Children's Gender-Based Reasoning about Toys - People Server at

Children's Gender-Based Reasoningabout Toys

Carol Lynn Martin, Lisa Eisenbud, and Hilary RoseArizona State University

MARTIN, CAHOL LYNN; EISENBUD, LISA; and ROSE, HILARY. Children's Gender-Based Reasoning

about Toys. CHILD DEVELOPMENT, 1995, 66,1453-1471. The goal of these studies was to investi-gate how preschool children use gender-based reasoning in making judgments about toy prefer-ences for themselves and for others. In Studies 1 and 2, children (n = 22, n = 71) were shownunfamiliar, non-sex-typed toys and asked to rate how much they, other girls, and other boyswould like each toy. As expected, children made gender-based inferences: "What I like, childrenof my sex will also like, and children of the other sex will not like." Study 3 was designed to assesshow children use gender-based reasoning to make decisions about attractive and unattractive toyswhen they are given gender labels. Children (n = 91) were shown unfamiliar toys varying inattractiveness that were given explicit gender labels (e.g., "this is a toy girls really like") or nolabel. With a different experimenter (to avoid demand characteristics), children rated their ownand others' liking of the toys. Children used gender labels to guide their own preferences andtheir expectations for others. Even with very attractive toys, children liked toys less if they werelabeled as being for the other sex, and expected other girls and boys to do the same. The roleof gender-based reasoning in cognitive theories of gender and on children's play preferences isdiscussed.

In some situations, children have access traditionally stereotyped for their own sexto explicit gender-related knowledge (e.g., more than toys stereotyped for the other sex.they may have been told that boys like to Sex-typed toy preferences have been appar-play with trucks), but in other situations, ent in studies of children's play (Eisenberg,they do not. The goal of the present studies Murray, & Hite, 1982; Fein, Johnson, Kos-was to investigate children's use of gender- son. Stork, & Wasserman, 1975; Goldman,based inferences in reasoning about toy Smith, & Keller, 1982), and when childrenpreferences for themselves, for other girls, have been asked to select favorite toys forand for other boys, in both kinds of situa- themselves or for others (Bradbard, 1985;tions. In studies 1 and 2, we explored Bradbard & Parkman, 1983; Goldman et al.,whether children would make gender-based 1982; Robinson & Morris, 1986; Ross &inferences when asked to make judgments Ross, 1972).about their own and others' liking of toys ^ r u -u i ,... .1 . . . . . u- u iU I, J • One consequence ot children selectingin the situation m which they had no prior . /* , ^ • j i ,. l. -r1 1 J r .1 4. • r 4.U t own-sex-stereotyped toys is that it limitsknowledge of the sex-typmg of the toys. . . . • /D in^r\ T>.-irC i j o J - j i i u-ij ' their experiences (Bem, 1975). DifferentStudy 3 was designed to explore children s . r. j - i- ^ j-rr a.

r j u j J. 1 - j types 01 toys encourage distinctly differentuse of gender-based reasoning to make judg- . r i j i • /r> B O I -

4- u 4. iu • J lu ' T-T • f types of play and learning (Connor & Serbm,ments about their own and others liking of /c^r, rSr, • s u i i n o o c i,- s. 1 iu • 1- -j. • r 1977; O Brien & Huston, 1983; Serbm &toys when they were given explicit mforma- ^ , ««r.\ TU -4. • • J ,. ^i. ^..•' 1 .,.1 . • rj. Gonnor, 1979). Thus, it is important that wetion about the sex-typing of toys. j i. j r u l. l -u l

• ^ " understand more fully how children makeChildren's Toy Choices decisions about what they want to play with.

A consistent finding in the develop- And, because children also influence theirmental literature is that children prefer toys peers' toy choices through their reactions to

Versions of these studies were submitted by Lisa Eisenbud (under the direction of CarolMartin) to Arizona State University in partial fulfillment of the requirements for the M.A. degree.The ideas for Studies 1 and 2 were developed in part from discussions with Charles Halverson.We appreciate the help of Jennifer Bowers, Caroline Descano, Pamela Donovan, Mariss Karbon,LaDan Helsing, and Michelle Tucci in data collection, and thank Richard Fabes, Iris Godes,and Gary Peterson for their helpful suggestions. Appreciation is also extended to the ASU ChildDevelopment Laboratory, and to the children who participated in the studies. Portions of thesestudies were presented at the meetings of the Society for Research in Child Development, NewOrleans, 1993. Requests for reprints should be sent to the first author. Department of FamilyResources and Human Development, Arizona State University, Tempe, AZ 85287-2502.

{Child Development, 1995,66,1453-1471. © 1995 by the Society for Research in Child Development, Inc.All rights reserved. 0009-3920/95/6605-0003$01.00]

1454 Child Development

them (e.g.. Fagot, 1977; Langlois & Downs,1980), we also need to investigate how chil-dren determine the toys they consider to beappropriate for others.

Reasoning about Sex-Typed ToysCognitive theories of gender-role acqui-

sition are based on the idea that childrensocialize themselves into gender roles. Gen-der stereotypes provide children with nor-mative information about how each sexshould look, act, and think, according to cul-tural beliefs (Kohlberg, 1966; Martin &Halverson, 1981). One way to communicatethese cultural gender stereotypes to childrenis by explicit gender labeling of objects andactivities (e.g., "boys like to play with cars")through television, books, peers, and adults.

According to Martin and Halverson's(1981) cognitive approach, children use gen-der to reason about toys following a specificpattern. First, they decide if the toy is "forboys" or "for girls," drawing on their pre-existing beliefs about which toys are typi-cally liked by girls and which toys are typi-cally liked by boys. Then, they comparetheir answer about who usually likes the toyto their knowledge of which sex they are.For instance, a girl will reason that a dollis something girls usually like, I am a girl,therefore I will probably like to play withthe doll. In some situations, this kind of rea-soning may become so well learned that itis done virtually automatically.

In most studies of toy preference, theinfluence of gender labels has been inferredfrom children's behavior. The problem isthat in these studies, children's familiaritywith toys and their prior reinforcement his-tories with them, rather than gender labels,may be accounting for their behavior. Forthis reason, the ideal method for assessingthe effects of gender labels involves usingnovel objects or toys given arbitrary sex-typed labels ("this is a toy girls like"). Be-cause novel toys have not been previouslyassociated with either sex, girls and boys donot have differential exposure to or experi-ence with them (Bradbard & Endsley, 1983).

Labeling toys as being for girls or forboys influences children's behavior in anumber of ways. First, children's explora-tion of toys varies depending on how a toyis labeled. Ghildren tend to explore same-sex-labeled toys more than other-sex-labeledtoys (;Bradbard & Endsley, 1983; Bradbard,Martin, Endsley, & Halverson, 1986). Labelsalso influence children's performance ongames. When a novel game is labeled as be-

ing for their own sex, children perform bet-ter than when a game is labeled as being forboth sexes, and much better than when it islabeled as being for the other sex (Monte-mayor, 1974). Recall of information also isinfluenced by gender labels. Ghildren betterremember the names of objects labeled asbeing for their own sex than names of objectslabeled as being for the other sex (Bradbard& Endsley, 1983; Bradbard et al., 1986).

Reasoning about Novel ToysOnce children learn gender labels, their

behavior often changes so that it matchesstereotypic expectations. However, whathappens when children are asked to maketoy choices when the sex-typing of a toy isunknown? Presumably their toy choiceswould be affected by the perceived attrac-tiveness of the toy. The interesting situationto consider is children's reasoning if theythen are asked to make judgments about toychoices for other girls and for other boys. Inthis case, there are two likely patterns of re-sponses. One pattern, an egocentric pattern,would involve children using their ownopinion about how attractive a toy is as theonly criterion for making judgments aboutall other children's interest in the toy. Inother words, a child could reason, "I like thistoy (because it is attractive); therefore othergirls and other boys also would like thistoy."

The other pattern, a gender-based or"gender-centric" pattern, would involvechildren drawing inferences based on theirown liking of toys combined with their be-liefs about boys and girls. For example, a girlmight reason, "I like this toy (because it isattractive); therefore other girls would likeit and other boys would not like it" (see Fig.1). This pattern involves the same informa-tion sources as in the examples presented byMartin and Halverson (1981), although theorder of use of the information is somewhatdifferent. In this case, even without explicitgender labels, the girl may be relying on anunderlying abstract theory of gender groupdifferences, thereby inferring that becauseshe likes a toy, children of the other sexmight not like the toy. Furthermore, she alsomay be relying on an abstract theory ofwithin-group similarity, thereby inferringthat because she likes the toy, children ofthe same sex also might like the toy.

The literature on conceptual develop-ment provides insights into why childrenmight form such theories about gendergroups. Theories underlying categories pro-

Martin, Eisenbud, and Rose 1455

- F o r

Not For Me

1 am a Girl

For Boys

FIG. 1.—Children's reasoning about novel toys

vide conceptual coherence to categories,even when group members seem to be dis-similar (Medin, 1989). These implicit theo-ries are based on assumptions that categorymembers share deeper properties, or es-sences (Gelman, 1989; Gelman & Markman,1986, 1987; Medin, 1989). Because groupmembers share a category label, perceiversassume that they also share other properties.For instance, after being taught a novel char-acteristic about boys, children assume thatan unfamiliar boy also will have the charac-teristic, even if he doesn't look like the otherboys (Gelman, Gollman, & Maccoby, 1986).

OverviewIn the first two studies, we examined

children's reasoning about toys when theyhad no explicit gender labels to provide in-formation about the sex-typing of the toys.To do this, children were asked to makejudgments about how much they and otherchildren would like a group of novel, non-sex-typed toys. We expected to find thatchildren would use gender-based reasoningto help them make judgments about toychoices for others because they assume acommon essence underlying the gender cat-egories.

In the third study, we explored the in-fluence of explicit gender knowledge (i.e.,gender labels) on children's judgments ofho^v much they and other children would

like a group of unfamiliar toys that varied inattractiveness. We were particularly inter-ested in investigating the kind of situationdescribed by a colleague in which he ob-served a young boy happily playing with atoy racing car until the helmet of the racecar driver fell off, revealing a female withblonde hair. The boy then dropped the carlike a "hot potato" (D. B. Garter, personalcommunication, April 1987). Although re-searchers have shown that children avoidplaying with sex-inappropriate toys (e.g.,Frey & Ruble, 1992; Hartup, Moore, &Sager, 1963), the question remains whethergender labels are so powerful that they alterchildren's desire for unfamiliar and very at-tractive toys.

Study 1To examine children's use of gender-

based inferences when they do not know thesex-typing of a toy, we simply asked childrento make judgments about ho\v much theyand others would like a set of unfamiliar,non-sex-typed toys. If children show an ego-centric pattern of responding, we shouldfind that their judgments of how much othergirls and boys like the toys will vary withtheir own liking of the toys. If children showthe expected gender-centric pattern, theirjudgments of how much children of the samesex like the toys should be similar to their


judgments of how much they like the toys,whereas their judgments of how much chil-dren of the other sex like the toys should notmatch their own preferences.

MethodSubjects.—Twenty-two children (11

boys, 11 girls) ranging in age from 50 to 67months (M = 59 months) participated in thestudy. Most were Gaucasian (77%) and frommixed-sex preschools in middle-class neigh-borhoods.

Materials.—Because of the importanceof using toys that were unfamiliar to the chil-dren, we bought unusual toys and toylikeobjects designed for adults (e.g., magneticball sculpture) from toy stores, zoos, and giftshops. We also included six familiar sex-typed objects. Two adult judges made thefinal selection from this group of 10 "toys"to be used in the study: One sex-typed mas-culine toy (a transformer) and one sex-typedfeminine toy (a doll house) were selected onthe basis of being familiar sex-typed toys,and eight objects and toys were selected thatwere judged to be unfamiliar to most chil-dren, interesting, and not sex-typed. Theeight novel objects (and the brief labels usedfor them in the tables) included a mag-netic balls and links sculpture (magnetstand), spinning bells on a stand (spinningbells), colored interlocking gears (wonderwheels), metal nail sculpture (pin pressi-ons), prism looking glass (looking glass), col-ored sand in a fhin transparent plastic box(magic hill), colored magnetic puzzlesquares (magnetix), and a create-a-creatureflip sections book (animal flip book).

A "cup" rating scale was used so chil-dren could indicate how much they liked thetoys. The scale was made up of Styrofoamcups cut into four heights (5, 4, 3, and 2inches) and glued upside down from tallestto smallest on a cardboard base. Similarscales have been used in other studies (e.g.,Martin, 1989) with children as young as 4years old.

Procedure.—A female experimentertested children individually. The experi-menter explained that she was interested infinding out how much the children likeddifferent things. The experimenter thenshowed the children how to use the four-point cup rating scale by using familiar foodsuntil the children could successfully indi-cate each level of the scale. She explainedthat when they pointed to the tallest cup, itmeant they liked something a lot (scored asa 4); the next tallest cup meant they liked

something pretty much (3); the second short-est cup meant they liked something okay (2);and the shortest cup meant they liked some-thing a little bit (1).

The experimenter then said she wantedto flnd out how much children like differenttoys. A toy was randomly selected, broughtout of a box, and put on the table. The exper-imenter demonstrated each toy (when ap-propriate), then pushed it within reach of thechildren. Then the children had 30 sec toinspect the toy. When time was up, the ex-perimenter moved the toy out of reach andassessed familiarity by asking if tlie childhad ever seen the toy before. The ratingscale was placed in front of the children andthey were asked to rate how much they likedthe toy (by pointing to the appropriate cup),how much they thought girls would like toplay with it, and how much they thoughtboys would like to play with it (the order ofasking about girls or boys varied). The sameprocedure was repeated for each of the 10toys.

ResultsThe familiar sex-typed toys were pre-

sented with the unfamiliar toys to allow chil-dren to rate and play with familiar toys, butthey were not included in any of the analy-ses. As a manipulation check, sex-typing ofthe novel toys was assessed by comparinghow much girls and boys said they wouldlike each toy. Only one of the unfamiliar toyswas found to be sex-typed by the children,and it was dropped from all further analyses(see Table 1). Familiarity ratings (see Table1) showed that, for the toys selected to beunfamiliar, most children had not seen thesetoys before (59% to 82%). Because some toyswere more familiar than expected, analyseswere done both including and excludingthese toys.

There are several ways to investigatechildren's patterns of responses for the toys.We chose the method of analyzing absolutedifference scores because it is the most di-rect way to assess whether children are us-ing a gender-centric pattern. Specifically,the absolute difference between children'sown liking of the toys and their predictionsfor same-sex peers was compared with theabsolute difference between their own lik-ing and their predictions for other-sex peers,using t tests. Evidence of a gender-centricpattern would be apparent if the magnitudeof difference between the own- and same-sex ratings was smaller than the magnitudeof difference between the own- and other-


TABLE 1

GIRLS' AND BOYS' RATINGS OF LIKING AND PERCENTAGE OFCHILDREN WHO ABE UNFAMILIAR WITH THE TOYS: STUDY 1

Toy Girls Boys % Unfamiliar

Magnet stand 3.45 2.82 68Spinning bells 3.27 3.27 64Wonder wheels 2.91 3.55 64Pin pressions 3.09 3.91 59Looking glass 3.27 3.64 59Magic hill 3.73 3.91 73Magnetix 2.64 3.37" 82Animal flip book 3.00 3.27 82

* Indicates that girls and boys differed on how much they liked thetoy, p < .05.

sex ratings. Using simple mean comparisonswould also be effective, but given that wewere not really interested in mean scoresand the crucial comparisons are the self andsame-sex versus self and other-sex differ-ences, it is more meaningful to compare di-rectly these differences. Subtracting from aconstant score (i.e., own-ratings) is equiva-lent to a direct comparison of the scores, butby doing the subtraction, we could assessthe degree of difference as well.

Analyses of all toys combined.—Themagnitude of differences was compared forthe novel toys, using all the toys except theone toy that was sex-typed by the children.For each toy, two scores were calculated.First, we calculated the absolute differencebetween each child's own liking of the toyminus the child's prediction for same-sexpeers. Second, we calculated the absolutedifference between each child's own likingof the toy minus the child's prediction forother-sex peers. Then these scores weresummed across toys. As expected, childrenused a gender-centric pattern: the absolutedifference between children's own likingminus predictions for same-sex peers (M =6.95) was found to be smaller than the abso-lute difference between children's own lik-ing minus their predictions for other-sexpeers (M = 11.77), t(21) = 4.66, p < .001.Analyses of mean scores also confirmed thispattern. Overall, children liked the toys (M= 3.34), and they thought that children ofthe same sex would like the toys more (M =3.26) than children of the other sex (M =2.21).

Because two of the toys were more fa-miliar than we expected, we conducted thesame analyses again but removed the scoresfrom the two most familiar toys (the metalsculpture and prism glass toys). The patterns

remained the same: the absolute differencebetween children's own liking minus pre-dictions for same-sex peers (M = 4.95) wassmaller than the absolute difference be-tween children's own liking minus their pre-dictions for other-sex peers (M = 8.23, p <.001).

Individual toy analyses.—The absolutedifference method also was used to analyzechildren's responses to each individual toy.As shown in Table 2, for four of the seventoys, the absolute differences between chil-dren's own liking of the toy and their predic-tions for same-sex peers were significantlysmaller than the absolute differences be-tween children's own liking of the toy andtheir predictions for other-sex peers. For theremaining toys, the differences were in theexpected direction. Ghildren's familiaritywith the toys did not appear to influencethe likelihood of using the gender-centricpattern.

DiscussionThe flndings suggest that children pre-

dicted others' liking of toys using gender-centric patterns. For these novel toys,children's judgments were influenced bygender. Ghildren matched their predictionsabout others' liking of toys to the sex of theperson and their own liking of the toy. Forexample, if a girl liked a toy, she assumedthat others of her own gender group (i.e.,girls) would like the toy, and that others notin her gender group (i.e., boys) would notlike the toy as much. The pattern \vas con-firmed on individual toys, although the dif-ferences were not always significant, proba-bly because of the small size of the sample.Even for these young children, gender ap-pears to be salient, and seems to provide im-portant information for making judgmentsunder conditions of uncertainty.


TABLE 2

CHILDREN'S ABSOLUTE DIFFERENCE SCORES FOR INDIVIDUALNOVEL TOYS: STUDY 1

Own/Same- Own/Other-Toy Sex" Sex'' p Value

Magnet stand 1.05 1.41 .30Spinning bells 1.18 1.32 .70Wonder wheels 1.14 1.82 .02Pin pressions .59 1.68 .001Looking glass 1.41 1.86 .31Magic hill .73 1.82 .004Animal flip book .86 1.86 .005

" Absolute difference score based on children's own liking of toy mi-nns their predictions for same-sex peers.

^ Absolute difference score based on children's own liking of toy mi-nus their predictions for other-sex peers.

Study 2

The goal of Study 2 was to extend thefindings from the first study, using a largersample and more toys. The toys in Study 1were all interesting and attractive. In Study2, we wanted to investigate whether chil-dren also would use gender-based matchingto make predictions for toys that were lessattractive than the toys usecl in Study 1. Also,because the sample size was increased,more sophisticated methods could be usedto assess the patterns of children's re-sponses.

MethodSubjects.—The subjects were 71 4- and

5-year-old children (38 males, 33 females;50-69 months, M = 58 months). Most of thechildren were Gaucasian (78%) and frommixed-sex preschools in middle-class neigh-borhoods.

Materials.—Two adult judges selected12 toys. Five "toys" were selected on thebasis of being unfamiliar and highly attrac-tive, according to the results of Study 1(spinning bells, magnet stand, looking glass,animal flip book, wonder wheels). Three ob-jects were selected from the set of objectspretested by Bradbard and Endsley (1983)and used in an earlier study (Bradbard et al.,1986). These objects included a metal garlicpress (dabble press), a metal shoe shaper(silver shape), and a pizza cutter (pizza cut-ter). Although these objects are familiar toadults, they have been found to be unfamil-iar, moderately attractive, and not sex-typedby young children. Judges also selected twosimilar objects, a nutcracker and a mellonbailer (wonder scoop), believed to be unfa-miliar to children. Finally, the judges se-

lected two toys on the basis of their be-ing sex-typed, one sex-typed for females (aBarbie doll) and one sex-typed for males (a"gross face" ball).

Procedure.—Each child was tested indi-vidually by a female experimenter. The pro-cedure was identical to the one used inStudy 1.

ResultsPreliminary analyses were done to en-

sure that the novel toys (the Barbie doll andgross face ball were not included in anyanalyses) were unfamiliar, not sex-typed,and that they varied in attractiveness. Onlytwo toys were found to be relatively familiarto the children, the nutcracker and the pizzacutter. Boys and girls said they liked thenovel toys about equally, suggesting thatthey are not sex-typed (see Table 3). Ghil-dren's own liking of toys was used to verifyjudges' expectations about attractiveness.The five toys selected by judges as beinghighly attractive (M = 3.56) were rated bythe children as being significantly more"liked" than the less attractive toys (M =2.41), t{68) = 9.99, p < .001.

Analyses of all toys combined.—Themajor analysis was a repeated-measuresANOVA with two between-subjects factors(sex, age) and two within-subject factors, at-tractiveness (high, moderate) and target(own liking compared with predictions forsame-sex peers, own liking compared withpredictions for other-sex peers). The depen-dent measure was the summed absolute dif-ference score. Specifically, the own liking/same-sex peers score was calculated as theabsolute difference between children's ownliking for each toy minus their predictions


TABLE 3

GIRLS' AND BOYS' RATINGS OF LIKING AND PERCENTAGE OFCHILDREN WHO WERE UNFAMILIAR WITH THE TOYS: STUDY 2

Girls Boys % Unfamiliar

Highly attractive toys:Spinning bells 3.81Magnet stand 3.27Looking glass 3.45Animal flip book 3.55Wonder wheels 3.59

Moderately attractive toys:Dabble press 2.13Silver shape 1.94Wonder scoop 2.06Nut cracker 2.42Pizza cutter 2.61

3.583.583.613.763.45

2.532.182.472.892.79

8675738278

7793815934

of how much same-sex peers would like thetoys, summed over toys. Similarly, the ownliking/other-sex peers score was calculatedas the absolute difference between chil-dren's own liking for each toy minus theirpredictions of how much other-sex peerswould like the toys, summed over toys. Ifchildren responded using gender-based rea-soning, then the target effect should be sig-nificant. The target effect was significant,F(l, 65) = 67.51, p < .001, indicating thatthe absolute difference between own likingand predictions for same-sex peers (M =5.47) was smaller than the absolute differ-ence between own liking and predictions forother-sex peers (M = 13.73). An age effectwas also found to be significant, F(l, 65) =4.71, p < . 04, in which younger children (M= 5.42) had larger absolute difference scoresthan older children (M = 4.36). An examina-tion of the mean scores across all toys con-firmed the gender-centric pattern in thatchildren's own liking of toys (M = 2.87) wasmore similar to predictions for same-sexchildren (M = 2.94) than to their predictionsfor other-sex children (M = 2.35).

Furthermore, there was a trend for theage X attractiveness x target effect, F(l, 65)= 3.64 , p < .07. Simple effects analyseswere done for each age group separately.Importantly, both younger and older chil-dren showed the expected target effect {ps< .001). Although younger and older chil-dren sho\ved somewhat different patterns,the attractiveness x target interaction ^vasnot significant for either group (for youngerp = .15; for older p = .36), and so no furthersimple effects analyses were conducted.

Because two of the toys were more fa-miliar than the rest, t tests were conducted

on the major comparison of interest (target)with and without inclusion of these two toysto indicate whether familiarity with the toysinfluenced children's responses. In bothcases, when the absolute difference scoreswere compared across toys, the absolute dif-ference between own liking and predictionsfor same-sex peers was significantly smallerthan the absolute difference between ownliking and predictions for other-sex peers(both ps < .001).

Individual toy analyses.—Becausethere were differences in the familiarity ofthe supposedly novel toys, analyses alsowere conducted separately for each toy.First, a simple comparison was made of chil-dren's own liking with their predictions forthe other two groups of children (same-sexpeers and other-sex peers). For each toy, wecalculated the percentage of children whogave themselves and others the same scoresversus lower and higher scores. As can beseen in Table 4, about 60% of the childrenpredicted that same-sex others would likethe toys equally as much as they liked thetoys, regardless of toy attractiveness. In con-trast, only 28% (for attractive toys) and 39%(for less attractive toys) of the children pre-dicted that other-sex peers would like thetoys equally as much as they liked the toys.

Second, for each toy, the absolute differ-ence scores were calculated as was done be-fore. As shown in Table 5, for each toy, theabsolute difference between children's ownliking of the toy and their predictions forsame-sex peers was significantly smallerthan the absolute difference between chil-dren's ow n liking of the toy and their predic-tions for other-sex peers, ps < .005.


TABLE 4

COMPARISON OF THE PERCENTAGE OF CHILDREN WHOSE LIKING OF TOYS WAS THE SAME, HIGHER,AND LOWER THAN THEIR PREDICTIONS FOR PEERS: STUDY 2

SAME-SEX PEERS OTHER-SEX PEERS

COMPARED WITH: Lower%

Equal Higher%

Lower Equal Higher

Highly attractive toys:Spinning bells 21 73 6 63 31 6Magnet stand 31 55 14 59 23 18Looking glass 25 63 11 52 35 13Animal flip book 29 65 7 72 24 4Wonder wheels 37 57 6 63 27 10Mean % 29 63 9 62 28 10

Moderately attractive toys:Dabble press 17 56 27 34 36 30Silver shape 13 59 29 31 29 40Wonder scoop 13 64 23 39 30 31Nutcracker 17 63 20 41 27 32Pizza cutter 14 65 21 42 32 24Mean % 15 61 24 37 39 31

For the purpose of comparison, in grouppresentations, 137 undergraduate students(119 women, 18 men) were asked to predicton a 7-point scale how much they and otheradult men and women would like one of theobjects (the magnetic sculpture) labeled an"adult toy." When the absolute differencescore was calculated, the pattern was verysimilar to children's responses. Specifically,the difference between the students' ownliking ratings minus ratings for same-sexpeers was significantly smaller (M = 1.10)

than the absolute difference between theirown liking ratings minus ratings for other-sex peers (M = 1.86), p < .001. When ana-lyzed separately, both women and menshowed this pattern.

Correlational analyses.—Two kinds ofcorrelational analyses were computed. Thefirst correlational analyses involved assess-ing the relation between children's own lik-ing of the toys and their predictions forsame-sex and other-sex peers, with scores

TABLE 5MEAN ABSOLUTE DIFFERENCES SCORES FOR INDIVIDUAL TOYS AS A FUNCTION

OF TOY ATTRACTIVENESS AND TARGET: STUDY 2

Own/Same-Sex" Own/Other-Sex' '

Highly attractive toys:Spinning bellsMagnet standLooking glassAnimal flip bookWonder wheels

Moderately attractive toys:Dabble pressSilver shapeWonder scoopNut crackerPizza cutter

.40

.65

.56

.48

.61

.66

.63

.53

.55

.48

1.34"1.45"1.25'=1.49'=1.49"=

1.21''1.34'=1.40'1.44'=1.3P

^ Absolute difference score based on children's own liking of toy minus their predictionsfor same-sex peers.

^ Absolute difference score based on children's own liking of toy minus their predictionsfor other-sex peers.

" Difference between scores significant p < .001.• Difference between scores significant p < .005.

summed across all toys. Three summedscores were calculated: children's own lik-ing, children's predictions for same-sexpeers, and children's predictions for other-sex peers. The correlations among summedscores showed that children's own liking rat-ings were highly positively correlated withtheir predictions for same-sex peers, r(68) =.73, p < .001, whereas their own liking rat-ings were unrelated to their predictions forother-sex peers, r(68) = —.07. Children'spredictions for same- and other-sex peerswere in the negative direction, although notsignificant, r(68) = - .19, p = .13.

The second correlational analysis in-volved assessing the relation among the pro-files for the toy preference scores for eachchild individually. Specifically, two p corre-lations were computed. The first repre-sented the extent to which each child's rat-ings of how much he or she liked each toywas correlated with the child's predictionsof how much girls would like each toy. Thesecond represented the extent to which thechild's ratings of how much he or she likedthe toy was correlated with the child's pre-dictions of how much boys would like eachtoy. Thus, for each child, two p values werecomputed, each across the 10 novel toys.Some children had to be dropped from theanalyses because of a lack of variability intheir scores. For example, several childrengave exactly the same ratings for themselves(e.g., 4) and for other-sex peers (e.g., 1)across all the toys and so no correlationcould be computed. To obtain average p val-ues, each child's score was converted to a zvalue (as in an r to z transformation), the zscores were averaged, and then convertedback to p values. The results are similar tothose found using the r correlation method.Specifically, for girls, their own liking rat-ings across toys correlated more highly withthe predictions they made of how much girlswould like the toys, p(24) = .77, than withtheir predictions of how much boys wouldlike the toys, p(25) = .18. Similarly, for boys,their own liking ratings across toys corre-lated more highly with the predictions theymade of how much boys would like the toys,p(25) = .87, than with their predictions ofhow much girls would like the toys, p(26)= .12.

DiscussionThe findings from Study 2 extend the

preliminary results reported in Study 1.Generally, children were found to use thegender-centric pattern. Children's own lik-ing of toys was more similar to their predic-


tions of how much same-sex peers wouldlike the toys than to their predictions ofhow much other-sex peers would like thetoys. By including attractive and less attrac-tive toys, it was possible to assess whethergender-centric patterns occurred with bothtypes of toys. Even when toys were not ex-tremely attractive or interesting, childrenshowed a gender-based pattem in whichthey inferred that their own gender groupwill like the same toys as they do, and thatthe other gender group will like them less.Adults also showed a gender-centric pattem,although the study is only a preliminary ex-amination of this effect. The correlationalanalyses showed that children make gender-based inferences more strongly for their ownsex than for the other. Specifically, childrenexpect others in the same gender group toprefer the same toys that they do more thanthey expect children in the other gendergroup to reject these toys. Finally, the corre-lational analyses showed the same gender-based pattems when profiles of liking werecompared across toys. That is, a child's ownliking of toys highly correlated with his orher predictions about same-sex peers, acrossall the toys. In contrast, a child's own likingof toys was not highly correlated with his orher predictions about other-sex peers acrossthe toys.

Study 3

Studies 1 and 2 illustrated how childrendraw gender-based inferences to make deci-sions about toys not known to be sex-typed.The goal ofthe third study was to investigatechildren's judgments of toy preferences forthemselves and for others when toys weregiven explicit gender labels.

This study differs from earlier studies ofgender labeling in three ways. First, thestudy involved judgments of toy preferencerather than exploration or memory. Althoughthe study does not assess actual toy play pat-terns, children's reports of toy preferencesshould provide insight into their toy behav-ior. Second, we minimized the demands in-herent in the labeling situations so that chil-dren could easily disregard the labels if theyso desired. This was done by setting up apretense in which the experimenter who la-beled the toys said she had to leave and an-other experimenter came in and asked chil-dren their toy preferences. Third, the toysvfe used in this study varied in attrac-tiveness, unlike the toys used in many oftheearlier gender labeling studies. It is reason-able to assume that the efFect of gender


labels may vary depending on toy attrac-tiveness. For instance, the infiuence of gen-der labels potentially could be counteractedby a toy's inherent attractiveness. If a girlfinds a toy extremely attractive, she may dis-regard information about who should playwith it (according to cultural stereotypes),especially if it contradicts her desire to playwith the toy. Alternatively, gender labelsmay be so powerful that they alter the per-ception of attractiveness. The case ofthe boydropping the racing car like a hot potato sug-gests that a very attractive toy may suddenlybecome extremely unattractive once its sex-typing becomes obvious.

The relation between attractiveness andsex-typing has been investigated in severalstudies. In a study to assess the avoidance ofcross-sex-typed toys, 3-8-year-old childrenwere given a choice between attractive toysconsidered to be neutral and unattractivetoys (e.g., broken toys) considered appro-priate for their own sex (Hartup et al., 1963).Civen this choice, children tended to selecton the basis of sex-typing rather than on thebasis of attractiveness, indicating that chil-dren actively avoid cross-sex-typed toys.

Recent evidence suggests that somechildren may be more infiuenced by sex-typed norms than others. Frey and Ruble(1992) assessed the relation between attrac-tiveness and sex-typing of toys when theseaspects were in confiict and not in confiictin children from 5 to 10 years old. In theconfiict situation, they were shown a video-tape of children of their same sex playingwith an unattractive neutral toy and childrenof the other sex playing with a highly attrac-tive neutral toy. In the nonconfiict situation,they were shown a videotape of same-sexchildren playing with an attractive toy andother-sex children playing with a similar at-tractive toy. Afterward, children's time play-ing with the toys and with other non-sex-typed toys was assessed. As expected, theconflict situation led to different toy prefer-ences than the nonconfiict situation. Spe-cifically, boys who understood gender con-stancy, but not girls, played with thesex-typed but unattractive toys more thanwith the attractive toys. Frey and Ruble(1992) suggested that gender constancy maypush children into having a stronger motiva-tion to follow sex-typed norms, even when itrequires restricting their behavior and playwith attractive toys.

In Study 3, we assessed the relation ofsex-typing and toy attractiveness, but weused a preference paradigm and extended it

to include attributions of others' liking ofthetoys. Children were shown novel toys, vary-ing in attractiveness, each of which was ei-ther labeled as being for the same sex, forthe other sex, or not labeled. They wereasked how much they liked each toy andhow much they thought other boys and othergirls would like the toy. We assessed chil-dren's memory for the gender labels be-cause, for cognitive theories of gender de-velopment, labels are assumed to exert theirinfiuence only if they are remembered. Spe-cifically, if children remembered labels forthe toys, we would expect their preferencesand behavior to be guided by these labels.If they do not remember the labels, no in-fluence should be found.

Children's liking of the novel toys andtheir attributions about others' liking of thetoys were expected to be infiuenced by theexplicit gender labels. Consistent with cog-nitive gender theories (e.g., Kohlberg, 1966;Martin & Haiverson, 1981), and with previ-ous findings using toys (Hcirtup et al., 1963)and novel objects (e.g., Bradbard & Endsley,1983), gender labels were expected to in-crease children's interest in toys labeled fortheir own sex over toys labeled for the othersex. Furthermore, we expected that chil-dren would attribute liking to others follow-ing the same pattern. Specifically, childrenshould predict that others will like toys la-beled for their own sex more than toys la-beled for the other sex.

Level of attractiveness of toys also wasexpected to infiuence liking of toys and attri-butions of liking. Specifically, when toyswere given same-sex labels, we expectedchildren to like attractive toys more than lessattractive toys, and to report that other chil-dren would like attractive toys more thanless attractive ones. When toys were givenother-sex labels, attractiveness was not ex-pected to influence liking. That is, childrenwere expected to show a "hot potato" effectby disliking all toys labeled for the other sex,and by expecting others to dislike toys la-beled as being for the other sex, regardlessof the toys' original attractiveness. Due tothe importance of avoiding things associatedwith the other sex (Frey & Ruble, 1992;Hartup et al., 1963), it is hypothesized thatno matter how attractive or interesting thetoys are initially, if they are given other-sexlabels they will be liked less.

MethodSubjects.—Ninety-one preschool chil-

dren (47 girls, 44 boys) ranging in age from48 to 70 months of age (M = 58 months)


participated in the study. Data from threegirls were dropped because of failures to un-derstand the rating scale, and data fromthree boys were dropped because of shortattention span and experimenter error. Themajority of the children were Caucasian(80%) and were enrolled in mixed-sex pre-schools in middle-class neighborhoods.

Materials.—The toys used in the studywere selected for their tiovel quality, lack ofsex-typing, and level of attractiveness basedon the results of the first two studies: threewere selected as being attractive (high lik-ing, M = 3.5—3.8) and three were selectedas being less attractive (moderate liking, M= 2.4-2.9). The attractive toys (referred toby the names they were given when pre-sented to the children) were the spinningbells, the magnet stand, and the lookingglass. The less attractive toys were the dab-ble press (garlic press), silver shape (shoeshaper), and wonder scoop (melon bailer).Three boxes covered in white paper wereused to hold the toys, each approximately 14X 9 X 8 inches. One box had pictures ofgirls glued on the front with the word "girls"written on it, the second had pictures of boysand the word "boys" written on it, and thethird had no pictures or Writing. A stopwatchwas used to time presentation of toys. Tomeasure children's liking of the toys, thefour-point cup rating scale was used (seeStudy 1).

Procedure.—To ensure that toys wereincluded in the label and the nonlabel con-ditions an equal number of times, six ver-sions of label-toy combinations were devel-oped. The versions were constructed byhaving each attractive toy randomly pairedwith a less attractive toy twice, forming twosets of three pairs each. Children were ran-domly assigned to one of the two sets.Within each set, the labels for each pair werevaried in three patterns so that each toy pairreceived each label (or no label) once. Foreach set, one-third of the children were pre-sented with each pattern of labels. Thus,each child was exposed to two toys in thenonlabel condition, two toys labeled as be-ing "for girls," and two toys labeled as being"for boys."

Two experimenters were used in thestudy. One experimenter told children thesex-typed labels and the other asked themtheir preferences. This was done to ensurethat children would not feel under pressureto conform to the sex-typed labels theywere given.

Experimenter 1 arranged toys in theirappropriate boxes under a table and Experi-menter 2 brought the child into the testingroom. Experimenter 2 then left the room say-ing she had forgotten something. Experi-menter 1 began the toy labeling task by tell-ing children that they would be shown sometoys. The children were told that later theywould be asked how much they liked eachof the toys. The three boxes were placed onthe table. Each toy was lifted out of its box,one at a time (random order). The experi-menter labeled it by describing who likes toplay with it most (boys, girls, or no label).If the toy was labeled, the experimenterpointed to the pictures and the words (boysor girls) on the box. She gave the toy's name(described in Materials section), had thechild repeat the name, and demonstrated itsuse. The child then had 30 sec to inspect thetoy. Afterward, the experimenter repeatedthe toy's name and label, and placed the toyin a different box (the general toy box underthe table). When the toy was unlabeled, theexperimenter only told the child the nameof the toy and did not mention a label.

To ensure that children learned the gen-der labels, their immediate memory was as-sessed by Experimenter 1 asking them to tryto remember which sex she said liked eachof the toys, or if she did not say if either sexliked the toy. If the child was incorrect, theexperimenter provided the correct informa-tion. Experimenter 2 came into the room andsaid she would finish playing with the child,and Experimenter 1 left the room. Experi-menter 2 taught the child how to use therating scale as was done in Studies 1 and 2.

Then, Experimenter 2 randomly chosea toy from the general toy box and placed iton the table. For each toy, she asked chil-dren to indicate how much they liked thetoy, how much they thought girls would likethe toy, and how much they thought boyswould like the toy (order varied). This proce-dure was used for all the toys. Experimenter2 was not aware of the labels Experimenter1 had given the toys.

Finally, children's memory for the gen-der labels was assessed. The experimenterplaced all three boxes on the table. Experi-menter 2 then randomly selected a toy andasked the children to try to remember whatExperimenter 1 had said about who likedeach of the toys; boys, girls, or she did notsay; children were then asked to place thetoy in the appropriate box. This procedurewas repeated for each of the toys.


ResultsMemory for labels.—Children remem-

bered many of the gender labels for the toys(or that they were unlabeled) (60%), al-though this percentage is lower than in oneof the previous studies (e.g., Bradbard & En-dsley, 1983, found 80% accuracy; Bradbardet al., 1986, found 60% accuracy). A few chil-dren did not remember any labels correctly(1%) and a few remembered all the labelscorrectly (11%). Due to the variability inmemory scores, the children were dividedinto groups of rememberers and nonremem-berers (rememberers were those who cor-rectly remembered at least four of six labels).It was expected that the effects of the genderlabels would occur only for children who re-membered them.

Children's own toy preferences.—Chil-dren's ratings of how much they liked thetoys were analyzed using a repeated-measures ANOVA with two between-subjectfactors (sex of subject; level of recall of gen-der labels—high, low) and two within-subject factors (toy attractiveness—attractive, less attractive; gender label—forsame sex, for other sex, no label). Age wasnot included in any of the analyses becausethe age range was relatively narrow and be-cause recall level correlated positively withage.

Not surprisingly, toy attractiveness in-fiuenced children's liking of toys, F(l, 87) =35.02, p < .001. Children preferred attrac-tive toys (M = 3.55) more than the less at-tractive toys (M = 3.05). The gender labeleffect also was significant, F(2, 174) = 6.41,p < .002. Children preferred toys labeled asbeing for their own sex (M = 3.43) more thantoys labeled for the other sex (M = 3.04) orunlabeled toys (M = 3.26). An unexpectedsex of subject effect was found to be margin-ally significant, F(l, 87) = 3.76, p < .06, withgirls tending to like the toys more than boys.

We might have expected that the genderlabel effect would be modified by inter-actions with recall level, but none of thehigher-order interactions were significant.Nonetheless, because of the predictions, fur-ther analyses were done for the childrenwith high and low levels of recall. The labeleffect was not significant for children whoshowed poor recall of the toys' gender la-bels, but it was significant for those childrenwho remembered the labels, F(2, 88) = 6.29,p < .003. Furthermore, as shown in Figure2, for children who remembered the genderlabels, attractiveness of toys infiuenced rat-

ings for the same-sex-labeled toys (p < .008)and for the nonlabeled toys (p < .001). Forthe other-sex-labeled toys, attractivenesshad less of an infiuence on children's liking,consistent with the hot-potato idea thatother-sex-labeled toys are avoided even ifthey are attractive. The ratings for attractivetoys were somewhat higher than for the lessattractive ones, although the difference wasnot significant (p = .14).

Children's attributions of others' toypreferences.—Children's predictions abouthow much other girls and boys would likethe toys was analyzed using a repeated-measures ANOVA with two between-subjectfactors (sex of subject; level of recall of gen-der labels) and three within-subject factors(gender labels; sex of target; toy attrac-tiveness).

Children predicted that other childrenwould like attractive toys (M = 3.00) morethan less attractive ones (M = 2.38), F(l, 87)= 4.84, p < .03. Whether children used at-tractiveness in attributing toy preferences toothers was modified somewhat by children'srecall levels, F(l, 87) = 3.58, p < .07. Chil-dren who recalled gender labels tended tobelieve others would like the attractive toysmore than the less attractive ones (p < .008),whereas children who recalled fewer genderlabels did not show an infiuence of toy at-tractiveness. Furthermore, this interactionwas subsumed by the significant interactionof recall level x toy attractiveness x sex ofsubject, F(l, 87) = 4.97, p < .03. For boys,attractive toys were given higher ratingsthan the unattractive ones, and recall leveldid not interact with attractiveness of toys,but for girls it did (p < .004). Cirls who didnot recall toy labels tended not to differenti-ate others' liking of toys based on attrac-tiveness, whereas girls who recalled moregender labels did (see Table 6).

Surprisingly, the sex of target effect wassignificant, F(l, 87) = 7.70, p < .007. Thiseffect was subsumed by the significant sexof subject X sex of target interaction, F(l,87) = 31.72, p < .001. Boys thought otherboys would like the toys more than othergirls (Ms = 3.10 vs. 2.86, p < .05), and girlsthought other girls would like the toys morethan other boys (Ms = 3.24 vs. 2.52, p <.001). Both the main effects of attractivenessand target and this two-way interaction weresubsumed by the significant sex of subjectX target X attractiveness interaction, F(l,87) = 4.43, p < .04. For girls, toy attrac-tiveness did not interact with sex of target.


Liking Ratings

Attractive Toys K ^ Less Attractive Toys

For Same Sex No LabelGender Labels

For Other Sex

FIG. 2.—Mean liking ratings given by children who remembered gender labels as a function oftoy attractiveness and gender labels: Study 3.

but for boys, it did marginally (p < .07). Boysused toy attractiveness to determine howmuch other boys would like toys (p < .03)but not to determine how much other girlswould like the toys (see Table 6).

The main effect of gender label was sig-nificant, F(2, 174) = 6.72, p < .003, withchildren generally believing others wouldlike toys labeled for the same sex (i.e., theybelieve girls will most like girls' toys) morethan toys labeled for the other sex (Ms =3.04 vs. 2.77). This main effect was sub-sumed by the significant interaction of gen-der label X recall level of labels, F(2, 174)= 4.21, p < .02. Not surprisingly, childrenwho recalled more labels used them to maketheir judgments (p < .001), whereas chil-dren who recalled fewer labels did not.

To assess the predictions concerningthe hot potato effect, further analyses wereconducted for children who recalled the la-bels. For these children, attractiveness oftoys infiuenced ratings for the toys givensame-sex labels, t{45) = 2.41, p < .02, andthere was a trend for the neutral toys, t(45)= 1.96, p < .06. For the toys labeled for theother sex, attractiveness did not influencechildren's attributions of how much otherswould like the toys, consistent with the hotpotato idea that they believe children willavoid toys labeled for the other sex, even ifthey are attractive (see Fig. 3).

DiscussionWhen making decisions about unfamil-

iar toys for themselves and for others, mostchildren used the explicit information pro-vided by gender labels. Consistent withKohlberg's cognitive developmental theoryand with Martin and Halverson's (1981) gen-der schema theory, it was the children whoremembered the labels who used them. Fur-thermore, children demonstrated the "hotpotato" effect in that, regardless of toy attrac-tiveness, they tended to like less toys thatwere labeled as being for the other sex. Thehot potato effect was even more apparent inchildren's judgments about others. They ex-pected other children of both sexes to be lessinterested in toys that were labeled for theother sex, regardless of their attractiveness.In neither the case of their own liking of toysnor in the predictions for others, however,did labeling a toy for the other sex com-pletely eliminate interest in the toy. Chil-dren still said they would have some interestin playing with the toy and expected otherswould as well.

Although the methods were quite differ-ent, it is possible to compare the presentfindings concerning children's own liking ofthe toys with the results reported by Freyand Ruble (1992). The young children in thepresent study showed an influence of sex-typed labels in that these labels infiuenced


TABLE 6

CHILDREN'S PREDICTIONS OF OTHERS' LIKING OFFOR LABELS, TOY ATTRACTIVENESS, TARGET, .

TARGET

Attractive toys:

Girls

TotalUnattractive toys:

Boys

Girls

Total

Attractive toys:Boys

Girls

TotalUnattractive toys:

Boys

Girls

Total

For

Boy

. 3.63(.68)

. 3.07(1.11)3.35

. 2.95(1.05)

. 2.74(1.29)2.85

TOYS AS A\ND SEX OI

FUNCTION OF MEMORY' SUBJECT: STUDY 3

Children Who Remembered Sex-Typed LabelsTOY LABELS

Same Sex

Girl

3.05(1.18)3.44(.85)3.25

2.95(1.03)3.11(.97)3.03

Children Who

For

Boy

. 3.28(.79)

. 2.30(1.13)

. 2.79

. 3.00(1.16)

. 3.00(.97)

. 3.00

Same Sex

Girl

2.68(1.07)3.20(.83)2.94

3.04(1.06)3.20(.83)3.12

For Other Sex

Boy

2.58(1.26)2.48

(1.09)2.53

2.79(1.23)2.26(1.16)2.53

Girl

2.32(1.29)3.04(1.02)2.68

2.53(1.17)2.82(.96)2.68

Did Not RememberTOY LABELS

For Other Sex

Boy

3.08(1.08).2.30

(1.13)2.69

2.96(1.02)2.35(1.35)2.66

Girl

3.12, (1.24)

3.30(.80)3.21

2.80(1.19)3.40(.82)3.10

Nonlabeled

Boy

3.42(.84)2.48

(1.05)2.95

3.00(1.25)2.26(1.13)2.63

Girl

2.90(1.20)3.63(.69)3.27

3.05(1.18)3.07

(1.11)3.06

Sex-Typed Labels

Nonlabeled

Boy

3.32(.80)2.50

(1.10)2.91

3.16(.99)2.45(1.32)2.81

Girl

3.04(1.17)3.45(.89)3.25

2.72(1.31)3.30(.87)3.01

their ratings of how attractive or desirable atoy was perceived to be. Attractive toys losttheir appeal when given other-sex labels butnot when given same-sex labels. Similarly,Frey and Ruble found that gender-constantboys often avoided playing with an attractivetoy demonstrated by a girl and instead pre-ferred playing with a less attractive toy dem-onstrated by a boy. However, we did not findsex-typed labels dominating attractivenessas strongly as they found, possibly due to thelack of confiict between these two dimen-sions in our study. Children said they wouldlike about equally well unattractive toys thatwere sex-typed for the same sex and attrac-tive toys that were sex-typed for the othersex. That is, they did not show a strongerpreference for using sex-typing over attrac-

tiveness, a pattern that would be expectedfrom children who have obtained genderconstancy, according to Frey and Ruble(1992). Further research is needed to deter-mine how cognitive development, such asgender constancy, and how individual dif-ference factors influence children's toy pref-erences.

Unlike other labeling studies, we alsoexamined children's attributions about oth-ers' liking of toys, and some unexpected pat-terns emerged. Both boys and girls showeda tendency to attribute more liking of toysto own-sex versus other-sex children. In asense, this also may be considered a gender-centric pattern in that children attributed toyliking, in part, based on the sex of the peer


Liking Ratings

3.5

2.5

1.5

I Attractive Toys E^^ Less Attractive Toys

For Same Sex No Label

Gender LabelsFor Other Sex

FIG. 3.—Mean predictions of others' liking given by children who remembered gender labels asa function of toy attractiveness, gender labels, and sex of target: Study 3.

group. Surprisingly, girls and boys used at-tractiveness somewhat differently for de-termining others' toy preferences. Cirls con-sidered toy attractiveness when attributinghow much other girls and boys would liketoys, such that they believed other peerswould generally like attractive toys morethan unattractive ones. In contrast, boysused toy attractiveness in that same manneronly when attributing toy liking to otherboys.

General Diseussion

The present studies were designed toinvestigate children's use of gender-basedreasoning when they are asked to make judg-ments about their own and others' toy pref-erences. The findings suggest that childrenuse both gender labels and gender-basedmatching as ways of making decisions aboutthemselves and about others.

Gender-Based Reasoning When Toys AreSex-Typed

A number of studies have been doneto assess the impact of gender labels onchildren's memory, performance, and ex-ploration (e.g., Bradbard & Endsley, 1983;Montemayor, 1974). We were interested inassessing the effects of gender knowledge innew domains, specifically, children's judg-ments about toy preferences for themselvesand for others.

The typical method in some of the ear-lier studies has been to use toys (novel ob-jects) that are of equal but moderate attrac-tiveness. By varying the attractiveness ofunfamiliar toys as we did in Study 4, wewere able to assess more directly the "hotpotato" idea in which children's desires foreven very attractive toys are diminished byother-sex labels.

As predicted based on cognitive gendertheories, for children who remembered thetoy labels, their own liking of toys varieddepending on attractiveness of the toys andtheir labels. Children who remembered thelabels used them to guide their preferences:they said they preferred toys labeled fortheir own sex more than toys labeled for theother sex, especially if they were attractive.Toy attractiveness played more of a role ininfluencing children's judgments for own-sex-labeled toys than for other-sex-labeledtoys, consistent with the "hot potato" ideathat toys lose their attractiveness when la-beled for the other sex.

Children's attributions of how muchother girls and boys would like the same un-familiar toys followed a similar but not iden-tical pattern to their own liking of the toys.As before, both attractiveness of toys andtheir gender labels influenced ratings. Notsurprisingly, children generally believedthat others would like the attractive toys


more than the less attractive ones, and thatthey would like own-sex-labeled toys morethan other toys. Attributions of how muchothers would like the toys, however, alsovaried according to both attractiveness andtoy labels. Children who remembered thetoy labels said that others would like attrac-tive toys more than less attractive ones onlyif they had been given same-sex labels. Fortoys given other-sex labels, the attrac-tiveness of toys did not offset the labels,again providing support for the hot potatoidea.

Consistent with other studies using realtoys (e.g., Hartup et al., 1963), our findingssuggest that gender labels provide childrenwith salient cues that can be used to influ-ence their own desires as well as their ex-pectations about others. Cender categoriza-tion may be useful for helping childrenorganize their social worlds, but they alsomay limit their experiences with certainkinds of toys. In terms of being able to mini-mize the negative consequences of gender-stereotypic toy choices, however, we wouldhope that children would relax their stereo-typic ideas when faced with particularly at-tractive and interesting toys. Our findingssuggest that children relax their standardsonly slightly when making decisions abouttheir own toy preferences. In general, thehot potato effect is evident: children seemto lose interest, and expect others to do thesame, when confronted with even very inter-esting toys labeled as being for the other sex.

Gender-Based Reasoning When Toys AreUnfamiliar

In studies 1 and 2, we explored chil-dren's use of gender-based inferences whenthey had no prior knowledge of the sex-typing of toys. We also investigated thetypes of patterns children would use to makedecisions about attractive and less attractiveunfamiliar toys. The results of the studiesconverge to form a consistent picture of howchildren use gender when making decisionsabout unfamiliar toys for themselves and forothers.

Children tended to show a gender-centric pattern of responding when makingdecisions about both unfamiliar and familiartoys ("what I like, peers of my sex will like;what I like, peers of the other sex will notlike"). Although the evidence is only prelim-inary, undergraduates showed a similar gen-der-based pattern of inferences.

The gender-centric pattern is intriguingfor several reasons. First, it suggests that

gender is salient for young children, and foradults, even in situations where it need notbe. Second, gender-based inferences aboutunfamiliar toys may contribute to the initialformation of gender stereotypes. Childrenmay add information to their gender stereo-types based on what they like (and dislike)merely by generalizing their own prefer-ences to members of their own group. Thisprocess may contribute to the early develop-ment of gender stereotypes.

The most intriguing aspect of the gen-der-centric pattern is what it implies aboutindividuals' beliefs about gender categories.Children and adults seem to develop ab-stract theories about gender that go beyondthe explicit gender knowledge they mayhave been given. Their abstract gender theo-ries seem to be of two forms. One form is atheory of group differences: what one sexlikes, the other does not. The other form isa theory of within-gender-group similarity:what a person of one sex likes, other peopleof the same sex also will like.

Abstract gender theories may derivefrom individuals' tendencies to endow somesocial categories, just as they do with naturalkind categories, with a core essence that isexpected to characterize every member ofthe group (see Rothbart & Taylor, 1990). In-dividuals may be particularly likely to as-sume an underlying essence for certain so-cial categories, such as gender and race.These categories may be treated as centralcategories because they are assumed to besalient and unalterable, even though theseaspects are in fact culturally determined andmany aspects of them are alterable (seeRothbart & Taylor, 1990). Furthermore, thecentrality of gender categories may be re-inforced by the perception that these cate-gories have strong functional significance(Bem, 1981), that is, they have a high induc-tive potential (Rothbart & Taylor, 1990).These ideas—that gender and race catego-ries are unalterable and have high inductivepotential—likely relate to beliefs that differ-ences between these categories are due tobiological factors (Rothbart & Taylor, 1990).Both adults (Martin & Parker, 1994) and chil-dren (Taylor & Celman, 1991) show evi-dence of holding beliefs that differences be-tween males and females are based, at leastin part, on biological factors.

Once individuals attribute essences tothe social categories of gender, they may rea-son on that basis. As such, they would em-ploy what we've called the within-gender-

group theory, which allows them to assumesimilarities among members of the samegender group. The basis of the other theory,the gender-group-differences theory, seemsless clear. Individuals often assume differ-ences once group boundaries have beenmarked. Research on intergroup relationshas shown that perceivers attribute differ-ences to members of different groups evenwhen the categories are relatively arbitrary(e.g., Tajfel, 1981). Also, research has shownthat more differences are attributed whenthere is an assumption of a biological basisfor group differences (Hoffnian & Hurst,1990). However, the number of groups in-volved also may influence individuals' judg-ments. For gender groups, because there areonly two mutually exclusive groups (at leastin popular beliefs), assumptions of group dif-ferences may be extreme. The reasoningmay be that if females have a particular char-acteristic, then males are unlikely to havethat characteristic.

Thus, these two gender-based theoriesappear to form because of the basic tenden-cies perceivers have to classify and simplifyinformation in their social worlds, and be-cause of the powerful tendencies of individ-uals to endow social categories with mean-ingful essences (Rothbart & Taylor, 1990).

There may be developmental trends inthe use of gender-based theories. When theyare younger than 5, children may apply theirgender-based theories across many situa-tions because they fail to understand aboutindividual differences among members ofone gender group (or across gender groups).Research has shown that young children as-sume similarity within one gender groupeven when given information that chal-lenges this assumption. For instance, chil-dren believe that an unknown boy with fem-inine interests (e.g., likes playing withkitchen sets) will be as interested in trucksas a boy with masculine interests, but olderchildren are more likely to modify theirjudgments based on information about thespecific individuals (Berndt & Heller, 1986;Martin, 1989). Once a child knows anotherperson well, the tendency to rely on gender-based theories may be somewhat offset.

Furthermore, although more research isneeded to definitively map the progressionof these theories, it appears that the within-group theory is stronger, and may developearlier, than the between-group theory. Twodifferent lines of research suggest that thebetween-group theory becomes stronger


with age. In one, Martin (1989) found thatyoung children assume there are some dif-ferences in interests between girls and boys,but older children showed more highly dif-ferentiated beliefs about the sexes. In theother, Biemat (1991) found that children'sbeliefs about masculinity and femininity ini-tially appear to be viewed as two separatedimensions, but as children grow older, theybecome treated more as a unidimensionalconstruct, with masculinity and femininitybeing negatively correlated.

Also, situations may vary in eliciting ab-stract gender theories. When given little in-formation upon which to make a judgment(as was done in the present studies), bothchildren and adults may use their abstracttheories. When more information is avail-able, reliance on gender may be minimized(Martin, 1989; Serbin & Sprafkin, 1986).

There are several limitations associatedwith the studies we conducted. First, weconstructed the rating scales in a way thatlimited the range of responses childrencould give. The scales ranged from liking atoy very much to liking a toy only a little. Weassumed that children would like anythinglabeled as a toy. However, by limiting theirresponses, we may have been less likely todetect beliefs about group differences (e.g.,children would not have been able to indi-cate that a toy would be disliked by other-sex children). Second, children may havebeen hesitant to predict how much otherswould like toys in the kinds of situations weused (in which they had extremely limitedinformation about the other children). Dur-ing testing, however, we did not notice chil-dren being indecisive about their predic-tions for others. Third, it was impossible toassess whether children from different ra-cial groups or from different socioeconomicbackgrounds Would respond differendy thanthe children we tested. Fourth, the situa-tions used in these studies emphasized gen-der categories to the exclusion of other cate-gories (except in Studies 2 and 3 in whichattractiveness also varied). In real world set-tings, children also may employ other kindsof information, such as age, as a basis formaking judgments. Finally, it is important toremember that the present studies assessedonly one aspect relevant to children's toychoices, that is, their stated preferences forthemselves. The relation between children'sstated preferences and their actual behaviorhas not been determined, nor would we ex-pect that they would be in perfect accord.Children's preferences are likely to be more


stereotypic than their actual behavior. None-theless, it is not unreasonable to expectthat children's stated preferences relate totheir actual toy choices. Further research isneeded to investigate how gender knowl-edge translates into behavioral choices.

In conclusion, children's toy choices areinfluenced by gender-based reasoning inseveral ways. Children's abstract theoriesabout gender groups may allow them tomake assumptions about who would like toplay with novel, non-sex-typed toys. Theyassume that if they like a toy, children oftheir same sex will also like the toy morethan children of the other sex. When avail-able, explicit gender labels also guide chil-dren's preferences and their expectationsabout others' preferences by making evenvery attractive toys lose some of their appealwhen labeled as being for the other sex.When children use gender-based reasoningin these ways, they may unknowingly limitthe range of their own and others' play expe-riences.

RefereneesBem, S. L. (1975). Sex role adaptability: One con-

sequence of psychological androgyny. Jour-nal of Personality and Social Psychology, 31,634-643.

Bem, S. L. (1981). Gender schema theory: A cogni-tive account of sex typing. Psychological Re-view, 88, 354-364.

Berndt, T., & Heller, K. A. (1986). Gender stereo-types and social inferences: A developmentalstudy. Journal of Personality and Social Psy-chology, 50, 889-898.

Biernat, M. (1991). Gender stereotypes and therelationship between masculinity and femi-ninity: A developmental analysis. Journal ofPersonality and Social Psychology, 61, 351-365.

Bradbard, M. R. (1985). Sex differences in adults'gifts and children's toy requests at Christmas.Psychological Reports, 56, 969-970.

Bradbard, M. R., & Endsley, R. C. (1983). Theeffects of sex-typed labeling on preschoolchildren's information-seeking and retention.Sex Roles, 9, 247-260.

Bradbard, M. R., Martin, C. L., Endsley, R. C.,& Halverson, C. F. (1986). Influence of sexstereotypes on children's exploration andmemory: A competence versus performancedistinction. Developmental Psychology, 22,481-486.

Bradbard, M. R., & Parkman, S. A. (1983). Genderdifferences in preschool children's toy re-quests. Journal of Genetic Psychology, 145,283-285.

Gonnor, J. M., & Serbin, L. A. (1977). Behaviorallybased masculine- and feminine activity-preference scales for preschoolers: Correlateswith other classroom behaviors and cognitivetests. Child Development, 48, 1411-1416.

Eisenberg, N. H., Murray, E., & Hite, T. (1982).Children's reasoning regarding sex-typed toychoices. Child Development, 53, 81—86.

Fagot, B. 1. (1977). Consequences of moderatecross-gender behavior in preschool children.Child Development, 48, 902-907.

Fein, B. L, Johnson, D., Kosson, N., Stork, L., &Wasserman, L. (1975). Sex stereotypes andpreferences in the toy choices of 20-month-old boys and girls. Developmental Psychol-ogy, 11, 527-528.

Frey, K. S., & Ruble, D. N. (1992). Gender con-stancy and the "cost" of sex-typed behavior: Atest of the conflict hypothesis. DevelopmentalPsychology, 28, 714-721.

Gelman, S. A. (1989). Children's use of categoriesto guide biological inferences. Human Devel-opment, 32, 65-71.

Gelman, S. A., CoUman, P., & Maccoby, E. E.(1986). Inferring properties from categoriesversus inferring categories from properties.Child Development, 57, 396-404.

Gelman, S. A., & Markman, E. M. (1986). Catego-ries and induction in young children. Cogni-tion, 23, 183-209.

Celman, S. A., & Markman, E. M. (1987). Youngchildren's inductions from natural kinds: Therole of categories and appearances. Child De-velopment, 58, 1532-1541.

Goldman, J. A., Smith, J., & Keller, E. D. (1982).Sex-role preference in young children: Whatare we measuring? Journal of Genetic Psy-chology, 141, 83-92.

Hartup, W. W., Moore, S. G., & Sager, G. (1963).Avoidance of inappropriate sex-typing byyoung children. Journal of Consulting Psy-chology, 27, 467-473.

Hoffman, C , & Hurst, N. (1990). Gender stereo-typing: Perception or rationalization?/ournaJof Personality and Social Psychology, 58,197-208.

Kohlberg, L. A. (1966). A cognitive-develop-mental analysis of children's sex-role con-cepts and attitudes. In E. E. Maccoby (Ed.),The development of sex differences (pp. 82-172). Stanford, CA: Stanford University Press.

Langlois, J. H., & Downs, A. G. (1980). Mothers,fathers, and peers as socialization agents ofsex-typed play behaviors in young children.Child Development, 51, 1237-1247.

Martin, C. L. (1989). Children's use of gender-related information In making social judg-ments. Developmental Psychology, 25, 80-88.

Martin, C. L., & Halverson, C. F. (1981). A sche-


matic processing model of sex typing andstereotyping in children. Child Development,52, 1119-1134.

Martin, C. L., & Parker, S. (1994). Folk theoriesabout sex and race differences. Personalityand Social Psychology Bulletin, 21, 45-57.

Medin, D. L. (1989). Concepts and conceptualstructure. American Psychologist, 44, 1469-1481.

Montemayor, R. (1974). Children's performancein a game and their attraction to it as a func-tion of sex-typed labels. Child Development,45, 152-156.

O'Brien, M., & Huston, A. C. (1983). Activity leveland sex-stereotyped toy choice in toddlerboys and girh. Journal of Genetic Psychology,146, 527-533.

Robinson, G. C , & Morris, J. T. (1986). The gen-der-stereotyped nature of Christmas toys re-ceived by 36-, 48-, and 60-month-old chil-dren: A comparison between nonrequestedand requested toys. Sex Roles, 15, 21-32.

Ross, D. M., & Ross, S. A. (1972). Resistance bypreschool boys to sex-inappropriate behavior.

Journal of Educational Psychology, 63,342-346.

Rothbart, M., & Taylor, M. (1990). Category labelsand social reality: Do we view social catego-ries as natural kinds? In G. Semin & K.Fiedler (Eds.), Language, interaction, and so-cial cognition (pp. 11-36). London: Sage.

Serbin, L. A., & Connor, J. M. (1979). Sex-typingof children's play preferences and patterns ofcognitive performance. Journal of GeneticPsychology, 134, 315-316.

Serbin, L. A., & Sprafkin, C. (1986). The salienceof gender and the process of sex typing inthree- to seven-year-old children. Child De-velopment, 57, 1188-1199.

Tajfel, H. (1981). Human groups and social cate-gories. Cambridge: Cambridge UniversityPress.

Taylor, M. C , & Gelman, S. A. (1991, April). Chil-dren's beliefs about sex differences: The roleof nature vs. nurture. Paper presented at themeetings of the Society for Research in GhildDevelopment, Seattle.

Documents

Children's Gender-Based Reasoning about Toys - People Server at