The Development of Gender Stereotype Components Author(s): Carol Lynn Martin, Carolyn H. Wood, Jane K.

Little Reviewed work(s): Source: Child Development, Vol. 61, No. 6 (Dec., 1990), pp. 1891-1904 Published by: Blackwell Publishing on behalf of the Society for Research in Child Development Stable URL: http://www.jstor.org/stable/1130845 . Accessed: 28/11/2011 08:31
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Blackwell Publishing and Society for Research in Child Development are collaborating with JSTOR to digitize, preserve and extend access to Child Development.

http://www.jstor.org

The Development of Gender Stereotype Components

Carol Lynn Martin
Arizona State University

Carolyn H. Wood and Jane K. Little

University of British Columbia LYNN; WOOD, CAROLYN and LITTLE,JANEK. The Development MARTIN, CAROL H.;

been limited by a narrowconcept of stereotypes.A more complexmodel is presented, and developmentalchanges in gender stereotypeswere investigatedusing the new model. In 2 studies, children were told about several sex-unspecifiedchildren, each described as having 1 masculine or 1 feminine characteristic.The children then predicted the likelihood of each story child having other masculine and feminine characteristics. Study 1, 56 children (4-6 years) were told about target In children who liked either a masculine or feminine toy, and then children predicted the targets' interests in other toys. In Study 2, 76 older children (6, 8, 10 years)were told abouttargetchildren with a masculine or feminine characteristic from 1 of 4 categories(appearance, personality,occupations, toys), and then they predicted the likelihood of targetshaving other masculine and feminine characteristicsfrom the same and from different categories as the cue. 2 developmental trends relevant to their own emerged: (a) children appearfirstto learn associationsamong characteristics sex and, later, to learn them for the other sex, and (b) older children's stereotypicjudgments are more extremethan those of youngerchildren.The implicationsof these results forthe development of stereotypes,assessing gender knowledge, and understandingsocialjudgments are discussed. Little is known about how children learn gender stereotypes because developmental researchers have used a reasonable, but narrow, concept of stereotypes. In this article, we present a more complex model of gender stereotypes and reconsider what we know about gender stereotype development using the new model. We also present two studies that were designed to investigate developmental changes in gender stereotypes. Gender stereotypes have been defined as beliefs about the behaviors and characteristics of each sex (Del Boca & Ashmore, 1980). Based on this conception, developmental researchers have determined levels of stereotype knowledge by using the relatively simple method of assessing when children are able to associate gender groups and sex-typed characteristics. For example, in popular stereotype tests, when children infer that a woman is more likely to cook than a man, we assume that they know and are using gender stereotypes. But this is only one way stereotypes are used. Adults also make stereotypic inferences based on knowledge of a sex-typed attribute rather than on a person's sex. For instance, when adults "guess" that a longhaired person is more likely to wear a dress than trousers, they are basing this prediction on their gender-related knowledge concerning the co-occurrence of masculine and feminine characteristics. A Component Model of Stereotypes To account for more complex forms of stereotyping, a broader perspective is needed. Our concept of gender stereotypes is

of Gender 1990, 61, 1891-1904. Developmental research has Stereotype Components. CHILDDEVELOPMENT,

The work reportedhere was funded by a Humanitiesand Social Science Grant(No. 5-57907) fromthe University of BritishColumbia. Study 1 is based on a master'sthesis done by the second authorand Study2 is based on a master'sthesis done by the thirdauthor,both underthe directionof the firstauthor.They were submittedto the Universityof BritishColumbiain partialfulfillmentof the requirementsfor the M.A. degree. Portionsof this work were presented at the meeting of the Society for Researchin Child Development, Baltimore,April 1987. We gratefullyacknowledge the assistance of Mary Etey, Pat Habich, Michelle McBride,Hilary Rose, and Deborah Victorin data collection (Study 2) and thank MargaretArcus, RichardFabes, Daniel Perlman, and Tannis Williams, for their helpful suggestions. Requests for reprintsshould be sent to Carol Martin,Department of Family Resourcesand Human Development, ArizonaState University,Tempe, AZ 852872502.
[Child Development, 1990, 61, 1891-1904. @ 1990 by the Society for Researchin Child Development, Inc. All rightsreserved.0009-3920/90/6106-0027$01.00]

1892

Child Development
TABLE 1
TYPES OF ASSOCIATIVELINKS IN GENDER STEREOTYPES

Gender label-component links: Associating person's sex with gender-related component information a) man-has short hair b) woman-wears a dress Within-component links: Associating informationwithin a content domain based on its "masculinity" or "femininity" a) wears a dress-wears high heels b) is assertive-is independent Between-component links: Associating informationbetween different content domains based on its "masculinity" or "femininity" a) has broad shoulders-is assertive b) wears a dress-is nurturant

acteristic about someone in one content domain, we can make inferences about the person in other content domains (e.g., knowing someone likes to cook leads to the inference that this person likely has long hair). Because of these associative links, we can infer much about a person, albeit sometimes erroneously, based on a single piece of information. Reconsidering Stereotype Development Most developmental theorists have focused on only the most simple stereotype association: how children learn to associate gender groups and sex-typed characteristics. For instance, in the schematic processing approach (Martin & Halverson, 1981, 1987), we argued that the "superordinate" gender schema allows children to fit information about people, toys, clothes, activities, traits, and jobs into either the male ("for boys") or female ("for girls") category, thereby initiating the formation of gender label-component links. Similarly, most of the studies on children's gender stereotypes have focused on documenting the existence of these simple associations. The typical procedure has been to assess whether children associate males or females with various toys, jobs, traits, and physical characteristics. For instance, research has shown that even young children associate gender labels with appearance cues (i.e., identify the sexes; Thompson, 1975), with role behaviors (i.e., activities and toys; Blakemore, LaRue, & Olejnik, 1979; Martin & Little, in press; Thompson, 1975), and with occupations (Kuhn, Nash, & Brucken, 1978; O'Keefe & Hyde, 1983; Papalia & Tennent, 1975). Moreover, Williams, Bennett, and Best (1973) found that even kindergarten children know the sex-typing of personality traits, and this knowledge increased with age. In summary, many investigators have found that very young children have developed the labelcomponent associations that exist in each content domain. Missing in the literature is research on children's understanding of the more complex types of stereotype associations. To date there are no studies assessing whether children can make stereotypic inferences either within or between gender-related components. Limiting research to only the simple associations has fostered the assumption that, for young children, stereotype knowledge has "reached ceiling," thereby making it a relatively uninteresting construct in understanding sex-role development. Acquiring Complex Stereotype Associations Young children clearly understand the relatively simple links between gender la-

based on the Deaux and Lewis (1984) component model (designed to account for adult stereotyping) where gender stereotypes are viewed as a set of associations between gender labels (i.e., "male," "female") and genderrelated, content-specific beliefs. Beliefs are organized by content areas (e.g., occupations) and are labeled "components" of the stereotype. Presumably for children, just as for adults, there are at least four main components: role behaviors, occupations, traits, and physical appearance. Each component has a masculine and a feminine version. For instance, within the occupation component, the masculine version would include "plumber," whereas the feminine version would include nurse. From the evidence obtained with adults, we know that there are three types of associations that link the information within gender stereotypes (see Table 1). The most simple type occurs between gender labels and information in each component. Even when we only know someone's sex, we can make predictions about the person's behavior, occupation, traits, and appearance (e.g., Ashmore & Del Boca, 1979; Deaux & Lewis, 1984). The second type of association occurs within components. By knowing a characteristic about someone in one content domain, for instance, an interest (e.g., likes cooking), we can make inferences about the person in the same content domain (e.g., likes sewing) based on the gender concepts of masculinity and femininity that link information within each content domain. The third type of association occurs between components. When we know a char-

Martin, Wood, and Little

bels and sex-typed characteristics. To match adults' thinking, however, they must also learn at some point to make the more complex component-based inferences within and between the different content domains. How do these complex associations develop? The most intuitively compelling and logical explanation is the "mediation hypothesis": complex associations are mediated by their simple associations to gender groups. That is, once children know how gender labels and component information are associated, they are able to draw all other logical stereotypic inferences. For instance, a child who knows that boys like trucks and boys like cars should be able to infer that someone who likes trucks will also like cars. Although appealing, the mediation hypothesis may not hold because it involves inferences that are difficult for children to make. To be able to use gender this way, children must first infer the person's sex once told an attribute and then make another inference based on their guess about the person's sex. However, children are able to infer a category from an attribute developmentally later than inferring an attribute from a category, at least using novel attributes (Gelman, Collman, & Maccoby, 1986). Similarly, young children do not use higher-order categories such as traits to mediate judgments of others. For instance, children may understand that sharing lunch and helping a friend are each associated with the trait "being kind," but they do not necessarily assume that a person who shares her lunch will also help her friends (Rholes & Ruble, 1984). Although no research has directly tested the mediation hypothesis for genderrelated judgments, there is some evidence to suggest that young children have difficulty making complex stereotypic judgments. Unlike older children (7-9 years) and adults, young children (4-5 years) are unable to use component information to make predictions when it conflicts with predictions made from gender labels (Berndt & Heller, 1986; Martin, 1989; Zucker, Wilson, & Stern, 1985). For instance, when told about a girl who likes playing football, young children predict that she would like dolls more than cars, that is, they assume that she would like girls' toys (because she is a girl) despite the evidence that she has one traditionally masculine interest. An explanation that may help to resolve these apparent inconsistencies in how children make social judgments concerns social experience. That is, the learning of complex associations requires either direct experience of the covariation (or lack of) among compo-

1893

nent items or simply repeated exposure and practice with the information before inferences can be drawn. If social experience influences learning, we would expect children to first learn associations relevant to their own sex and later learn associations relevant to the other sex. This process is consistent with schematic processing views concerning the development of gender schemas. Specifically, once children are able to identify their own gender group, they become more motivated to approach, attend to, and to learn about own-sex activities and interests than about other-sex activities and interests (Martin & Halverson, 1981). There is some evidence of this selective learning: children have more indepth information about own-sex than othersex toys (Halverson & Martin, 1985) and they learn more about novel objects labeled as being for their sex than about those labeled as being for the other sex (Bradbard, Martin, Endsley, & Halverson, 1986). To investigate how children learn complex gender associations, two studies were conducted. The first study was designed to test the generality of the mediation hypothesis. The second study was designed to explore age differences in children's complex gender-based judgments. Study 1

We assessed whether young children make stereotypic predictions about others when they know the person's interest (i.e., a favorite toy) but do not know the person's sex. Children were told stories about sex-unspecified fictitious children with either traditionally masculine or feminine toy interests. They were asked to predict how much the fictitious children would be interested in the toys used in the stories, as well as in new masculine and feminine toys. Predictions would be considered to be stereotypic when children expect others to be more interested in toys that are consistent with their stated interests than those that are not (e.g., predicting that a child who is said to like a masculine toy will like other masculine toys more than feminine ones). The results should help us understand whether gender labels mediate component judgments and why, in earlier studies, young children were unable to use component information when making predictions about others. There are several possible outcomes. One outcome could be that children consistently make stereotypic predictions. This would suggest that they are either using gen-

1894

Child Development
The majority of children were Caucasian (88%) and from middle-class homes. total of 18 sex-typed Toy selection.-A toys (nine feminine toys, nine masculine toys) were required for the study (see Appendix A). Toys were selected from a previously tested sample of 113 toys considered to be sex-typed by at least 75% of children (Halverson & Martin, 1985). Procedure.-Each child was tested individually in a small room or in an area away from other children. Testing involved two 10min sessions held between 4 and 7 days apart. In the first session, the experimenter explained that she wanted the child to decide how much other children liked certain toys. To indicate liking, children viere trained in using the faces rating scale. The scale consisted of three face pictures (big smile, little smile, big frown) drawn on 5 x 8-inch white cards. The children were instructed to point to the face with a big smile to indicate that the child liked the toy a lot, the face with a little smile to indicate that the child liked the toy a little, and the face with a big frown to indicate that the child did not like the toy at all. Then, the gender-plus-interest task was administered. This task was used to assess whether children could make stereotypic predictions when given both gender and interest information. Half of the children were told one brief story and half were given four brief stories (two of girls and two of boys), randomly selected from Martin's (1989) counter-stereotypic conditions.' In each story, a picture of the target child was shown, and the child was described as having a best friend of the opposite sex and an interest in a traditionally sexinconsistent toy (e.g., "Jason's best friend is a girl. Jason likes to play with makeup kits"). Each story also contained neutral information (e.g., "Jason lives down the street"), and the target child's name, age, and sex. The age of the target child corresponded to the age of the child being tested. After hearing each story, children then rated how much they thought the target child would like the same toy as mentioned in the story, two toys consistent with the target child's gender, and two toys consistent with the target child's interest (order of toys was randomized). Presentation of male and female story versions was counterbalanced. Next, children were tested for their knowledge of stereotypes for the toys used (either 16 or 18 toys were used depending on

der to mediate their judgments, or that they have had enough social experience to learn all the stereotypic associations among toys. The failure to use interests to make predictions in previous studies would then be due to reasons unrelated to their gender-related knowlege (e.g., developmental constraints). The second outcome could be that children consistently fail to make stereotypic predictions, suggesting that they are unable to use gender to mediate their judgments or that they have not yet learned the complex stereotypic associations among toys. The failure to use interests to make predictions in the earlier studies would then likely be due to the lack of predictive value of interests. Because of children's tendencies to first learn about own-sex information, we expected a third outcome, that children would make stereotypic predictions ,electively (i.e., only for others with own-sex-relevant interests). This would suggest that gender is not used to mediate judgments, and that children have learned more associations among own-sex-typed toys than among other-sex-typed toys. The failure to use interests in previous studies, then, has to be explained as a combination of lack of knowledge about associative links (for othersex-relevant predictions) and some kind of developmental constraint, such as being unable to consider multiple sources of conflicting information (for own-sex-relevant predictions). We also investigated whether young children's use of component information is influenced by its salience. Interest information was not made salient in previous studies. To manipulate salience, some stories included pictures of the favorite toys (higher salience) and some included only verbal descriptions (lower salience). Because mentioning only one toy interest may not make the child's interest salient, we also varied the number of toys used in the stories (two toys or one toy). Moreover, to ensure that children in this study were similar to those from previous studies, they were given a task to assess whether, when given both gender and interest information, they based their inferences on gender labels. Finally, to ensure that children knew the simple associations between gender labels and masculine/feminine interests, a standard stereotype knowledge assessment was conducted.

Method
Subjects.-Fifty-six children (26 girls, 30 boys), ranging in age from 38 to 73 months, were recruited from local day-care centers.

1The number of stories was changed to improve the reliabilityof these results.

Martin, Wood, and Little

the number of gender-plus-interest stories used). Children were shown, one at a time, pictures of toys and were asked to indicate whether only girls, only boys, or both girls and boys most like to play with the toy by placing each toy in front of one of three rating cards. After sorting all pictures, a forcedchoice procedure was used for the items placed in the category "for boys and girls." For these items, the experimenter said, "You said both boys and girls like [toy name], but who likes it most, boys or girls?" The toys were presented in one predetermined order. The gender-plus-interest task always preceded the stereotype task so as not to make stereotypes salient to the child. In the second testing session, the experimenter reviewed the faces rating scale. Then the interest-only task was administered. This task tested children's abilities to make predictions when they know about a toy someone likes but they do not know the person's sex. Three conditions (single interest, multiple interests/no pictures, multiple interests/pictures) were used to manipulate the salience of toy interest information. In each condition, there was a story about someone with a masculine interest and a story about someone with a feminine interest. This resulted in stories about six fictitious children (three with feminine interests, three with masculine interests). Children were first presented with a story from the single-interest condition in which they were told about a gender-unspecified child's interest in one sex-typed toy (e.g., "A child I know really likes to play with dolls"). Then the experimenter said, "Now I want you to help me decide what other toys this child would like to play with. Remember, this is not about what you like, but about what the other child that I told you about would like." Using the faces scale, children then predicted how much they thought the target child would be interested in each of five toys. Two of the toys were interest consistent, that is, they had the same sex-typing as the toys in the story (e.g., if the story toy was a football, then an airplane), two of the toys were interest inconsistent (e.g., if the story toy was a doll, then a race car), and one toy was simply the toy used in the story (i.e., the same toy). Toys were presented in random order. The next story the child heard was the other version (masculine or feminine) with only a single interest. Again, the child made predictions about how much the other child would like five toys. This general procedure was then repeated for each condition with a few minor changes. In the multiple-interest/no-

1895

pictures condition, children were told about a child interested in two toys from the same sex-typing category. For the multiple-interest/ pictures condition, children were told about a child interested in two toys from the same category, and pictures of these two toys were shown. Because the multiple-interest conditions each involved stories of children with two toy interests, a total of six predictions, instead of five, were made for each description (two same, two consistent, two inconsistent). The single-interest stories were always administered first to ensure that the stories with multiple interests and pictures would not influence children's predictions in the single-interest condition. Orders of presentation of the stories in the two multiple-interests conditions were counterbalanced. Order of presentation of the masculine and feminine versions within each of the three conditions was also counterbalanced. ratings of targets' Scoring.--Children's toy interests were derived from the 3-point faces scale (score of 3 = "likes a lot," 2 = "likes a little," and 1 = "not like at all"). For the interest-only task, mean scores were computed for consistent toys, inconsistent toys, and same toys. For the gender-and-interest task, mean scores were computed for toys consistent with the gender of the target, toys consistent with the interests of the target, and the same toy as stated in the story. For the stereotyping measure, a score of 1 was given for children who either originally gave a stereotyped answer (e.g., they said only girls play with doll houses) or when forced to choose gave a stereotyped answer (e.g., they said both boys and girls like to play with a tool kit, but when forced said only boys). A score of 0 was given for children who originally gave a counterstereotypic answer (e.g., they said only boys play with dollhouses) or when forced to choose gave a counterstereotypic answer (e.g., they said both boys and girls like to play with a tool kit, but when forced said only girls). Results Judgments made when gender labels assess whether the chilwere present.-To dren in this sample responded similarly to children in earlier studies (e.g., Martin, 1989), we analyzed their predictions when they heard stories in which gender labels and component information were discrepant (e.g., a boy with a feminine interest). Planned comparisons revealed that children tended to base their inferences on gender labels more than

1896

Child Development
< .01 for both comparisons). The two multiple-interest conditions did not differ. If children consistently make stereotypic predictions, we would expect to find an unqualified significant target-type effect that would indicate that children believed others would prefer toys consistent with their interests more than toys inconsistent with their interests. The target-type effect was found, F(2,108) = 61.79, p < .001, but it was subsumed by the significant interaction of cue type and target type, F(2,108) = 15.16, p < .001. Simple effects analyses revealed that the target effect was significant for same-sex cues and for opposite-sex cues (p < .001). When judgments were made from same-sex cues or from opposite-sex cues, Newman-Keuls analyses revealed that children would like the same toys most of all (see Table 2). This is hardly surprising given that they were told that these toys were liked by the target children. The more interesting comparisons concerned whether children predicted that others would like toys that were consistent with the target's favorite toy more than inconsistent toys. When judgments were made from same-sex cues, Newman-Keuls analyses revealed that children made stereotypic predictions (i.e., they said targets would like consistent toys more than inconsistent toys) (p < .05). When judgments were made from opposite-sex cues, however, they did not make stereotypic predictions. This pattern suggests that children make stereotypic predictions selectively rather than consistently. During the interest-only task, children's spontaneous remarks about the sex of target children were recorded. Three girls and four boys mentioned sex. For example, when the experimenter said, "I know a child who really likes to play with a purse and comb set a lot and also with a dollhouse a lot," two girls and one boy asked if the target child was a girl (the experimenter answered that she did not know and repeated the statement). Also, when the experimenter said she knew a child

on toy interests. For each of the fictitious target children,2 young children predicted that they would be more interested in toys consistent with the gender of the target (M = 2.49) than for toys consistent with the interests of the target (M = 1.99), t(29) = 4.13, p < .0001. No significant differences were found between attributed interest in toys consistent with the gender of the target and those previously stated in the story (same toys) (M = 2.39). Thus, the young children in this study predicted others' interests in the same way as did the young children in earlier studies: they used the sex of the person more than using their toy interests to make predictions.

Judgments made when gender labels

were absent.-The main analysis involved assessing whether children consistently make stereotypic predictions when they are only given information about targets' interests. Their predictions were dependent measures in a repeated-measures analysis of variance (ANOVA), with one between-subject factor, sex of subject, and three within-subject factors, condition (single interests, multiple interests/no pictures, and multiple interests/ pictures), cue type (same-sex and oppositesex), and target type (same, consistent, and inconsistent).3 If increasing the salience of interests makes stereotypic predictions more likely, then a significant condition x target type interaction should be found. But the interaction was not significant, indicating that salience of interests did not influence stereotypic judgments. However, the interaction of sex of subject x condition was significant, F(2,108) = 7.98, p < .001. Newman-Keuls analysis revealed that boys attributed to the target children an equal interest in toys regardless of their salience, whereas girls predicted that others would like the toys in the singleinterest condition (M = 2.49) more than toys in the multiple-interests/pictures condition (M = 2.32) or the toys in the multipleinterests/no-pictures condition (M = 2.29) (p

2 Means for same, gender-consistent,and interest-consistenttoy ratingswere summed across the four replication stories. Because half of the subjects were tested with only one story and the other half were tested with four stories,this meant a decrease in subjectsby 50%.However, when t tests for rating scores were individually computed for each story, it was found that differences between all means (except as expected between same and gender-consistenttoys) were significant at at least the .03 level. 3 An analysis of variance selecting for children whose ages ranged between 43.0 and 73.0 months was also calculated because this age range was the most accuratematch to Martin's(1989)

young children. This meant dropping six children between 38.5 and 43.0. The results of this analysis did not reveal significant effects different from the analysis using all children. Selecting for different ages was therefore not used in any further data analysis; results are reported only from the analysis using all ages.

Martin, Wood, and Little

TABLE 2
MEAN PREDICTIONS OF OTHERS' TOY PREFERENCES AS A FUNCTION OF CUE TYPE AND TARGET CONSISTENCY: STUDY 1 SAME-SEX CUES OPPOSITE-SEX CUES

1897

Target Same Boys (n = 30) ......... Girls (n = 26) ......... Total (n = 56) .........
NOTE.-Standard

Target Consistent 2.41 (.39) 2.40 (.43) 2.41

Target Inconsistent 1.89 (.46) 1.92 (.41) 1.91

Target Same 2.57 (.49) 2.73 (.32) 2.65

Target Consistent 2.13 (.40) 2.15 (.37) 2.14

Target Inconsistent 2.29 (.46) 2.15 (.54) 2.22

2.68 (.44) 2.83 (.27) 2.76

deviations are in parentheses.

who liked to play with a tool kit a lot and also with an airplane a lot, two boys spontaneously remarked, "That's a boy." All seven gender remarks "correctly" identified the sex of the target child who would traditionally be associated with the sex-typed toy (e.g., purse and comb sets are associated with girls). These findings suggest that some children may easily draw these gender-label inferences. However, the percentage of these spontaneous gender inferences is still low (2%) given the number of predictions children made. showed Stereotyping scores.-Children high levels of stereotypic knowledge for the 16 toys presented in the two testing sessions (M = 12.78).4 Only five children answered 50% or more of the questions "incorrectly," indicating they did not give traditional sextyping attributions. These findings indicate that most of the children knew the simple stereotypic associations between gender labels and the toys used. Discussion The purpose of this study was to investigate whether children can make complex stereotypic predictions when they know only about someone's interests. We found that children were able to make stereotypic inferences about others, but they did not make these judgments consistently. Instead, children demonstrated selective stereotyping: they made stereotypic judgments for others with toy interests like their own but not for others with interests unlike their own. That is, boys made stereotypic judgments when told about an unknown child with masculine interests but not when they were told about an

unknown child with feminine interests. Conversely, girls made stereotypic judgments when told about an unknown child with feminine interests but not when the other had masculine interests. When interests were made more salient by presenting multiple examples and by showing pictures of toys, stereotypic judgments were not more likely to occur. It is unlikely that gender labels mediated children's judgments. If they did, stereotypic predictions would have been made consistently rather than selectively. Given that children demonstrated own-sex selectivity, a more likely explanation concerns social experience. Because from birth most children are encouraged to engage in traditionally "sexappropriate" activities (see Huston, 1983), they may acquire more direct experience of the associations among own-sex activities than they do for other-sex activities. It is intriguing, however, that some children spontaneously guessed the sex of the target children, indicating that they may have used gender to mediate their judgments. Perhaps gender is very salient for these children. Further research is needed to explore individual differences in using gender to mediate judgments. The present findings reflect on earlier research concerning how children use genderrelated information to make social judgments. In the earlier studies in which children were given conflicting information about someone's sex and an interest (e.g., a boy who likes feminine toys), young children ignored the information about interests in their judgments.

4 For the stereotype measure, children were tested with unequal numbers of toys (due to unequal stories in testing of the replication task). Therefore, to equalize children's toy ratings when calculating the stereotype scores, two toys were randomly dropped from analysis. This resulted in a total of 16 instead of 18 toys.

1898

Child Development
prior to associations between components because it is a simpler cognitive task to learn associations within a category than between categories. Finally, if social experience influences the learning of gender-related associations, we would expect to find that the strength of the association among gender-related items will increase with age. In other words, as children grow older, they will differentiate more between masculine and feminine characteristics. There is evidence of this pattern in one study where children made judgments about how much boys and girls will like masculine and feminine toys. Older children (6-10) were more extreme than younger children (35): they predicted, for example, that boys would like masculine toys more and feminine toys less than did the younger children (Martin, 1989). The ability to use within- and betweencomponent links was assessed in 6-, 8-, and 10-year-old children. Children were told about eight unknown children, each of whom was described as having a masculine or feminine characteristic (cue item) from one of the components. After each description, children made judgments about how much the target child would like other masculine and feminine items from the same and from different components (target items). Method Subjects.- Seventy-six children (38 boys and 38 girls) attending after-school care programs, Sunday schools, and soccer clubs in a large urban area were interviewed. Children from three age groups (6, 8, and 10 years) were tested. The mean ages of the 6-year-old group was 77 months (SD = 3.9), the mean age of the 8-year-old group was 103 months (SD = 3.6), and the mean age of the 10-yearold group was 124 months (SD = 4.3). Most of the children (91%) in the sample were Caucasian, and half were from two-parent families. Materials and procedure.-To make the stories and the questions for the judgment task, three masculine and three feminine items were chosen to represent each of the four components (traits, clothing, occupational aspirations, and toy preferences). For an item to be chosen, it had to be demonstrated as being sex-typed by at least 75% of the children in previous studies. From the set of 24 items, one masculine and one feminine item was selected from each component to be used as the cue in a story (see Appendix B). This resulted in eight stories. The remaining

Given the present findings, their failure to use interests has to be explained as a combination of the lack of knowledge about gender associations (when making opposite-sex-relevant predictions) and some kind of developmental constraint that limits the number of information sources young children can use (when making own-sex-relevant predictions).

Study 2
The results from Study 1 suggest the importance of investigating the development of the complex stereotypic associations that constitute children's gender stereotypes. Just because children have label-component links does not indicate that they can make the full range of associations within stereotypes. Instead, there appear to be developmental changes in learning gender-based associations. One developmental trend is learning first about information that is relevant to one's own sex. The 4- and 5-year-olds in Study 1 understood own-sex- (and not other-sex-) relevant associations within the one component that was tested (i.e., toys). This pattern is consistent with the idea that children learn more in-depth information about own-sex schemas than other-sex schemas (Martin & Halverson, 1981). Despite this tendency, we would expect that, as children get older and acquire more experience, they should also acquire associations among other-sex-relevant information. Thus, the first goal of Study 2 was to further investigate age differences in understanding of own- and other-sex-relevant information by testing children older than those used in Study 1. In Study 1, we assessed how children make complex stereotypic predictions within the domain of toys. But adults are able to use their gender schemas to make predictions across content domains. No research has been conducted to investigate when children begin to make these between-component judgments. Thus, the second goal of Study 2 was to examine the development of these between-component links by using four content domains. We expected that, similar to withincomponent links, children will first begin to make judgments based on own-sex information because of the influence of own-sex selectivity. At a later age, children should begin to make between-component judgments based on opposite-sex information. No direct evidence exists concerning whether the associations within and between components differ. We might assume, however, that withincomponent associations would be learned

Martin, Wood, and Little eight feminine and eight masculine items were used to develop the questions. For each story, eight questions were formulated: two were target items from the same component as the cue (one was same-sex relevant and the other was opposite-sex relevant) and six were targetitems from each of the other three components (one fromeach component was samesex relevant and one from each was oppositesex relevant).Thus, children were told stories about eight children and made eight predictions about each child. Five female undergraduateresearch assistantsinterviewed the children. Each experimenter underwent extensive training and was periodically monitored to maintain as much uniformity in interview techniques as possible. The experimenters were not familiar with the hypotheses of the study. Each child was tested individually in a quiet area away from other children. Before data collection began, children were trained to use the face scale to indicate their predictions of others' preferences. Then the judgment task was begun. First, children were told about a target child of unspecified sex who likes an item (cue item) from one of the components (e.g., "I know a child who really likes to play with tool kits"). Then children were asked to indicate on the face scale how much the target child would like a series of masculine and feminine items (e.g., "How much would this child want to wear a dress?") (target items), two of which were fromthe same component (within-component judgments), and six of which were from the other three components (between-component judgments). Cue information was repeated during questioning to ensure that children did not forgetthe cue. If a child indicated that he or she did not understandthe meaning of a targetitem, the experimenterwould give him or her a standard definition that the child could understand.

1899

sex; and target type: consistent and inconsistent with the sex-typing of the cue item). Although the pattern of results differed slightly owing to the content domain, componenttype was not included as a factorin the final analysis so as to increase the reliability of the data. The dependent measure was the preference ratings given to the target items. The first prediction was that children would make stereotypicjudgments fromownsex-relevant cues prior to opposite-sex-relevant cues. To test this prediction, children's judgments when given same- and oppositesex cues were comparedacrossages. The predicted developmental trend was found (see Table 3). The 6-year-oldchildren were able to make stereotypic predictions (i.e., the target effect was significant) when- given own-sex cues but not when given opposite-sex cues. In contrast, the 8- and 10-year-olds made stereotypic predictions when given either same-sex or opposite-sex cues (p's < .001). Developmental changes were expected for the extremity of stereotypic judgments. Overall, the analysis showed that children made stereotypic predictions (i.e., the target effect was significant),F(1,70) = 350.48, p < .001, and the main effect for age was significant, F(2,70) = 5.95, p < .01. But, as expected, both of these effects were subsumed by higher-orderinteractions.Specifically, we expected to find that judgments would be more extreme for older children than for the younger children. This prediction was supported by the significant age x target type interaction,F(2,70) = 19.51, p < .001. Simple effects analyses demonstratedthat the targettype effect was significant at each age level (p's < .001), indicating that children in each age produced stereotypic judgments. Stereotypic judgments were more extreme, however, for the older children than for the younger ones. According to simple effects analyses, the increase in extremitywas due to age changes in judgments about target charThree predetermined random orders of acteristics that were inconsistent with the within- and between-component target items cues ratherthan for those that were consistent and cue items were constructed. Each order with cues. Newman-Keuls analyses showed was given to at least 24 children (at least four that the 6-year-olds predicted that others males and four females from each of the three would like target characteristicsthat were inage groups). consistent with cues more than did the 8-yearolds and 10-year-olds.Similaranalyses of conResults Children's judgments were analyzed us- sistent target characteristicsshowed no ageing a repeated-measures ANOVA. There related differences. Within these content dowere two between-subject factors (sex and mains, as children get older, they appear to which characteristics not do age: 6, 8, and 10 years) and three within- learn more about factors (judgment-type: within- and co-occur. subject The degree of stereotyping varied debetween-component judgments; cue type: same and opposite sex-typing to the subject's pending on whether children's judgments

1900

Child Development
TABLE 3
AS MEAN PREDICTIONS CONCERNING OTHERS' CHARACTERISTICS A FUNCTION OF AGE, CUE TYPE, AND TARGET CONSISTENCY: STUDY 2 SAME-SEX CUE OPPOSITE-SEX CUE

Target
AGEANDSEX Consistent

Target
Inconsistent

Target
Consistent

Target
Inconsistent

Six-year-olds: Girls ............... Boys................. Total ............... Eight-year-olds: Girls ............... Boys ................ Total ............... Ten-year-olds: Girls................ Boys................ Total ............... Girls.................. Boys..................

2.30 (.32) 2.20 (.35) 2.25 2.32 (.24) 2.11 (.30) 2.23 2.34 2.24 (.18) 2.29 2.32 2.19

1.78 (.38) 1.63 (.28) 1.71 1.32 (.20) 1.51 (.35) 1.41 1.38 1.38 (.15) 1.38 1.49 1.51

2.06 (.29) 2.10 (.39) 2.08 2.08 (.40) 2.13 (.33) 2.10 2.06 2.32 (.17) 2.19 2.07 2.18

1.96 (.39) 2.00 (.45) 1.98 1.48 (.33) 1.58 (.41) 1.53 1.55 1.41 (.19) 1.48 1.65 1.66

(.26)

(.21)

(.18)

(.30)

NOTE.-Standarddeviations are in parentheses.

were based on same or opposite cues, F(1,70) = 20.79, p < .001. Girls and boys showed slightly different patterns, as indicated by the significant sex x cue type x target type interaction, F(1,70) = 4.14, p < .05. Simple effects analyses showed that the cue type x target type interaction was significant for girls (p < .001) and marginally significant for boys (p < .11). Further analyses showed that both girls and boys made stereotyped predictions: they predicted that others would like consistent items better than inconsistent items regardless of cue type (p's < .001). As can be seen in Table 3, however, stereotypic predictions tended to be less extreme when made from opposite-sex cues than when made from same-sex cues. One reason that the ratings were less extreme for girls was that, for consistent items, they predicted that others would like the items more when following a same-sex cue than an opposite-sex cue. For both sexes, the ratings were less extreme because, for inconsistent items, they predicted that others would like the items less when following a same-sex cue than an opposite-sex cue (p's < .01). We suspected that stereotypic judgments (i.e., a target effect) would occur for withincomponent judgments prior to betweencomponent judgments, yet this was not borne

out. Within- and between-component judgments showed somewhat different patterns, however. The sex x judgment type x cue type interaction was significant, F(1,70) = 21.99, p < .001. Simple effects analyses showed that the cue x sex interaction was significant for judgments made between components and not for judgments made within components. Further analyses showed that, for between-component judgments, the cue type effect was significant for both sexes (p's < .001) but the patterns differed. As can be seen in Table 3, boys tended to predict that others would like the items more when given same- versus opposite-sex cues. In contrast, girls tended to predict that others would like the items more when given opposite-sex rather than same-sex cues. Because this finding refers to children's predictions about overall liking (i.e., for consistent and inconsistent characteristics combined), it is not apparent how it relates to developmental changes in stereotypes. Discussion The purpose of Study 2 was to investigate the ability of children of three ages to make gender-related associations within and between four content domains. Two major developmental trends were apparent. First, children appear to learn the associations among

Martin, Wood, and Little

own-sex characteristics before learning the associations among opposite-sex characteristics. The 6-year-olds in this study were similar to the 4-year-olds in Study 1 in making stereotypic judgments selectively: they made stereotypic judgments for others with interests like their own and not for others with interests unlike their own. This selectivity suggests that gender labels are not simply mediating stereotypic judgments. By the age of 8, children demonstrated understanding of all types of associative links. They made stereotypic predictions about others' interests in one content domain given only one piece of information from a different content domain, and they could do this both when the other had own-sex- or opposite-sex-relevant characteristics. The second major developmental trend concerned stereotype extremity: older children were more extreme in their stereotypic judgments than the younger children. This difference was particularly due to older children expecting stronger negative associations among stereotype-inconsistent characteristics than younger children. For example, when told about someone who likes airplanes (i.e., a masculine interest), older children thought the person would dislike dolls (i.e., a feminine interest) more than did younger children. The strength of positive associations among sex-consistent characteristics did not differ for younger and older children. One unexpected finding was that stereotype extremity was more pronounced for judgments based on own-sex-relevant cues (vs. opposite-sexrelevant cues), and this tendency was more apparent in girls than in boys. It was also somewhat surprising to find so few differences between judgments made within versus between components. Withincomponent judgments were not learned first. Instead, once children could make one type of judgment, they could also make the other type. Apparently, the gender concepts of masculinity and femininity are powerful enough to overwhelm differences based on specific content domains.

1901

General Discussion
Implications for the Development of Gender Stereotypes Contrary to the popular assumption that children's stereotype knowledge reaches ceiling in early childhood, the present findings suggest that stereotypes continue to develop into middle childhood. By using the component model to consider the present results and the results of earlier studies, we can elab-

orate our knowledge about stereotype development. Specifically, stereotypes appear to develop through a series of stages. Children in the first stage (the one that has been emphasized by most researchers) learn what kinds of things are directly associated with each sex, such as "boys play with cars," and "girls play with dolls." At least for the domains we assessed, around the ages of 4-6, children seem to move to the second stage, where they begin to develop the more indirect and complex associations for information relevant to their own sex but have yet to learn these associations for information relevant to the opposite sex. By the time they are 8, children move to the third stage, where they have also learned the associations relevant to the opposite sex. These children have mastered the gender concepts of masculinity and femininity that link information within and between the various content domains. In addition to changes in the types of associations, the quality of these associations also changes as children get older. Specifically, older children make more extreme stereotypic judgments than do younger children when they know only someone's sex (Martin, 1989) or when they know only someone's interests (the present studies). This tendency may be due to expansions in the breadth of children's gender-related knowledge. As more information becomes associated with gender labels and/or gender concepts, perceived distinctions may increase (see Martin, 1989). The origins of these developmental trends are unclear. Social experience influences the types of associations children learn, although the present results suggest that social experience does not simply aid children's learning of which things co-occur, it also helps children learn which things do not cooccur. The timing of these developmental trends may also depend on the nature of the information. It may be that more concrete and obvious characteristics, such as toys and appearance, are learned more quickly than abstract characteristics (e.g., traits). Implications for Assessing Children's Knowledge These findings demonstrate that gender concept learning does not proceed as expected, and, furthermore, suggest that the learning of other concepts may follow similar unexpected patterns. Specifically, the mediation hypothesis was not supported: children did not automatically associate two pieces of information, even when each was associated with a gender group. For instance, even when

1902

Child Development
being female. Thus, categorical information about gender subsumes all other information. Children in Stage 2 make perplexing social judgments. When given conflicting category and individuating information, they make category-based predictions (i.e., based on person's sex). But, when they are not told someone's sex, they can make predictions based on individuating (i.e., component) information, at least when predicting about others with interests like their own. This suggests that own-sex-relevant information is diagnostic whereas opposite-sex-relevant information is not. Thus, it is not surprising that in previous studies when children have been asked to make predictions about others with interests unlike their own, they used sex as a basis for making predictions because only sex had predictive value. Interests would not be predictive because these associations have not yet been learned. What is surprising is what happens when children have been asked to make predictions about others with interests like their own. In these cases, they have two sources of information with predictive value: the person's sex and the sex typing of their interests. Given either of the two information sources, children can and do make stereotypic judgments but, when these two sources of information lead to conflicting predictions, these children default to the predictions based on sex. If interests are diagnostic for these children, why are they ignored when information about sex is present? Perhaps this can be explained as a general cognitive constraint. Young children may base inferences on only one piece of information due to their inability to integrate multiple pieces of information at the same time (see Higgins, 1981). This type of cognitive constraint helps explain how much but not what kind of information children use. If constrained to only one type of information, the most salient information may be used (see Higgins, 1981). As children grow older, they tend to change from using concrete, external categories to using more abstract and internal categories to describe others (Livesley & Bromley, 1973; Peevers & Secord, 1973; Scarlett, Press, & Crockett, 1971). Similarly, for these young children, gender labels (that represent external categories) may be perceived as being more salient than sex-typed interests (that represent internal categories) (Martin, 1985; Martin & Halverson, 1981). Further research is needed to understand the kinds of social judgments Stage 2 children make.

children understood that boys play with cars and boys play with airplanes, they did not necessarily understand that those children who play with cars are likely also to play with airplanes (and not dolls). The formation of these label-component links does not guarantee the formation of other logical links among items. Instead, an incomplete associative network forms and is used for information processing. Further research is necessary to demonstrate whether similar patterns of learning occur when children acquire other kinds of concepts. Given that children's gender stereotypes continue to develop during middle childhood, we should not assume that research on gender stereotype knowledge is uninteresting. For instance, we now need to reconsider how this knowledge influences behavior. Are children's behaviors more likely to be influenced by simple label-focused concepts of gender or by the more complex, multifaceted, and multipredictive (i.e., from either labels or components) fully formed concepts? Possibly both influence behavior, although in different ways. Additional research, using more sophisticated methods, is needed to assess how children's gender-related knowledge influences behavior.

To understand how judgments are made, it is necessary to assess the predictiveness or diagnosticity of the available information (Krueger & Rothbart, 1988). Researchers disagree about the predictive value of categorical (e.g., person's sex) and individuating (e.g., person's interest) information when making judgments about others. In some cases, individuating information appears to have stronger predictive value than categorical information (Locksley, Borgida, Brekke, & Hepburn, 1980) whereas, in other cases, both types of information have predictive value (Rasinski, Crocker, & Hastie, 1985). Diagnosticity of information is an especially important issue in the developmental literature. How does the predictive value of information change as children's stereotype knowledge develops from simple to more complex types of associations? In Stage 1, when children know only the simple gender label-component associations, their range of gender-based predictions is limited: gender labels have predictive value, whereas other types of gender-related information do not. Essentially, masculinity is equivalent to being male and femininity is equivalent to

Implicationsfor UnderstandingChildren's Social Judgments

Martin, Wood, and Little

By the time children have reached Stage 3, they have at their disposal the full range of stereotype knowledge. This marks the complete emergence of the psychological constructs of masculinity and femininity which are distinct (but not entirely independent) from actual gender labels. Developing an understanding of the concepts of femininity and masculinity does not, however, signal a lessening of gender stereotyping. Older children simply have more levels of information (i.e., gender labels and gender concepts) upon which judgments can be made, and they clearly use both of these types of information when making social judgments (Berndt & Heller, 1986; Martin, 1989). Having an understanding of gender concepts, however, may effectively decrease the salience and accessibility of gender labels because children can develop subtypes of individuals within gender groups (e.g., masculine girls) that have more predictive value than either the label or the concept alone. To understand how children make social judgments, developmental researchers need to consider both the predictive value of information as well as specific cognitive constraints that may limit its use even when it has diagnostic value. As children get older, the range of gender-related information with diagnostic value increases. Categorical information, at least in the case of gender, also remains diagnostic but may be outweighed by more diagnostic individuating information. Summary In two studies, we investigated developmental changes in the complexity of gender stereotypes. As children grow older, their ste-

1903

reotypes become increasingly elaborated as more associative links form within and between the content domains traditionally associated with gender groups. Potentially, these extensive associations provide cognitively economical, although often inaccurate, ways to process social information. We must remember, however, that having stereotype knowledge does not guarantee its use. By adulthood, we all have learned cultural gender beliefs, and yet we do not always use these beliefs to make judgments. There is more to using gender stereotypes than just having gender-related associations; for example, personal values or the salience of gender also influence whether we access or ignore our gender knowledge base. Nonetheless, it is important to understand more fully how and when children come to acquire the complex set of associations that our culture deems relevant to gender.

Appendix A Toys Used in Study 1

Masculine trucka army tanka train engineb carsb
motorbikea

Feminine doll clothesa purse and comb seta sewing machineb dollsb makeup kita tea set with traya baby doll and criba iron and ironing board dollhouse

airplane G.I. Joe soldiera tool kita baseball

aThese toys were used as cue items. b These toys were used as response items.

Appendix B Items Used in Study 2

Masculine Items Traits. ................. Appearance............ Occupations ........... Toys .................. gets into fights (aggressive) is a strong person (strong)a says bad words (coarse) suit shirt and tie football shirt and trousersa construction workera stockbroker(works with money) plumber truck airplane tool kita Feminine Items cries a lot (emotional) always says "thankyou" (appreciative)a is a shy person (meek, mild) blouse and skirt skirta dress nursea hairdresser seamstress (sews clothes) baby doll and crib makeup kit Barbie dolla

a These items were used as cues.

1904

Child Development
Martin, C. L. (1985, April). The influence of sex tion. Paper presented at the biennial meeting of the Society for Research in Child Development, Toronto. Martin, C. L. (1989). Children's use of genderrelated information in making social judg-

References
Ashmore, R. D., & Del Boca, R. K. (1979). Sex stereotypes and implicit personality theory: Toward a cognitive-social psychological conceptualization. Sex Roles, 5, 219-248. Berndt, T., & Heller, K. A. (1986). Gender stereotypes and social inferences: A developmental

stereotypes on children's impression forma-

study. Journal of Personality and Social Psychology, 50, 889-898. Blakemore, J. E., LaRue, A. A., & Olejnik, A. B. (1979). Sex appropriate toy preference and the ability to conceptualize toys as sex-role related.

ments. Developmental Psychology,25, 80-88.

Martin, C. L., & Halverson, C. F. (1981). A schematic processing model of sex typing and stereotyping in children. Child Development, 52, 1119-1134. Martin, C. L., & Halverson, C. F. (1987). The roles of cognition in sex roles acquisition. In D. B.

Developmental Psychology, 15, 339-340.

Bradbard, M. R., Martin, C. L., Endsley, R. C., & Halverson, C. F. (1986). Influence of sex stereotypes on children's exploration and memory: A competence versus performance distinction. Developmental Psychology, 22, 481-486. Deaux, K., & Lewis, L. L. (1984). Structure of gender stereotypes: Interrelationships among components and gender label. Journal of Personal-

Carter(Ed.), Current conceptions of sex roles and sex typing: Theoryand research(pp. 123137). New York: Praeger. Martin, C. L., & Little, J. K. (in press). The relation of gender understanding on children's sextyped preferences and gender stereotypes.

Child Development.
Martin, C. L., & Wood, C. H. (1987). Children's sex-

ity and Social Psychology,46, 991-1004.

Del Boca, F. K., & Ashmore, R. D. (1980). Sex role stereotypes through the life cycle. In L.

typed interest attributions. Paper presented at

the biennial meeting of the Society for Research in Child Development, Baltimore. O'Keefe, E. S., & Hyde, J. S. (1983). The development of occupational sex-role stereotypes: The effects of gender stability and age. Sex Roles, 9, 481-492. Papalia, D. E., & Tennent, S. S. (1975). Vocational aspirations in preschoolers. Sex Roles, 1, 197199. Peevers, B. H., & Secord, P. F. (1973). Developmental changes in attribution of descriptive concepts to persons. Journal of Personality and

Wheeler (Ed.), Review of personality and social psychology (Vol. 1, pp. 163-192). Beverly Hills, CA: Sage. Gelman, S. A., Collman, P., & Maccoby, E. E. (1986). Inferring properties from categories versus inferring categories from properties: The case of gender. Child Development, 57, 396-404. Halverson, C. F., & Martin, C. L. (1985). Sex-typed

knowledge: Who remembers what? Unpublished manuscript. (Available from School of Family Resources and Human Development, Arizona State University, Tempe.) Higgins, E. T. (1981). Role taking and social judgment: Alternative developmental perspectives and processes. In J. H. Flavell & L. Ross

Social Psychology,27, 120-128.

(Eds.), Social cognitive development:Frontiers

and possible futures (pp. 119-153). Cambridge: Cambridge University Press. Huston, A. C. (1983). Sex-typing. In E. M. Hetherington (Ed.), P. H. Mussen (Series Ed.),

Rasinski, K. A., Crocker, J., & Hastie, R. (1985). Another look at sex stereotypes and social judgments: An analysis of the social perceiver's use of subjective probabilities. Journal of

Personality and Social Psychology, 49, 317-

Handbookof child psychology: Vol. 4. Socialization, personality, and social development

(pp. 387-467). New York: Wiley. Krueger, J., & Rothbart, M. (1988). Use of categorical and individuating information in making inferences about personality. Journal of Per-

sonality and Social Psychology,55, 187-195.

Kuhn, D., Nash, S. C., & Brucken, L. (1978). Sex role concepts of two- and three-year-old children. Child Development, 49, 445-451. Livesley, W. J., & Bromley, B. D. (1973). Person perception in childhood and adolescence. London: Wiley. Locksley, A., Borgida, E., Brekke, N., & Hepburn, C. (1980). Sex stereotypes and social judgment. Journal of Personality and Social Psychology, 39, 821-831.

326. Rholes, W. S., & Ruble, D. L. (1984). Children's understanding of dispositional characteristics of others. Child Development, 55, 550-560. Scarlett, H. H., Press, A. N., & Crockett, W. H. (1971). Children's description of peers: A Wernerian developmental analysis. Child Development, 42, 439-453. Thompson, S. K. (1975). Gender labels and early sex-role development. Child Development, 46, 339-347. Williams, J. E., Bennett, S. M., & Best, D. L. (1973). Awareness and expression of sex stereotypes in young children. Developmental Psychology, 11, 635-642. Zucker, J. K., Wilson, D. N., & Stern, A. (1985,

April). Children's appraisals of sex-typed behavior in their peers. Paper presented at the biennial meeting of the Society for Research in Child Development, Toronto.