Ca

Aa

b

c

d

a

ARRA

KEPSF

i(it2a11(cc&p

taHf

UQ

((

0d

Early Childhood Research Quarterly 27 (2012) 137– 146

Contents lists available at ScienceDirect

Early Childhood Research Quarterly

oncurrent and longitudinal links between children’s and their friends’ antisocialnd prosocial behavior in preschool

reana R. Eiversa,b,∗, Mara Brendgenc, Frank Vitarod, Anne I.H. Borgea

Department of Psychology, University of Oslo, P.O. Box 1094, Blindern, 0317 Oslo, NorwaySchool of Psychology and Counselling, Queensland University of Technology, Kelvin Grove Campus, Victoria Park Rd., Kelvin Grove, Qld. 4059, AustraliaDépartement de psychologie, Université du Québec à Montréal, C.P. 8888 succursale Centre-ville, Montréal, Québec, Canada H3C 3P8University of Montreal, Groupe de recherche sur l’inadaptation psychosociale chez l’enfant (GRIP), 3050 blvd. Edouard Montpetit, Montréal, Québec, Canada H3T 1J7

r t i c l e i n f o

rticle history:eceived 26 March 2010eceived in revised form 16 May 2011ccepted 23 May 2011

a b s t r a c t

Concurrent and longitudinal links between children’s own and their nominated best friends’ antisocialand prosocial behavior were studied in a normative sample of 3–5-year-olds (N = 203). Moderating effectsof age and gender were also explored. Subscales of the Strength and Difficulties Questionnaire (SDQ)were used to obtain teacher ratings of behavior for each target child and his/her nominated best friends.

eywords:arly childhoodeer influenceocial adjustmentriendship

Nomination of best friends with higher levels of antisocial behavior and lower levels of prosocial behaviorwas concurrently linked to more antisocial behavior in boys. Nomination of highly prosocial best friendswas concurrently linked to more prosocial behavior in both boys and girls. However, the study found nolongitudinal effects of best friends’ behavior on target child’s behavior over a one-year period. A group ofchildren who nominated no best friends at T1 were generally perceived as less prosocial, but not moreantisocial, than other children.

The important role of peers in shaping and socializing behav-or and adjustment throughout childhood is widely recognizedBagwell, Newcomb, & Bukowski, 2000; Ladd, 1999). Peers maynfluence one another by acting as agents of mutual socializa-ion (Aboud and Mendelson, 1996; Ladd, Herald, & Andrews,006), by reinforcing already existing similarities in social cognitionnd behavior (Burleson, 1994; Kupersmidt, DeRosier, & Patterson,995; Poulin et al., 1997; Rubin, Lynch, Coplan, & Rose-Krasnor,994), and by providing social support to one another when neededDunn, 2004; Laursen, Bukowski, Aunola, & Nurmi, 2007). When ahild chooses a friend, it is the mark of a voluntary desire to be espe-ially affiliated with and liked by that person (Bukowski, Newcomb,

Hartup, 1996; Dunn, 2004). Consequently, close friends may exertarticular influence on a child’s social development.

Early theories regarding children’s social development doubtedhe capacity of preschool children to develop stable and identifi-

ble friendships (Mannarino, 1995; Selman, 1971; Sullivan, 1965).owever, observations of young children in social settings have

ound that even toddlers show preferences for certain playmates

∗ Corresponding author at: School of Psychology and Counselling, Queenslandniversity of Technology, Kelvin Grove Campus, Victoria Park Rd., Kelvin Grove,ld. 4059, Australia. Tel.: +61 7 3138 4728; fax: +61 7 3138 0486.

E-mail addresses: a.eivers@qut.edu.au, areana eivers@yahoo.com.auA.R. Eivers), brendgen.mara@uqam.ca (M. Brendgen), frank.vitaro@umontreal.caF. Vitaro), a.i.h.borge@psykologi.uio.no (A.I.H. Borge).

885-2006/$ – see front matter © 2011 Elsevier Inc. All rights reserved.oi:10.1016/j.ecresq.2011.05.001

© 2011 Elsevier Inc. All rights reserved.

over others, form relatively stable affiliative networks, and engagein different kinds of behavior with preferred versus non-preferredpeers (Boivin, Vitaro, & Poulin, 2005; Hay and Cook, 2007; Howesand Lee, 2006; Strayer and Santos, 1996). Whilst there is now alarge body of literature concerned with the nature and character-istics of friendship in early childhood and its outcomes (Howes,2009; Parker, 1986), most research regarding the concurrent andlongitudinal associations between their friends’ and children’s ownbehavior has focused on older children and adolescents (see Vitaro,Boivin, & Bukowski, 2009). Notable exceptions to this are describedin the following sections.

1. Links between friends’ and children’s own behavior:existing research with young children

Research on peer affiliation and the potential influence of friendsin early childhood provides evidence of both behavioral similarityand mutual socialization within groups of affiliated peers of this age(Boivin et al., 2005; Farver, 1996; Hanish, Martin, Fabes, Leonard,& Herzog, 2005; Snyder, Horsch, & Childs, 1997; Snyder et al.,2005). For example, a cross-sectional study examining associationsbetween aggressive behavior and social networking in a sample of

4-year-olds observed that children formed social “cliques” accord-ing to similarities in levels of aggressive activity, social competence,and behavioral style (Farver, 1996). These findings provide impor-tant evidence of behavioral homophily among affiliated peers, even

1 Resea

ipnsoaset3caaestei(

sabCttafiCtsr(aaieppqwtl

ymitipasaadiiFt

2b

e

38 A.R. Eivers et al. / Early Childhood

n very young children, but the cross-sectional nature of the datarecludes any conclusions with regard to the longer term sig-ificance of such homophily. However, at least two longitudinaltudies have provided evidence of short-term longitudinal effectsf aggressive or externalizing peers on preschool-aged children. Inn observational study of preschoolers’ peer affiliations and aggres-ive behavior, Snyder et al. (1997) found that association withither moderately or substantially aggressive peers was relatedo increases in both observed and teacher-rated aggression over a-month period. In comparison, those who had only minimal asso-iation with aggressive peers at Time 1 showed no change in theirggressive behavior over the same period. Snyder and colleagueslso observed that both aggressive and non-aggressive preschool-rs tended to actively seek out and affiliate with peers who had aimilar behavioral style. Another study (Hanish et al., 2005) foundhat preschool children who spent more time interacting withxternalizing peers – based on observed behavior – showed greaterncreases in teacher-rated aggression over the course of a semesterSeptember to December).

Other studies of the effects of aggression on children’s friend-hip patterns in early childhood indicate that aggressive childrenre less well-liked, have fewer mutual friends and have less sta-le friendships than nonaggressive children (Burr, Ostrov, Jansen,ullerton-Sen, & Crick, 2005; Johnson and Foster, 2005). Impor-antly, however, this pattern appears to differ according to theype of aggression under study, with some evidence that relationalggression may in fact be associated with having more mutualriends (Burr et al., 2005). Moreover, gender appears to play a rolen moderating this association (Burr et al., 2005; Sebanc, 2003).ollectively, these studies of preschool-aged children suggest thathe impact of social behavior on processes of peer affiliation andocialization by close peers is already evident in the earliest peerelationships. Yet even the longitudinal findings of Snyder et al.1997) were based on a period of only three months. Preschoolersre subject to a great degree of influence from other sources, suchs the regulation of parents and teachers, and the effects of peers’nfluence may thus be unstable or short-lived at this age (Boivint al., 2005; Ladd, 1992). There is a need, therefore, to study theotential effects of preschoolers’ friendship affiliations over longereriods of time in order to establish their significance for subse-uent behavior development. Hence, the current study investigatedhether associations between preschoolers’ behavior and that of

heir nominated best friends could be found both concurrently andongitudinally, after a one year interval.

Most of the aforementioned studies of friends’ influence onoung children employed either observational or teacher-reportethods for determining friendship affiliation. However, child

nterviews and peer nominations have been employed in otherypes of studies on similar themes, for example in several stud-es examining qualities of friendship and their associations withositive and negative social behaviors (Burr et al., 2005; Johnsonnd Foster, 2005; Sebanc, 2003). Children’s self reports of friend-hip have also been used effectively in studies of friends’ influencemong older children (Adams, Bukowski, & Bagwell, 2005; Berndtnd Keefe, 1995; Mrug, Hoza, & Bukowski, 2004). Asking young chil-ren whom of their peers they consider to be their friends adds an

mportant, subjective dimension to understanding friendship affil-ations in early childhood that is not gained through observation.or this reason, the current study used individual child interviewso obtain information regarding participants’ friendship relations.

. Links between friends and children’s own behavior:

eyond aggression

Most research has been concerned with the putative influ-nce of friends’ behavioral characteristics on children’s aggressive

rch Quarterly 27 (2012) 137– 146

behavior, or on related externalizing difficulties such as antisocialand delinquent behavior (Bowker, Rubin, Rose-Krasnor, & Booth-LaForce, 2007; Snyder, 2002; Snyder et al., 2005; Vitaro, Brendgen,& Tremblay, 2000). A smaller body of research has also beenconcerned with the predictive effect of friends’ behavioral charac-teristics on children’s levels of internalizing behavior (Bagwell et al.,2000; Brendgen, Lamarche, Wanner, & Vitaro, 2010; Brendgen,Vitaro, & Bukowski, 2000; Pedersen, Vitaro, Barker, & Borge, 2007).In contrast, relatively little attention has been given to the poten-tially positive influence of friends’ prosocial behavior. However, onestudy by Sebanc (2003) found that prosocial behavior was a signif-icant predictor of supportive friendships among preschoolers, andsupportive friendships have, in turn, been found to lead to positivesocial outcomes in older children (Berndt, Hawkins, & Jiao, 1999).

Hay and Cook (2007) divide prosocial behavior into threedomains: feeling for another (friendliness, affection, empathic con-cern), working with another (cooperative activity and goal-setting,sharing resources, helping another to accomplish tasks) and min-istering to another (nurturing, comforting, providing resources,responding to another’s wishes and needs). Researchers haveincreasingly established the importance of prosocial behavior forchildren’s social and emotional adjustment and, more particularly,their peer relations (Arsenio and Lemerise, 2001; Nelson, Robinson,& Hart, 2005; Phillipsen, Bridges, McLemore, & Saponaro, 1999;Vitaro, Gagnon, & Tremblay, 1990). Furthermore, prosocial behav-ior explains unique variance in children’s peer relations over andabove that explained by antisocial behavior. This is shown in astudy by Vitaro et al. (1990), who found that stability in peer rejec-tion from kindergarten to grade one was more dependent uponrejected children’s lack of prosocial behavior than upon their lev-els of antisocial behavior. Another study, investigating longitudinalassociations between overt and relational aggression, prosocialbehavior and social adjustment, observed that prosocial behaviorcontributed unique information (beyond that provided by overt andrelational aggression) to the prediction of future social adjustment(Crick, 1996).

In one of the few studies to examine the putative influence ofhighly prosocial friends, Barry and Wentzel (2006) found a longitu-dinal association between friends’ and adolescents’ own prosocialbehavior. Specifically friends’ prosocial behavior was related toadolescents’ prosocial behavior one year later, albeit not concur-rently. To our knowledge, however, there have been no studies ofthe direct predictive effects of friends’ prosocial behavior on chil-dren’s own prosocial behavior in early childhood. Furthermore, wedo not know how the prosocial behavior of friends might be relatedto levels of antisocial behavior in children or, conversely, whetheraffiliation with antisocial peers is related to children’s expressionof prosocial behavior. Consequently, the current study simultane-ously examined concurrent and longitudinal links between friends’antisocial and prosocial behavior and children’s own antisocial andprosocial behavior.

3. Gender and age differences

Some studies have found gender differences in rates of anti-social, primarily aggressive behavior among preschool children.Specifically, boys have been found to engage in more physicalaggression than girls (Crick et al., 2006; Juliano, Werner, & Cassidy,2006; Romano, Tremblay, Boulerice, & Swisher, 2005). There isalso some, but far more limited evidence that girls in this agegroup show more prosocial behavior than boys (Van Leeuwen,Meerschaert, Bosmans, De Medts, & Braet, 2006). Previous stud-

ies have further suggested that gender may play a moderating rolein the link between the behavior of the peers children affiliatewith and children’s own social behavior. For example, in Han-ish et al.’s (2005) study of the effects of observed exposure to

Resea

efpaas(aclilb

eficpsf

4

tclfmmsaltfg

ibn1alitdpirwfchifa

Hdmbcttg

A.R. Eivers et al. / Early Childhood

xternalizing peers during early childhood, different effects wereound for girls versus boys. Whilst exposure to externalizing peersredicted multiple problems in girls – higher levels of teacher-ratedggressive behavior, hyperactivity and anxiety – it was only associ-ted with higher levels of teacher-rated aggression in boys. Anothertudy, conducted with children in the first two grades of schoolEstell, Cairns, Farmer, & Cairns, 2002), showed that affiliating withggressive peers was associated with aggression in boys, both con-urrently and one year later. In contrast, girls’ peer affiliations andevels of overt aggression were unrelated. In light of these find-ngs, child gender was examined as a potential moderator of theink between friends’ and children’s own antisocial and prosocialehavior in the current study.

In addition to gender-related differences, age-related differ-nces may also impact upon the quantity and quality of children’sriendships, leading to differences in the nature and extent of theirnfluence. Developmental differences among 3–5-year-olds can beonsiderable and may thus have an impact on the extent to whicheers reinforce or shape behavior. With this in mind, the currenttudy included age as a potential moderator of the link betweenriends’ and children’s own social behavior.

. The present study

To summarize, the first aim of the current study was to comparehe prosocial and antisocial behavioral outcomes for 3–5-year oldhildren whose nominated best friends demonstrate high versusow levels of prosocial and antisocial behavior. The term “bestriends” – defined as those peers whom children like to play with

ost – was used to distinguish close friends from playmates whoay be referred to generically as “friends” in some group day care

ettings. Associations between nominated best friends’ behaviornd children’s own behavior were examined both concurrently andongitudinally (i.e., after a one-year interval). The second aim waso investigate whether any links found between nominated bestriends’ behavior and children’s own behavior were moderated byender or by age.

Based on the findings of extant literature, we expected signif-cant links between nominated best friends’ and children’s ownehavior. Specifically, it was predicted that children who nomi-ated best friends with higher levels of antisocial behavior at Time

(T1) would themselves be rated as more antisocial at T1. It waslso expected that children who nominated best friends with higherevels of antisocial behavior at T1 would show a greater increasen their own antisocial behavior one year later (T2), compared tohose who nominated less antisocial best friends. Moreover, chil-ren with antisocial best friends at T1 were expected to be lessrosocial at T1 and to show a greater decrease in prosocial behav-

or from T1 to T2 than other children. Although previous findingsegarding the potential effects of prosocial friends are sparse, itas expected that children who nominated highly prosocial best

riends at T1 would be rated as more prosocial and less antiso-ial themselves at T1. It was further predicted that children withighly prosocial best friends at T1 would show a greater increase

n prosocial behavior, and a greater decrease in antisocial behaviorrom T1 to T2, compared to those with less prosocial best friendst T1.

In accordance with results of previous studies (Estell et al., 2002;anish et al., 2005), both main and moderating effects of gen-er were expected. Overall, boys were expected to be rated asore antisocial and girls as more prosocial. Furthermore, the link

etween their nominated best friends’ antisocial behavior at T1 and

hildren’s own antisocial behavior at both T1 and T2 was predictedo be stronger for boys than for girls. Finally, it was also expectedhat effects would be weaker among children in the younger ageroup compared to older children given that the stability of friend-

rch Quarterly 27 (2012) 137– 146 139

ships, and hence the potential strength of their influence, increasesas children develop toward middle childhood.

5. Method

5.1. Sample

The current study was part of a larger, ongoing longitudinalstudy investigating the importance of friendship and peer rela-tions in early childhood for subsequent behavioral, emotional andsocial adjustment. All children from 33 day care centers, across twosemi-rural Norwegian communities – situated 30–90 km north ofNorway’s largest city, Oslo – were invited to participate in the study.Out of an eligible 996 children, 619 children (62% participation rate)were given the active, written consent of their parents/guardiansto participate. Ethical clearance for the study was obtained fromthe respective Norwegian Regional Ethics Committee and the Nor-wegian Data Inspectorate. As is common in Norwegian semi-ruralareas, the sample was exclusively White Caucasian. The averagehousehold income for participants in the current study (approx.$96,684USD) was lower than the national Norwegian average forhouseholds with children aged 0–5 years (approx. $114,733USD),but slightly higher than the regional average for the communi-ties sampled (approx. $93,183USD) (Statistics Norway, 2010b). Theaverage maternal education level of mothers in the study washigher than the Norwegian national average: 45% of mothers inthe current sample had completed some form of tertiary educa-tion compared with 28% in Norway as a whole in 2007 (StatisticsNorway, 2010a). This may reflect a possible social selection biasamong parents choosing to send their children to group day care.Furthermore, average levels of education in Norway are higher thanthe mean for Organization for Economic Co-operation and Devel-opment (OECD) countries (OECD, 2010b). Over 90% of children inNorway aged six years and under are enrolled in formal day care,and between the ages of three and five years, children in Norway areexpected to spend an average of 2.7 years in a formal child care oreducation setting. These rates are also above the average for OECDcountries (OECD, 2010a).

To be eligible for the current study, children needed to be agedthree years or over at T1 and still in day care (i.e., not have startedschool) one year later, at T2. Of the 290 children who met thesecriteria, 209 children met the additional inclusion criterion of hav-ing complete teacher data at both T1 and T2. Six children were laterexcluded from the analyses on the basis of their friendship nomina-tions (see section on measurement of best friends’ behavior below),bringing the final sample size to N = 203 (50% female). Participantsranged in age from 36 to 64 months (M = 50, SD = 7.84). There was nosignificant difference between children with and without completeteacher data at both time points in terms of age, t(282) = 1.78, p = .08,or gender, �2(1, 203) = 3.07, p = .08. There were also no significantdifferences between children included in the study versus excludedbecause of missing teacher data with regard to T1 outcome vari-ables, i.e., teacher ratings of antisocial behavior, t(258) = .26, p = .80,and prosocial behavior, t(258) = .24, p = .81, (see description of mea-sures below). The median age at which children who participatedin the study had started in group day care was 19 months; approx-imately 80% of children had been attending group day care sincethe age of 30 months. Among the youngest children in the sample,those aged 36–42 months, over 90% had begun attending group daycare at least six months prior to T1.

5.2. Procedure

Children were interviewed individually in their day care center.The interviews were conducted by interviewers previously trained

1 Resea

itiitofrpdtas

tctfb

5

5

erfaed3aea(tcfi2amc“p“si

wtsppKtSndd(SSyac

40 A.R. Eivers et al. / Early Childhood

n the use of the instrument. Interviews took, on average, 20–25 mino complete. Children were withdrawn from their current activ-ty for this period and interviewed alone in a separate room, orn a quiet area away from other children. Participants respondedo the nomination items in the interview orally and by pointingut individuals from photographs of their day care group placed inront of them. Ethical guidelines applied to the study did not allowesearchers to take photographs of the children and so individualhotographs of children (faces and upper torso) were provided byay care centers. To check their understanding of the nominationask, children were first asked to identify their own photographmong those placed before them. The same procedure and mea-ures were employed 12 months later, at T2.

Teacher data were gathered via questionnaire. Day care cen-er teachers were asked to rate the behavior of each participatinghild in their group/class, with an average of eight children pereacher. Again, the same procedure and measures were employedor the teacher ratings at T2. The teachers providing ratings of childehavior were all female.

.3. Measures

.3.1. Child prosocial and antisocial behaviorBecause teachers were thought to have better access than par-

nts to children’s behavior in peer-oriented social contexts, teacheratings were used to evaluate children’s and their nominated bestriends’ behavior. To this end, a teacher version of the Strengthsnd Difficulties Questionnaire (SDQ; Goodman, 1997, 2001) wasmployed. The SDQ is a brief behavioral screening instrumentesigned to be completed by parents or teachers of children aged–16 years. Other studies report the use of the SDQ with preschool-ge samples (Hughes, White, Sharpen, & Dunn, 2000; Van Leeuwent al., 2006) and a Norwegian translation of the instrument wasvailable via previous studies conducted with Norwegian childrenHeiervang, Goodman, & Goodman, 2008; Obel et al., 2004). Onlywo of the five SDQ subscales were used in the current study: Proso-ial Behavior and Conduct Problems. Each SDQ subscale consists ofve items measured on a three-point scale—1 “does not apply,”

“applies somewhat,” and 3 “certainly applies.” Individual itemsre summed to create a subscale score with a minimum of 5 and aaximum of 15. The Prosocial subscale includes items such as, “is

onsiderate of others’ feelings,” “shares with other children,” andoften offers to help others.” The Conduct Problems subscale com-rises items including “often has temper tantrums or bad mood,”fights often with other children,” and “lies or cheats often.” Thisubscale can thus be perceived as a measure of antisocial behav-or.

Cronbach’s internal consistency coefficients of ̨ = .75 and ̨ = .83ere obtained for antisocial and prosocial behavior scales, respec-

ively, at T1; ̨ = .73 and ̨ = .84 for antisocial and prosocial behaviorcales, respectively, at T2. Descriptive statistics for target childrosocial behavior and antisocial behavior scores were as follows:rosocial behavior scores at T1 (M = 11.41, SD = 2.28, Skewness = .02,urtosis = −.66) and T2 (M = 11.29, SD = 2.40, Skewness = −.12, Kur-

osis = −.53); antisocial behavior scores at T1 (M = 5.61, SD = 1.35,kewness = 3.43, Kurtosis = 15.33) and T2 (M = 5.83, SD = 1.54, Skew-ess = 2.34, Kurtosis = 6.00). To correct for positive skew in theistribution of target child antisocial behavior scores at T1, theseata were transformed using an inverse and reflect procedureTabachnick and Fidell, 2007). The transformed data (T1: M = .82,D = .03, Skewness = 1.65, Kurtosis = 2.80; T2: (M = .82, SD = .03,

kewness = 1.40, Kurtosis = .80) was used in all subsequent anal-ses involving antisocial behavior scores. Target child scores forntisocial and prosocial behavior were significantly and negativelyorrelated at T1, r = −.31, p < .01, and T2, r = −.37, p < .01. T1 and

rch Quarterly 27 (2012) 137– 146

T2 antisocial behavior scores were significantly correlated, r = .52,p < .01, as were T1 and T2 prosocial behavior scores, r = .58, p < .01.

5.3.2. Children’s friendship nominationsDuring the individual child interview at T1, each participant was

first asked if they had any friends at day care. Those who respondedin the affirmative were asked to identify their friends using theprocedures described earlier, i.e., by pointing out their friend(s)from photographs of their day care group placed in front of themand naming their friend(s) at day care. Children were then asked ifthey had any “best friends” at day care, defined as friends “you liketo play with the most.” Drawing on Howes (1987) developmentalmodel, this definition was used because it was anticipated that theyounger children in the sample may have difficulty distinguishingclose friends from playmates unless cued to do so. Moreover, inthe context of group day care, terms such as “friends” and “play-mates” may be used interchangeably by staff so that the distinctionis unclear to young children unless prompted to consider whomof their peers they like or play with the most. Participants whoresponded positively were asked to identify their best friend(s),by pointing to the friend’s photograph and naming the friend. Chil-dren’s best friend nominations were used in subsequent analyses asthese were thought to represent the peers with whom participantsaffiliated most strongly.

Children could nominate up to four best friends. Best friendscould be of either gender as there is evidence from previousresearch that cross-sex friendships in early childhood are common(Howes and Phillipsen, 1992). Of a total 278 best friend nomina-tions, 82 (29%) were for children of the opposite sex and 32 children(12%) nominated only opposite sex friends. Children were also freeto nominate best friends from their day care group who were notparticipants in the study. As even the non-participating children ineach day care center were assigned a unique identification number,‘non-participant’ friends were included in the calculation of howmany friendship nominations were made per target child. Therewere 160 children (76% of the total sample at T1) who nominatedone or more best friends at T1, whilst 49 children (23% of sample)in the current study did not nominate any best friends at T1. Forty-seven (17%) of a total 278 best friend nominations were for childrenwho did not participate in the study. However, all participating chil-dren nominated at least one other child who was in the study asone of their best friends, i.e., there was valid best friend behaviordata available for each participating child. Children who nominatedno best friends were, on average, significantly younger than thosewith one or more nominated best friends, t(201) = 3.89, p < .001,d = .55. However, there was no significant difference between thetwo groups in terms of gender composition, �2(1,203) = .02, p = .90.With regard to maintenance of nominated best friendships from T1to T2: of the 160 children who nominated at least one best friendat T1, 59 children (37%) nominated at least one of the same bestfriends again at T2, six children nominated two of the same bestfriends.

Using the criterion of mutual nomination, 48 children (23% ofthe sample) who participated in the study had at least one recipro-cal best friendship with another participating child; four of thesechildren had two reciprocal best friendships. This rate of recipro-cated friendship nominations is similar to that obtained in Hayes,Gershman, and Bolin’s (1980) study that established reciprocity offriendship in preschool children through the use of both child andparent interviews and observation of children’s play. However, asthere were no data available for non-participating children in thecurrent study, the rates of reciprocated best friendship should be

treated with caution, i.e., a participating child may indeed have hada mutually recognized best friendship, but with a child who did notparticipate and hence for whom nomination data were not avail-able. In such cases, friendship reciprocity could not be established

Resea

bn

5

suto1fiiocaesApifcmcbnwnctoi

iciortStwsSmdbaihstvebfpfgs(ac(n

A.R. Eivers et al. / Early Childhood

ased on the available best friendship nominations and so couldot be included in the count of reciprocated best friendships.

.3.3. Nominated best friends’ behaviorDue to the low rate of reciprocated best friendship in the current

tudy, the construct of friends’ behavior was based on participants’nilateral best friend nominations. Indeed, some theorists arguehat many preschool children do not yet have an understandingr appreciation of reciprocity in friendship (Furman and Bierman,983; Gottman and Graziano, 1983; Selman, 1971). Thus, unilateralriendships may well reflect young children’s conceptions of whats important in friendship (e.g., physical proximity, the friend hav-ng characteristics or possessions that the child likes). One studyf the development of children’s friendships found that preschoolhildren’s experiences of unilateral versus reciprocal friendshipsre in many ways similar (Hayes et al., 1980). Moreover, the unilat-rally nominated friendships of older children have been shown toignificantly predict children’s own behavior (Adams et al., 2005;loise-Young, Graham, & Hansen, 1994; Mrug et al., 2004). For theurposes of the current study, a measure of best friends’ behav-

or was obtained by calculating the mean of their nominated bestriends’ antisocial and prosocial behavior scores for each targethild. For example, for a child who nominated three friends, theean of the nominated best friends’ behavior scores would be cal-

ulated by first adding the scores on a particular subscale of all threeest friends and then dividing the total score by three. It should beoted that in some cases, where children nominated best friendsho were not participants in the study, data regarding one or moreominated best friends may not have been available. Six pairs ofhildren exclusively nominated one another as best friends. Hence,o avoid redundancy in the calculation of best friends’ behavior,nly one child from each of these pairs was randomly chosen to bencluded in the analyses.

The mean scores for best friends’ antisocial and prosocial behav-or were converted to nominal variables comprising two categories:hildren for whom the mean of their nominated best friends’ behav-or scores was in the low range, and children for whom the meanf their nominated best friends’ behavior scores was in the highange. In the case of prosocial behavior, which was normally dis-ributed (M = 11.66, SD = 2.04), the cut-off between low (M = 9.91,D = 1.05) and high (M = 13.27, SD = 1.24) categories was based onhe median score (11.50). Due to the low level of behavior problemsithin our normative sample, means for nominated best friends’

cores on the antisocial measure were positively skewed (M = 5.48,D = 1.06), with the minimum value of the subscale accounting forore than 60% of scores. Hence, for this subscale, the value imme-

iately above the minimum score (5.33) was used as the cut-offetween groups. As a consequence of the positive skew, the meanntisocial behavior scores of those nominated best friends fallingnto the ‘high antisocial’ (M = 6.48, SD = 1.36) category were notigh in absolute terms, but only relative to those in the ‘low anti-ocial’ (M = 5.00, SD = .14) category. For this reason, in contrast tohe nominated best friends’ prosocial behavior categories of ‘low’ersus ‘high’, nominated best friends’ antisocial behavior was cat-gorized as ‘low’ versus ‘moderate’. Categorization of best friends’ehavior in this way allowed for the further classification of bestriends’ behavior according to concomitant levels of antisocial androsocial behavior. Moreover, children who nominated no bestriends could also be included in the analyses as a comparisonroup. Thus, the resultant independent variable created to repre-ent nominated best friends’ T1 behavior consisted of five groups:1) moderate antisocial/high prosocial best friends, (2) moderate

ntisocial/low prosocial best friends, (3) low antisocial/high proso-ial best friends, (4) low antisocial/low prosocial best friends, and5) no nominated best friends. Of the total sample: 8.9% (n = 18)ominated moderate antisocial/high prosocial best friends; 16.3%

rch Quarterly 27 (2012) 137– 146 141

(n = 33) nominated moderate antisocial/low prosocial best friends;30.5% (n = 62) nominated low antisocial/high prosocial friends;20.2% (n = 41) nominated low antisocial/low prosocial friends; and24.1% (n = 49) nominated no friends.

6. Results

6.1. Analytic strategy

Associations between nominated best friends’ behavior at T1 –including those who nominated no best friends at T1 as an addi-tional comparison group – and target child’s behavior at T1 wereinvestigated using multivariate analysis of variance (MANOVA),with the target child’s antisocial and prosocial scores entered as mul-tivariate dependent variables. Longitudinal associations betweennominated best friends’ behavior at T1 and changes in target child’santisocial and prosocial behavior from T1 to T2 were also inves-tigated using MANOVA. To this end, the variance in T2 targetchild behavior scores explained by T1 target child scores for thesame behavior was first partialed out by saving the unstandardizedresiduals from a regression of T2 behavior on T1 behavior. Theseresidual values were then entered as the T2 dependent variablesin the subsequent MANOVA. In both the concurrent and longitu-dinal analyses, nominated best friends’ behavior was entered as a5-level between-subjects independent variable (no friends, mod-erate antisocial/high prosocial, moderate antisocial/low prosocial,low antisocial/high prosocial, and low antisocial/low prosocial).Gender and age were also included as between-subjects indepen-dent variables in both equations, with age dichotomized into ayounger (n = 97) versus older (n = 106) group by dividing T1 agein months at the median (51 months). A customized model includ-ing main effects and all two-way interactions was tested. It wasoriginally intended that a three-way interaction between sex, age,and behavior of nominated best friends at T1 be included in themodel. However, a power analysis using a hypothetical effect sizeof .05 for a three-way interaction (estimated based on the effectsize obtained for the significant two-way interaction) indicatedthat a sample size of N = 320 was required to adequately test thethree-way interaction. This interaction was consequently droppedfrom the final analysis. Significant main and interaction effects ofbetween-subject factors were probed using Gabriel’s test of posthoc comparisons. This test was chosen because of its suitability insituations where sample sizes are unequal. All analyses were con-ducted using the Statistical Package for the Social Sciences (SPSS)v.18.

6.2. Effects of nominated best friends’ behavior at T1 on childbehavior at T1

There were no significant multivariate main effects of eithergender or age on child antisocial and prosocial behavior at T1.However, nominated best friends’ behavior at T1 had a significantmultivariate main effect on target child’s antisocial and prosocialbehavior at T1: � = .79, F(8,372) = 5.69, p < .001, �2 = .11. In addi-tion, there was a significant multivariate interaction effect of childgender and nominated best friends’ T1 behavior on target child’sT1 behavior: � = .92, F(8,372) = 2.08, p = .05, �2 = .04. SubsequentANOVAs revealed that this interaction effect only concerned tar-get children’s T1 antisocial behavior, F(4,187) = 3.20, p < .05, �2 = .06.With regard to target children’s T1 prosocial behavior, a main effectof nominated best friends’ T1 behavior was found that was inde-

pendent of child gender, F(4,187) = 10.85, p < .001, �2 = .19.

ANOVAs conducted separately for boys and girls regardingprediction of target children’s T1 antisocial behavior showed a sig-nificant main effect of nominated best friends’ T1 behavior for boys,

1 Resea

FPctonw

Gccap(Thaci

6c

i�AbyTsna

7

ctbatolnawnnwfwmit

7c

rtsfi

42 A.R. Eivers et al. / Early Childhood

(4,91) = 4.68, p < .01, �2 = .17, but not for girls, F(4,92) = .86, p = .49.ost hoc Gabriel tests revealed that boys with moderate antiso-ial/low prosocial friends were significantly more antisocial at T1han boys with low antisocial/high prosocial friends (p < .01). Nonef the remaining mean comparisons with regard to the effect ofominated best friends’ T1 behavior on boys’ antisocial behaviorere significant.

Regarding target children’s T1 prosocial behavior, post hocabriel tests indicated that children with low antisocial/high proso-ial friends at T1 were more prosocial at T1 than any otherhildren in the study, i.e., more prosocial than those with moderatentisocial/high prosocial friends (p < .05), moderate antisocial/lowrosocial friends (p < .001), low antisocial/low prosocial friendsp < .001), and those with no nominated best friends (p < .001) at1. No significant differences in T1 prosocial behavior emerged,owever, between any other groups. The means and standard devi-tions of target children’s T1 antisocial and prosocial behavior byhild gender and nominated best friends’ behavior at T1 are shownn Table 1.

.3. Effects of nominated best friends’ behavior at T1 on change inhild behavior from T1 to T2

There was a significant multivariate main effect of age on changen target child’s antisocial and prosocial behavior from T1 to T2:

= .95, F(2,186) = 5.28, p < .01, �2 = .05. A subsequent univariateNOVA revealed that this effect applied to change in antisocialehavior only: F(1,187) = 10.60, p < .01, �2 = .05. Specifically, theounger group decreased their antisocial behavior slightly from1 to T2 whereas the older group’s antisocial behavior increasedlightly. There were no significant main effects of either gender orominated best friends’ behavior at T1 on change in target child’sntisocial or prosocial behavior from T1 to T2.

. Discussion

The aim of the current study was twofold: (1) to investigate theoncurrent and longitudinal links in 3–5-year old children betweenhe antisocial and prosocial tendencies of the children’s nominatedest friends and children’s own antisocial and prosocial behavior;nd (2) to investigate age and gender as possible moderators ofhese links. As hypothesized, the study found that children’s levelf antisocial and prosocial behavior was significantly related to theevels of antisocial and prosocial behavior demonstrated by theirominated best friends. The type and magnitude of these associ-tions varied according to behavior type. An unexpected findingas the relatively large number of children who did not nomi-ate any best friends at T1. Contrary to our hypothesis, there wereo main effects of gender in the current study. However, genderas found to moderate concurrent links between nominated best

riends’ behavior and target child’s antisocial behavior—this linkas found for boys only. In contrast, age of the target child did notoderate the associations between nominated best friends’ behav-

or and children’s own behavior. A discussion of these findings andheir possible implications follows.

.1. Links between nominated best friends’ behavior and targethild’s antisocial behavior

Boys in the current study who nominated best friends withelatively high (moderate) levels of antisocial behavior tended

o show concurrently higher levels of antisocial behavior them-elves. This was consistent with the results of previous studies thatound young children with aggressive and externalizing behav-or tend to seek out and affiliate mostly with other aggressive or

rch Quarterly 27 (2012) 137– 146

externalizing children (Farver, 1996; Hanish et al., 2005; Snyderet al., 1997). Since there was no main effect of gender (no genderdifference) with respect to antisocial behavior at T1, the gender-specific finding does not seem to be due to a very low level ofsuch behavior in girls. A possible explanation, therefore, couldbe sex-specific socialization: boys have been shown to displaymore aggressive, confrontational behavior in their social interac-tions with friends and with nonfriends than girls (Denton andZarbatany, 1996; Maccoby and Jacklin, 1980). Moreover, girls rateaggressive behavior as more unacceptable than boys, even whendisplayed in response to an aggressive provocation from peers(Goldstein, Tisak, & Boxer, 2002). To the extent that antisocial formsof behavior may be perceived as more acceptable among boys,boys may thus also more readily perceive antisocial peers as anacceptable friendship choice – and emulate these peers’ behavior– than girls. Importantly, the current study found that the effectof more antisocial friends on target boys’ antisocial behavior wasonly significant where there was a concomitant low level of friends’prosocial behavior. Boys whose nominated best friends receivedhigher ratings for both antisocial and prosocial behavior were notperceived by teachers as being more antisocial than other chil-dren. These results are in line with findings from a study by Vitaro,Tremblay, and Bukowski (2000) that the presence of non-antisocialfriends can counterbalance the negative effect of antisocial friendson early adolescents’ own behavior. Nevertheless, there were rela-tively few children in the present study sample whose nominatedbest friends were high in both antisocial and prosocial behavior.Caution is therefore warranted when drawing conclusions regard-ing the comparative influence of antisocial versus prosocial friendswhere both are present. We also do not know whether thesemoderate antisocial/high prosocial friends were individuals whodemonstrated both types of behavior or whether they were sep-arate individuals with different behavior characteristics. Futurestudies should examine the potentially different consequences ofhaving friends who demonstrate relatively high levels of both anti-social and prosocial behavior versus having some friends who arehighly antisocial and others who are highly prosocial.

7.2. Links between nominated best friends’ behavior and targetchild’s prosocial behavior

There is support in the current findings for prosocial homophily.Children who nominated best friends who were highly prosocial atT1 were themselves highly prosocial. The current study is the first tostudy the role of prosocial behavior in young children’s friendships.It was interesting to find, therefore, that highly prosocial children,like antisocial or aggressive children, like to stick together. Impor-tantly, high levels of nominated best friends’ prosocial behavior atT1 were only linked with high levels of prosocial behavior in thetarget child at T1 when friends’ concomitant levels of antisocialbehavior were low. Again, this finding may indicate that divergentsources of social influence compete with each other in their effecton children’s behavior. It is also possible that children who are nei-ther markedly aggressive nor specifically prosocial are more willingto affiliate with friends who possess a wider variety of behavioralcharacteristics than are children with pronounced positive or neg-ative behavior profiles. However, as stated in the previous section,the low rate of moderate antisocial/high prosocial friendship nom-inations makes it difficult to compare the effects of different levelsof best friends’ antisocial behavior in this instance.

7.3. Links between perceived friendlessness and target child’s

behavior

Somewhat unexpectedly, a large group emerged in the cur-rent study comprised of children who did not nominate any best

A.R. Eivers et al. / Early Childhood Research Quarterly 27 (2012) 137– 146 143

Table 1Means and standard deviations for target children’s T1 antisociala and prosocial behavior by child gender and nominated friends’ behavior at T1 (N = 203).

Nominated friends’ behavior at T1 Target child: T1 antisocial behavior Target child: T1 prosocial behavior

M SD M SD

Boys (n = 101)Moderate antisocial/high prosocial friends 5.45 .52 10.82 2.09Moderate antisocial/low prosocial friends 6.74 2.26 10.42 1.90Low antisocial/high prosocial friends 5.07 .47 12.81 2.30Low antisocial/low prosocial friends 5.80 1.51 11.05 1.43No nominated friends 5.96 2.10 10.00 1.96

Girls (n = 102)Moderate antisocial/high prosocial friends 5.57 .98 12.00 2.00Moderate antisocial/low prosocial friends 5.29 .61 11.00 2.25Low antisocial/high prosocial friends 5.34 .84 12.91 1.87Low antisocial/low prosocial friends 5.76 1.30 10.62 2.38No nominated friends 5.36 .81 11.16 2.29

Total sampleModerate antisocial/high prosocial friends 5.50 .71 11.28 2.08Moderate antisocial/low prosocial friends 6.12 1.88 10.67 2.04Low antisocial/high prosocial friends 5.23 .71 12.87 2.05

orme

fpsftwpmsfstbncPdnbofdaurasyogmc

tcpbosdyf(

Low antisocial/low prosocial friends 5.78

No nominated friends 5.65

a Note. Original antisocial behavior scores are presented in table. However, transf

riends at T1. This category of children has not been reported inrevious studies of preschool or older children. Its presence in thistudy could be due to participants being asked to identify “best”riends, a concept that implies a friendship quality and intensityhat some young children may be unable or unwilling to identifyith. The relative instability of friendships during the preschooleriod (Newcomb and Bagwell, 1995), compared to later childhood,ight also have resulted in some children being ‘between’ friend-

hips at the time they were interviewed. They may also have been inriendships that were, at that stage, too brief and fleeting to be con-idered “best” friendships. Importantly, the ‘friendless’ category inhe current study did not necessarily imply that a child was rejectedy her/his peers. Rather, it indicated that s/he was unable or choseot to nominate anyone as a best friend. Thus, in addition to rejectedhildren, this group may include children referred to by Coplan,rakash, O’Neil, and Armer (2004) as socially disinterested: chil-ren who are satisfied playing on their own and who neither needor seek out the company of close friends. Alternatively, these maye shy or socially inhibited children who find it difficult to approachthers and who are largely excluded from friendship networks. Theact that there were more of the younger children than older chil-ren within the ‘friendless’ category also suggests a developmentalspect to the findings. Perhaps the concept of “best friend” is lessnderstood by or less relevant to younger preschoolers, althoughesearch has shown that even toddlers show signs of selective peerffiliation (Dunn, 2004; Hay and Cook, 2007; Howes, 1996). A pos-ible, alternative explanation for the age difference is simply thatounger children have had less time in day care and hence lesspportunity to develop strong affiliations. Nevertheless, the emer-ence of this group in the current study raises questions about theeanings and consequences of peer affiliation, or lack thereof, for

hildren who do not perceive themselves as having close friends.In the current study, perceived friendlessness at T1 was linked

o less prosocial behavior, but not concurrently to more antiso-ial behavior, when compared to children with low antisocial/highrosocial friends. In terms of these findings, this group of childrenehaves similarly to the “friendless” children described in studiesf friends’ influence among older children and adolescents. In onetudy, early adolescents with deviant friends were found to be more

elinquent than youth with non-deviant friends and friendlessouth. However, there was no difference in delinquency betweenriendless adolescents and adolescents with non-deviant friendsBrendgen et al., 2000). Similarly, another study found that chron-

1.39 10.83 1.961.59 10.59 2.19

d data (inversed and reflected) were used in analyses.

ically friendless children in grade 5 demonstrated lower levels ofpeer-reported social competence and prosocial behavior than otherchildren, but were not viewed by their peers as more aggressive(Wojslawowicz Bowker, Rubin, Burgess, Booth-Laforce, & Rose-Krasnor, 2006). Children without close friends are perhaps lessantisocial due to a lack of peer modeling and/or reinforcement ofantisocial behavior. At the same time, given that the current find-ings suggest that prosocial preschoolers affiliate with one another,a lack of highly prosocial friends may be a result of these chil-dren engaging in less prosocial behavior themselves. However,the extent to which adjustment outcomes for young children whonominate no best friends are the same as for children regarded asfriendless on the basis of other measures (e.g., observation, peer andteacher report) is not clear from the current findings, and requiresfurther investigation.

7.4. Longitudinal influences of nominated best friends’ behavior

The current study did not find any significant longitudinal effectsof nominated best friends’ behavior at T1 on change in targetchild’s antisocial or prosocial behavior one year later. This lackof apparent socializing influence contrasts with previous stud-ies by Snyder et al. (1997) and Hanish et al. (2005) that foundmore frequent association with aggressive and externalizing peers,respectively, predicted increases in observed and teacher-ratedaggression among preschoolers over a 3-month period. There areseveral possible reasons for the lack of longitudinal findings inthe current study. One is the length of time between the originaland follow-up assessments. There have been no previous studiesof young children that examined friends’ influence over periodslonger than three months. The instability of young children’s friend-ships relative to older children (Newcomb and Bagwell, 1995), andthe finding of the current study that only one third of the bestfriends nominated at T1 were nominated by the same childrenagain at T2, suggests that young children do not maintain a con-sistent perception of their close friendships over long periods oftime. This may result, potentially, in a dilution of the long-terminfluence of specific relationships.

Furthermore, with regard to the more antisocial children in our

sample at least, research indicates that the friendships of youngchildren with aggressive tendencies are less stable than those ofnonaggressive children and the impact of specific friends maytherefore be limited (Johnson and Foster, 2005; Snyder et al., 1997).

1 Resea

Icnteopdhdatisld

7

astasprpo1omtesosbhvbm1b

crtywmlwawAntntevvbsr

44 A.R. Eivers et al. / Early Childhood

n addition, the rate of antisocial behavior was generally low in theurrent sample, whilst the rate of prosocial behavior was high. Theumber of highly antisocial children may have been too few andhe level of antisocial behavior too low to exert a strong influence,specially one that persisted over time. Studies of normal devel-pmental progression have also shown that aggressive behavioreaks in early childhood at around 2.5 years and subsequentlyeclines during the preschool years (Tremblay, 2000). This mayave resulted in the natural reduction of antisocial behavior ten-encies of both target children and their nominated best friendscross the one year time period. Conversely, one possible explana-ion for the lack of longitudinal effects of friends’ prosocial behaviors that the generally high rate of prosocial behavior within theample led to a ceiling effect. As a result, there might have beenittle room for improvement among young children who alreadyemonstrated high levels of prosocial behavior.

.5. Limitations and future directions

There are several limitations to the current study, suggestingreas where further research is needed. For instance, the currenttudy used unilateral rather than reciprocated friendship nomina-ions to determine peer affiliation. There is debate in the literaturebout the extent to which very young children are capable of under-tanding or appreciating reciprocity in friendship. Some authorsostulate that, although young children may appear to engage ineciprocal friendships, their conception of reciprocity is limited andossibly less important to their perception of friendship than it is tolder children (Furman and Bierman, 1983; Gottman and Graziano,983; Mannarino, 1995). Moreover, there is evidence from previ-us studies that unilateral friendship nominations tell us just asuch, if not more about the dynamics of peer/friend influence

han studies of reciprocated friendships (Adams et al., 2005; Mrugt al., 2004). Nonetheless, future studies should compare the pos-ible influences of unilaterally and reciprocally nominated friendsn young children’s behavior. It is conceivable, for example, thatignificant longitudinal effects of best friends’ behavior may haveeen found if the rate of mutual (reciprocal) friendships at T1ad been high enough to make studying their particular influenceiable. Reciprocated friendships have a tendency to be more sta-le (Gershman and Hayes, 1983) and young children may spendore time in dyadic interaction with reciprocal friends (Hayes et al.,

980) resulting, potentially, in stronger longitudinal influences onehavior.

The presence in the current study of a relatively large group ofhildren who did not nominate any best friends, despite havingegular access to a stable peer group, supports the argument foraking into account children’s own perceptions when describingoung children’s friendship networks—or lack thereof. However,e do not know to what extent children’s friendship nominationsatch their observable patterns of peer association. The relatively

ow participation rate in the current study, although not unusualithin the context of similar studies (Johnson and Foster, 2005),

lso meant that there was a considerable proportion of childrenhose relationships with those in the study were unaccounted for.lthough all of the target children in the current study who nomi-ated a best friend nominated at least one other study participant,heir perceived friendships with non-participating children couldot be included or considered in the current study’s analysis otherhan to determine the number of best friend nominations made byach child. Thus, future studies that combine self report with obser-ation are warranted. Future studies should also consider similarity

ersus difference in children’s and their friends’ baseline levels ofehavior as a possible moderator of longitudinal effects, as well astudying potential mediating factors of friends’ influence, such aselationship quality and social cognitive processes.

rch Quarterly 27 (2012) 137– 146

The findings of the current study indicate that, even in preschool,children choose close friends with similar behavior characteristics.However, although the possibility that young friends socialize oneanother cannot be ruled out, the findings also indicate that youngchildren’s choice of close friends is fluid. Moreover, the influencethat these friends have on one another’s behavior appears to berelatively weak and not long-lasting and is possibly overshadowedby the stronger socializing influence of parents and teachers. Thereis great potential, therefore, for parents and teachers to monitorand shape young children’s exposure to peers, more so perhapsthan at later points in development (Kokko, Tremblay, Lacourse,Nagin, & Vitaro, 2006). Parents have a direct influence on theiryoung children’s friendship choices by organizing peer play groupsand choosing day care settings. Moreover, parents and other adultssuch as day care teachers can influence especially young children’sfriendships by monitoring peer interactions and offering support,instructions, and guidance (Ladd, 1992).

The indication in the current study that homophily with respectto both antisocial and prosocial behavior begins already at a youngage has potential implications for policy and practice of early child-hood education. Preschool teachers and other day care providershave an opportunity to intervene early when antisocial childrenaffiliate with one other, thereby minimizing the risk of a contin-uation or further increase of such behavior. Yet the current studyalso shows that even when teachers encourage and facilitate affilia-tion between young children with differing behavioral tendencies,the children still show a preference for peers with similar levels ofantisocial and prosocial behavior to their own. Thus, policies thatseek to prohibit antisocial children from associating with otherantisocial children in order to modify their behavior are likely tobe unsuccessful and, furthermore, potentially counterproductive.Children with aggressive and antisocial tendencies find it more dif-ficult to make friends generally and are at greater risk of rejectionby their peers, which may lead to poorer psychosocial adjustmentsubsequently (McDougall, Hymel, Vaillancourt, & Mercer, 2001).Their friendships – or perceived friendships – with other anti-social children may, thus, provide psychosocial benefits, such asprotection from loneliness, that friendless children are deprived of(Vitaro et al., 2009). As the current study showed that nominat-ing both antisocial and prosocial friends seemed to be associatedwith lower levels of antisocial behavior in the nominating child,a more beneficial practice may be to facilitate additional – ratherthan alternative – affiliations with more prosocial children in orderto balance the potential socializing impact of having only antisocialfriends. Jointly, these strategies may be conducive to encouragingand supporting healthy peer relationships among young children.

Acknowledgements

This study was supported by grants from the NorwegianResearch Council (Grant 142592), the Norwegian Ministry of Childand Family Affairs, and the Faculty of Social Sciences at the Uni-versity of Oslo. Additional support was provided by the Fonds deRecherche sur la Société et la Culture to Frank Vitaro and by theFonds de Recherche enanté du Québec via a senior research fellow-ship to Mara Brendgen. These funding bodies were not involved inthe collection, analysis or interpretation of data used in the currentstudy, or in the preparation of the written report. The authors wishto acknowledge the work of interviewers who gathered data foruse in this report and the children, parents and teachers who gavetheir time and attention to our study.

References

Aboud, F. E., & Mendelson, M. J. (1996). Determinants of friendship selection andquality: Developmental perspectives. In W. M. Bukowski, A. F. Newcomb, & W.

Resea

A

A

A

B

B

B

B

B

B

B

B

B

B

B

C

C

C

D

D

E

F

F

G

G

G

G

G

H

A.R. Eivers et al. / Early Childhood

W. Hartup (Eds.), The company they keep: Friendship in childhood and adolescence(pp. 87–112). Cambridge, U.K.: Cambridge University Press.

dams, R. E., Bukowski, W. M., & Bagwell, C. (2005). Stability of aggression dur-ing early adolescence as moderated by reciprocated friendship status andfriend’s aggression. International Journal of Behavioral Development, 29, 139–145.doi:10.1080/01650250444000397

loise-Young, P. A., Graham, J. W., & Hansen, W. B. (1994). Peer influence on smok-ing initiation during early adolescence: A comparison of group members andgroup outsiders. Journal of Applied Psychology, 79, 281–287. doi:10.1037/0021-9010.79.2.281

rsenio, W. F., & Lemerise, E. A. (2001). Varieties of childhood bullying: Val-ues, emotion processes and social competence. Social Development, 10, 59–73.doi:10.1111/1467-9507.00148

agwell, C. L., Newcomb, A. F., & Bukowski, W. M. (2000). Preadolescent friendshipand peer rejection as predictors of adult adjustment. In W. Craig (Ed.), Childhoodsocial development: The essential readings (pp. 86–112). Oxford, U.K.: Blackwell.

arry, C. M., & Wentzel, K. R. (2006). Friend influence on prosocial behavior: The roleof motivational factors and friendship characteristics. Developmental Psychology,42, 153–163. doi:10.1037/0012-1649.42.1.153

erndt, T. J., Hawkins, J. A., & Jiao, Z. (1999). Influences of friends and friendships onadjustment to junior high school. Merrill-Palmer Quarterly, 45, 13–41.

erndt, T. J., & Keefe, K. (1995). Friends’ influence on adolescents’ adjustment toschool. Child Development, 66, 1312–1329. doi:10.2307/1131649

oivin, M., Vitaro, F., & Poulin, F. (2005). Peer relationships and the developmentof aggressive behavior in early childhood. In R. E. Tremblay, W. W. Hartup, & J.Archer (Eds.), Developmental origins of aggression (pp. 376–397). New York, NY:Guilford.

owker, J. C., Rubin, K., Rose-Krasnor, L., & Booth-LaForce, C. (2007). Good friend-ships, bad friends: Friendship factors as moderators of the relationship betweenaggression and social information processing. European Journal of DevelopmentalPsychology, 4, 1–20. doi:10.1080/17405620701632069

rendgen, M., Lamarche, V., Wanner, B., & Vitaro, F. (2010). Links between friendshiprelations and early adolescents’ trajectories of depressed mood. DevelopmentalPsychology, 46, 491–501. doi:10.1037/a0017413

rendgen, M., Vitaro, F., & Bukowski, W. M. (2000). Deviant friends and early adoles-cents’ emotional and behavioral adjustment. Journal of Research on Adolescence,10, 173–189. doi:10.1207/sjra1002 3

ukowski, W. M., Newcomb, & Hartup, W. W. (1996). Friendship and its significancein childhood and adolescence: Introduction and comment. In W. M. Bukowski,A. F. Newcomb, & W. W. Hartup (Eds.), The company they keep: Friendship inchildhood and adolescence (pp. 1–15). Cambridge, MA: Cambridge UniversityPress.

urleson, B. R. (1994). Friendship and similarities in social-cognitive and commu-nication abilities: Social skill bases of interpersonal attraction in childhood.Personal Relationships, 1, 371–389. doi:10.1111/j.1475-6811.1994.tb00071.x

urr, J. E., Ostrov, J. M., Jansen, E. A., Cullerton-Sen, C., & Crick, N. R. (2005). Relationalaggression and friendship during early childhood: “I won’t be your friend!”. EarlyEducation & Development, 16, 161–183. doi:10.1207/s15566935eed1602 4

oplan, R. J., Prakash, K., O’Neil, K., & Armer, M. (2004). Do you “want” to play? Distin-guishing between conflicted shyness and social disinterest in early childhood.Developmental Psychology, 40, 244–258. doi:10.1037/0012-1649.40.2.244

rick, N. R. (1996). The role of overt aggression, relational aggression, and prosocialbehavior in the prediction of children’s future social adjustment. Child Develop-ment, 67, 2317–2327. doi:10.2307/1131625

rick, N. R., Ostrov, J. M., Burr, J. E., Cullerton-Sen, C., Jansen-Yeh, E., &Ralston, P. (2006). A longitudinal study of relational and physical aggres-sion in preschool. Journal of Applied Developmental Psychology, 27, 254–268.doi:10.1016/j.appdev.2006.02.006

enton, K., & Zarbatany, L. (1996). Age differences in support processes in conversa-tions between friends. Child Development, 67, 1360–1373. doi:10.2307/1131705

unn, J. (2004). Children’s friendships: The beginnings of intimacy. Oxford, U.K.: Black-well.

stell, D. B., Cairns, R. B., Farmer, T. W., & Cairns, B. D. (2002). Aggression in inner-cityearly elementary classrooms: Individual and peer-group configurations. Merrill-Palmer Quarterly, 48, 52–76. doi:10.1353/mpq.2002.0002

arver, J. A. M. (1996). Aggressive behavior in preschoolers’ social networks: Dobirds of a feather flock together? Early Childhood Research Quarterly, 11, 333–350.doi:10.1016/S0885-2006(96)90011-3

urman, W., & Bierman, K. L. (1983). Developmental changes in young children’s con-ceptions of friendship. Child Development, 54, 549–556. doi:10.2307/1130041

ershman, E. S., & Hayes, D. S. (1983). Differential stability of reciprocal friendshipsand unilateral relationships among preschool children. Merrill-Palmer Quarterly,29, 169–177.

oldstein, S. E., Tisak, M. S., & Boxer, P. (2002). Preschoolers’ normative and pre-scriptive judgments about relational and overt aggression. Early Education andDevelopment, 13, 23–39. doi:10.1207/s15566935eed1301 2

oodman, R. (1997). The Strengths and Difficulties Questionnaire: A research note.Journal of Child Psychology and Psychiatry, 38, 581–586. doi:10.1111/j.1469-7610.1997.tb01545.x

oodman, R. (2001). Psychometric properties of the Strengths and Difficulties Ques-tionnaire. Journal of the American Academy of Child and Adolescent Psychiatry, 40,

1337–1345. doi:10.1097/00004583-200111000-00015

ottman, J. M., & Graziano, W. G. (1983). How children become friends. Monographsof the Society for Research in Child Development, 48, 1–86. doi:10.2307/1165860

anish, L. D., Martin, C. L., Fabes, R. A., Leonard, S., & Herzog, M. (2005).Exposure to externalizing peers in early childhood: Homophily and peer

rch Quarterly 27 (2012) 137– 146 145

contagion processes. Journal of Abnormal Child Psychology, 33(3), 267–281.doi:10.1007/s10802-005-3564-6

Hay, D. F., & Cook, K. V. (2007). The transformation of prosocial behavior from infancyto childhood. In C. A. Brownell, & C. B. Kopp (Eds.), Socioemotional development inthe toddler years: Transitions and transformations (pp. 100–131). New York, NY:Guilford Press.

Hayes, D. S., Gershman, E., & Bolin, L. J. (1980). Friends and enemies: Cognitive basesfor preschool children’s unilateral and reciprocal relationships. Child Develop-ment, 51, 1276–1279. doi:10.2307/1129572

Heiervang, E., Goodman, A., & Goodman, R. (2008). The Nordic advantage inchild mental health: Separating health differences from reporting style in across-cultural comparison of psychopathology. Journal of Child Psychology andPsychiatry, 49, 678–685. doi:10.1111/j.1469-7610.200801882.x

Howes, C. (1987). Social competence with peers in young children Devel-opmental sequences. Developmental Review, 7, 252–272. doi:10.1016/0273-2297(87)90014-1

Howes, C. (1996). The earliest friendships. In W. M. Bukowski, A. F. Newcomb, & W.W. Hartup (Eds.), The company they keep: Friendship in childhood and adolescence(pp. 66–86). Cambridge, MA: Cambridge University Press.

Howes, C. (2009). Friendship in early childhood. In K. H. Rubin, W. M. Bukowski,& B. Laursen (Eds.), Handbook of peer interactions, relationships, and groups (pp.180–194). New York, NY: Guilford Press.

Howes, C., & Lee, L. (2006). Peer relations in young children. In L. Balter, & C. S. Tamis-LaMonda (Eds.), Child psychology: A handbook of contemporary issues (2nd ed., pp.135–151). New York, NY: Taylor and Francis.

Howes, C., & Phillipsen, L. (1992). Gender and friendship: relationships within peergroups of young children. Social Development, 1, 230–242. doi:10.1111/j.1467-9507.1992.tb00126.x

Hughes, C., White, A., Sharpen, J., & Dunn, J. (2000). Antisocial, angry andunsympathetic: “Hard to manage” preschoolers’ peer problems and possiblecognitive influences. Journal of Child Psychology and Psychiatry, 41, 169–179.doi:10.1111/1469-7610.00558

Johnson, D. R, & Foster, S. L. (2005). The relationship between relational aggression inkindergarten children and friendship stability, mutuality, and peer liking. EarlyEducation & Development, 16, 141–160. doi:10.1207/s15566935eed1602 3

Juliano, M., Werner, R. S., & Cassidy, K. W. (2006). Early correlates ofpreschool aggressive behavior according to type of aggression and mea-surement. Journal of Applied Developmental Psychology, 27, 395–410.doi:10.1016/j.appdev.2006.06.008

Kokko, K., Tremblay, R. E., Lacourse, E., Nagin, D. S., & Vitaro, F. (2006). Trajectories ofprosocial behavior and physical aggression in middle childhood: Links to ado-lescent school dropout and physical violence. Journal of Research on Adolescence,16, 403–428. doi:10.1111/j.1532-7795.2006.00500.x

Kupersmidt, J. B., DeRosier, M. E., & Patterson, C. J. (1995). Similarity as thebasis for children’s friendships: The roles of sociometric status, aggres-sive and withdrawn behavior, academic achievement and demographiccharacteristics. Journal of Social and Personal Relationships, 12, 439–452.doi:10.1177/0265407595123007

Ladd, G. W. (1992). Themes and theories: Perspectives on processes in family–peerrelationships. In R. D. Parke, & G. W. Ladd (Eds.), Family–peer relationships: Modesof linkage (pp. 3–34). Hillsdale, NJ: Erlbaum.

Ladd, G. W. (1999). Peer relationships and social competence during earlyand middle childhood. Annual Review of Psychology, 50, 333–359.doi:10.1146/annurev.psych.50.1.333

Ladd, G. W., Herald, S. L., & Andrews, R. K. (2006). Young children’s peer rela-tions and social competence. In B. Spodeck & O. N. Saracho (Eds.), Handbookof research on the education of young children (2nd ed., pp. 23–54). Mahwah, NJ:Erlbaum.

Laursen, B., Bukowski, W. M., Aunola, K., & Nurmi, J. E. (2007). Friendship moder-ates prospective associations between social isolation and adjustment problemsin young children. Child Development, 78, 1395–1404. doi:10.1111/j.1467-8624.2007.01072.x

Maccoby, E. E., & Jacklin, C. N. (1980). Sex differences in aggression: A rejoinder andreprise. Child Development, 51, 964–980. doi:10.2307/1129535

Mannarino, A. P. (1995). The development of children’s friendships. In H. C. Foot,A. J. Chapman, & J. R. Smith (Eds.), Friendship and social relations in children (pp.45–63). New Brunswick, NJ: Transaction Publishers.

McDougall, P., Hymel, S., Vaillancourt, T., & Mercer, L. (2001). The consequencesof childhood peer rejection. In M. R. Leary (Ed.), Interpersonal rejection (pp.213–247). New York, NY: Oxford University Press.

Mrug, S., Hoza, B., & Bukowski, W. M. (2004). Choosing or being chosen byaggressive–disruptive peers: Do they contribute to children’s externalizingand internalizing problems? Journal of Abnormal Child Psychology, 32, 53–65.doi:10.1023/B:JACP.0000007580.77154.69

Nelson, D. A., Robinson, C. C., & Hart, C. H. (2005). Relational and physical aggres-sion of preschool-age children, Peer status linkages across informants. EarlyEducation Development, 16, 115–140. doi:10.1207/s15566935eed1602 2

Newcomb, A. F., & Bagwell, C. L. (1995). Children’s friendship relations: Ameta-analytic review. Psychological Bulletin, 117, 306–347. doi:10.1037/0033-2909.117.2.306

Obel, C., Heiervang, E., Rodriguez, A., Heyerdahl, S., Smedje, H., Sourander, A., et al.

(2004). The Strengths and Difficulties Questionnaire in the Nordic countries.European Child and Adolescent Psychiatry, 13, ii32–ii39, doi:10.1007/s00787-004-2006-2.

Organisation for Economic Cooperation and Development. (2010a).Enrolment in child care and preschools. OECD Directorate for

1 Resea

O

P

P

P

P

R

R

S

S

S

S

adolescence (pp. 346–378). New York, NY: Cambridge University Press.Wojslawowicz Bowker, J. C., Rubin, K. H., Burgess, K. B., Booth-Laforce, C., & Rose-

46 A.R. Eivers et al. / Early Childhood

Employment, Labour and Social Affairs—Social Policy Division. Retrievedfrom http://www.oecd.org/els/social/family/database.

rganisation for Economic Cooperation and Development. (2010b). To what levelhave adults studied? Highlights from Education at a Glance 2010. Retrievedfrom http://www.oecd-ilibrary.org/education/highlights-from-education-at-a-glance-2010 eag highlights-2010-en.

arker, J. G. (1986). Becoming friends: Conversational skills for friendship formationin young children. In J. Gottman, & J. G. Parker (Eds.), Conversations of friends:Speculations on affective development (pp. 103–138). Cambridge, U.K.: CambridgeUniversity Press.

edersen, S., Vitaro, F., Barker, E. D., & Borge, A. I. H. (2007). The timingof middle-childhood peer rejection and friendship: Linking early behav-ior to early-adolescent adjustment. Child Development, 78, 1037–1051.doi:10.1111/j.1467-8624.2007.01051.x

hillipsen, L. C., Bridges, S. K., McLemore, T. G., & Saponaro, L. A. (1999). Perceptionsof social behavior and peer acceptance in kindergarten. Journal of Research inChildhood Education, 14, 68–77.

oulin, F., Cillessen, A. H. N., Hubbard, J. A., Coie, J. D., Dodge, K. A., & Schwartz, D.(1997). Children’s friends and behavioral similarity in two social contexts. SocialDevelopment, 6, 224–236. doi:10.1111/j.1467-9507.1997.tb00103.x

omano, E., Tremblay, R. E., Boulerice, B., & Swisher, R. (2005). Multilevel correlatesof childhood physical aggression and prosocial behavior. Journal of AbnormalChild Psychology, 33, 565–578. doi:10.1007/s10802-005-6738-3

ubin, K. H., Lynch, D., Coplan, R., & Rose-Krasnor, L. (1994). Birds of a feather. . .:Behavioral concordances and preferential personal attraction in children. ChildDevelopment, 65, 1778–1785. doi:10.2307/1131293

ebanc, A. M. (2003). The friendship features of preschool children, Links with proso-cial behavior aggression. Social Development, 12, 249–268. doi:10.1111/1467-9507.00232

elman, R. L. (1971). Taking another’s perspective: Role-taking development in earlychildhood. Child Development, 42, 1721–1734. doi:10.2307/1127580

nyder, J. (2002). Reinforcement and coercion mechanisms in the development ofantisocial behavior: Peer relationships. In J. B. Reid, G. R. Patterson, & J. Snyder(Eds.), Antisocial behavior in children and adolescents: A developmental analysis

and model for intervention (pp. 101–122). Washington D.C.: American Psycho-logical Association.

nyder, J., Horsch, E., & Childs, J. (1997). Peer relationships of young childrenaffiliative choices the shaping of aggressive behavior. Journal of Clinical ChildPsychology, 26, 145–156. doi:10.1207/s15374424jccp2602 3

rch Quarterly 27 (2012) 137– 146

Snyder, J., Schrepferman, L., Oeser, J., Patterson, G., Stoolmiller, M., Johnson, K., et al.(2005). Deviancy training and association with deviant peers in young children:Occurrence and contribution to early-onset conduct problems. Development andPsychopathology, 17, 397–413. doi:10.1017/S0954579405050194

Statistics Norway. (2010a). Education: Level of education. Retrieved fromhttp://www.ssb.no/english/subjects/04/01.

Statistics Norway. (2010b). Personal economy and housing: Income, property andtaxes. Retrieved from http://www.ssb.no/english/subjects/05/01.

Strayer, F. F., & Santos, A. J. (1996). Affiliative structures in preschool peer groups.Social Development, 5, 117–130. doi:10.1111/j.1467-9507.1996.tb00075.x

Sullivan, H. S. (1965). Personal psychopathology. New York, NY: Norton.Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston,

MA: Allyn & Bacon/Pearson Education.Tremblay, R. E. (2000). The development of aggressive behaviour during childhood:

What have we learned in the past century? International Journal of BehavioralDevelopment, 24, 129–141. doi:10.1080/016502500383232

Van Leeuwen, K., Meerschaert, T., Bosmans, G., De Medts, L., & Braet, C. (2006).The Strengths and Difficulties Questionnaire in a community sample of youngchildren in Flanders. European Journal of Psychological Assessment, 22, 189–197.doi:10.1027/1015-5759.22.3.189

Vitaro, F., Boivin, M., & Bukowski, W. M. (2009). The role of friendship in childand adolescent psychosocial development. In K. H. Rubin, W. M. Bukowski, &B. Laursen (Eds.), Handbook of peer interactions, relationships and groups (pp.568–585). New York, NY: Guilford Press.

Vitaro, F., Brendgen, M., & Tremblay, R. E. (2000). Influence of deviant friends ondelinquency: Searching for moderator variables. Journal of Abnormal Child Psy-chology, 28, 313–325. doi:10.1023/a:1005188108461

Vitaro, F., Gagnon, C., & Tremblay, R. E. (1990). Predicting stable peer rejectionfrom kindergarten to grade one. Journal of Clinical Child Psychology, 19, 257–264.doi:10.1207/s15374424jccp1903 9

Vitaro, F., Tremblay, R. E., & Bukowski, W. M. (2000). Friends, friendship and con-duct disorder. In J. Hill, & B. Maughan (Eds.), Conduct disorders in childhood and

Krasnor, L. (2006). Behavioral characteristics associated with stable and fluidbest friendship patterns in middle childhood. Merrill-Palmer Quarterly, 52,671–693. doi:10.1353/mpq.2006.0000