15
Journal of Adolescence 1997, 20, 645–659 Behaviour problems and relationships with family and peers during adolescence W. H. BUYSSE This study examines the relationship between (a) characteristics of adolescents’ social networks, personal resources, and environmental risks; and (b) adolescents’ behaviour problems. A socio-ecological perspective was used to determine whether social support and perceived conflict operate as risk or protective factors. Three samples of adolescents were studied: 63 adolescents in residential care; 29 in day treatment; and 63 who had had no contact with professional care. Interrelations were explored by means of structural equation modelling. Social support in family and peer group may operate as a risk factor or a protective factor depending on other risk factors in these subsystems. Perceived conflict is related to behaviour problems, although the influence differs depending on environmental risk factors and the type of behaviour problems. 1997 The Association for Professionals in Services for Adolescents Introduction For the majority of youth, movement into the adolescent years involves an expansion of their social contacts: for instance, the nature of relationships with parents changes and peers begin to occupy a more central role (Jackson and Rodriguez-Tome ´ , 1993). In adolescence three important categories of network members can be distinguished: peers, kin and non-related adults (Blyth al., 1982; Bo ¨ , 1989; Cotterel, 1992). The largest portion of the social network is made up of peers and kin. During adolescence the characteristics of the relationships with these network members changes. From middle childhood to adolescence support from peers increases, support from teachers decreases and support from parents or family remains somewhat more stable (Cauce et al., 1990, 1994; Furman and Buhrmester, 1992). By the end of adolescence, parents cease to be the most prominent providers of support, but in an absolute sense the amount of support perceived from parents remains great; meanwhile peers emerge as significant sources of support (Blyth and Traeger, 1988; Berndt, 1989; Meeus, 1989, 1993; Cauce et al., 1990; Furman and Buhrmester, 1992). Relationships are not exclusively supportive, however. They can be conflictual in nature and even in supportive relationships adolescents can also experience conflict. With the onset of adolescence, conflict with parents increases (Furman and Buhrmester, 1992), but relationships with parents are not conflict-ridden overall, they are supportive as well as conflictual (Cauce et al., 1990; Compas and Wagner, 1991; Smetana et al., 1991; Buysse and Van der Ploeg, 1992). However, the amount of conflict is much lower than has long been Reprint requests and correspondence should be addressed to W. H. Buysse, Centre for Special Education and Child Care, Leiden University, P.O. Box 9555, 2300 RB Leiden, The Netherlands. 0140-1971/97/060645+ 15/$25·00/0/ad970117 1997 The Association for Professionals in Services for Adolescents

Behaviour problems and relationships with family and peers during adolescence

Embed Size (px)

Citation preview

Journal of Adolescence 1997, 20, 645–659

Behaviour problems and relationships with family andpeers during adolescence

W. H. BUYSSE

This study examines the relationship between (a) characteristics of adolescents’social networks, personal resources, and environmental risks; and (b) adolescents’behaviour problems. A socio-ecological perspective was used to determine whethersocial support and perceived conflict operate as risk or protective factors. Threesamples of adolescents were studied: 63 adolescents in residential care; 29 in daytreatment; and 63 who had had no contact with professional care. Interrelationswere explored by means of structural equation modelling. Social support in familyand peer group may operate as a risk factor or a protective factor depending onother risk factors in these subsystems. Perceived conflict is related to behaviourproblems, although the influence differs depending on environmental risk factorsand the type of behaviour problems.

1997 The Association for Professionals in Services for Adolescents

Introduction

For the majority of youth, movement into the adolescent years involves an expansion oftheir social contacts: for instance, the nature of relationships with parents changes andpeers begin to occupy a more central role (Jackson and Rodriguez-Tome, 1993). Inadolescence three important categories of network members can be distinguished: peers, kinand non-related adults (Blyth al., 1982; Bo, 1989; Cotterel, 1992). The largest portion ofthe social network is made up of peers and kin. During adolescence the characteristics ofthe relationships with these network members changes. From middle childhood toadolescence support from peers increases, support from teachers decreases and support fromparents or family remains somewhat more stable (Cauce et al., 1990, 1994; Furman andBuhrmester, 1992). By the end of adolescence, parents cease to be the most prominentproviders of support, but in an absolute sense the amount of support perceived from parentsremains great; meanwhile peers emerge as significant sources of support (Blyth and Traeger,1988; Berndt, 1989; Meeus, 1989, 1993; Cauce et al., 1990; Furman and Buhrmester, 1992).

Relationships are not exclusively supportive, however. They can be conflictual in natureand even in supportive relationships adolescents can also experience conflict. With theonset of adolescence, conflict with parents increases (Furman and Buhrmester, 1992), butrelationships with parents are not conflict-ridden overall, they are supportive as well asconflictual (Cauce et al., 1990; Compas and Wagner, 1991; Smetana et al., 1991; Buysse andVan der Ploeg, 1992). However, the amount of conflict is much lower than has long been

Reprint requests and correspondence should be addressed to W. H. Buysse, Centre for Special Education andChild Care, Leiden University, P.O. Box 9555, 2300 RB Leiden, The Netherlands.

0140-1971/97/060645+15/$25·00/0/ad970117 1997 The Association for Professionals in Services for Adolescents

646 W. H. Buysse

thought. In general, adolescents do not have solely conflictual relationships with theirparents. In several Dutch surveys (Van der Linden, 1991; Granefski and Diekstra, 1993) itwas found that the majority of adolescents reported getting along well with their parents. Inmost studies, relationships with siblings are found to be the most conflictual duringadolescence (Cauce et al., 1990; Buysse and Van der Ploeg, 1992; Furman and Buhrmester,1992; Barrera et al., 1993). Cauce et al. (1990) suggested that sibling rivalry is quite normalin childhood and adolescence. Conflict with peers or friends is more rare (Berndt, 1989;Claes, 1992; Barrera et al., 1993).

Characteristics of the relationships with family and peers, i.e. perceived support andperceived conflict, may operate as risk or protective factors in the development ofbehaviour problems. In general, most studies have found low family support to be associatedwith internalizing and externalizing behaviour problems (Forehand et al., 1991; Windle,1991, 1992; Barrera et al., 1993) and high conflicts in the family to be related to bothbehaviour problems (Barrera et al., 1993).

The findings regarding the effects of peer support are more contradictory. Studies havereported connections between indices of the absence of behaviour problems andsupportiveness of peers (Burke and Weir, 1978; Cauce, 1986; Hirsch and Rapkin, 1987;Slavin and Rainer, 1990; Vernberg, 1990; Claes, 1992; Hirsch and Dubois, 1992). Indices ofpeer support have also been found to be unrelated to externalizing behaviour (Barrera et al.,1993) or to be positively related to externalizing behaviour (Cauce et al., 1990; Windle,1991), i.e. high peer support being related to higher levels of externalizing behaviour. Ininvestigations of special risk groups such as children with an ill or depressed parent (Hirschand Reischl, 1985), adolescent inpatients in a psychiatric clinic (Barrera and Garrison-Jones, 1992) and adolescents after discharge from residential care (Wells et al., 1991), highpeer support was found to be positively related to symptomatology, depressive symptomsand psychopathology, respectively. In general, most studies demonstrating positive effects ofpeer support examined a general outcome variable or internalizing behaviour, whereas moststudies showing a negative effect used externalizing behaviour as outcome variables. Highconflict with peers has been found to be related to maladjustment and depressive effect(Vernberg, 1990; Claes, 1992). On the other hand, conflict with best friend was not relatedto externalizing behaviour in Barrera et al.’s (1993) study.

The above studies mostly examined the role of family and peer group separately. Windle(1992) suggested that it is useful to focus on the simultaneous influence of family and peersupport, because of the increased salience of peer influence in adolescence and because ofthe restructuring of parent–child relations. Several authors have suggested that the absenceof family support places adolescents at-risk for negative peer influence (Garmezy, 1987; Vander Ploeg and Scholte, 1990; Barrera and Garrison-Jones, 1992).

In order to explain the influence of relationships with family and peers on behaviourproblems, we need a theoretical framework which encompasses characteristics ofrelationships with family and peers as well as an explanation for the development ofbehaviour problems. In the present study we used a socio-ecological perspective as aframework from which the relationship between characteristics of adolescents’ relationshipswith family and peers and behaviour problems can be studied. Van der Ploeg and Scholte’s(1990) socio-ecological model of the development of behaviour problems considersbehaviour problems as the outcome of an interaction between risk factors in the person andrisk factors in the social environment. The category of behaviour problems is subdividedinto internalizing and externalizing behaviours.

647Family and peer relationships

Van der Ploeg and Scholte (1990) distinguished three main areas in their model: thefamily, the peer group and the school. Risk factors and protective factors may originate ineach of these areas of the environment as well as in the person. The principle risk factors inthe family are: (1) severe family conflict; (2) insecure attachment and poor familycommunication; (3) poor supervision; and (4) non-responsive or permissive child-rearingpractices. The main risk factors in school include: (1) poor scholastic motivation andacademic performance; (2) conflict with teachers; (3) an impoverished school culture withlow standards of achievement; and (5) undemocratic teaching styles. Important risk factorsin the peer group are: (1) antisocial behaviour in friends; (2) risky leisure time behaviour,like hanging around the streets combined with alcohol and drug abuse; and (3) relationshipproblems with peers. There are also a number of personality traits of the adolescent whichheighten the risk of behaviour and emotional disorders. Most of these traits have to do withpoorly developed cognitive emotional skills such as poor ego-control, low self-esteem,external locus of control, and ineffective coping skills. The risk factors are interrelated withone another, can cause one another and can enhance one another. In this manner theyresult in a developmental process that reinforces deviance. The model presents acombination of factors which can lead to behaviour problems, but does not specifyexclusive causes of behaviour problems.

The socio-ecological model also assumes that a well-functioning family, a favourableschool career, a positive peer group, and a positive personal disposition may have arestraining influence on the development of behaviour problems. For instance, a healthyadjustment in the family and school may discourage participation in a delinquent peergroup.

Not all areas in the social environment have an equal effect on the socialization ofadolescents. The stronger the bond with the representatives of a sub-system, the strongerthe influence of the sub-system will be. For instance, contact with a delinquent peer groupwill contribute less to the development of structural delinquent behaviour if the bond withthe family and/or school is strong (Steinberg, 1987). The inhibiting influence is, however,only present when the norms and values in the family and school are not deviant as well.

The aim of the present study is to explore the relationship between characteristics ofadolescents’ relationships with family and peers and behaviour problems and whether thisrelationship is influenced by risk factors in the person or the environment. In line with thesocio-ecological perspective we consider social support as an indication of the degree ofbonding with family and peers. We assume that the more adolescents feel supported by thepersons in the family or peer group, the stronger the influence will be. Furthermore, weassume that social support may have positive as well as negative effects on the adolescents.For instance, social support from a deviant peer group may reinforce the adolescents deviantbehaviour (Emler, 1993; Dishion et al., 1995; Buhrmester et al., 1996). Thus social supportmay be a risk factor as well as a protective factor. Furthermore, we assumed perceivedconflict with family and peers to be a risk factor in the development of behaviour problems.Again in line with the socio-ecological perspective, we assumed that risk factors in theperson and the environment might influence the association between the characteristics ofthe relationships with family and peers and behaviour problems. Finally, because perceivedenvironment has been proven to be a better predictor of later adjustment than moreobjective measures of the environment (Jessor, 1981; Van der Ploeg, 1983) we focus on theadolescents’ perceptions.

648 W. H. Buysse

Method

SubjectsThree samples of adolescents were studied: 63 adolescents admitted to residential treatmentcentres (35 boys and 28 girls, mean age 14·3 years); 29 adolescents admitted to daytreatment centres (21 boys and 8 girls, mean age 13·5 years); and 63 adolescents who hadhad no contact with professional care and were used as a reference group (24 boys and 29girls, mean age 14·2 years). All adolescents were Dutch; adolescents from ethnic minoritieswere not included in the samples. The adolescents in the three samples had a comparableeducational background: most of them were following a lower level secondary educationprogramme such as vocational training. The socio-economic status of the family was alsocomparable between the samples and was, on average, low to middle.

ProcedureThe adolescents in the residential sample were recruited in 13 residential treatment centresand the adolescents in the day-treatment sample were recruited in 12 day treatment centres.After mutual consultation between the researcher and the contact persons in eachinstitution, the latter approached adolescents between 12 and 16 years of age who had beenadmitted to the institution within the preceding 3 months, and also their legal guardiansabout participation in the study (consent of both the adolescent and the guardian wasnecessary). A brochure about the research project and a letter of introduction were madeavailable. The adolescents were interviewed at their institution between October 1992 andApril 1993. These interviews lasted for one and a half to two hours and focused on theirsituation in the 6 months prior to admission. The adolescents received a gift certificate forparticipating in the interview.

The adolescents in the reference sample were recruited in seven schools for lower levelsecondary education. Only schools for lower level secondary education were approachedbecause most adolescents in the two care samples were following an educational programmeat this level. After an in-class introduction to the research project, adolescents who wereinterested in participating received a brochure about the research project and a letterrequesting the consent of their parents. The adolescents were interviewed individually atschool between October and December 1993. The interview took one to one and a halfhours and focused on the 6 months preceding the interview. The adolescents received a giftin turn of their participation.

Measures

Risk factors in the environment. Standardized scores were derived from theinformation obtained in the interview concerning the following risk factors: (1) high-riskfamily background (for instance, parents divorced, parents had emotional problems, parentsfought a lot; 6 items, α=0·74); (2) high-risk school career (for instance, changed schoolsoften, repeating a class, felt less comfortable at school; 9 items, α=0·67); (3) high-riskleisure time activities (for instance, hanging around on the street; 5 items, α=0·52); (4)reporting non-delinquent vs. delinquent peers (for instance, reporting to have peers whoconduct vandalism, shoplifting, fights, who have had contact with police; 6 items, α=0·58).

649Family and peer relationships

One question in the interview concerned attitudes towards delinquent activities(delinquent activities are alright vs. delinquent activities are strictly forbidden, i.e. “loose”vs. “strict” attitudes toward delinquent activities).

Characteristics of relationships with family and peers. Perceived support fromand perceived conflict with family members and peers were measured by means of anadapted version of the Social Network Grid (Tracy and Whittaker, 1990). Perceivedconflict was measured on a 3-point scale. Perceived support was a mean composite score offour items (family support, α=0·79; peer support, α=0·76). An item concerning howimportant adolescents found their family members’ and peers’ opinion of them was alsoincluded in the Social Network Grid.

Scores concerning the family were mean composite scores derived from scores for mother,father, siblings and stepparents (when applicable). Scores concerning peers were meancomposite scores of scores concerning friends, romantic partner (when applicable) andpeers from the living-group (in the previous residential or day treatment institution).

Behaviour problems. Internalizing and externalizing behaviour problems weremeasured with the Dutch version of the Youth Self Report Form of the Child BehaviourChecklist (YSR) (Verhulst et al., 1990). In the analyses standardized scores for internalizingbehaviour and delinquent behaviour were used. We decided not to use the composite forexternalizing behaviour because delinquent behaviour and aggressive behaviour, whichform the two subscales of externalizing behaviour, showed opposite relationships withcertain network characteristics (such as conflicts with peers).

We chose not to consider aggressive behaviour, because operationalization of aggressionin the aggressive behaviour subscale has been criticized (Kohnstamm, 1994). Furthermore,we decided to label the behaviour in the delinquent behaviour subscale as antisocialbehaviour. Delinquent behaviour comprises activities against criminal law, whereasantisocial behaviour involves activities against social norms. The delinquent behavioursubscale of the CBCL consists of behaviours which are more antisocial, such as lying,running away, and swearing. The items that refer to delinquent behaviour concern minorcriminal activities such as vandalism and stealing.

Risk factors in the person. Personal resources, i.e. self-esteem, locus of control andcoping, were examined by means of the Self Report List (ZelfBeoordelingslijst, ZBL, Van dePloeg, 1992). The ZBL is a self-report questionnaire of 37 items scored on a 5-point scaledeveloped by Van der Ploeg and covers the three concepts of the person as defined in thesocio-ecological model (Van der Ploeg and Scholte, 1990): self-esteem (15 items, α=0·71),locus of control (10 items, α=0·51), and coping (12 items, α=0·65).

Loneliness. Loneliness can be considered as a construct which is directly related tocharacteristics of the social network and to internalizing behaviour problems (Buysse,1997). For this reason loneliness was included in the current study. Loneliness wasmeasured by means of a 6-item loneliness scale (α=0·82) developed by Van der Ploeg (foradolescents in residential care). Items were scored on a 5-point scale.

650 W. H. Buysse

Results

The direct and indirect influences of (a) characteristics of relationships with family andpeers, risk factors in the environment and risk factors in the person on (b) behaviourproblems were investigated with the use of the structural equation modelling program EQS(Bentler, 1989). This approach distinguishes two types of variables: (1) manifest variables,which are observed or measured directly; (2) latent variables, which are derived frommanifest variables and represent “true” measures free of measurement error. Latent variablescan be estimated from multiple manifest variables or indicators with the use of confirmatoryfactor analysis. In the construction of latent variables it is strongly recommended to use atleast three indicators for each latent variable (Dunn et al., 1993). In the present study wedid not have three indicators for each latent variable. Therefore, we used measuredvariables instead of latent variables. The use of structural equation modelling in this study isequivalent to path analysis. Measured variables are not free of measurement error. Thisimplies that interpretation of the results should be undertaken with caution and the pathsfound should be tested in further research.

Hypotheses can be tested using structural equation modelling. However, in the presentstudy we used the technique in a more explorative way. We wanted to investigate theindirect influences of variables in our model which cannot be determined by multipleregression analysis. Furthermore, we wanted to test the influence on the two behaviourproblems simultaneously. The arrows in the model (see Figure 1) were determined on thebasis of theoretical considerations and of the results of regression analyses (see Buysse,1997).

Internalizing behaviour problems and antisocial behaviour were significantly correlated(r=0·37, p<0·01). Therefore we constructed a model in which both behaviour problemswere included. In this model we allowed the error terms of the behaviour problems to becorrelated. The correlation between the error terms was 0·43.

We considered the risk factors in the environment and the risk factors in the person aspredictive variables. These variables were allowed to correlate. The network characteristicswere used as intervening variables.

The EQS analysis established whether or not the model fitted the observed data. If the χ2

representing the fit of the model was non-significant, the model was accepted because thismeant that there were relationships between the variables in the model. If the χ2 wassignificant the model was rejected. The χ2 depends strongly on the size of the sample.Therefore two other fit indices are mentioned: the comparative fit index (CFI), whichavoids the overestimation of fit in small samples, and the non-normed fit index (NNFI),which has the advantage of reflecting model fit very well at all sample sizes. These indicesneed to have values above 0·9 before the corresponding model can be consideredmoderately adequate.

Because of missing values only 142 subjects were used in the analyses. The influence ofbeing in the residential or day treatment samples on the relationships between thedependent and independent variables was examined in regression analyses by including twodummy variables. None of the relationships found were altered by including these dummyvariables (see Buysse, 1997). In view of this, EQS modelling was applied to the threesamples together.

The model fitted the data very well, with χ2(105)=95·26, p=0·74, NNFI=1·00 and CFI=

651Family and peer relationships

HSC

HRF

LOC

COP

SE

AT

LEI

DP

OF CF

SF

LON

SP

OP CP

IB

AB

0.29 –0.25–0.19

0.41

0.25

0.23

0.25

–0.19

–0.43

0.14

–0.250.200.35

0.21

0.23

0.36

0.20

–0.07

–0.16

0.17

0.18

0.350.19

0.12

0.36

0.18

0.43

0.77

0.80

0.10 0.24 –0.20 0.25

Figure 1. HSC=high-risk school career; HRF=high-risk family background; LOC=internal locus ofcontrol; COP=effective coping; SE=high self-esteem; AT=loose attitudes vs. strict attitudesconcerning delinquency; LEI=high-risk leisure time; DP=delinquent peer group; OF=opinion family;CF=perceived conflict family; SF=perceived support family; LON=loneliness; SP=perceived supportpeers; OP=opinion peers; CP=conflict peers; IB=internalizing behaviour problems; AB=antisocialbehaviour. χ2(105)=95·26, p=0·74, NNFI=1·00 and CFI=1·00. The correlations between theindependent variables at the left side of the model are not presented in the figure for reasons of clarityof presentation. The arrows represent the influence of the predicting and intervening variables inregression coefficients.

1·00, and was therefore accepted as an adequate representation of the data. Four of thehypothesized paths, i.e. from perceived conflict with family members to internalizingbehaviour, antisocial behaviour and perceived support from peers, and from loose attitudesconcerning delinquency to antisocial behaviour were not significant (see the dotted arrowsin Figure 1). Deletion of these paths from the model did not result in a more significant fitof the model.1 Furthermore, considering the relative small size of our sample it was notadvisable to change the model as a function of the data by eliminating the paths. In smallsamples there is a bigger chance that coincidental characteristics of the sample also play arole in the determination of the model (MacCallum et al., 1992). Therefore, it was decidedto retain the four paths in the model.

The model (see Figure 1) shows both direct and indirect contributions to bothinternalizing behaviour problems and antisocial behaviour. A direct contribution is anunmediated relation between two variables (e.g. between loneliness and internalizingbehaviour problems). An indirect contribution is a relation between two variables that ismediated by one or more other variables (e.g. the influence of high-risk family backgroundon internalizing behaviour is mediated by perceived conflict in the family).

1The difference between the two chi-squares was 7·264 (df=4). This difference is not significant which impliesthat the improvement of the model is not significant.

652 W. H. Buysse

Correlational relationships between the independent variables in the modelBefore discussing the factors which directly and indirectly influence internalizing behaviourand antisocial behaviour we will first describe the correlations between the independentvariables in the model. The independent variables are listed at the left side of the model(see Figure 1 and Table 1). The correlations indicated that all the personal resources,except for coping and self-esteem, were moderately associated with one another. High-riskleisure time was only associated with low self-esteem and reporting more delinquent peers.Reporting delinquent peers was associated with all the background variables, except forlocus of control. This implies that reporting more delinquent peers was associated withhigh-risk school career, high-risk family background, ineffective coping, positive self-esteem, loose attitudes toward delinquency and high-risk leisure time. Finally, high-riskschool career was associated with high-risk family background and ineffective coping.

Direct influences on internalizing behaviourSix variables demonstrated a significant direct influence (p<0·05) on internalizingbehaviour: feelings of loneliness (β=0·36), strict attitudes towards delinquency (β=0·27),high-risk school career (β=0·24), low self-esteem (β=−0·20), a delinquent peer group (β=0·18), and high perceived conflict with peers (β=0·18). As mentioned earlier, the directinfluence of perceived conflict with family members was not significant (β=0·10).

Indirect and total influences on internalizing behaviourTable 2 summarizes the indirect effects on internalizing behaviour along with direct andtotal effects. Indirect effects were calculated by summing the products of direct effectsassociated with corresponding paths of influence. All indirect influences were small. Lowself-esteem had the largest indirect influence on internalizing behaviour (β=−0·15), mainlythrough its effect on loneliness (β=−0·16, or −0·43×0·36). Although low perceived supportfrom family had no direct effect on internalizing behaviour, it had a very small indirecteffect (β=−0·07), mediated through feelings of loneliness (β=−0·07, or −0·19×0·36). Asimilar pattern was found for high-risk family background: it had a very small indirect effect(β=0·06), through its effect on perceived support from family members and feelings ofloneliness (β=0·02, or 0·25×0·19×0·36) and also through its effect on perceived conflict(β=0·04, or 0·41×0·10). Finally, low perceived support from peers had a small indirect effect

Table 1 Correlations among the independent variables in the EQS-model, p<0·051

HSC HFR LOC COP SE AT LE DP

HSC —HRF 0·30 —LOC 0·18 —COP −0·23 0·24 —SE 0·29 —AT 0·32 —LEI −0·22 —DP 0·23 0·14 −0·17 0·20 −0·25 0·17 —

1The positive side of the following variables mean HSC=high-risk school career; HRF=high-riskfamily background; LOC=internal locus of control; COP=effective coping behaviour; SE=high self-esteem; AT=strict attitudes; LEI=high-risk leisure time; DP=more delinquent peers.

653Family and peer relationships

(β=−0·04), mediated through high perceived conflict with peers (β=−0·05, or −0·25×0·18).Most of the indirect effects were mediated through loneliness, which is thus an important

mediating variable for the development of internalizing behaviour problems. The onlyvariables which were not mediated through feelings of loneliness were low perceivedsupport from peers, finding peers’ opinion of oneself less important and high-risk leisuretime. These variables were mediated through perceived conflict with peers. Effects of factorsin the person of the adolescents and aspects of the family were mediated through feelings ofloneliness and aspects of the peer group were mediated through perceived conflict withpeers.

The total effects in Table 2 indicate the combined contributions of risk factors in theperson, risk factors in the environment and aspects of relationships with peers and familytoward internalizing behaviours. The magnitudes vary considerably, and many variables didnot make a significant contribution. Feelings of loneliness and low self-esteem made thegreatest contribution, though these were still moderate. High-risk family background, strictattitudes concerning delinquent activities, more delinquent peers, and high perceivedconflict with peers and with family all made a small contribution. The contributions of theother variables were so low that they can be regarded as essentially insignificant.

Direct influences on antisocial behaviourFour variables demonstrated a direct and significant contribution to antisocial behaviour, p<0·05: delinquent peers (β=0·37), high perceived support from peers (β=0·19), ineffectivecoping behaviour (β=−0·18) and high-risk school career (β=0·17). Perceived conflict withfamily members had a non-significant direct effect on antisocial behaviour (β=0·12).

Indirect and total effects on antisocial behaviourThe indirect effects, along with direct and total effects, on antisocial behaviour are

Table 2 Standardized effects on internalizing behaviour problems for the EQS model

Variable Direct effect Indirect effect Total effect

High-risk school career 0·24 — 0·24High-risk family background — 0·06 0·06Internal locus of control — −0·01 −0·01Effective coping — −0·02 −0·02High self-esteem −0·20 −0·15 −0·35Attitudes (loose to strict) 0·27 — 0·27High-risk leisure time — 0·00 0·00Delinquent peers 0·18 −0·01 0·17Opinion family — −0·03 −0·03Conflict family 0·10 0·01 0·11Support family −0·08 −0·08Loneliness 0·36 — 0·36Support peers — −0·04 −0·04Opinion peers — −0·02 −0·02Conflict peers 0·18 — 0·18

654 W. H. Buysse

summarized in Table 3. Similar to the direct effects on internalizing behaviour, the indirecteffects on antisocial behaviour were very small. Finding peers’ opinion important had thelargest indirect effect on antisocial behaviour (β=0·07), through its effect on high perceivedsupport from peers (β=0·07, 0·35×0·19). An effect of similar magnitude was found for high-risk family background (β=0·06), although due to the fact that it was mediated throughseveral variables the standardized value was non-significant (β=0·01). The delinquent peersfactor had a direct as well as an indirect effect (β=0·04) on antisocial behaviour, the latterthrough it’s influence on high perceived support from peers (β=0·13, or 0·20×0·37). Theindirect influence of aspects of the relationships with family members on antisocialbehaviour are in the opposite direction to what was expected, although the influences werevery small, high family support and finding family’s opinion important contributed toantisocial behaviour. These indirect influences went together with high levels of perceivedconflict. We found neither a direct nor an indirect influence of loose attitudes on antisocialbehaviour, although this variable was significantly correlated with a delinquent peer group.

The total effects in Table 3 indicate that aspects of the peer group make the largestcontribution to antisocial behaviour: a supportive, non-conflictuous delinquent peer group,whose opinion is considered important showed the largest effect on antisocial behaviour.

Discussion

Before drawing some conclusions it is important to point out that our sample is notrepresentative for Dutch youth. We studied adolescents at a lower level of education, with alower socio-economic background and with a higher-risk (family) background than Dutchyouth in general. Two of our samples had been admitted to professional care. Furtherresearch has to demonstrate whether the interrelations found in this study are also valid inother samples of adolescents. Furthermore, structural equation modelling was used in anexplorative way and with measured variables. This means that the findings in the currentstudy should be considered as hypotheses which should be tested in further research,preferably with latent variables.

Table 3 Standardized effects on antisocial behaviour for the EQS model

Variable Direct effect Indirect effect Total effect

High-risk school career 0·17 — 0·17High-risk family background — 0·06 0·06Internal locus of control — 0·01 0·01Effective coping −0·18 0·02 −0·16High self-esteem — — —Attitudes (loose vs. strict) — — —High-risk leisure time — 0·02 0·02Delinquent peers 0·37 0·04 0·41Opinion family — 0·04 0·04Conflict family 0·12 0·02 0·19Support family — 0·04 0·07Loneliness — — —Support peers 0·19 — 0·15Opinion peers — 0·07 0·07Conflict peers — — —

655Family and peer relationships

The aim of the study was to explore the relationship between characteristics ofadolescents’ relationships with family and peers and behaviour problems and whether theserelationships were influenced by risk factors in the person and the environment. The resultsof the EQS modelling indicated that when adolescents experience multiple risk factors intheir family (high-risk family background, high perceived conflict and low perceivedsupport) and school career, this may lead to feelings of loneliness. These feelings incombination with low self-esteem, strict attitudes toward delinquent activities, highperceived conflict with a delinquent peer group whose opinion is less valued, lead to highlevels of internalizing behaviour. If the multiple risk factors in the family and school areaccompanied by an ineffective coping style, loose attitudes toward delinquent activities,high-risk leisure time activities and high perceived support from a delinquent peer groupwhose opinion is valued, this leads to higher levels of antisocial behaviour.

Literature results often report a protective effect of family support on behaviour problems(see review of Barrera and Li, 1996). This result was not found in the current study. Thismay be due to the fact that we have aggregated scores for the whole family, whereas moststudies have examined parental support. Our aggregated scores may also have maskeddistinct differences between parents and siblings. We did find an indirect effect of familysupport on internalizing behaviour mediated through feelings of loneliness.

A delinquent peer group proved to be a risk factor for behaviour problems. However, ahigh supportive delinquent peer group was more associated with antisocial behaviour,whereas a conflictuous delinquent peer group was more associated with internalizingbehaviour. There are several possible explanations for the first finding. First, adolescentsselect peers who are similar to themselves in multiple characteristics (Cairns et al., 1995;Buhrmester et al., 1996). This implies that adolescents with antisocial behaviour will selectpeers who are similar in behaviour. Second, high peer support from deviant peers may bedue to a projection mechanism, i.e. adolescents attribute their own behaviour to thebehaviour of their peers (Aseltine, 1995). This is also in line with the finding that looseattitudes toward delinquent activities were associated with reporting membership of adelinquent peer group. Third, high support from a delinquent peer group may enhance thedeviant behaviour of the adolescents (Cauce et al., 1990; Emler, 1993; Dishion et al., 1995).

The finding that a conflictous peer group in combination with strict attitudes concerningdelinquent activities was related to internalizing behaviour problems may be explained asfollows. Adolescents who have delinquent peers, but who think delinquent activities arestrictly forbidden, perceive conflict with these peers concerning these matters. Instead ofjoining their peers in antisocial behaviours, they internalize their problems.

Furthermore, we found a small indirect effect of high support from family on antisocialbehaviour. This high family support was combined with a high-risk family background andhigh perceived conflict with family. We assumed that bonding, i.e. high perceived support,from a deviant subsystem could enhance behaviour problems. In high-risk families, parentsmay also have deviant behaviours. Support in these relationships may enhance antisocialbehaviour in the adolescents. Another possible explanation may be that high perceivedsupport from a high-risk family may indicate deficits in parenting behaviour. Becauseparents have problems of their own, their control attempts on their children’s behaviourmay decrease or the problem behaviour of the adolescents may have the effect of raisingparental tolerance (Stice et al., 1993) and this could result in decreased parental controlattempts. Adolescents may perceive this decrease in parental control as social support. Italso seemed that antisocial behaviour was closely related to more ambivalent, i.e. supportive

656 W. H. Buysse

and conflictuous, relationships with family members. Other studies have also indicated thatthe receipt of social support from conflicted or ambivalent relationships plays a negativerole in personal adjustment (Barrera, 1981; Sandler and Barrera, 1984; Rook, 1990).

Loneliness was an imporant mediating variable for internalizing behaviour problems. Weconsidered loneliness as a construct which is directly related to characteristics of the socialnetwork and to internalizing behaviour problems. The results of the present study indicatethat loneliness is a direct indicator of internalizing behaviour problems. Feelings ofloneliness were also associated with characteristics of the relationships with family, but notwith peers. Finding family members’ opinion important, high perceived conflict with andhigh perceived support from family were related to more feelings of loneliness. Brennan(1982) notes that the reorganization of relationships in adolescence disrupts personal socialrelationships and that these disruptions presumably lead to loneliness. The results in thisstudy indicate that for the adolescents in our sample it seems that it is disruption ofrelationships with family, in particular, which may lead to these feelings of loneliness.Goswick amd Jones (1982) suggest that one of the primary causes for these disruptions inrelationships is relatively inadequate social skills. In this study an indirect relationship wasfound between feelings of loneliness and ineffective coping skills. These findings need to bestudied in further research which may focus more explicitly on the relationship of the abovevariables with loneliness.

Finally, the variables in our model only explained a small portion of the variance inbehaviour problems. This implies that other variables, such as parenting styles, may beincorporated in the study of the influence of social support and conflict on behaviourproblems.

In conclusion, exploring the relationships in the socio-ecological model demonstratedthat social support can operate as a protective or as a risk factor for behaviour problemsdepending on other characteristics of the subsystems involved. High support from a non-deviant system may prevent the development of behaviour problems. High support from orhigh bonding with a deviant subsystem may enhance behaviour problems. High perceivedconflict may be considered a risk factor for the development of behaviour problems,although the way in which it influences the behaviour is also dependent on the subsysteminvolved. The results of this study point clearly to the need for more focused attention tothe influence of environmental and personal risk factors in the study of social support andconflict in relation to adolescents’ behaviour problems.

Acknowledgements

The author would like to thank Prof. Dr J. D. van der Ploeg and Dr E. M. Scholte and thereviewer for their comments on earlier drafts of this manuscript.

References

Aseltine, R. H. (1995). A reconsideration of parental and peer influences on adolescent deviance.Journal of Health and Social Behaviour, 36, 103–121.

Barrera, M. Jr. (1981). Social support in the adjustment of pregnant adolescents. Assessment issues. InSocial Networks and Social Support, Gottlieb, B. H. (Ed.). Beverly Hills, CA: Sage, pp. 69–96.

657Family and peer relationships

Barrera, M. Jr. and Garrison-Jones, C. (1992). Family and peer social support as specific correlates ofadolescent depressive symptoms. Journal of Abnormal Child Psychology, 20, 1–16.

Barrera, M. Jr. and Li, S. A. (1996). The relation of family support to adolescents’ psychologicaldistress and behavior problems. In Handbook of Social Support and the Family, Pierce, G. R.,Sarason, B. R. and Sarason, I. G. (Eds). New York: Plenum Press, pp. 313–343.

Barrera, M. Jr., Chassin, L. and Rogosch, F. (1993). Effects of social support and conflict onadolescent children of alcoholic and nonalcoholic fathers. Journal of Personality and SocialPsychology, 64, 602–612.

Bentler, P. M. (1989). EQS: Structural Equations Manual. Los Angeles: BMDP Statistics Software.Brennan, T. (1982). Loneliness at adolescence. In Loneliness: A sourcebook of current theory, research

and therapy, Peplau, L. A. and Perlman, D. (Eds). New York: Wiley-Intersience.Berndt, T. J. (1989). Obtaining support from friends during childhood and adolescence. In Children’s

Social Networks and Social Supports, Belle, B. (Ed.). New York: Wiley & Sons, pp. 308–331.Blyth, D. A. and Traeger, C. (1988). Adolescent self-esteem and perceived relationships with parents

and peers. In Social Networks of Children, Adolescents and College Students, Salzinger, S.,Antrobus, J. and Hammer, M. (Eds). Hillsdale, NJ: Lawrence Erlbaum, pp. 171–194.

Blyth, D. A., Hill, J. P. and Thiel, K. S. (1982). Early adolescents’ significant others: Grade andgender differences in perceived relationships with familial and nonfamilial adults and youngpeople. Journal of Youth and Adolescence, 11, 425–450.

Bo, I. (1989). The significant people in the social networks of adolescents. In The Social World ofAdolescents, Hurrelman, K. and Engel, U. (Eds). Berlin: De Gruyter, pp. 141–165.

Buhrmester, D., Yin, J. and Tao, R. (1996). A longitudinal study of friends’ influence on adolescents’adjustment. Poster presented at the 1996 Conference of the International Network on PersonalRelationships, June 29–July 3, Seattle, WA.

Burke, R. J. and Weir, T. (1978). Sex differences in adolescent life stress, social support, and well-being. The Journal of Psychology, 98, 277–288.

Buysse, W. H. (1997). Personal social networks and behavior problems in adolescence. An explorative studyin three samples: a residential sample, a day treatment sample and a reference sample. Doctoraldissertation, Leiden University.

Buysse, W. H. and Van der Ploeg, J. D. (1992). Het Sociaal netwerk van jongeren in tehuizen [Socialnetworks of adolescents in residential institutions]. Unpublished document, Leiden University,Department of Special Education and Child Care.

Cairns, R. B., Leung, M. C., Buchanan, L. and Cairns, B. D. (1995). Friendship and social networksin children and adolescents: Fluidity, reliability and interrelations. Child Development, 66,1330–1345.

Cauce, A. M. (1986). Social networks and social competence: Exploring the effects of earlyadolescent friendships. American Journal of Community Psychology, 14, 607–628.

Cauce, A. M., Reid, M., Landesman, S. and Gonzales, N. (1990). Social support in young children:measurement, structure, and behavioral impact. In Social Support: An interactional view, Sarason,B. R., Sarason, I. G. and Pierce, G. R. (Eds). New York: Wiley, pp. 64–94.

Cauce, A. M., Mason, C., Gonzales, N., Hiraga, Y. and Liu, G. (1994). Social support duringadolescence: Methodological and theoretical considerations. In Social Networks and Social Supportin Childhood and Adolescence, Nestmann, F. and Hurrelmann, K. (Eds). Berlin, Germany: deGruyter, pp. 89–108.

Claes, M. E. (1992). Friendship and personal adjustment during adolescence. Journal of Adolescence,15, 39–55.

Compas, B. E. and Wagner, B. M. (1991). Psycho-social stress during adolescence: Intrapersonal andinterpersonal processes. In Adolescent Stress: Causes and consequences, Colten, M. E. and Gore, S.(Ed.). New York: Aldine de Gruyter, pp. 93–110.

Cotterel, J. L. (1992). The relation of attachments and supports to adolescent well-being and schooladjustment. Journal of Adolescent Research, 7, 28–42.

Dishion, T. J., French, D. C. and Patterson, G. R. (1995). The development and ecology of antisocialbehavior. In Developmental Psychopathology, Vol. 2: Risk, disorder and adaptation, Cichetti, D. andCohen, D. J. (Eds). New York: John Wiley & Sons, pp. 421–471.

Dunn, G., Everitt, B. and Pickles, A. (1993). Modelling Covariances and Latent Variables using EQS.London: Chapman & Hall.

658 W. H. Buysse

Emler, N. (1993). The young person’s relationship to the institutional order. In Adolescence and ItsSocial Worlds, Jackson, S. and Rodriguez-Tome, H. (Eds). Hove, U.K.: Lawrence Erlbaum,pp. 229–250.

Forehand, R., Neigbors, B. and Wierson, M. (1991). The transition to adolescence: The role ofgender and stress in problem behavior and competence. Journal of Child Psychology and Psychiatry,32, 929–937.

Furman, W. and Buhrmester, D. (1992). Age and sex differences in perceptions of networks ofpersonal relationships. Child Development, 63, 103–115.

Garmezy, N. (1987). Stress-resistent kinderen: Op zoek naar protectieve factoren [Stress resistantchildren: The search for protective factors]. In Protectieve Fatoren in de Ontwikkeling van Kinderenen Adolescenten, Groenendaal, H. J., Meijer, R. W. J., Veerman, J. W. and de Wit, J. (Eds). Lisse,The Netherlands: Swets and Zeitlinger, pp. 17–38.

Goswick, R. A. and Jones, W. H. (1982). Components of loneliness during adolescence. Journal ofYouth and Adolescence, 11, 373–383.

Granefski, N. and Diekstra, R. F. W. (1993). Scholierenonderzoek 1992: Landelijke uitkomsten rond eenonderzoek naar gedrag en gezondheid van ca. 11·000 scholieren van het voortgezet onderwijs[Studentresearch 1992: National results of a study concerning behavior and health of ca. 11,000students in secondary education]. Leiden, The Netherlands: Leiden University, Department ofClinical and Health Psychology.

Hirsch, B. J. and Dubois, D. L. (1992). The relation of peer social support and psychologicalsymptomatology during the transition to junior high school: A two-year longitudinal analysis.American Journal of Community Psychology, 20, 333–347.

Hirsch, B. J. and Rapkin, B. D. (1987). The transition to junior high school: A longitudinal study ofself-esteem, psychological symptopathology, school life, and social support. Child Development,58, 1235–1243.

Hirsch, B. J. and Reischl, T. M. (1985). Social networks and developmental psychopathology: Acomparison of adolescent children of a depressed, arthritic, or normal parent. Journal of AbnormalPsychology, 94, 272–281.

Jackson, S. and Rodriguez-Tome, H. (1993). Adolescence: Expanding social worlds. In Adolescenceand its Social Worlds, Jackson, S. and Rodriguez-Tome, H. (Eds). Hove, U.K.: Lawrence Erlbaum,pp. 1–13.

Jessor, R. (1981). The perceived environment in psychological theory and research. In Toward aPsychology of Situations, Magnusson, D. (Ed.). Hillsdale, NJ: Lawrence Erlbaum.

Kohnstamm, G. A. (1994). Hoe voorspelbaar zijn individuele verschillen in agressie? [Thepredictability of individual differences in aggression.] Kind en Adolescent, 12, 27–37.

MacCallum, R. C., Roznowski, M. and Necowitz, L. B. (1992). Model modifications in covariancestructure analysis: The problem of capitalization on chance. Psychological Bulletin, 11, 490–504.

Meeus, W. (1989). Parental and peer support in adolescence. In The Social World of Adolescents,Hurrelman, K. and Engel, U. (Eds). Berlin: De Gruyter, pp. 167–183.

Meeus, W. (1993). De psychosociale ontwikkeling van adolescenten [The psychosocial developmentof adolescents]. In Jongeren in Nederland. Een nationaal survey naar ontwikkeling in de adolescentieen naar intergenerationale overdracht, Meeus, W. and ’t Hart, H. (Eds). Amersfoort, TheNetherlands: Academische Uitgeverij, pp. 31–55.

Rook, K. S. (1990). Parallels in the study of social support and social strain. Journal of Social andClinical Psychology, 9, 118–132.

Sandler, I. N. and Barrera, M. Jr. (1984). Toward a multimethod approach to assessing the effects ofsocial support. American Journal of Community Psychology, 12, 37–52.

Slavin, L. A. and Rainer, K. L. (1990). Gender differences in emotional support and depressivesymptoms among adolescents: A prospective analysis. American Journal of Community Psychology,18, 407–421.

Smetana, J. G., Yau, J., Restrepo, A. and Braeges, J. L. (1991). Conflict in adolescence.Adolescent–parent conflict. In Adolescent Stress: Causes and consequences, Colten, M. E. andGore, S. (Eds). New York: Aldine de Gruyter, pp. 43–65.

Steinberg, L. D. (1987). Single parents, stepparents, and the susceptibility of adolescents to antisocialpeer pressure. Child Development, 58, 269–275.

Stice, E., Barrera, M. Jr. and Chassin, L. (1993). Relation of parental support and control to

659Family and peer relationships

adolescents’ externalizing symptomatology and substance abuse: A longitudinal examination ofcurvilineair effects. Journal of Abnormal Child Psychology, 21, 609–629.

Tracy, E. M. and Whittaker, J. K. (1990). The Social Network Map: Assessing social support inclinical practice Families in Society, 71, 461–470.

Van der Linden, F.J. (1991). Adolescent Lifeworld. Theoretical and empirical orientations in socializationprocesses of Dutch Youth. Amsterdam: Swets and Zeitlinger.

Van der Ploeg, J. D. (1983). De Omgeving in (Ortho)Pedagogisch Perspectief. Een beschouwing over hetkind als factor van de omgeving ter onderscheiding van de omgeving als factor van het kind [TheEnviromnent in Orthopedagogical Perspective]. Alphen a/d Rijn, The Netherlands: Samsom.

Van der Ploeg, J. D. (1992). Zelfbeooedelingslyst (ZBL) [Self Repoet List]. Unpublishedquestionniree. Leiden University: Vakgrsep Oethopedozoziek.

Van der Ploeg, J. D. and Scholte, E. M. (1990). Lastposten of Slachtoffers van de Samenleving[Troublemakers or Victims of Society]. Rotterdam, The Netherlands: Lemniscaat.

Verhulst, F. C., Koot, J. M., Akkerhuis, G. W. and Veerman, J. W. (1990). Practische Handleiding voorde CBCL [Manual for the CBCL]. Assen, The Netherlands: van Gorcum.

Vernberg, E. M. (1990). Psychological adjustment and experiences with peers during earlyadolescence: reciprocal, incidental, or unidirectional relationships? Journal of Abnormal ChildPsychology, 18, 187–198.

Wells, K., Wyatt, E. and Hobfoll, S. (1991). Factors associated with adaptation of youths dischargedfrom residential treatment. Children and Youth Services Review, 13, 199–216.

Windle, M. (1991). Temperament and social support in adolescence: Interrelations with depressivesymptoms and delinquent behaviors. Journal of Youth and Adolescence, 21, 1–21.

Windle, M. (1992). A longitudinal study of stress buffering for adolescent problem behaviors.Developmental Psychology, 28, 522–530.