to confi-dence for not me judgments, the four groupsdid not differ in their judgments to feminineor neutral words, although, feminine sub-jects were significantly more confident abouttheir not me judgments to masculine wordsthan were any of the other three groups ofsubjects.

Since confidence and latency are usuallyhighly correlated, these findings are note-worthy. They suggest that this correlationis not invariant across individuals. The lowandrogynous subjects were relatively lessconfident about their self-descriptions, eventhough they initially responded "me" just asquickly as the high androgynous subjects.The confidence judgment can be seen as anindication of the subjects' second thoughtsabout the initial judgment. As suggested ear-lier it may be that the low androgynous sub-jects do not have well-developed self-sche-mas of masculinity and femininity and thuscannot easily justify their initial endorse-ments. The lack of available clearly cate-gorized behavioral evidence for their mas-culine or feminine endorsements may wellbe revealed in lower levels of confidence.

Task 2: Supplying Behavioral Evidencefor Self-Description

We expected that individuals with self-schemas would be able to provide behavioralevidence relevant to their schema. Thus, forexample, if a woman has previously thoughtof herself as affectionate, she would be ableto readily recall some specific instances ofbehavior that were categorized by herself orby others as affectionate.

The data reveal that there were markeddifferences among the four groups in whichadjectives they chose to supply examples forand in the number of examples provided forthose adjectives. As shown in Table 2, mas-culine schematics provided significantly morebehavioral descriptions for a given masculineword than did any of the other groups ofsubjects. (For the purpose of this analysis,the androgynous subjects were once againdivided into high and low androgynousgroups.) Masculine schematics also providedsignificantly more examples of masculine

WORDS:•Masculine E22Feminine

87.0 77.5 76.0 S2.5 74.S 72.SMEAN % "ME"RESPONSES

Androgynous-High

rogynous-Low

SUBJECTS

Figure 3. A comparison of high adrogynous and lowandrogynous subjects on endorsement, endorsement la-tency, and confidence of endorsement of masculine, fem-inine, and neutral trait adjectives.

behaviors than of feminine behaviors, t(9) =3.46, p < .01. Feminine schematics suppliedsignificantly more behavioral descriptionsfor a given feminine word than did any ofthe other groups of subjects. Feminine sche-


Table 2Mean Number of Behavioral Examples for Masculine and Feminine Words

Type of subject

Word type

MasculineFeminine

Total

Masculine"

11.303.00

14.30

Feminine6

4.528.62

13.14

Highandrogynous0

7.646.71

14.35

Lowandrogynous*1

6.006.88

12.88

10. b « = 21. c n = 14. "« = 16.

matics also recalled significantly more ex-amples of feminine behavior than of mas-culine behavior, f(20) = 4.61, p < .001. Incontrast, the high androgynous and the lowandrogynous subjects did not differentiatein the number of behavioral examples pro-vided for masculine and feminine attributes,although the low androgynous subjects sup-plied relatively fewer examples overall.

These data could be used to support thenotion that these individuals have knowledgestructures varying in size or differentiation.For a given masculine word that was en-dorsed, the masculine subjects supplied moreexamples than did the feminine or androg-ynous subjects, implying that they haveavailable a relatively greater store of rele-vant knowledge. It is also important to notethat the high androgynous subjects, thoughendorsing many of these words just asquickly as the masculine schematics, for ex-ample, did not provide as many examples fora given masculine word as the masculineschematics.12

Discussion

These studies revealed systematic differ-ences in cognitive performance among groupsof subjects identified as masculine sche-matic, feminine schematic, or androgynous.Individuals identified as feminine schematics(by their pattern of responses to the BSRIor by endorsement of a subset of self-de-scriptive adjectives) remembered more fem-inine than masculine attributes, endorsedmore feminine qualities, required shorterprocessing times for me judgments to theseattributes than to other types of attributes,and were more confident of their judgments.In addition, these people were able to supplyrelatively more examples of past femininebehavior than any other group that endorsed

feminine words. A parallel pattern of resultswas fou/nd for masculine stimuli for thoseindividuals identified as masculine sche-matics.

Taken as a whole the results of these stud-ies indicate that the masculine schematicshave self-schemas relevant to masculinityand the feminine schematics have self-sche-mas relevant to femininity. Overall they ar-gue against the idea that these subjects haveschemas relevant to gender as a whole. Forexample, if masculine schematics were sen-sitive not only to masculinity but to femi-ninity as well, they should have exhibitedfaster and more confident judgments to fem-inine items than to neutral items. The op-posite is in fact true; the masculine sche-matics were slow to endorse feminine itemsand not at all certain when they did so. Per-haps, then, subjects did not have the appro-priate behavioral evidence to support theseself-descriptions and thus responded "me"on some other basis, such as social desira-bility. This inference gains support fromTable 2 of Study 2, which indicates that themasculine schematics supplied relatively fewexamples of feminine behavior, even thoughsome feminine attributes were previouslyendorsed. A masculine schematic probablyhas some understanding of and sensitivity tofemininity, as do all subjects, but it does notappear to be a particularly thorough under-

12 When the examples were coded for how well theymatched the stereotypic meanings of the words and forthe level of detail found in the examples, however, theschematics and the high androgynous subjects did notdiffer. Overall, across all the groups less than 10% ofthe examples were judged by an independent group ofcoders as not corresponding to the stereotypic definitionsof the trait attributes, and this did not vary systemati-cally by group. The four groups also did not differ inhow much detail they supplied for masculine, feminine,and neutral adjectives.


standing. The masculine schematics were noquicker to respond "not me" to masculinewords than were the androgynous subjects.If we assume that self-schemas are sum-maries and constructions of past behavior,it is likely that the schematics will have agreat deal of information and experienceabout one aspect of gender and relativelylittle about the other. It is probably true thatknowing what you are will sensitize you towhat you are not, but it is unlikely that youwill have a very well-developed or clearlyorganized understanding of the character-istics that do not describe you. We assumethat one who develops a self-schema withrespect to femininity becomes an expert infemininity, but there is little reason to as-sume that this person simultaneously be-comes an expert in masculinity—no morereason, in fact, than to assume that a thinperson will have a well-developed sense ofobesity or that an extrovert will have a well-developed understanding of introversion.

In contrast to the masculine and feminineschematics, subjects identified as androgy-nous did not differentiate in their processingof masculine and feminine attributes. Theyrecalled as many masculine attributes asfeminine attributes, and they showed no dif-ference between masculine and femininewords with respect to latency or confidence.Two types of androgynous subjects could bedistinguished, however. We found that thehigh androgynous subjects did not differ intheir performance from the masculine sche-matics with respect to masculine attributesor from the feminine schematics with respectto feminine attributes. It seems, then, thatthese individuals do define themselves interms that are gender relevant, but that theyare equally responsive to both masculine andfeminine attributes. As suggested earlier,high androgynous subjects may have mas-culinity schemas that allow them to respondto masculine stimuli as do the masculineschematics and femininity schemas that al-low them to respond to feminine stimuli asdo the feminine schematics.

It is important to recognize, however, thateven if the high androgynous individual andthe masculine schematic are both assumedto have some type of masculinity schema,they are quite likely to have very differentoverall views of themselves and might be

expected to behave in different ways. Themasculine schematic individual and the highandrogynous individual may both think ofthemselves as assertive and attach very sim-ilar meanings and examples to that attribute,but the high androgynous individual is alsolikely to have the attributes "understanding"and "compassionate" as defining features ofthe self-concept. These other attributes ofthe self may well modify or constrain therange of behavioral alternatives that followfrom thinking of one's self as assertive.Somewhat more subtly these additional fea-tures of the self-concept may influence thestyle and manner in which one acts in ac-cordance with the attribute assertive. Theflexibility that has been attributed to thehigh androgynous individual in recent stud-ies (e.g., Wiggins & Holzmuller, in press)may be both an antecedent and consequenceof this multiple categorization or definitionof the self. The high androgynous individual,according to these data, has knowledgeabout both masculine and feminine behav-ior. Presumably both of these knowledgestructures are equally available for process-ing of information. Whether one, the other,or both will drive processing is dependent onthe circumstances and the nature of the sit-uation. These multiple views of the self, ormultiple self-schemas, may also be asso-ciated with the higher levels of self-esteemfound to characterize the high androgynoussubjects.

The high androgynous subjects, then, ap-pear androgynous by virtue of the fact thatthey have incorporated both masculinity andfemininity schemas in their self-concepts.The results for the low androgynous subjectsform a different picture, however. Althoughmany of our results indicate that the highand low androgynous subjects are similar,on closer analysis it is evident that despitethese general similarities, the two groupshandle gender-relevant information in dif-ferent ways and that they may be cognitivelydistinct from each other. The low androgy-nous subjects do not seem to have integratedmany potentially stereotypic masculine orfeminine terms into their self-structures.They appear to be without knowledge struc-tures, or aschematic, with respect to manyof the concepts, qualities, and features thatare often culturally tied to masculinity or


femininity. As suggested in the introduction,if these subjects have knowledge structuresof masculine or feminine domains that arerelevant to the self, they are probably notwell developed or elaborated. It seems, forexample, that with regard to the masculinecharacteristics of aggressive and competitiveor the feminine qualities of warmth and ten-derness, the low androgynous individuals donot have particularly well-articulated viewsof themselves. They seem to be truly withoutself-schemas relevant to these aspects ofmasculinity and femininity. The relativelyimpoverished knowledge structures assumedto characterize the low androgynous subjectsis suggested by several other findings. Thesesubjects supplied substantially fewer behav-ioral examples, recalled somewhat fewer, sex-typed words than did the high androgynoussubjects, showed more clustering of neutralwords than of masculine or feminine words,and showed less confidence in me judgmentsthan the high androgynous subjects.

The data indicate that even though boththe high and low androgynous subjects donot differentiate between the masculine andfeminine adjectives in describing themselves,it is only the low androgynous subjects whoshould be viewed as aschematic with respectto gender. It is only these individuals whodo not use the domains of masculinity orfemininity in thinking about themselves. Thehigh androgynous individuals, in contrast,have multiple gender self-schemas. They aresensitive to and importantly concerned withboth masculine and feminine aspects of theirself-concepts.

Together these findings suggest that thereare important differences among individualsin how gender-relevant knowledge may beorganized in memory and how it has beenincorporated into the self-concept. As wouldbe the case for any other cognitive model ofthe self, our data, even though they are con-sistent with the self-schema model, cannotin themselves provide unequivocal evidenceabout the precise nature and form of the self-schema. Nevertheless a self-schema accountprovides a single, coherent, and plausibleexplanation of the variety of findings re-ported here. In these studies the concept ofa self-schema as a knowledge structure thatsummarizes and integrates the representa-tions of one's own attributes and behavior

in important domains has been particularlyuseful in organizing and describing the re-sults.

Reference Notes

1. Markus, H., Crane, M., & Siladi, M. Cognitive con-sequences of androgyny. Paper presented at the an-nual meeting of the Midwestern Psychological As-sociation, Chicago, May 1978.

2. Markus, H., Hamill, R., & Sentis, K. P. Thinkingfat: Self-schemas for body weight and the processingof weight relevant information. Unpublished manu-script, University of Michigan, 1979.

References

Anderson, J. R., & Bower, G. H. Human associativememory. Washington, D.C.: Winston, 1973.

Bern, S. L. The measurement of psychological androg-yny. Journal of Consulting and Clinical Psychology,1974, 42, 155-162.

Bern, S. L. On the utility of alternative procedures forassessing psychological androgyny. Journal of Con-sulting and Clinical Psychology, 1977,45, 196-205.

Bern, S. L. Gender schema theory: A cognitive accountof sex typing. Psychological Review, 1981, 88, 354-364.

Bousfield, A. K., & Bousfield, W. A. Measurement ofclustering and sequential constancies in repeated freerecall. Psychological Reports, 1966, 19, 935-942.

Pagan, J. P. Measuring verbal recall II: The ITR scoreexpressed as a ratio. Psychonomic Science, 1968, 11,205.

Green, R. Sexual identity conflict in children andadults. New York: Basic Books, 1974.

Kuiper, N. A., & Rogers, T. B. Encoding of personalinformation: Self-other differences. Journal of Per-sonality and Social Psychology, 1979, 37, 2014-2024.

Markus, H. Self-schemata and processing informationabout the self. Journal of Personality and SocialPsychology, 1977, 35, 63-78.

Rogers, T. B., Kuiper, N. A., & Kirker, W. S. Self-reference and the encoding of personal information.Journal of Personality and Social Psychology, 1977,35, 677-688.

Sears, R. R., Rau, L., & Alpert, R. Identification andchild rearing. Stanford, Calif.: Stanford UniversityPress, 1965.

Shuell, T. J. Clustering and organization in free recall.Psychological Bulletin, 1969, 72, 353-374.

Spence, J. T., Helmreich, R., & Stapp, J. Ratings ofself and peers on sex role attributes and their relationto self-esteem and conceptions of masculinity andfemininity. Journal of Personality and Social Psy-chology, 1975, 32, 29-39.

Wiggins, J. S., & Holzmuller, N. Further evidence onandrogyny and interpersonal flexibility. Journal ofResearch in Personality, in press.

Received June 9, 1980Revision received May 18, 1981 •

self-schemas and gender hazel markus, marie crane, stan

Documents