19
Blehavioral, Affective, and Social Correlates of Involvement in Cross-Sex Friendship in Elementary School Donna M. Kovacs, Jeffrey G. Parker, aitid Lois W. Hoffman University of Michigan KOVACS, DONNA M.; PARKER, IEFFBEY G.; and HOFFMAN, LOIS W. Behavioral, Affective, and Social Correlates of Involvement in Cross-Sex Friendship in Elementary School. CHILD DEVELOPMENT, 1996, 67, 2269-2286. Tbe purpose of tbis study was to compare cbildren witb and witbout cross-sex friends on measures of social and cognitive competence, endorsement of sex-role ste- reotypes, and family composition. Subjects were 723 tbird and fourtb graders (377 girls, 346 boys) from diverse socioeconomic backgrounds; 35% were African American. Measures included sociometric assessments of peer acceptance, friendsbip, and bebavioral reputation, as well as self-reports of perceived self-competence and endorsement of sex-role stereotypes. In addition, teacbers completed ratings of cbildren's social and cognitive competence. In all, 92 cbildren, about 14% of tbe sample, bad one or more reciprocal opposite-sex friends; for 21 of tbese cbildren, tbeir cross-sex friendsbips were tbeir primary or only friendsbips. African American cbil- dren were more likely tban European American cbildren to bave opposite-sex friends. Involve- ment in cross-sex friendsbips was unrelated to tbe gender make-up of tbe classroom, but was related to family structure. Comparisons of tbe cbildren wbo bad primarily or only cross-sex friends to matcbed groups of cbildren wbo bad only same-sex friends and to cbildren wbo bad cross-sex friends secondarily to same-sex ones revealed a number of differences between tbe groups in social competence and relationsbips witb peers. Overall, cbildren witb primarily oppo- site-sex friends bad poorer social skills tban otber cbildren witb friends, altbougb tbey were less stereotyped about sex roles tban otber cbildren, and were better adjusted tban cbildren witb no friends on most measures. In contrast, cbildren involved in opposite-sex friendsbip secondarily to same-sex friendsbip were as well adjusted socially as cbildren witb only same-sex friendsbips. Tbese results suggest tbat cbildren witb cross-sex friends differ among tbemselves, depending on tbe primacy of tbe cross-sex relationship. Children's social networks become in- peers as to opposite-sex peers in early ele- creasingly gender segregated with age, such mentary school (e.g., Maccoby & Jacklin, that by elementary school most children en- 1987). Likewise, studies of children's socio- gage in only minimal and highly ritualized metric choices and preferences (e.g.. Hay- interaction with peers of the opposite sex. den-Thomson, Rubin, & Hymel, 1987) indi- This increasingly strong tendency of chil- cate that children in elementary school dren to affiliate with other children of the typically give much higher ratings to class- same sex has been observed in a number of mates of their own gender than they give to diverse settings and cultures, and at many classmates of the opposite sex. Sex segrega- important descriptive levels (Golombok & tion is thought to be especially strong with Fivush, 1994; Leaper, 1994). For example, respect to children's friendships, with some wide disparities exist in the amount of time authors reporting rates of cross-sex friend- children spend in the company of same-ver- ship so low as to be virtually zero at ages sus opposite-sex peers, with some estimates above 7 or 8 years (e.g., Cottman, 1986). suggesting that children may devote as Cender segregation in children's peer rela- much as 11 times as much time to same-sex tionships in middle childhood is so salient Tbis researcb was supported by grant no. 91-1437-91 from tbe William T. Grant Foundation to tbe tbird autbor and by a NICHD training grant to tbe first autbor. Portions of tbis paper were presented at tbe 1995 biennial meeting of tbe Society for Researcb in Cbild Development in Indianapolis. Tbe autbors wisb to tbank Rebekab Coley, Allison Fuligni, lobn Seal, and Mari- anne Taylor for tbeir belp in developing measures and in collecting and coding data for tbis project. Correspondence conceming tbis article sbould be addressed to Donna Dumm Kovacs, Developmental Psycbology, University of Micbigan, 525 E. University, Ann Arbor, MI 48109-1109. [Child Development, 1996,67,2269-2286. © 1996 by the Society for Research in Child Development, Inc. Allrightsreserved. 0009-3920/96/6705-0023S01.00]

Behavioral, Affective, and Social Correlates of Involvement in Cross-Sex Friendship in Elementary School

Embed Size (px)

Citation preview

Blehavioral, Affective, and Social Correlatesof Involvement in Cross-Sex Friendshipin Elementary School

Donna M. Kovacs, Jeffrey G. Parker,aitid Lois W. HoffmanUniversity of Michigan

KOVACS, DONNA M.; PARKER, IEFFBEY G.; and HOFFMAN, LOIS W. Behavioral, Affective, and SocialCorrelates of Involvement in Cross-Sex Friendship in Elementary School. CHILD DEVELOPMENT,1996, 67, 2269-2286. Tbe purpose of tbis study was to compare cbildren witb and witboutcross-sex friends on measures of social and cognitive competence, endorsement of sex-role ste-reotypes, and family composition. Subjects were 723 tbird and fourtb graders (377 girls, 346boys) from diverse socioeconomic backgrounds; 35% were African American. Measures includedsociometric assessments of peer acceptance, friendsbip, and bebavioral reputation, as well asself-reports of perceived self-competence and endorsement of sex-role stereotypes. In addition,teacbers completed ratings of cbildren's social and cognitive competence. In all, 92 cbildren,about 14% of tbe sample, bad one or more reciprocal opposite-sex friends; for 21 of tbese cbildren,tbeir cross-sex friendsbips were tbeir primary or only friendsbips. African American cbil-dren were more likely tban European American cbildren to bave opposite-sex friends. Involve-ment in cross-sex friendsbips was unrelated to tbe gender make-up of tbe classroom, but wasrelated to family structure. Comparisons of tbe cbildren wbo bad primarily or only cross-sexfriends to matcbed groups of cbildren wbo bad only same-sex friends and to cbildren wbo badcross-sex friends secondarily to same-sex ones revealed a number of differences between tbegroups in social competence and relationsbips witb peers. Overall, cbildren witb primarily oppo-site-sex friends bad poorer social skills tban otber cbildren witb friends, altbougb tbey were lessstereotyped about sex roles tban otber cbildren, and were better adjusted tban cbildren witb nofriends on most measures. In contrast, cbildren involved in opposite-sex friendsbip secondarilyto same-sex friendsbip were as well adjusted socially as cbildren witb only same-sex friendsbips.Tbese results suggest tbat cbildren witb cross-sex friends differ among tbemselves, dependingon tbe primacy of tbe cross-sex relationship.

Children's social networks become in- peers as to opposite-sex peers in early ele-creasingly gender segregated with age, such mentary school (e.g., Maccoby & Jacklin,that by elementary school most children en- 1987). Likewise, studies of children's socio-gage in only minimal and highly ritualized metric choices and preferences (e.g.. Hay-interaction with peers of the opposite sex. den-Thomson, Rubin, & Hymel, 1987) indi-This increasingly strong tendency of chil- cate that children in elementary schooldren to affiliate with other children of the typically give much higher ratings to class-same sex has been observed in a number of mates of their own gender than they give todiverse settings and cultures, and at many classmates of the opposite sex. Sex segrega-important descriptive levels (Golombok & tion is thought to be especially strong withFivush, 1994; Leaper, 1994). For example, respect to children's friendships, with somewide disparities exist in the amount of time authors reporting rates of cross-sex friend-children spend in the company of same-ver- ship so low as to be virtually zero at agessus opposite-sex peers, with some estimates above 7 or 8 years (e.g., Cottman, 1986).suggesting that children may devote as Cender segregation in children's peer rela-much as 11 times as much time to same-sex tionships in middle childhood is so salient

Tbis researcb was supported by grant no. 91-1437-91 from tbe William T. Grant Foundationto tbe tbird autbor and by a NICHD training grant to tbe first autbor. Portions of tbis paper werepresented at tbe 1995 biennial meeting of tbe Society for Researcb in Cbild Development inIndianapolis. Tbe autbors wisb to tbank Rebekab Coley, Allison Fuligni, lobn Seal, and Mari-anne Taylor for tbeir belp in developing measures and in collecting and coding data for tbisproject. Correspondence conceming tbis article sbould be addressed to Donna Dumm Kovacs,Developmental Psycbology, University of Micbigan, 525 E. University, Ann Arbor, MI48109-1109.

[Child Development, 1996,67,2269-2286. © 1996 by the Society for Research in Child Development, Inc.All rights reserved. 0009-3920/96/6705-0023S01.00]

2270 Child Development

and so robust that Maccoby (1990) and oth-ers (e.g., Thome, 1986) have proclaimed thatboys and girls live in "separate worlds" dur-ing this period.

The origins and consequences of thegender segregation in children's groupshave been of long-standing interest to socialscientists. Many clinicians and develop-mental theorists, including Douvan and Ad-elson (1966), Havighurst (1953), Sullivan(1953), and Sroufe (Sroufe, Bennett, En-glund. Urban, & Shulman, 1993), have re-garded gender segregation of peer groups asan appropriate and understandable out-growth of normal development. However,some authors, including Archer (1992), haveworried that gender segregation of chil-dren's peer groups constrains sex-role devel-opment. Additionally, Maccoby (1990) spec-ulates that, because boys and girlsexperience such "separate worlds" in child-hood, they confront greater difficulties whenthey become involved in romantic relation-ships later in life (see also Leaper, 1994).

Despite the pervasiveness of gendersegregation in peer relationships, there isalso evidence that a small but stable subsetof children maintain cross-sex friendshipswell beyond the age at which such relation-ships are normative (Bukowski, Gauze,Hoza, & Newcomb, 1993). The existence ofthese cross-sex friendships raises interestingquestions for developmentalists: What arethe personal characteristics of children in-volved in cross-sex friendships? Do childrenwith cit)ss-sex friends also have same-sexfriends? ^Are they more or less sociallyskilled than Other children? Are they moretolerant of others and less sex-stereotyped?Involvement in cross-sex friendship at thisage may have implications for these chil-dren's cross-sex relationships as adolescentsand adults. Yet, to date, the nature of chil-dren's involvement in cross-sex friendships(as opposed to same-sex friendships) at dif-ferent ages has largely been unexamined.

The literature that does exist presents avery mixed portrait of children who engagein cross-sex friendships, and, in fact, two op-posing hypotheses have been formulated.The first hypothesis proposes that childrenwho have cross-sex friendships are better ad-justed socially than those who do not. Thisview suggests that children who have cross-sex friendships may be better liked by peersof both sexes and better able to communi-cate and cooperate with other children thanchildren without cross-sex friends. For these

children, participation in cross-sex friend-ships may be linked with high social skills,either as a cause or as a result of their inter-actions with opposite-sex peers.

The alternate view is that children withcross-sex friends may be less socially skilledin some respects than other children theirage. That is, some authors have reasonedthat because cross-sex friendship is a devel-opmental anomaly, the children who be-come involved in cross-sex friendships arethose who are otherwise lacking the socialskills required to form same-sex friendships.If involvement in cross-gender friendship isassociated with problems in same-sex peerrelationships, there are at least two possiblecausal pathways: children who are rejectedby same-sex peers may turn to cross-sexpeers for friendship; or, alternatively, chil-dren who choose to have cross-sex peers asfriends may subsequently be rejected bytheir classmates.

As noted, the literature to date providesa very mixed picture regarding these twocompeting hypotheses. Howes (1988) ob-served young children, ages 1 to 6, duringfree-play periods in a child-care center. Al-though the number of cross-sex friends de-creased with age, at all ages children whohad cross-sex friendships engaged in moresocial play and showed more positive affectthan children without cross-sex friends. Atthe preschool ages (ages 4 to 6), childrenwith cross-sex friends engaged in more co-operative social pretend play, which wasfound to be a general marker of competentpeer interaction. On the basis of these re-sults, Howes (1988) concluded that pre-school children with cross-sex friends maybe more socially skilled than other children.However, one limitation to Howes's study isthat friendship was defined based on obser-vations of children's play patterns, ratherthan on direct reports from children them-selves. Thus, it remains uncertain in thisstudy whether the children who interactedpreferentially with specific opposite-sexpeers in fact regarded these relationships asfriendships.

In addition, the subjects in Howes'sstudy were children of preschool age. Gross-gender friendship and interaction are lessanomalous in preschool than in later years,and other evidence suggests that children'sreactions to cross-gender behaviors by peersbecome increasingly negative with age (seeGolombok & Fivush, 1994). For example.Garter and McGloskey (1984) reported that.

Kovacs, Parker, and Hoffman 2271

relative to kindergartners, older elementaryschool children reacted more negatively to-ward displays of cross-gender activity andfriendship preferences in hypothetical vi-gnettes. Moreover, although these childrendid not see cross-gender behavior as wrongper se, they indicated that they would prefernot to associate with children who violatedsex-role norms. If children's attitudes towardcross-gender interaction and friendship be-come increasingly negative with age, itseems possible that the link between cross-sex friendship and social competence ob-served by Howes (1988) could disappear atlater ages, or even reverse direction.

Offering results that are consistent withsuch a developmental view, Ladd (1983) ob-served a sample of third- and fourth-gradechildren on the playground, and then exam-ined the relationship between children'speer interactions and their acceptance bypeers, assessed sociometrically. In compari-son with popular or average boys, rejectedboys were found to interact more with girlsand also preferred younger children as play-mates. Ladd speculates that rejected boysmay be drawn to interactions with cross-sexand younger peers because they are moreinfiuential with these groups than with boystheir age; another possibility is that the playpatterns and interaction styles of rejectedboys are more compatible with those of girls.

Likewise, Sroufe and colleagues (1993)observed the patterns of croSs-gender inter-action among 10- and 11-year-olds in a sum-mer camp setting. Of particular interest waschildren's "border work," or interactionsthat serve to reinforce, proclaim, and makesalient the distinction between same-sexand cross-sex peer relationships. Reliable in-dividual differences were observed in chil-dren's tendencies to maintain the boundarybetween the sexes by actively calling atten-tion to it, and in children's willingness toviolate the gender boundary by interactingwitli peers of the opposite sex. Importantly,and consistent with Ladd's (1983) findings,children who were low in maintaining gen-der boundaries and high on gender bound-ary violation were rated as less socially com-petent by camp counselors and were lesspopular with peers.

In Sroufe et al.'s (1993) study, as inLadd's (1983) earlier study, children werecharacterized on the basis of their observedbehaviors and interactions with same- andopposite-sex peers, not necessarily on thebasis of their professed friendships. Al-

though children who violate norms for gen-der boundaries behaviorally are obviouslyinvolved in more cross-sex interaction, thisfact does not necessarily imply that they suealso more likely than other children to haveopposite-sex friendships per se. More re-cently, Bukowski and colleagues (1993) ex-amined the relation between elementaryschool-aged children's social and behav-ioral characteristics and the pattern of theirsame- versus opposite-sex sociometricfriendship choices. For an individuaJ child,a bias toward same-sex peers was indicatedby the presence of a wide disparity betweenthe number of same-sex peers and the num-ber of opposite-sex peers he or she nomi-nated as friends. Bukowski et al. found that,over a period of more than 1 year, childrentended to be stable in the extent of theirpreference for same-sex peers over opposite-sex peers. Moreover, boys who preferred ac-tivities involving gross motor skills showeda stronger same-sex preference than otherboys, and girls who disliked gross motor ac-tivities were more likely than other girls toshow a same-sex preference. The extent ofchildren's same-sex preference was unre-lated to their same-sex acceptance and to re-jection by both same- and opposite-sexpeers. It should be noted, however, that thisstudy was not an exploration of children'sfriendships per se. Rather, the study reliedon unilateral nominations of friendship, andthus is more aptly characterized as a studyof liking and popularity. Further, children inthis study could achieve a high preferencescore either by having only a few same-sexfriends, or by combining a large number ofsame-sex friends with a few opposite-sexfriends; conversely, children could obtain alow score by having no friends at all, or byhaving a large number of friends of both gen-ders. Thus, while Bukowski et al. (1993) re-ported that those children who showed thegreatest preference for same-sex peers wereleast liked by peers of the opposite sex, theydid not actually compare children who hadonly same-sex friends with those childrenwho had friends of both genders.

Overall, then, the literature to date pre-sents conflicting empirical and theoreticalviews of the adaptiveness of involvement incross-sex friendship during the middlechildhood years. Whereas some authorshave tended to view such forms of involve-ment as healthy, other authors have ex-pressed concern, worrying that childrenwith cross-sex friends may be lacking theskills necessary to form same-sex friend-

2272 Child Development

ships. And, whereas some studies have sug-gested links between cross-sex friendshipinvolvement and difficulties in peer rela-tionships and social skills, other studies donot, or have reported findings in the oppo-site direction. In part, methodological andother differences may explain some of theseinconsistencies among studies. For example,because children's attitudes toward cross-sex friendship may grow increasingly nega-tive with age, the link between cross-sexfriendship involvement and children's ac-ceptance by same-sex peers may changewith age. Another difficulty is that much ofthe research to date has relied on genderpreferences that children demonstrate intheir social interaction or in unilateral socio-metric choices and liking ratings. Thesemeasures are important barometers of chil-dren's experiences and attitudes, but they donot unambiguously index friendship.

Finally, it appears that few authors haveconsidered that involvement in cross-sexfriendship is a heterogeneous phenomenon.In particular, there appears to have been lit-tle attention to the possibility that the posi-tion of cross-sex friendships within thelarger friendship network may contribute tovariability in the experiences of childrenwith cross-sex friends. Ghildren's friendshipnetworks ordinarily include both peripheraland central members. Ghildren are closer tocentral members of their friendship net-works than to peripheral members, and pastresearch has demonstrated differences in thefeatures, stability, and influence of friend-ships according to their centrality (Epstein,1983; Furman, 1996; Savin-Williams &Berndt, 1990). In the context of cross-sexfriendship involvement, it seems likely thatthe significance, social stimulus value, andinfiuence of a cross-sex friend may dependon the centrality or closeness of that friendwithin the larger peer network. Ghildrenwho are primarily involved with friends ofthe opposite sex may differ in some respectsfrom children whose cross-sex friendshipsare secondary to same-sex relationshipswithin the peer network.

Accordingly, a primary purpose of thepresent study was to extend previous re-search on children's cross-sex friendshipinvolvement in middle childhood by moreclearly describing the characteristics of chil-dren with cross-sex friends. Toward thisend, children involved in reciprocal cross-sex friendships are identified using socio-metric assessments, and are compared tochildren without cross-sex friends in terms

of gender, ethnicity, and two family compo-sition variables: parental marital status andthe presence or absence of opposite-sex sib-lings in the home. In a recent meta-analyticliterature review, Stevenson and Black(1988) reported that boys in father-absenthomes were less sex-role stereotyped thanboys in father-present homes. Marital statuswas therefore examined in the present studyin order to explore whether boys in father-absent homes might also show greaterinvolvement in cross-sex friendship. Like-wise, Toman (1976) theorized that individu-als with opposite-sex siblings would showgreater involvement in cross-sex friendshipthan other persons. Previous research (e.g.,Burker, Goldstein, & Gaputo, 1981) has sofar provided mixed support for Toman's the-ory; the present study thus offered an oppor-tunity to explore Toman's thesis further.

The link between cross-sex friendshipinvolvement and the gender balance withinclassrooms was also examined in the currentstudy. Glassroom contextual features havebeen shown to exert powerful infiuences onthe formation of friendships between chil-dren (see Epstein & Karweit, 1983; Halli-nan, 1979). For example, children in largerclassrooms have been found to make fewerfriendship choices in sociometric assess-ments, and greater heterogeneity in aca-demic skills has been found to infiuence theformation of friendships between children ofdifferent academic abilities (Epstein, 1983).Hallinan and Tuma (1978) suggest that anopportunity hypothesis may explain suchfindings: the greater the opportunity for chil-dren with diverse characteristics to interact,the greater their tendency to becomefriendly and to maintain close friendships.In the context of the present study, itseemed likely that children in classroomswith more opposite-sex peers might show agreater likelihood to form cross-sex friend-ships.

Finally, examining the significance ofheterogeneity in the nature of children'sinvolvement in cross-sex friendship was thesecond main goal of this study. In particular,a distinction is drawn between children whoare primarily or only involved in cross-sexfriendship and those children who are in-volved in cross-sex friendship secondarily totheir involvement in same-sex friendship.These two groups of children with cross-sexfriends are compared with each other, tochildren with only same-sex friends, and tochildren without friends entirely, in orderto determine whether or not they differ on a

Kovacs, Parker, and Hoffman 2273

number of measures of social and cognitivecompetence, self-esteem, and adherence tosex-role stereotyping. On the basis of pastresearch, it was predicted that this distinc-tion in the centrality of cross-sex friendshipinvolvement would capture an importantdifference among the children participatingin these relationships.

MethodSubjects

Seven-hundred and twenty-three chil-dren (377 girls, 346 boys) participated assubjects. The participating subjects repre-sented all but four children enrolled in 27third- and fourth-grade classrooms in eightelementary schools. The schools were lo-cated in a large Midwestern city, and thesocioeconomic status of the children rangedfrom lower to upper-middle class. Participat-ing children were 61.2% European Ameri-can, 34.6% African American, and 4.2%Asian, Arabic, and Hispanic. Due to theirrelatively small numbers, Asian, Arabic, andHispanic children were included with Euro-pean American children in analyses involv-ing ethnicity. Classrooms varied greatly intheir ethnic composition, from entirely Eu-ropean American to entirely African Ameri-can, and they ranged in size from 22 to 32children (M = 26.78). The number of girlsin each classroom ranged from 9 to 20 (M =13.96); the number of boys ranged from 8 to22 (M = 12.81). The average ratio of girls toboys in each classroom was approximately1:1, ranging from about 1:2 to about 2:1.

MeasuresLevel of acceptance.—A "roster-and-

rating" sociometric procedure (Singleton &Asher, 1977) was used to assess children'slevel of acceptance by their peers. Childrenwere given rosters of all classmates' namesand were asked to rate how much they likedeach child on a 5-point scale, from "Like aLot" to "Dislike a Lot." Two acceptancescores were calculated for each child: (1) theaverage rating received from peers of thesame sex, and (2) the average rating receivedfrom peers of the opposite sex. Higher scores

on these indices indicate greater acceptanceand popularity with peers.

Identification of friendships.—A two-step sociometric nomination procedure wasused to identify children's friendships. First,children were asked to indicate their most-liked classmate from a roster of all childrenin their class. Following this, children weregiven a second (duplicate) roster of theirclassmates and asked to circle the names oftheir "best friends in the class." An unlim-ited number of choices was permitted onthis second part.

To identify children's friends, the ma-trix of all choices made within the classroomwas examined for instances of mutual nomi-nations. Two levels of exclusivity/closenesswere considered. First, pairs of childrenwere considered to be best friends if eachincluded the other among his or her choicesof friends on the unlimited nomination mea-sure; in other words, children were consid-ered best friends if their choices were mutu-ally reciprocal at the unlimited level. Of the683^ children in the sample for whom rele-vant data were available, 19, or 2.8%, had noreciprocal friends. Each of the 664 re-maining children reported at least one recip-rocal best friendship, with a range of 1 to 13friends, and a mean of 4.01 (SD = 2.44).These prevalence rates are comparable tothose reported in other studies with similaroperationalizations (e.g., Berndt & Hoyle,1985; Epstein & Karweit, 1983).

Second, a child's very best friend wasidentified by examining whether the class-mate he or she designated as the person"liked most" included the target child, inturn, among his or her friends on the unlim-ited nomination measure. Thus, a child'svery best friend did not necessarily need toreciprocate at the very best level (but didneed to reciprocate at the "best friend"level). Complete data on reciprocity at thislevel were available for 666 children. Out ofthese 666 children, 475 children, or 71.3%,had a reciprocal very best friend.̂ This pro-portion is consistent with that reported by

^ Due to absences on tbe day of testing and cbanges in classroom enrollment, reciprocalityof friendsbips could not be determined for all cbildren in tbe sample.

^ Altbougb tbey were instructed not to do so, some cbildren selected more tban one class-mate as tbe person liked most; as a result, 14 cbildren bad two very best friends; in all 14 cases,tbe two friends were tbe same gender; tbese 14 cases were included in tbe analyses. Oneadditional subject bad tbree very best friends, but, as tbese tbree friends were not all tbe samegender, tbis subject was dropped from all analyses involving very best friends. Additionally, 20cbildren were excluded from tbe analyses because tbe person tbey named as tbe very best friendwas absent, and reciprocity of tbe relationsbip could not be determined.

2274 Child Development

other authors using similarly restrictivedefinitions (e.g., Parker & Asher, 1993).

Behavioral reputation with peers.—A"guess who" type peer nomination behav-ioral assessment (Hymel & Rubin, 1985) wasused to assess children's behavioral reputa-tion with their classmates. Ghildren nomi-nated classmates for four behavioral items:(1) "Who says mean things to other kids?"(2) "Who hits or shoves other kids?"(3) "Who helps others a lot?" and (4) "Whois shy and quiet?" For each child, the pro-portion of all same-sex classmates and theproportion of all opposite-sex classmateswho nominated the child were calculated foreach of the four behavioral items. These pro-portion scores were then transformed usingan arcsine square-root transformation.

Perceived self-competence.—The HarterPerceived Gompetence Scale (Harter,1983) was used to assess children's self-esteem and self-perceptions. This 36-itemmeasure assesses five domains of chil-dren's self-perceptions: scholastic compe-tence, social acceptance, athletic compe-tence, physical appearance, and behavioralconduct. A sixth subscale measures chil-dren's global self-worth or self-esteem.Items are presented as dichotomous state-ments (e.g., "Some kids are happy with theway they look . . . but other kids eire nothappy with the way they look"). Ghildrenfirst decide which part of the dichotomy istrue for them, and then indicate whether thestatement is "really" versus "sort of" true ofthem. Harter (1982, 1983) presents informa-tion regarding the validity and reliability ofthe Perceived Gompetence Scale. In the cur-rent study, the reliabilities for the subscaleswere: scholastic competence, alpha = .72;social acceptance, alpha = .70; athletic com-petence, alpha = .74; physical appearance,alpha = .75; behavioral conduct, alpha =.74; and global self-worth, alpha = .73.

Sex-role stereotypicality.—To assessthe degree to which children agreed withtraditional beliefs about sex roles, a 30-itemmeasure of sex-role stereotyping was admin-istered. The measure of sex-role stereotyp-ing was based on a questionnaire developedby Miller (1975) and later adapted for olderchildren by Signorella and Liben (1985).Each item of the questionnaire presented anactivity, such as fixing a sink or washingdishes, and children were asked to indicatewhether the activity was appropriate to menonly, women only, or both men and women.To assist children with responses, each page

included a small line drawing of a man, awoman, and a man and woman at the top.Ghildren were requested to mark only oneresponse per item.

To score the measure, a stereotypicaland nonstereotypical response was first es-tablished for each item. The stereotypicalityof each item was based on previous research(Miller, 1975; Signorella & Liben, 1985) andempirically tested (i.e., items were includedin the subscales only if a majority of the sam-ple agreed that only men or only womencould perform the activity described). Spe-cifically, 13 items reflected traditional men'sroles, as follows: fix a sink, build a house,shovel snow, be a doctor, fly a plane, fix acar, mow the lawn, play football, be a law-yer, collect garbage, climb a mountain, racea car, and tell jokes. Twelve items refiectedtraditional women's roles, including: set thetable, go grocery shopping, wash dishes,bake cupcakes, use a sewing machine, takecare of children, be a secretary, cook a meal,wash the family clothes, teach school, be anurse, and clean house. The remaining fiveitems were "filler" items, including: ride abike, work in an office, go to the beach, goto the movies, and write a book.

In scoring the measure, two subscaleswere formed: a measure of stereotyped atti-tudes toward masculine roles and a measureof stereotyped attitudes toward feminineroles. Scoring was based on the frequencywith which a child chose the stereotyped re-sponse (i.e., chose a response of "men only"for a men's-role item). Thus, on the scale formasculine roles, scores range from 0 to 13,with higher scores indicating a greater en-dorsement of stereotypicality toward tradi-tionally male activities. The reliability forthe masculine subscale was alpha = .83. Onthe scale for feminine roles, scores rangefrom 0 to 12, with higher values indicatinga greater acceptance of stereotyping abouttraditionally female activities. The reliabil-ity for the feminine subscale was alpha =.84. Details on the development and valida-tion of the stereotyping measure are pre-sented in Hoffman and Kloska (1995).

Teacher ratings of child social and cog-nitive competence.—Ghildren's classroomteachers were asked to complete theTeacher-Ghild Rating Scale (T-GRS), devel-oped by Hightower et al. (1986). Items onthis scale tap children's behavioral problemsand social and academic competencies. Forbehavioral problems, teachers rate each itemusing a scale from 1 ("Not Problem") to 5

Kovacs, Parker, and Hoffman 2275

("Very Serious Problem"). For competenceitems, teachers rate each item using a scaleranging from 1 ("Not at all") to 5 ("Verywell").

The 38 T-CRS items form seven distinctsubscales: (1) acting out behavior, (2) shy-ness and anxiety, (3) learning skills, (4) frus-tration tolerance, (5) assertive social skills,(6) task orientation, and (7) peer social skills.Higher scores on the first three scales (actingout, shyness, learning skills) indicate greaterdifficulty, whereas higher scores on the re-maining four scales (frustration tolerance, as-sertive social skills, task orientation, peerskills) indicate greater competence in thatarea. The reliability for each of the sevensubscales was as follows: acting out behav-ior, alpha = .94; shyness-anxiety, alpha =.85; learning skills, alpha = .95; frustrationtolerance, alpha = .92; assertive socialskills, alpha = .90; task orientation, alpha =.94; peer social skills, alpha = .95.

ProcedurePrior to the administration of the mea-

sures reported in this study, each schoolprincipal mailed a letter describing thestudy, the measures, and the assessment pro-cedures to the parents of each of the childrenin the selected classrooms. The letter gavethe phone numbers of the principal and ofthe research project staff, and invited theparents to call if they had any questions. Par-ents were asked to inform the school if theydid not want their child to participate in thestudy either by phone or by returning a formenclosed with the letter. Of the 727 childrenwhose parents were contacted, three wereexcluded from the study at parental requestand one child received parental permissionto participate but declined.

Testing took place in class in a single1-hour session. At the beginning of the ses-sion, the research staff introduced them-selves and briefly described the project.Children were assured that their participa-tion was entirely voluntary, that they coulddecline to participate at any time, and thatall their responses would be confidential.After this introduction, the children were ad-ministered the sociometric measures of peeracceptance, friendship status, and behav-ioral reputation; the Harter Perceived Com-petence Scale; and the measure of sex-rolestereotypes. Afrer the session had con-cluded, teachers were given copies of theT-CRS measure and were requested to com-plete one form for each child. Teachers re-

ceived $50 for completing their question-naires.

At the end of each classroom session,forms were distributed to the children to de-liver to their parents, asking parents for con-sent to include their child in individual in-terviews that are not relevant to the presentstudy. These consent forms also asked par-ents to indicate their employment and mari-tal status, and the ages and genders of allchildren in the home. In the present study,information obtained from these consentforms was used in analyses involving demo-graphic variables (see below). For thosechildren who failed to return the active con-sent form, information on parental maritalstatus was obtained from classroom teachersand school records. Children received asmall prize as an inducement to return theconsent form, regardless of whether or notthe parent agreed to further participation inthe study.

ResultsData Reduction

Because of the relatively large numberof dependent measures, preliminary analy-ses were undertaken to address data reduc-tion. Data reduction was guided by concep-tual criteria, with empirical tests used toconfirm the grouping of variables. First, theseven subscales of the T-CRS were groupedconceptually into two domains: academicskills and social competence. Academicskills consisted of three T-CRS subscales:learning skills, frustration tolerance, andtask orientation. A principal components fac-tor analysis was performed to determine ifthere was support for this grouping, and theanalysis supported the presence of a singlelatent variable. As expected, learning skillsloaded negatively on this factor, while frus-tration tolerance and task orientation loadedpositively on the factor. Accordingly, a sin-gle academic skills factor score was createdfor each child; higher scores indicate greaterlevels of academic competence (and fewerlearning difficulties) as rated by the teacher.

The second conceptual grouping, socialcompetence, consisted of the T-CRS scoresfor acting out behavior, shyness and anxiety,assertive social skills, and peer social skills.Again, principal components factor analysissupported an interpretation of a single latentconstruct, and a single social competencefactor score was derived. The acting out andshyness-anxiety subscales loaded negativelyon this social competence factor, whereas

2276 Child Development

the assertive social skills and peer skills sub-scales loaded positively on this factor. Thus,higher scores on this social competence fac-tor indicate greater social skills (and fewerbehavior problems) as perceived by theclassroom teacher.

A third grouping consisted of the sixsubscales of the Harter Perceived Gompe-tence Scale: scholastic competence, socialacceptance, athletic competence, physicalappearance, behavioral conduct, and globalself-worth. Principal components analysissupported a single factor underlying thesesubscales, with all scores loading positivelyon this factor. Accordingly, a single self-esteem factor was formed, with higher scoresrefiecting a more positive self-image and ahigher estimation of one's competence.

Ghildren's reputations among peers forsaying mean things and hitting others werehighly correlated for both same-sex and op-posite-sex ratings, r = .78 and .77, respec-tively (p < .001). Thus, separate same-sexand opposite-sex aggressiveness scores werecomputed by averaging the respective same-and opposite-sex nominations the subject re-ceived from peers for meanness and hitting.

Overview of Data AnalysisAs initial steps, the prevalence of oppo-

site-sex friendship was examined using therestrictive ("very best") and less restrictive("best") criteria, and the links betweeninvolvement in opposite-sex friendship andgender, ethnicity, family structure, and thegender composition of the classroom wereexamined using parametric and nonpara-metric procedures. Subsequently, sets offocused comparisons were conducted, com-paring the social and behavioral characteris-tics of two conceptually interesting groupsof children involved in opposite-sex friend-ships with those of control children involvedin only same-sex friendships. In the first setof analyses, children involved in cross-sexfriendships along with and secondarily tosame-sex friendships are compared to chil-dren with same-sex friendships only. In thesecond set of analyses, children with primar-ily or only cross-sex friendships are com-pared to matched groups of children withonly same-sex friends, to children with op-posite-sex friendships secondary to same-sexfriendships, and to children without friendsentirely.

Descriptive Information: Prevalenceof Opposite-Sex Friendship

As expected, opposite-sex friendships(N = 63, or 4.7%) constituted only a small

fraction of the total number (N = 1331) ofreciprocal best friendships in the sample. Interms of individual children, 92 children, or13.5% of all children in the sample, had oneor more opposite-sex reciprocal best friends.More than three-fourths of these children {N= 71, or 77.2%) had only one opposite-sexfriend, whereas the remaining 22.8% (IV =21 children) had two or more. Importantly,there were 18 children for whom their cross-sex friend was their very best friend, andthere were seven children who had onlycross-sex friendships. Thus, consideringthese criteria jointly, there were 21 instancesof children involved primarily or only in op-posite-sex friendships, and 71 instances ofcnildren involved in cross-sex friendshipssecondarily to their same-sex friendships.An additional small sample of 19 childrenhad no reciprocal friendships at all, leavinga pool of 572 children who were involved inonly same-sex friendships.

Sex and ethnic differences.—The pro-portion of children with cross-sex friendswas approximately equivalent for boys(12.5%) and girls (15.1%), x^(l, N = 664) =.91, N.S. However, African American chil-dren (22.3%) were twice as likely as Euro-pean American children (9.4%) to have op-posite-sex friends, x^(l, N = 664) = 20.74,p < .001. The greater involvement of AfricanAmerican children in cross-sex friendshipwas part of a larger tendency for AfricanAmerican children to have more friends ingeneral than European American children,as follows: a 2 (sex) x 2 (ethnicity) analysisof variance found no difference betweenboys and girls in numbers of opposite-sexfriends, F(l, 660) = 0.27, N.S., and no inter-action of sex and ethnicity in predictingnumbers of opposite-sex friends, F(l, 660) =0.12, N.S. However, African American chil-dren reported more opposite-sex friends (M= 0.32) than European American childrendid (M = 0.12), F(l, 660) = 20.33, p < .001.In a similar analysis, a 2 (sex) X 2 (ethnicity)analysis of variance showed no interactionof sex and ethnicity in predicting numbersof same-sex friends, F(l, 660) = 0.00, N.S.However, African American children re-ported more friends of the same sex (M =4.35) than did European American children(M = 3.54), F(l, 660) = 18.60, p < .001.

Classroom make-up.—The number ofopposite-sex friendships children had wasunrelated to the gender make-up of theirclassroom. Specifically, number of opposite-sex friends did not significantly correlatewith number of opposite-sex children in the

classroom (r = .05, N.S.). Likewise, the pro-portion of opposite-sex friends in a child'sfriendship network did not significantly cor-relate with the proportion of opposite-sexchildren in the class (r = .06, N.S.).

Family structure and sibling status.—Information on siblings was available for 550of the 664 children with friends: of thesechildren, 335 (60.9%) had opposite-sex sib-lings. Children with opposite-sex siblingswere slightly more likely to have at least onecross-sex friend than children with no sib-lings or with only same-sex siblings, x^(l, N== 550) = 3.47, p < .10. Whereas 15.8% ofchildren with opposite-sex siblings hadcross-sex friends, only 10.2% of childrenwithout opposite-sex siblings did.

Information on parents' current maritalstatus was available for 646 of the 664 chil-dren with friends: of these children, 394(6il.0%) were living in two-parent (as op-posed to father-absent) homes. Childrenfrom single-parent homes were more likelyto have at least one cross-sex friend thanchildren living in two-parent homes, x^(l, N= 646) = 7.81, p < .01. Whereas 19.0% ofchildren of single parents had cross-sexfriends, only 11.2% of children from two-parent homes did. Children in single-parenthomes also had more cross-sex friends (M =.29) than did children in two-parent homes(M = .14). However, a 2 (sex) x 2 (ethnicity)X 2 (marital status) analysis of variance indi-cated that the effect of marital status on thenumber of opposite-sex friends was quali-fied by the presence of a significant interac-tion with ethnicity, F(l, 638) = 8.52, p <.01. For European American children, therewas little difference between children fromsingle- and two-parent homes in the num-bers of opposite-sex friends (M = .11 and.13, respectively). However, African Ameri-can children from single-parent homes hadmore opposite-sex friends (M = .43) thanthose from two-parent homes (M = .17).There was also a trend (p < .10) for a sex xethnicity x marital status interaction. Thepattern of means showed that African Ameri-can boys in single-parent homes had thegreatest number of opposite-sex friends (M= .52), followed by African American girlsin single-parent (M = .37) and two-parent(M = .28) homes, European American boysin two-parent (M = .15) and single-parent(M = .12) homes, European American girlsin two-parent (M = .11) and single-parent(M = .09) homes, and African American boysin two-parent homes (M = .07). Based onthese means, it appears that there is a ten-

Kovacs, Parker, and Hoffman 2277

dency for the impact of parental marital sta-tus on involvement in cross-gender friend-ship to be greatest for African Americanboys.

Primary versus Secondary Involvementin Opposite-Sex Friendship: Gender,Ethnicity, and Friendship Network Size

Chi-square analyses were performed inorder to compare the gender and ethnicityof children primarily or only involved incross-sex friendship ("primary-opposites"),children involved in cross-sex friendshipsecondarily to same-sex friendship ("sec-ondary-opposites"), and children with onlysame-sex friendships ("same-sex onlys").Consistent with prior analyses, no sex differ-ences were found in the composition of thethree groups, x^(2, 2V = 664) = 3.28, N.S.,and African Americans were more likelythan European Americans to belong to bothof the groups with cross-sex friends, x^(2, N= 664) = 20.85, p < .001. Of children pri-marily involved in opposite-sex friendship,52.4% were African American; of childreninvolved secondarily in opposite-sex friend-ship, 56.3% were African American; and ofchildren with only same-sex friends, 31.1%were African American.

In order to determine whether childrenwith primarily or only cross-sex friendships,children with cross-sex friendships second-arily to same-sex friendships, and childrenwith only same-sex friendships differed inoverall network size (i.e., number of friendsof either sex), in addition to differing in thegender composition of those networks, a 2(sex) X 2 (ethnicity) x 3 (group) analysis ofvariance was performed. The secondary-opposites had more friends (M = 6.21) thanthe primary-opposites {M = 3.86) and thesame-sex onlys (M = 3.74), F(2, 652) =31.82, p < .001. For both sexes, secondary-opposites had the most friends {M = 7.15 forboys, M = 5.64 for girls); however, a signifi-cant interaction with sex, F(2, 652) = 3.44,p< .05, indicated that this effect was moder-ately stronger for boys than for girls. Tukeypost-hoc tests for each sex indicated that,among boys, primary-opposites (M = 3.42)and same-sex onlys (M = 3.94) did not differsignificantly from each other. For girls, pri-mary-opposites (M = 4.44) did not differ sig-nificantly from secondary-opposites or same-sex onlys (M = 3.56). A statisticallysignificant difference between the groupswas also found for numbers of same-sexfriends, F(2, 652) = 13.98, p < .001; the sec-ondary-opposites group had more same-sexfriends (M = 4.94) than the same-sex onlys

2278 Child Development

group (M = 3.74), who had more same-sexfriends than the primary-opposites group (M= 2.14). Again, this effect was stronger forboys than for girls, F(2, 652) = 5.11, p <.01. These analyses indicate that the largeroverall size of secondary-opposites' friend-ship networks cannot be attributed solely tothe addition of opposite-sex friends to an av-erage-size same-sex network. Instead, sec-ondary-opposites have more friends of bothsexes than other groups.

Focused Comparisons: Correlatesof Secondary Involvementin Cross-Sex Friendship

In order to compare secondary-op-posites to same-sex onlys on the social andcognitive functioning variables, a series of 2(sex) X 2 (ethnicity) x 2 (group) analyses ofvariance were conducted. Dependent mea-sures for these analyses included the mea-sures of behavioral reputation with peers,peer acceptance, teacher ratings of socialand academic competence, self-esteem, andadherence to sex-role stereotypes.

Behavioral reputation with peers.—Secondary-opposites and same-sex onlys didnot differ significantly in their reputationswith peers for aggressiveness or shyness.However, secondary-opposites were viewedby children of both sexes as more helpfulthan same-sex onlys. A significant three-wayinteraction of sex, ethnicity, and friendshipgroup, F(l, 634) = 4.68, p < .05, indicatesthat the effects of opposite-sex friendship onsame-sex helpfulness were strongest for Eu-ropean American boys (M = .79 for second-ary-opposites; M = .54 for same-sex onlys)and for African American girls (M = .72 forsecondary-opposites; M = .66 for same-sexonlys). These differences were negligibleamong African American boys (M = .58 forsecondary-opposites; M = .59 for same-sexonlys) and European American girls (M =.79 for secondary-opposites; M = .77 forsame-sex onlys). In a similar manner, a sig-nificant ethnicity X friendship group inter-action indicates that the effect of cross-sexfriendship on opposite-sex helpfulness isstronger for European Americans than forAfrican Americans, F(l, 634) = 7.97, p < .01.Among European Americans, secondary-opposites have a mean of .68 on opposite-sex

helpfulness, as opposed to .42 for same-sexonlys; corresponding figures for AfricanAmericans are .52 and .40.

Peer acceptance.—In comparison withsame-sex onlys, secondary-opposites werebetter accepted by peers of both sexes. Theeffect of friendship group on acceptance bysame-sex peers was stronger for boys thanfor girls, F(l, 634) = 4.93, p < .05. Boys in-volved in cross-sex friendship secondarilyreceived a mean same-sex rating of 4.49, asopposed to a mean rating of 3.99 for boyswith only same-sex friends; for girls, the cor-responding means are 4.20 and 4.05. Like-wise, the effect of friendship group on ac-ceptance by opposite-sex peers wasqualified by a significant interaction withethnicity, F(l, 634) = 5.28, p < .05, with theeffect being stronger for European Ameri-cans (M = 3.60 for secondary-opposites; 2.87for same-sex onlys) than for African Ameri-cans (M = 3.40 for secondary-opposites; 3.02for same-sex onlys).

Teacher ratings of social and academiccompetence.—No significant main effects offriendship group were found for the teacherratings of social and academic competence.However, the sex x friendship group inter-action fell just short of significance for theteacher ratings of academic competence,F(l, 621) = 3.79, p < .06. Boys with second-ary cross-sex friends (M = .02) tended toscore higher than boys with only same-sexfriends (M = — .22); however, girls with sec-ondciry cross-sex friends (M = .00) tendedto score lower than girls with only same-sexfriends (M = .31).

Self-reports of self-esteem and sex-rolestereotyping.—No significant differenceswere found between the groups in self-esteem; however, significant main effects offriendship group were found for attitudesabout sex-role stereotypes. Gontrary to ex-pectations, secondary-opposites were morestereotyped about traditional female role ac-tivities (M = 6.04) than were children withonly same-sex friends (M = 4.54), F(l, 625)= 10.21, p < .01. No significant differenceswere found for stereotypes about masculineactivities.^

^ Because the differences that were found between the secondary-opposites and same-sexonlys in terms of overall numbers of friends might explain the pattern of findings in the ANOVAsin tiiis section, this set of analyses was repeated using the number of reciprocal friends as acovariate. Almost all the previous findings were replicated, with the following exceptions: Thethree-way interaction of sex, ethnicity, and friendship group for same-sex helpfulness fell fromsignificance to the level of a trend (p < .10). The main effect of friendship group and the interac-

Kovacs, Parker, and Hoffman 2279

Focused Comparisons: Correlatesof Primary Involvementin Cross-Sex Friendship

The primary-opposites group included21 subjects: six European American boys, sixAfrican American boys, four EuropeanAmerican girls, and five African Americangirls. The 21 subjects in this group werematched to 21 children from the secondary-opposites group and to 21 children from thesame-sex only group. All children werematched by gender and ethnicity, and by thepresence of a very best reciprocal friend-ship. To the extent that it was possible to doso, children were also matched by class orschool and by the total number of friends."*

In order to compare the three groups onthe social and cognitive' functioning vari-ables, a series of planned contrast analyseswere performed. Dependent measures forthese planned contrasts included the mea-sures of behavioral reputation with peers,peer acceptance, teacher ratings of socialand academic competence, self-esteem, andadherence to sex-role stereotypes. For eachdependent measure, two a priori contrastswere conducted. First, the group of primary-opposites were contrasted with the matchedgroup of children with only same-sex friendsto determine how primary opposite-sexfriendship involvement differs from the typi-cal pattern of exclusively same-sex involve-ment. Second, the group of primary-opposites were contrasted with the matchedsecondary-opposites group to determinehow the primacy of opposite-sex friendshipinvolvement relates to adjustment and atti-tudes. The results of these planned contrastanalyses are shown in Table 1, and summa-rized below.

Behavioral reputation with peers.—Incomparison to both secondary-opposites andsame-sex onlys, the primairy-opposites wereperceived by peers of the same sex as higher

in aggressiveness and lower in helpfulness.Primary-opposites were also rated by same-sex peers as less shy than secondary-opposites. When assessed by peers of the op-posite sex, the primary-opposites were ratedas more aggressive, with a trend (p < .10)toward being less helpful, than secondary-opposites. None of the contrasts revealed asignificant difference between the groups inopposite-sex shyness.

Peer acceptance.—As shown in Table 1,children in the primary-opposites groupwere less popular with peers of both thesame and opposite sex than matched chil-dren in the secondary-opposites group. Theprimary-opposites were also less popularthan matched children with only same-sexfriends, but only when rated by members ofthe same sex.

Teacher ratings of social and academiccompetence.—In comparison to the second-ary-opposites group, the primary-oppositesreceived lower ratings from teachers for bothacademic skills and social competence.

Self-reports of self-esteem and sex-role stereotyping.—The primary-oppositestended (p < .10) to have lower self-esteemthan both the secondary-opposites andsame-sex onlys. Primary-opposites were alsoless stereotyped about traditional male rolesthan were same-sex onlys, and they tended(p < .10) to be less stereotyped about tradi-tional female roles than children in both thesecondary-opposites and same-sex onlygroups.

Because the preceding planned con-trasts did not test for the possibility of inter-actions with gender and ethnicity, a seriesof supplemental 2 (sex) x 2 (ethnicity) x 2(group) analyses of variance were per-formed: one set compared primary-oppositeswith secondary-opposites; a second set com-pared primary-opposites with same-sex on-

tion of sex and friendsbip group for same-sex acceptance were no longer statistically significant;and tbe interaction of sex and friendsbip group in predicting teacber ratings of academic skillsno longer reacbed tbe level of a trend. Thus, differences between tbe groups in terms of numbersof friends migbt provide an alternative explanation for tbe observed differences in same-sexacceptance. On tbe wbole, bowever, differences in numbers of friends do not seem sufficient toexplain most findings in tbis section.

•* A manipulation cbeck was performed to assess wbetber the groups differed in terms oftbe total number of reciprocal friends. Recause tbe tbree groups bad been matcbed for tbischaracteristic, no differences were expected. A one-way analysis of variance yielded no signifi-cant differences between tbe tbree groups for number of reciprocal friends, F(2, 60) = 1.29, N.S.A cbi-square test also indicated tbat tbe tbree matcbed groups did not differ significantly inparental marital status (i.e., two-parent vs. single-parent bomes), x^(2, N = 63) = 2.06, N.S.Altbougb tbe groups were not specifically matcbed for marital status, tbey were matcbed byethnicity and scbool, wbicb effectively controlled for marital status.

2280 Child Development

TABLE 1

ADJUSTMENT AND ATTITUDES OF MATCHED GROUPS OF CHILDREN WITH PRIMARY VERSUS SECONDARYVERSUS No INVOLVEMENT IN OPPOSITE-SEX FRIENDSHIPS

Outcome Measure

Opposite-sex aggressionSame-sex helpfulnessOpposite-sex helpfulnessSame-sex shynessOpposite-sex shynessSame-sex acceptanceOpposite-sex acceptanceTeacher rating: academic

skillsTeacher rating: social

competenceSelf report: self-esteemStereotyped about masculine

rolesStereotyped about feminine

roles

+ p < .10.• p < .05.** p < .01.*** p < .001.

Primary-Oppositesin = 21)

.60

.52

.56

.5125.26

3.58. 3.10

. - .49

. - .35

. - .31

4.64

. 3.16

Secondary-Opposites(n = 21)

.25

.27

.86

.67

.41

.374.473.64

.46

.35

.23

5.52

5.00

Same-SexOnlys

(n = 21)

37.41.73.40.27.23

4.103.14

-.19

- .03.27

7.25

4.89

Gontrast 1:Primary-Ops

vs.Secondary-Ops

it)4 49***2.71**

-3.46**-1.83+- 2 03*

1.61-5.04***-3.12**

- 3.30**

2.59*-1.75+

-.86

-1.86+

Gontrast 2:Primary-Ops

vs.Same-Sex

Onlys(*)

2 93**1.16

-2.05*1.16

26.46

-2.91**- .20

-1.04

-1.19-1.90+

-2.53*

-1.74+

lys. The dependent measures for these 2 x2 x 2 analyses of variance were the same asin the planned comparisons: same-sex ag-gression, opposite-sex aggression, same-sexhelpfulness, opposite-sex helpfulness, same-sex shyness, and opposite-sex shyness;same-sex acceptance and opposite-sex ac-ceptance; teacher ratings of academic skillsand social competence; self-esteem; and ad-herence to sex-role stereotypes for men andwomen. There were several significant maineffects of sex and ethnicity on the dependentmeasures, as well as a few significant sex xethnicity interactions. These sex and eth-nicity effects were not of central concernwithin the current study, and thus will notbe described. Of greater interest to the cur-rent study, in the comparisons of the pri-mary-opposites and same-sex onlys, therewere no significant interactions of groupwith sex or with ethnicity, nor were thereany significant three-way interactions ofgroup, sex, and ethnicity for any of the de-pendent measures. In the comparisons of theprimary-opposites with secondary-oppo-sites, there were two significant interactions,as follows. For opposite-sex helpfulness,there was a significant ethnicity x group in-

teraction, F(l, 34) = 5.25, p < .05, and asignificant sex X ethnicity x group interac-tion, F(l, 34) = 5.12, p < .05. Primary-opposites received lower ratings of help-fulness from opposite-sex peers thansecondary-opposites among EuropeanAmerican boys (.44 vs. .86), European Amer-ican girls (.75 vs. .92), and African Americangirls (.37 vs. .59). However, among AfricanAmerican boys, primary-opposites receivedhigher ratings of helpfulness from opposite-sex peers (M = .52) than did secondary-opposites (M = .38).

Children with Primarily Opposite-SexFriends versus Children with No Friends

Finally, to the extent that it was possibleto do so, the primary-opposites werematched by sex and ethnicity to a group ofchildren with no friends. There were 19children without friends in the sample, and15 of them (six European American boys,three European American girls, three Afri-can American boys, and three African Ameri-can girls) could be matched perfectly by sexand ethnicity to 15 children in the primary-opposites group. Where possible, the chil-dren were also matched by class or school.

Kovacs, Parker, and Hoffman 2281

TABLE 2

ADJUSTMENT AND ATTITUDES OF MATCHED GROUPS OF CHILDREN wrrH PRIMARILY OPPOSITE-SEXFRIENDS VERSUS CHILDREN WITH NO FRIENDS

Outcome Measure

Same-sex aggressionOpposite-sex aggressionSame-sex helpfulnessOpposite-sex belpfiilnessSame-sex sbynessOpposite-sex sbynessSame-sex acceptanceOpposite-sex acceptanceTeacber rating: academic skills ....Teacber rating: social competenceSelf report: self-esteemStereotyped about masculine rolesStereotyped about feminine roles

* p < .05.**p< .01.

Primary-Opposites(n = 15)

No Friends(n = 15)

.64

.47

.51

.48

.22

.203.483.10-.41-.27-.244.563.23

.61

.66

.35

.27

.25

.222.822.21-.95

-1.30-.846.634.42

.25-1.47

1.492.25*-.42-.162.28*3.55**1.422.89**1.45

-1.50-1.05

A series of matched-groups t tests wasthen performed to determine if there wereany differences on the battery of social andcognitive outcome measures between thechildren with no friends and the primary-opposites. As shown in Table 2, primary-opposites were better accepted by peers ofboth sexes than the children with no friends,and they were perceived by their opposite-sex peers as more helpful. Additionally, theprimary-opposites were more socially com-petent than children with no friends, asrated by the classroom teachers. Differencesbetween the two groups were not statisti-cally significant for any other outcomemeasure.

Discussion

Consistent with previous research (Feir-ing & Lewis, 1991; Cottman, 1986; Hartup,1983; Maccoby & Jacklin, 1987), when con-sidered in the context of the total numberof friendships in the sample population, theprevalence of cross-sex friendships at thisage was very low, less than 5%. Moreover,most children involved in cross-sex friend-ships tended to have other and closer same-sex friends. Nevertheless, although cross-sexfriendships were rare when considered inthe broadest context, it was still the case thatclose to 14% of all children with friends re-ported at least one cross-sex friendship. Itmight be noted, too, that this prevalence isfor verified, reciprocal friendships. Presum-ably, additional children may believe them-selves to have a friendship with someone of

the opposite sex (i.e., have unilateral friend-ships). Likewise, other children may alsohave close reciprocal relationships with op-posite-sex children that they resist labelingas a "friendship." Thus, although overallinvolvement in cross-sex friendships is low,it seems that a fair number of children maybe gaining at least some experience in cross-sex relationships.

As might be expected, neither boys norgirls had a greater likelihood of being in-volved in cross-sex friendships. In a less ex-pected manner, the proportion of opposite-sex peers in the classroom did not appearto infiuence the propensity to form cross-sexfriendships. Although authors such as Ep-stein and Karweit (1983), Hallinan (1979),and Kistner, Metzler, Gatlin, and Risi (1993)have stressed the importance of structuraland contextual features of the classroom indetermining children's friendships, it ap-pears, at least in this sample, that individualcharacteristics of children carry greaterweight in predicting who will form cross-sexfriendships than do classroom structuralcharacteristics.

One unexpected finding of the presentstudy was that African American childrenhad more friends of both the same and oppo-site sex than did European American chil-dren. Although children of both ethnicgroups showed a greater involvement insame-sex friendship than cross-sex friend-ship, the proportion of African Americanchildren who were involved in cross-sex

2282 Child Development

friendship was higher than that for Euro-pean American children. This finding of eth-nic differences in friendship involvementsuggests that there may be some subculturaldifferences that infiuence children's atti-tudes toward friendship. Perhaps AfricanAmerican children apply a broader defini-tion to the term "friend" than do EuropeanAmerican children, or perhaps AfricanAmerican children are encouraged by theirfamilial, neighborhood, and school environ-ments to form wider and more diverse socialnetworks with peers. Future research onthese issues is obviously warranted. In thisconnection, one issue that remains for futurestudies is to consider how the ethnicmake-up of the classroom infiuences AfricanAmerican children's involvement in cross-sex reciprocal friendship. Although this is-sue has not been directly examined in pastresearch, Kistner and colleagues (1993)found that, regardless of their ethnic iden-tity, elementary school girls who were in thenumerical minority in their classrooms weremore likely to be rejected by peers. To theextent that these findings for acceptance canbe generalized to reciprocal friendship, itmight be predicted that African Americanchildren will have more friends of the sameand opposite sex only in classrooms wherethey belong to the numerical majority.

In the present study, African Americanchildren from single-parent homes had moreopposite-sex friends than African Americanchildren from two-parent homes, althoughthis pattern tended to be stronger for boysthan for girls. Wilson (1989) has suggestedthat African American children with singleparents are likely to live in extended-familysettings with many female family membersand few adult males. The present findingsmight, therefore, be explained by the argu-ment that within such a setting, boys wouldbe especially likely to develop skills for in-teracting with members of the opposite sex.It should be noted, however, that in thisstudy it was not possible to separate parentalmarital status per se from other related fam-ily factors; family size, parental education,family income, and geographic mobility areoften correlated with parental marital status,and thus may underlie the differences fount!in this study between children with andwithout cross-sex friends. In future research,effort should be directed toweird disentan-gling marital status from these other familyfactors.

Toman (1976) has theorized that peoplewith siblings of the opposite sex are more

likely to form cross-sex friendships than arepeople with no opposite-sex siblings. Burkerand colleagues (1981) reported mixed sup-port for Toman's hypothesis with adult sub-jects: they found that women with brotherswere more comfortable with male friendsthan were women without brothers; how-ever, men with sisters were slightly lesscomfortable with female friends than weremen without sisters. The present study alsoprovides only weak support for Toman's hy-pothesis. Specifically, there was a nonsig-nificant trend for children with opposite-sexsiblings being more likely than other chil-dren to have a cross-sex friend. Because it isdifficult to draw strong generalizations onthe basis of a trend, further research isneeded to determine whether Toman's hy-pothesis can be fully supported.

One of the central aims of the currentstudy was to determine whether childrenwith cross-sex friends would differ fromother children in terms of several adjustmentmeasures and in their adherence to sex-rolestereotypes. As noted, previous research hasoffered confiicting hypotheses and data onthis point. One hypothesis has been thatthose children who have cross-sex friendsare less socially skilled than other children.And, indeed, at least one subgroup of chil-dren showed just such a profile. Specifically,when they were compared to other childrenwith friends, children with primarily oppo-site-sex friends were rated as more aggres-sive and less helpful by peers, were lesswell liked by peers of both the same andopposite sex, tended to report lower self-esteem, and were judged to be less skilledacademically and socially by teachers.

However, in this study there was also asecond group of children involved in cross-sex friendship who displayed a quite differ-ent profile. Unlike children with primarycross-sex friends, those children who formedcross-sex friendships secondarily to same-sex friendships did not seem to be less so-cially skilled than other children. In theoverall sample, children involved second-arily in cross-sex friendship had morefriends than children involved primarily incross-sex friendship and than children withonly same-sex friends. In comparison to chil-dren with only same-sex friends, childrenwith secondary cross-sex friendships wereviewed as more helpful and were better ac-cepted by peers of both sexes, althoughthese effects tended to be stronger for Euro-pean Americans than for African Americans.Thus, support for the opposing hypothesis

that emphasizes the more positive aspectsof cross-sex friendship involvement was alsoevident within the present study.

Overall then, the findings support theidea that there are two distinct groups ofchildren with cross-sex friends, which differgreatly among themselves, depending uponwhether their cross-sex friendships are pri-mary or secondary in importance to them.Children whose very best friend or onlyfriend was an opposite-sex peer showed apattern of lower functioning on several ofthe measures of social adjustment; however,children who had a same-sex very bestfriend, but who had additional (secondary)friendships with opposite-sex peers, did notshow this pattern of lower functioning.

Children whose cross-sex friendshipsare secondary to same-sex ones seem to fitwithin the descriptions of cross-sex friend-ships provided by Cottman (1986) andHowes (1988). Among preschoolers, Howes(1988) suggests that children with cross-sexfriends are more socially skilled and showmore positive affect than other children.Coittman (1986) suggests that cross-sexfriendships may be special types of relation-ships, which promote social development.In the overall sample of the current study,children involved secondarily in cross-sexfriendship did seem to be more sociablethan other children; they even had a greaternumber of same-sex friends than did chil-dren with only same-sex friends.

In contrast, children with primarily op-posite-sex friendships seem to fit a descrip-tion that is more consistent with the predic-tions of Carter and McCloskey (1984) andLadd (1983). The work of Carter and Mc-Closkey (1984) suggests that children whoexhibit cross-gender friendship preferenceswill be more rejected by their peers. Ladd(1983) speculates that it is those childrenwho are rejected by their same-sex peerswho seek friends of the opposite gender. Inthe current study, children involved primar-ily in cross-sex friendship did have fewersamie-sex friends than other children in theoverall sample. In the matched group analy-ses, children primarily involved in cross-sexfriendship exhibited a pattern of adjustmentwhich suggests that they are less sociallyskilled than other children with friends.

However, it should also be noted thatchildren with only or primarily opposite-sexfriendships were more well liked by peersof both the same and opposite sex, and wererated as more socially competent by their

Kovacs, Parker, and Hoffman 2283

teachers, than were children with no friends.Thus, children with primarily opposite-sexfriends are better adjusted, socially, than arechildren with no friends at all. Hence, it ispossible that primary involvement in cross-sex friendship might prove to be a form ofresilient adaptation, relative to having nofriends at all. In this context, it is importantto distinguish between the short-term andlong-term implications. Although childrenwith primarily opposite-sex friends appearto be less socially skilled than other childrenwith friends in elementary school, the long-term consequences of this form of friendshipinvolvement remain unknown.

One further hypothesis tested by thecurrent study was whether children withand without cross-sex friends would differin their endorsement of sex-role stereotypes.Cohen, D'Heurle, and Widmark-Petersson(1980) have suggested that adherence to sex-role stereotypes may infiuence children'sdecisions to participate in cross-sex friend-ships, and Martin (1994) demonstrated thatyoung children with greater knowledge ofthe gender-typed qualities of play exhibitmore gender segregation in their own play.In contrast, Maccoby (1990) suggests thatsex-role stereotypicality bears little relationto engagement in cross-sex friendships. Inthe current study, children with primarilyopposite-sex friends reported less stereo-typed views of traditional male roles thandid children with only same-sex friends;they also tended to be less stereotyped abouttraditional female roles than other children,supporting arguments that traditional sex-role attitudes discourage cross-sex involve-ment at this age. However, this relation be-tween sex-role attitudes and friendshipinvolvement does not appear to hold forthose children who are secondarily involvedin cross-sex friendship. Clearly, future re-search is warranted. In future studies, therelation of cross-sex friendship and stereo-typing might be enhanced by the use of ad-ditional measures of stereotyping, both ob-servational and self-report.

One limitation of the current study isthat it is not possible to determine the causaldirections of the relations among the veiri-ables. In this study, children with primarycross-sex friends were less socially success-ful in certain respects than other childrenwith friends. However, as noted previously,there are at least two alternative pathwaysthat could account for such a pattern. Onepossibility is that children form friendshipswith opposite-sex peers only after their

2284 Child Development

same-sex peer relationships prove unsuc-cessful. Following this progression, childrenwith primarily opposite-sex friends wouldbe those children who, at a young age, lacksocial skills, are rejected by their same-sexpeers, experience low self-esteem, and thenturn to opposite-sex peers for friends (asLadd, 1983, suggests). Once children havedeveloped friendship relationships withmembers of the opposite sex, their stereo-typed ideas about the opposite gender maychange as well. An opposite developmentalprogression holds that primarily havingfriends of the opposite sex in middle child-hood, because it violates normative behav-ior, contributes to peer rejection (as Garter& McGloskey, 1984, propose). On this view,children who act in ways that break tradi-tional sex-typed norms (primarily havingfriends of the opposite sex, holding nonste-reotyped views) may, in turn, suffer peer re-jection. It also seems possible that teacherscould view these atypical patterns of socialinteraction as indicators of low social skills.As a consequence of the negative reactionsof others, the self-esteem of children withprimarily opposite-sex friendships may suf-fer. Either of these interpretations is consis-tent with the current data.

A similar ambiguity surrounds the ob-served pattern of findings concerning thelink between behavior and popularity withrespect to involvement in cross-sex friend-ship. Because the data in this study are cor-relational in nature, it remains unknown, forexample, whether children primarily in-volved in cross-sex friendship are less wellaccepted by peers because they display ag-gressive, socially unskilled behaviors, or be-cause of their involvement in cross-sexfriendship per se. Longitudinal research isnecessary in order to determine the causaldirections of these relationships. Under-standing the causal links among these vari-ables may prove especially useful for de-termining the most effective ways tointervene on the behalf of children who ex-perience peer rejection.

One area for future research that mightaid in an interpretation of the current resultswould be a comparison of the quality of chil-dren's friendships for primary-opposites,secondary-opposites, and same-sex onlys. lichildren involved primarily in opposite-sexfriendships are those who lack social skills,their friendships may be of lower quality(perhaps marked by greater levels of confiictand lower intimacy) than those of other chil-dren. It is possible, for example, that chil-

dren with primarily cross-sex friends showlower self-esteem than other children be-cause their friendship relationships are lesssatisfying to them. The quality of theirprimary friendship may also moderateself-esteem within this group of children;primary-opposites with higher-quality re-lationships may not show the less adaptivepattern typical of other primary-opposites.This line of reasoning is analogous to thefindings of Parker and Asher (1993), who re-port that rejected children with higher-quality friendships are less lonely than otherrejected children.

Further research is also needed to ascer-tain the long-term developmental sequelae.Will children with cross-sex friendships bemore adept in later romantic relationships,as Gottman (1986) and Maccoby (1990) havesuggested? Perhaps children who havecross-sex friends learn skills for opposite-sexinteraction at an earlier age than other chil-dren do. If so, children who have cross-sexfriends in elementary school may later ex-hibit advanced skills in social interactions,especially in romantic relationships. Futureresearch is also necessary to examine thefriendships that children form outside oftheir classrooms and schools. Ghildren whohave close opposite-sex friendships outsidethe school environment could maintain theprivacy of their relationships, and thus avoidthe potential ostracism associated withbreaking sex-role norms.

Howes (1988) suggested that the cross-sex friendships of preschoolers should beencouraged by parents and teachers. Whatsuggestions would the current study offer toadults who seek to infiuence children'sfriendship patterns? First, it seems clear thata cross-sex friend is better than no friend atall. If children can maintain primary rela-tionships with same-sex peers, then cross-sex friendships do not appear to be detri-ments to children's social skills and maypossibly have a beneficial impact. However,having a primary same-sex friend is relatedto greater social and behavioral adjustmentthan is having a primary opposite-sex friend.It should be stressed, nonetheless, that fur-ther research is necessary to determine thelong-term implications of these findings.

RefereneesArcher, J. (1992). Ghildhood gender roles: Social

context and organization. In H. McGurk (Ed.),Social development: Contemporary perspec-tives (pp. 31-62). Hillsdale, NJ: Erlbaum.

Kovacs, Parker, and Hoffman 2285

Rerndt, T. J., & Hoyle, S. C. (1985). Stability andcbange in cbildbood and adolescent friend-sbips. Developmental Psychology, 21,1007-1015.

Bukowski, W. M., Cauze, C , Hoza, B., & New-comb, A. F. (1993). Differences and consis-tency between same-sex and otber-sex peerrelationsbips during early adolescence. De-velopmental Psychology, 29, 255-263.

Burker, E. J., Goldstein, M. W., & Caputo, C. C.(1981). Effects of siblings of tbe opposite sexon friendsbip. Psychological Reports, 48,190.

Carter, D. B., & McCloskey, L. A. (1984). Peersand tbe maintenance of sex-typed bebavior:Tbe development of cbildren's conceptions ofcross-gender bebavior in tbeir pears. SocialCognition, 2, 294-314.

Cohen, J. J., D'Heurle, A., & Widmark-Petersson,V. (1980). Cross-sex friendsbip in cbildren:Gender patterns and cultural perspectives.Psychology in the Schools, 17, 523-529.

Douvan, E., & Adelson, J. (1966). The adolescentexperience. New York: Wiley.

Epstein, J. L. (1983). Selection of friends in differ-ently organized scbools and classrooms. InJ. L. Epstein & N. Karweit (Eds.), Friends inschool: Patterns of selection and influencein secondary schools (pp. 73-92). New York:Academic Press.

Epstein, J. L., & Karweit, N. (Eds.). (1983).Friends in school: Patterns of selection andinfluence in secondary schools. New York:Academic Press.

Feiring, C , & Lewis, M. (1991). Tbe transitionfrom middle cbildbood to early adolescence:Sex differences in the social network and per-^ceived self-competence. Sex Roles, 24,489-509.

Furman, W. (1996). Tbe measurement of friend-ship perceptions: Conceptual and method-ological issues. In W. M. Bukowski, A. F.Newcomb, & W. W. Hartup (Eds.), The com-pany they keep: Friendship in childhood andadolescence (pp. 41-65). Cambridge: Cam-bridge University Press.

Golombok, S., & Fivusb, R. (1994). Gender devel-opment. New York: Cambridge UniversityPress.

Gottman, J. M. (1986). Tbe world of coordinatedplay: Same- and cross-sex friendsbip in youngcbildren. In J. M. Gottman & J. G. Parker(Eds.), Conversations of friends: Specula-tions on affective development (pp. 139-191).New York: Cambridge University Press.

Hallinan, M. T. (1979). Structural effects on cbil-dren's friendsbips and cliques. Social Psy-chology Quarterly, 42, 43-54.

Hallinan, M. T., & Tuma, N. B. (1978). Classroomeffects on cbange in cbildren's friendsbips.Sociology of Education, 51, 270-282.

Harter, S. (1982). Tbe Perceived CompetenceScale for Cbildren. Child Development, 53,87-97.

Harter, S. (1983). Revision of the Perceived Com-petence Scale for Children. Unpublishedmanuscript. University of Denver.

Hartup, W. W. (1983). Tbe peer system. In E. M.Hetberington (Ed.), P. H. Mussen (SeriesEd.), Handbook of child psychology: Vol. 4.Socialization, personality, and social devel-opment (pp. 103-196). New York: Wiley.

Havigburst, R. J. (1953). Human development andeducation. New York: McKay.

Hayden-Tbomson, L., Rubin, K. H., & Hymel, S.(1987). Sex preferences in sociometriccboices. Developmental Psychology, 23,558-562.

Higbtower, A. D., Work, W. C , Cowen, E. L., Lo-tyczewski, B. S., Spinell, A. P., Guare, J. C ,& Robrbeck, C. A. (1986). Tbe Teacber-CbildRating Scale: A brief objective measure of el-ementary cbildren's scbool problem bebav-iors and competencies. School PsychologyReview, 15, 393-409.

Hoffman, L. W., & Kloska, D. D. (1995). Parents'gender-based attitudes toward marital rolesand cbild rearing: Development and valida-tion of new measures. Sex Roles, 32,273-295.

Howes, C. (1988). Same- and cross-sex friends:Implications for interaction and social skills.Early Childhood Research Quarterly, 3,21-37.

Hymel, S., & Rubin, K. (1985). Cbildren witb peerrelationsbip and social skills problems: Con-ceptual, metbodological, and developmentalissues. In G. J. Wbiteburst (Ed.), Annals ofchild development (Vol. 2, pp. 251-297).Greenwicb, CT: JAI.

Kistner, J., Metzler, A., Gatlin, D., & Risi, S.(1993). Classroom racial proportions and cbil-dren's peer relations: Race and gender ef-fects. Journal of Educational Psychology, 85,446-452.

Ladd, G. (1983). Social networks of popular, aver-age, and rejected cbildren in scbool settings.Merrill-Palmer Quarterly, 29, 283-307.

Leaper, C. (Ed.). (1994). New directions for childdevelopment: Vol. 65. Childhood gender seg-regation: Causes and consequences. SanFrancisco: Jossey-Bass.

Maccoby, E. E. (1990). Gender and relationsbips:A developmental account. American Psychol-ogist, 45, 513-520.

Maccoby, E. E., & Jacklin, C. N. (1987). Gendersegregation in cbildbood. In H. W. Reese(Ed.), Advances in child development and be-havior (pp. 239-288). New York: AcademicPress.

Martin, C. L. (1994). Cognitive influences on tbedevelopment and maintenance of gender seg-

2286 Child Development

regation. New Directions for Child Develop-ment, 65, 35-51.

Miller, S. M. (1975). Effects of maternal employ-ment on sex role perception, interests, andself-esteem in kindergarten girls. Develop-mental Psychology, 11, 405-406.

Parker, J. G., & Asher, S. R. (1993). Friendshipand friendship quality in middle childhood:Links with peer group acceptance and feel-ings of loneliness and social dissatisfection.Developmental Psychology, 29, 611-621.

Savin-Williams, R. G., & Berndt, T. J. (1990).Friendship and peer relations. In S. S. Feld-man & G. R. Elliott (Eds.), At the threshold(pp. 277-307). Gambridge: Gambridge Uni-versity Press.

Signorella, M. L., & Liben, L. S. (1985). Assessingchildren's gender-stereotyped attitudes. Psy-chological Documents, 15, 7.

Singleton, L. G., & Asher, S. R. (1977). Peer pref-erences and social interaction among third-grade children in an integrated school district.Journal of Educational Psychology, 69,330-336.

Sroufe, L. A., Bennett, G., Englund, M., Urban,J., & Shulman, S. (1993). The significance ofgender boundaries in preadolescence: Gon-temporary correlates and antecedents ofboundary violation and maintenance. ChildDevelopment, 64, 455-466.

Stevenson, M. R., & Black, K. N. (1988). Paternalabsence and sex-role development: A meta-analysis. Child Development, 59, 793-814.

Sullivan, H. S. (1953). The interpersonal theoryof psychiatry. New York: Norton.

Thome, B. (1986). Girls and boys together . . . butmostly apart: Gender arrangements in ele-mentary schools. In W. W. Hartup & Z. Rubin(Eds.), Relationships and development (pp.167-184). Hillsdale, NJ: Erlbaum.

Toman, W. (1976). Family constellation (3d ed.).New York: Springer.

Wilson, M. N. (1989). Ghild development in thecontext of the black extended family. Ameri-can Psychologist, 44, 380-385.