Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

Embed Size (px)

Citation preview

  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    1/17

    Holism, Contextual Variability, and the Study of Friendships in Adolescent DevelopmentAuthor(s): Robert Crosnoe and Belinda NeedhamReviewed work(s):Source: Child Development, Vol. 75, No. 1 (Jan. - Feb., 2004), pp. 264-279Published by: Wiley on behalf of the Society for Research in Child Development

    Stable URL: http://www.jstor.org/stable/3696580 .Accessed: 13/01/2013 16:07

    Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

    .JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

    content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

    of scholarship. For more information about JSTOR, please contact [email protected].

    .

    Wiley and Society for Research in Child Developmentare collaborating with JSTOR to digitize, preserve and

    extend access to Child Development.

    http://www.jstor.org

    http://www.jstor.org/action/showPublisher?publisherCode=blackhttp://www.jstor.org/action/showPublisher?publisherCode=srcdhttp://www.jstor.org/stable/3696580?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/3696580?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=srcdhttp://www.jstor.org/action/showPublisher?publisherCode=black
  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    2/17

    Child Development, January/February 2004, Volume 75, Number 1, Pages 264-279

    Holism, ContextualVariability,and the Study of Friendshipsin Adolescent DevelopmentRobertCrosnoeand BelindaNeedham

    This study treated a key relationship in the developmental ecology of adolescence, friendships, asmultidimensional and context specific. First, it examined 4 characteristics of friends (academic achievement,alcohol use, emotional distress, and extracurricular participation) as independent factors and as components inholistic friendship group profiles. Longitudinal analyses of 9,224 adolescents (ages 12-20) revealed thatmultiple characteristics of friends predicted adolescent behavioral problems, as did membership in the bestadjusted group profile. Second, the study examined whether the associations between friendship factors andadolescent behavior varied as a function of the larger peer network and school context, finding that networkcentrality, school academic press, and intergenerational bonding in schools conditioned the role of friends'characteristics and group profiles in positive and negative ways.

    Human development occurs within a web of inter-personal relations. Friendships are an importantcomponent of this ecology throughout life, but theyplay an especially vital role during adolescence, astage characterized by increased peer orientationand gradual autonomy from family control (Anto-nucci & Akiyama, 1995; Hartup & Stevens, 1997).Reflecting this, peer relations have featured moreprominently in developmental research on adoles-cence than on other life stages. The present studycontinued this rich tradition by combining anecological perspective (Bronfenbrenner & Morris,

    RobertCrosnoeand BelindaNeedham,Departmentof Sociol-ogy and Population Research Center,University of Texas atAustin.The authors acknowledge the support of grants from theNationalInstituteof Child Healthand HumanDevelopment R01HD40428-02,PI: Chandra Muller) and the National ScienceFoundation (REC-0126167,Co-PI: Chandra Muller and PedroReyes) to the PopulationResearchCenter,Universityof Texas atAustin.Opinionsreflect those of the authors and not necessarilythose of the grantingagencies.This researchuses datafromAddHealth,a programprojectdesigned by J.RichardUdry,Peter S.Bearman,and KathleenMullan Harris,and funded by a grantP01-HD31921 rom the National Instituteof Child Health andHuman Development,with cooperative funding from 17 otheragencies.Specialacknowledgments due RonaldR. RindfussandBarbaraEntwisle for assistance in the original design. Personsinterested n obtainingdata files fromAdd Health shouldcontactAdd Health,CarolinaPopulationCenter,123 W. FranklinStreet,Chapel Hill, NC 27516-2524 www.cpc.unc.edu/addhealth/con-tract.html).

    Correspondenceoncerning his articleshould be addressed toRobertCrosnoe,Departmentof Sociology and Population Re-searchCenter,Universityof Texasat Austin,1 UniversityStationA1700, Austin, TX 78712-1088.Electronicmail may be sent [email protected].

    1998) with exciting themes in research on interper-sonal relations.First, the microsystem level of ecology encom-passes the direct, developmental effects of interper-sonal contexts. Our microsystem-level treatment offriendships examined the association between friend-ship group characteristics and adolescent behavioralproblems. On the basis of developmental sciencedemonstrating the value of characterizing friendshipgroups by multiple, interrelated factors (Xie, Cairns,& Cairns, 2001), we examined four characteristics offriends (academic achievement, alcohol use, mentalhealth, and extracurricular participation) as indepen-dent factors and as components of holisticallydefined friendship group profiles. Second, themesosystem-level of ecology concerns the develop-mental significance of interactions among contexts.On the basis of recent sociological research demon-strating how larger contexts moderate friendshipdynamics (Crosnoe, Cavanagh, & Elder, 2003; Hay-nie, 2001), our mesosystem-level treatment of friend-ships examined whether the association between thefriendship group and adolescent behavior is condi-tioned by the peer network and school.To pursue these two goals, this study drew on the

    National Longitudinal Study of Adolescent Health(Add Health), which is nationally representative,allows the measurement of friendship characteristicsby the reports of friends themselves, provides con-structed measures of peer networks, and includesschool-level information. Coupling our ecologicalapproach with this data source enabled us to deepenan already rich literature and point toward new

    ? 2004by the Societyfor Research n ChildDevelopment,Inc.All rightsreserved.0009-3920/2004/7501-0018

    This content downloaded on Sun, 13 Jan 2013 16:07:49 PMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    3/17

    Adolescent Friendship 265avenues of inquiry. On a more general level, suchresearch has implications for the study of interper-sonal relations throughout life.

    FriendshipsDuring AdolescenceAcross the life course, friendships direct devel-opment through support, modeling, and assistance,but their significance is heightened in adolescence(Antonucci & Akiyama, 1995; Hartup & Stevens,1997). In this stage, friendships enable adolescents tomeet a key developmental task-establishing theirown lives independent from their families-byhelping them develop identities, test conventionalboundaries, and gain autonomy from parents(Brown, Clasen, & Eicher, 1986; Giordano Cernko-vich, & Pugh, 1986). Consequently, friendship hasbeen central to research on adolescence.A good deal of research in this area has focusedon the associations between the characteristics offriends and the functioning of adolescents (Crosnoe,2000). A large portion of these observed associationsarises because of selection, the tendency of youngpeople to be attracted to others who are similar tothem or have characteristics that they desire (Kan-del, 1996). Yet, these associations persist even aftertaking selection into account, indicating that friendssocialize each other through modeling, reinforce-ment, and coercion (Aseltine, 1995; Matsueda &Anderson, 1998). Thus, the friendship group is an

    important microsystem-level context in the develop-mental ecology.This study examined these microsystem-leveleffects as multidimensional by focusing on fourcharacteristics of friends. Friends' alcohol use andfriends' academic achievement both predict adoles-cent functioning. Adolescents do better when theirfriends avoid drinking and make good grades(Aseltine, 1995; Mounts & Steinberg, 1995). Friends'emotional distress and friends' extracurricular par-ticipation have less often been studied directly.Of course, the distress and extracurricular participa-tion of adolescents in general have been studiedextensively; therefore, we drew on these relatedliteratures. Adolescent emotional distress is a well-documented correlate of poor individual adjust-ment, including problems in interpersonal relations(Hussong, Hicks, Levy, & Curran, 2001), whichsuggests that distressed friends will be less activein adolescents' lives and less likely to serve asprosocial models of behavior. Past research has alsodemonstrated that adolescents who participate inextracurricular activities are more integrated intoschool, have closer relations with school personnel,

    and are more likely to avoid problem behavior (withthe exception of athletes and alcohol) and that theytend to be friends with those who share their sameactivity profiles (Eccles & Barber, 1999; Eder &Parker,1987;Holland & Andre, 1987). These patternssuggest that friends who participate in such activ-ities may play a more positive developmental role.This study pursued a multidimensional treatmentof friendship in two ways. In the variable-centeredapproach (Magnusson & Cairns, 1996), we examinedwhether each of these characteristics is associatedwith adolescent behavioral problems. Thus, ratherthan investigate how a specific behavior of friends isassociated with adolescents' engagement in thatsame behavior (e.g., friends' achievement - ado-lescent achievement), which has been demonstratedsufficiently in past research, we examined whetherspecific characteristics of friends are associated witha more general dimension of adolescent behavior. Byexamining these diverse characteristics in tandem,moreover, we were able to compare and contrasttheir general developmental significance.In the person-oriented approach (Magnusson &Cairns, 1996), we examined how these four char-acteristics come together to form general friendshipgroup climates that predict adolescent behavior.Thus, we looked at the configuration of character-istics that identify a friendship group rather thanviewing these characteristics as independent factors.For example, the variable-oriented approach mayreveal that friends' alcohol use is directly related toadolescent behavioral problems but that friends'achievement is inversely related to such problems.The person-oriented approach, on the other hand,may reveal that adolescents do fine, or at least notbadly, in groups that are high in both drinking andachievement, as many high school athletic groupsare (Crosnoe, 2002a). In other words, the first multi-dimensional approach tells us which friendshipcharacteristics matter, controlling for other charac-teristics, and the second tells us how these char-acteristics come together to form a unified whole(Xie et al., 2001).Friendships n Context

    Social context has become increasingly prominentin research on adolescent friendships. Typically, suchresearch has examined how various contexts predictfriendship dynamics (Crosnoe, 2000; Kubitschek &Hallinan, 1998; Moody, 2001). This study took analternative approach by examining how contextscondition the significance of adolescent friendships.In ecological terms, such conditioning, or modera-

    This content downloaded on Sun, 13 Jan 2013 16:07:49 PMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    4/17

    266 Crosnoe and Needhamtion, occurs at the mesosystem level of the develop-ment ecology. Studying conditioning effects iscertainly not new in research on human relation-ships (see Astone, Nathanson, Schoen, & Kim, 1999),but it is less common in research on adolescentfriendships. Two recent studies, however, have demon-strated the value of this approach (Crosnoe et al.,2003; Haynie, 2001). We built on this work byexamining the mesosystem-level interaction of thefriendship group with two larger contexts in which itis embedded.The first of these two contexts is the peer network.Small groups are often units in networks of socialties (Ennett & Bauman, 1996). In adolescence, forexample, cliques are subgroups of diffuse peercrowds (Brown, 1990). The characteristics of thesenetworks predict behavior beyond the characteristicsof friends (Gold, 1970). This study was less con-cerned with the direct effects of peer networks thanwith their potential to moderate associations be-tween the characteristics of friends and adolescents.Recent research suggests that such moderation doesoccur for delinquency. Building on prior researchthat delinquent friends have greater effects onadolescent delinquency when the friendships arehigher in attachment and contact (Agnew, 1991;Giordano et al., 1986),Haynie (2001)found that charac-teristics of the larger peer network also conditionedthe associations between the delinquency of friendsand adolescents.

    To extend Haynie's (2001) network framework,this study examined whether peer network densityand centrality condition associations between friend-ship group characteristics and adolescent behavioralproblems. We expected that these associationswould be stronger when density-the degree towhich network members are connected to eachother-is high because dense networks increaseinteraction among network members and blockexternal influence. We also expected that theseassociations would be stronger when centrality-amember's ties to all others in a network-is highbecause more central network members are moreconnected to other individuals and more exposed tobehavioral models in the network (Haynie, 2001;Wasserman & Faust, 1994).The second context is the school. Again, pastresearch has effectively shown that characteristics ofschools (e.g., structure, composition, climate) predictfriendship formation, the characteristics of friend-ship groups, and participation in peer-based activ-ities (Kubitschek & Hallinan, 1998; Moody, 2001;Quiroz, Gonzalez, & Frank, 1996). Our focus,however, was on whether school factors condition

    the associations between the characteristics offriends and adolescents. Recent research has demon-strated that this process does occur for academicachievement, substance use, and delinquency. Suchresearch suggests that the general climate of theschool is the most important factor in this process(Cleveland & Weibe, 2003; Crosnoe, 2002b; Crosnoeet al., 2003).Building on this line of research, this studyexamined how two climate-related characteristicsof schools moderate the associations between multi-

    ple dimensions of adolescent friendship groups(independently and holistically) and adolescentbehavioral problems. First, academic press refers tothe academic rigor of schools (Shouse, 1996). Astrong emphasis on conventional achievement mayreinforce more positive influences from friends but,by increasing the potential costs of problem behav-ior, counterbalance more negative influences fromfriends (Crosnoe et al., 2003; Hirschi, 1998). Second,intergenerational bonding in school refers to thegeneral level of closeness between the student bodyand the teaching staff. Our interest in this schoolfactor arose from past research that had demon-strated that young people benefit from positiveparenting-not just the parenting they receive butalso the general level of parenting that characterizestheir peer group and their school (Coleman, 1988;Fletcher, Darling, Steinberg, & Dornbusch, 1995).Given the evidence that student-teacher bondingalso promotes adolescent adjustment (Sanders &Jordan, 2000), we argue that a collective treatment ofthis factor may also serve as an important aspect ofschool climate, one that may condition friendshipeffects in the same way as academic press.

    MethodData and Sample

    This study drew on Add Health, a nationallyrepresentative, ongoing study of adolescents inGrades 7 through 12 in 1994 (Bearman, Jones, &Udry, 1997). Sample schools were selected by region,urbanicity, school type, racial composition, and size.Each school was then matched to a feeder school onthe basis of the number of its students coming fromthe feeder school. From September 1994 throughApril 1995, available students in these schools(n = 90,118) completed the In-School Survey. A sub-sample (n = 20,745 in 132 schools), selected evenlyacross school pairs, then participated in the In-HomeInterview at Wave 1 (April-December 1985) andWave 2 (April-August 1996). These data can be

    This content downloaded on Sun, 13 Jan 2013 16:07:49 PMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    5/17

    Adolescent Friendship 267linked to interviews with parents and schooladministrators.Measures

    The dependent variable was measured with datafrom the Wave 1 In-Home Interview, and all othervariables with data from the In-School Survey thatpreceded it. Descriptive statistics for all studyvariables are included in Table 1.Adolescent behavioral problems.The dependentvariable was constructed to move beyond thefriends' characteristics studied here. For example,alcohol use, academic achievement, and emotionaldistress were all used to gauge characteristics of thefriendship group. Employing any one as the out-come would prioritize that characteristic of friendsand hamper our ability to examine how differentfriends' characteristics come together. To avoid this,we drew on the work of Cook, Herman, Phillips, andSettersten (2002), who argued that a measuresummarizing involvement in multiple domains ofbehavior better gauges overall adjustment in a waythat allows the comparison of diverse predictors ofadjustment. This approach is reflected in risk scalesin developmental and educational research (Crosnoeet al., 2003; Furstenberg, Cook, Eccles, Elder, &Sameroff, 1999). Thus, adolescent behavioral prob-lems is an index of four binary items: whether, inthe past year, the adolescent had stolen somethingworth more than $50 (1 = yes), engaged in sexualintercourse (1 = yes), had a record of truancy(1 = had skipped school five or more times), andgotten into a physical fight (1 = yes). These itemsload onto one factor in a factor analysis.Friends' characteristics.In the In-School Survey,each adolescent was asked to list up to five femaleand five male friends. The adolescent's data could bematched to the data of any friend who had alsoparticipated in the In-School Survey. Thus, allfriendship measures represent the mean of the self-reported characteristics of all listed friends in thesample and not the respondents' estimation of thecharacteristics of their friends. The former measuresare superior to the latter because adolescents over-estimate the extent to which their friends are similarto them and lack complete knowledge about whattheir friends think and do, a bias that has beenrepeatedly targeted as a primary weakness of re-search on peer dynamics (Billy, Rodgers, & Udry, 1984;Kandel, 1996).Items on the incidence and prevalence of alcoholconsumption in the past year were combined intoone scale, ranging from 0 (none) to 7 (everyday).

    Table1DescriptivetatisticsorAll StudyVariables

    M (SD) %Characteristics of adolescents

    Gender (female)Non-Hispanic WhiteAfrican AmericanHispanic AmericanAsian AmericanOther race and ethnicityFamily structure (two-parent)Age (years)Parent educationAlcohol useEmotional distressAcademic achievementExtracurricularparticipation

    Adolescent outcomesBehavioral problemsStealingSexual intercoursePhysical fightTruancyCharacteristics of friendship groupFriends' alcohol useFriends' emotional distressFriends' academic achievementFriends' extracurricular participation

    Peer network factorsNetwork densityNetwork centrality

    School factorsAcademic pressIntergenerational bonding

    0.540.540.190.160.090.030.55

    16.62(1.40)4.97

    (1.80)1.00

    (1.33)1.11(0.86)2.81(0.78)1.55(1.22)1.04

    (1.02) 0.200.420.280.14

    1.08(0.93)1.10(0.51)2.82(0.54)1.57(0.79)0.29

    (0.15)0.95(0.59)0.01(0.62)3.63

    (0.14)Note. n = 9,234.

    Adolescents' self-reported grades in four subjects(math, science, English, and social studies) wereaveraged and converted to a standard 4-point scale.The reports of all friends on these two measures,

    This content downloaded on Sun, 13 Jan 2013 16:07:49 PMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    6/17

  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    7/17

  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    8/17

    270 Crosnoe and NeedhamSecond, our sample contained only adolescentswho attended high school, including 7th and 8thgraders enrolled in comprehensive (7th- 12th grades)schools. Several behaviors of friends studied here are

    age graded, increasing significantly across the transi-tion to high school. At the same time, the schoolrepresents the market for friendship formation, withadolescents typically drawing their friends from theirschools. Considering both of these factors, 7th and8th graders who attend traditional middle schoolswill typically have friends with different character-istics than those of 9th through 12th graders becausetheir pool of potential friends is characterized bylower engagement in risk behavior and feweropportunities for activity participation in school.This difference in the pool of friends would likelydrive the assignment of these younger adolescents tofriendship group profiles in our clustering method.Yet, Add Health does contain some comprehensiveschools, and we argue that, because of the focus ofthis study, the 7th and 8th graders in these schoolsare qualitatively different from those in traditionalmiddle schools. Because they mix with olderadolescents on a regular basis, their pool of potentialfriends is likely to be more diverse. If we wereperforming a cluster analysis on the behaviors ofadolescents, therefore, we would separate all 7th and8th graders from older adolescents. Because we werefocusing on the behaviors of friends, however, wedecided to separate our cluster analysis by schoollocation (high school students and younger adoles-cents in comprehensive schools vs. younger adoles-cents in traditional middle or junior high schools). Inthis study, we focused on the first of these groups,and we return to the second group in the future.Third, the oversampling of some groups in AddHealth means that the raw data are not nationally rep-resentative (Chantala & Tabor, 1999). Samplingweights were required to correct for this. Thus, ouranalytical sample included only adolescents who hada valid sampling weight assigned to them. In doingso, we eliminated all students in the two schools inwhich sample weights could not be calculated.Fourth, the construction of friendship measuresrequired that respondents nominate at least onefriend who participated in In-School Survey andprovided valid information on the characteristic ofinterest. The portion of the sample missing onfriends' characteristics because of these require-ments was excluded. Of those excluded by thisfilter, most did not nominate any friends, and asmaller percentage nominated friends who were notin an Add Health school. Thus, this study is bestthought of as an examination of school friendships

    among adolescents who have friends. This is clearlydifferent from using a less specialized sample and isa drawback of using network data to study friend-ship dynamics. Yet, the value of using non-ego-based measures of friends' characteristics has beenso well documented and alternatives approaches soheavily criticized (Kandel, 1996) that we argue thatthe narrowing of the sample that results from usingnetwork data is balanced by other methodologicaladvantages of these data.The application of these filters resulted in a studysample of 9,234 adolescents in 71 schools (the samplesize for all multivariate analyses). These filters didbias the sample toward greater social advantage andmore conventional behavior (see Table 2). Thesebiases must be remembered in the interpretation ofresults, although we argue that these filters werenecessary to examine the dynamics at the heart of thisstudy and that the study sample had the advantagesof being longitudinal, multilevel, and multisource.

    ResultsFour AdolescentFriendshipGroups

    A key aim of this study was to treat the adolescentfriendship group in a multidimensional way. Wemeasured, therefore, four different characteristicswhere each represented one dimension of the friend-ship group. These four characteristics could also beused together to identify holistic friendship groupprofiles. From our cluster analysis, we selected foursuch friendship group profiles.These four profiles are presented graphically inFigure 1 (with standardized values, M = 0). Withineach profile, adolescents had similar types of friends.Across the four profiles, adolescents had differenttypes of friends. Because these profiles represent theconfiguration of various characteristics of friends,two profiles may have similar levels of one char-acteristic (e.g., friends' achievement), but no twoprofiles have the same configuration of friends' char-acteristics (e.g., the profiles similar in achievementdiffered in friends' alcohol use). In the following, wegive a general description of each friendship groupprofile. Descriptive statistics on the demographic,network, school, and friends' characteristics of ado-lescents in each friendship group profile are given inthe Appendix.Maladjusted. Adolescents who fit this profile hadfriends with the lowest level of adjustment. Theirfriends drank much more and were more emotion-ally distressed than the friends of adolescents inother profiles. Conversely, their friends were rela-

    This content downloaded on Sun, 13 Jan 2013 16:07:49 PMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    9/17

    Adolescent Friendship 271

    O Friends'Drinking ? Friends'EmotionalDistress0 Friends'Achievement ? Friends' articipation1.

    v:

    0tN

    0.5-

    0

    -0.5

    Maladjusted Disengaged Engaged High-FunctioningFriendshipGroupProfiles

    Figure1. Four adolescent friendship group profiles.

    tively disconnected from the schooling process, hadlower grades, and were less likely to participate inschooling activities than the friends of adolescents inall but one other profile.Disengaged. In this profile, the adolescents hadfriends who were low on all factors, those typicallyassociated with lower adjustment (e.g., drinking,emotional distress) as well as those with better func-tioning (e.g., academic achievement, extracurricularparticipation). In other words, their friends appearedto be withdrawn and removed from social life.

    Engaged. This friendship group profile was themirror image of the disengaged profile. The adoles-cents who fit this profile were high on all factors,those associated with both lower and higher adjust-ment. In other words, their friends were active inmultiple domains, both positive and negative. Theywere highly involved in schooling but, at the sametime, drank at a high frequency and had moreemotional problems.High functioning. The final profile consisted ofadolescents with friends who showed the highestlevel of overall adjustment. These friends werehighly involved in schooling, avoided alcohol use,and had the least emotional problems.DevelopmentalSignificance of AdolescentFriendshipGroups

    The next objective of this study was to examinewhether multiple dimensions of the friendship

    group were related to adolescent behavioral prob-lems. Table 3 presents the results of a series ofPoisson regression models relevant to this phenom-enon. In the base model (Model 1), the adolescentoutcome was regressed on sociodemographic con-trols as well as measures of the adolescents' level ofthe same characteristics used to measure friends'characteristics. As expected, adolescents who didwell in school and participated in school activitieshad fewer behavioral problems the next year, butthose who were emotionally distressed or drankalcohol had more.For the variable-centered analysis, we added tothis base model the four friends' characteristics(Model 2). Friends' characteristics predicted adoles-cent behavioral problems in the same way as thecorresponding adolescent characteristics in Model 1,except that the coefficient for friends' extracurricularparticipation was only marginally significant. Togauge the size of these effects, we first exponentia-ted and then standardized all Poisson coefficients.Doing so allowed us to interpret these transformedcoefficients as the amount of change in the outcome(expressed in terms of percentage of the standarddeviation of the outcome) for each standard devia-tion change in the independent variable and tocompare all coefficients in the model with eachother.Examination of these exponentiated, standardizedcoefficients in Model 2 of Table 3 (not shown)revealed that adolescents were .90 times less likely

    This content downloaded on Sun, 13 Jan 2013 16:07:49 PMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    10/17

    272 Crosnoeand NeedhamTable3Results fPoissonRegressionredicting ehavioralroblemsyFriend-shipFactors

    Model 1 Model 2 Model 3 Model 4Characteristics of adolescents

    Gender (female) -.28***(.03)Age (years) .08***(.01)Family structure -.16***

    (two-parent) (.03)Parent education -.04***(.01)

    African American .27***(.04)

    Hispanic American .11*(.05)Asian American .03

    (.09)Other race and .19*ethnicity (.08)Alcohol use .18**

    (.01)Emotional distress .12**

    (.02)Academic - .18** -

    achievement (.02)Extracurricular - .03*

    participation (.01)Characteristics of friendsFriends' alcohol useFriends' emotional

    distressFriends' academic

    achievementMissing on friends'academic

    achievementFriends'

    extracurricularparticipation

    Friendship group profilesMaladjustedDisengagedEngaged

    Intercept - .81Note. Unstandardized b coefficients aerrors in parentheses.tp

  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    11/17

    Adolescent Friendship 273among the friendship group profiles. In the finalmodel, the high-functioning friendship group profilehad a significantly lower level of behavioral prob-lems than the other three profiles. Compared withthis profile, membership in the three other profileswas related to roughly 5% of a standard deviationincrease in behavioral problems, an effect size, instandardized form, about equal to race and ethnicityand family socioeconomic status. Rotating thereference group revealed no other significant differ-ences among the four friendship group profiles.Intersectionof AdolescentFriendshipGroupsand SocialContext

    Having examined the developmental significanceof friendship factors, we now turn to the mesosys-tem level of the developmental ecology: the inter-section of the friendship group with other socialcontexts. Specifically, we examined whether aspectsof the peer network and the school moderated thelongitudinal associations between friendship factorsand adolescent behavioral problems.Table 4 presents the results for peer networks.Model 1 added the main effects of the two networkfactors to the final comprehensive model from theprevious table. Neither the density of adolescents'peer networks nor their centrality in these networkswere related to their level of behavioral problems.Model 2 in Table 4 added the interactions ofnetwork density with the four friends' characteris-tics and with the three friendship group profiles.None of these interactions was significant. Model 3added interactions of network centrality with thefour friends' characteristics and with the three friend-ship group profiles. Two of these interactions werestatistically significant. To interpret these interac-tions, we wrote out multiple equations-alternatingthe values of the friendship factors (1 SD below andabove the mean for friends' characteristics, 1 and 0for friendship group profiles) and the network factor(1 SD below and above the mean) and holdingall other variables in the models to their samplemeans.

    Beginning with the interaction between friends'alcohol use and network centrality, writing out theequations revealed that friends' alcohol use differ-entiated, in terms of behavioral problems, adoles-cents regardless of their location in their peernetworks. Adolescents, both central and noncentral,drank more when they had friends who drankabove-average amounts of alcohol. This increaseassociated with having friends who drank, however,was much greater for centrally located adolescents

    Table4Results From Poisson Regression PredictingProblemsby Friendshipand NetworkFactors

    Adolescent Behavioral

    Model 1 Model 2 Model 3Network factors

    Network density -.09(.08)Network centrality -.02(.03)

    Characteristics of friendsFriends' alcohol use .07**

    (.02)Friends' emotional .04

    distress (.03)Friends' academic -.15**

    achievement (.04)Missing on friends' .21tacademic achievement (.10)Friends' extracurricular -.03

    participation (.02)Friendship group profiles

    Maladjusted .13*(.06)Disengaged .11*(.05)Engaged .10*(.04)

    Interaction terms (Network x Friend)Density x Friends' AlcoholUseDensity x Friends EmotionalDistressDensity x Friends' AcademicAchievementDensity x Friends'

    ParticipationDensity x MaladjustedDensity x DisengagedDensity x Engaged

    -.28(.20)-.02(.03).07**

    (.02).04

    (.03)-.15***(.04).23*

    (.10)-.04(.02).13*

    (.06).11*

    (.05).10*

    (.04)

    - .08(.08).08

    (.06).10***

    (.02).03

    (.03)- .17'**(.04).30*

    (.11)-.06*(.03).06

    (.07).08

    (.05).10*

    (.05).06

    (.09).07

    (.12).01

    (.14).10

    (.11).15

    (.33).43

    (.32).18

    (.30)Centrality x Friends' Alcohol .09**Use (.03)Centrality x Friends' - - -.01Emotional Distress (.05)Centrality x Friends' - - -.06Academic Achievement (.06)Centrality x Friends' - - -.07

    Participation (.05)Centrality x Maladjusted - - .27*(.10)Centrality x Disengaged - - -.09(.08)Centrality x Engaged - - -.13(.09)

    Intercept -.44 -.39 -.44Note. Unstandardized b coefficients are presented, with standarderrors in parentheses. All models controlled for gender, age,family structure, race and ethnicity, parent education, adolescentacademic achievement, adolescent alcohol use, adolescent emo-tional distress, and adolescent extracurricular participation.tp

  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    12/17

    274 Crosnoeand Needham(157% increase in the predicted level of behavioralproblems) than for those who were more peripheralin their networks (64%).The interaction term for themaladjusted profile essentially reflected the samephenomenon-centrality strengthening the implica-tions for adolescent behavior of being in maladjustedfriendship groups compared with high-functioningfriendship groups. In general, centrally located adoles-cents in friendship groups high in alcohol use or inmaladjusted friendship groups had the highest ratesof behavioral problems, generally twice that of otheradolescents.Table 5 presents the results for school context. Asseen in Model 1, the main effect of school academicpress on adolescent behavioral problems was notsignificant. On the other hand, adolescents hadfewer behavioral problems when attending schoolshigher in intergenerational bonding: A 1 SD increasein intergenerational bonding was associated with10% of a standard deviation decrease in behavioralproblems. The magnitude of this association, instandardized form, was larger than all friendshipfactors. Models 2 and 3 added interaction termsbetween the two school factors and the two sets offriendship factors. We followed the same proceduredescribed earlier to interpret the multiple significantinteraction terms in these models.For academic press, three interactions terms weresignificant (note that one significant interaction wasfound when the engaged friendship group profileserved as the reference category and therefore is notseen in Table 5). First, all adolescents had fewerbehavioral problems when they had friends who didwell academically, but this decrease in the predictedlevel of behavioral problems related to friends'achievement was greater in academically rigorousschools (25%vs. 5%). Adolescents with high-achiev-ing friends in schools that emphasized academicsuccess had the lowest rates of behavioral problems.Second, a different pattern emerged for friends'alcohol use. The behavioral problems of adolescentswhose friends drank below-average amounts didnot fluctuate by level of academic press in theschool. On the other hand, adolescents who hadfriends who drank typically had more behavioralproblems (11% ncrease) when they attended academ-ically rigorous schools than when they attendedschools low in academic press. Third, in schools lowin academic press, adolescents in the disengagedfriendship group profile had more behavioral prob-lems than those in the engaged profile, but themirror image of this difference was found for schoolshigh in academic press (e.g., those in the engagedprofile had more problems).

    Table 5Results From Poisson Regression Models Predicting AdolescentBehavioralProblemsby FriendshipFactors and SchoolFactors

    Model1 Model 2 Model 3School factors

    School academic pressSchool intergenerational

    bondingCharacteristics of friendsFriends' alcohol use

    Friends' emotional distressFriends' academic

    achievementMissing on friends'academic achievementFriends' extracurricular

    participationFriendship group profiles

    MaladjustedDisengagedEngaged

    .02(.05)-.52*(.25).07**

    (.02).04

    (.03)-.13**(.04).17t

    (.09)-.03(.02).12*

    (.06).10t

    (.05).08t

    (.04)Interaction terms (School x Friend)

    Press x Friends' AlcoholUse

    Press x Friends'Emotional Distress

    Press x Friends'Academic Achievement

    Press x Friends'ParticipationPress x Maladjusted

    Press x DisengagedPress x Engaged

    .03(.08)

    .62*(.26).07**

    (.02).03

    (.03)-.15'**(.02).17t

    (.09)-.04'(.02).09

    (.06).06

    (.05).07

    (.04)

    -.01(.04)-.62(.38).08***

    (.02).04

    (.03)- .16***(.04).16'

    (.08)-.04'(.02).08

    (.06).04

    (.05).06

    (.04).06'

    (.03).00

    (.05)-.13*(.05)

    -.03(.04)

    -.07(.11)

    -.11(.08).06

    (.08)Bonding x Friends' .45**Alcohol Use (.16)Bonding x Friends' - - -.06Emotional Distress (.15)Bonding x Friends' -.62**Academic Achievement (.20)Bonding x Friends' - - .01

    Participation (.10)Bonding x Maladjusted - - -.24(.54)Bonding x Disengaged - - -.34(.33)Bonding x Engaged - - .61(.33)

    Intercept 1.54 1.83 1.85Note. Unstandardized b coefficients are presented, with standarderrors in parentheses. All models controlled for gender, age,family structure, race and ethnicity, parent education, adolescentacademic achievement, adolescent alcohol use, adolescent emo-tional distress, and adolescent extracurricular participation.tp

  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    13/17

    Adolescent Friendship 275For intergenerational bonding in the school, fiveinteraction terms were significant. First, attendanceat schools characterized by high levels of intergen-erational bonding was related to fewer behavioralproblems for all adolescents, regardless of theirfriends' characteristics, but this decrease was slightlystronger for adolescents who had friends who werelow achievers. Adolescents who had low-achievingfriends in schools with low levels of intergenera-tional bonding had the most behavioral problems,and those with high-achieving friends in schoolswith high levels of bonding had the least. Second,attendance at schools high in intergenerationalbonding was related to fewer behavioral problems,but only for adolescents who had friends who didnot drink (250% increase in predicted behavioralproblems across schools in this group). If adolescentshad friends who drank, their rate of behavioral

    problems did not vary by the level of bonding intheir school. Third through fifth, the maladjusted,disengaged, and high-functioning friendship groupprofiles all interacted with intergenerational bond-ing in school when compared with the engagedprofile (results for maladjusted and disengagedprofiles are not shown in Table 5). In all of theseprofiles, behavioral problems were lower in schoolswith higher levels of intergenerational bonding, butthis was less true of the adolescents in the engagedprofile.Discussion

    Friendships have long been a central focus of theliteratures on early development and lifelong inter-personal relations. This study, guided by theecological perspective and built on recent research,addressed both literatures by focusing on twoaspects of adolescent friendships: their multidimen-sional nature and sensitivity to social context. Inother words, this study attempted to capture thecomplexity of the adolescent friendship group andto locate it within the larger developmental ecology.First, we viewed the adolescent friendship groupon the microsystem level, as an interpersonal contextwith direct developmental significance for youngpeople. In doing so, we sought to cast a wider net byconsidering multiple dimensions of the friendshipgroup as independent factors and as components ofholistic friendship group profiles. The variable-centered approach revealed the additive value ofconsidering multiple characteristics of friends. Notsurprising, alcohol use, which is also considered aproblem behavior in the early life course, predictedgreater adolescent behavioral problems. Friends'

    achievement, however, predicted less behavioralproblems even after accounting for the level ofdrinking in the friendship group, supporting socialcontrol perspectives on the value of conventionalnorms (Hirschi, 1998). At the same time, friends'emotional distress, an understudied phenomenon,also predicted behavioral problems even after thesemore commonly studied aspects of the friendshipgroup were taken into account, suggesting that themental health of friends likely affects their ability toserve as prosocial models or social controls.The person-centered approach, on the other hand,demonstrated some value in using diverse charac-teristics of friends as interrelated parts of a largergroup climate rather than as competing predictors.For example, comparison of the disengaged andengaged friendship group profiles revealed thatfriends' drinking was less problematic if it occurredin tandem with more prosocial behaviors. Asanother example, friends' emotional distress waslowest in two profiles, disengaged and high func-tioning, which diverged sharply in adolescentbehavioral problems. In other words, friends' char-acteristics can balance each other in positive ornegative ways. Although many of these distinctionswere washed out by controls for the friends' char-acteristics that constituted the profiles, the signifi-cantly better adjustment of adolescents who fit thehigh-functioning friendship group profile indicatedthat, beyond the role of any one friendship factor,having friends who were well adjusted across theboard was more developmentally significant com-pared with having friendship groups that were moreinconsistent or were consistently maladjusted.In other words, multiple factors each has its ownrole in adolescent development, but these factorsalso come together in holistic packages to predictdevelopmental outcomes. In this study, the variable-centered and person-oriented approaches have beenset up as competitors, and in this sense, the variable-centered approach was the "winner." Yet, we arguethat each approach offers something different anddifferently valuable to the study of interpersonalrelationships and that they should be used intandem or even interchangeably, depending on theresearch focus. As discussed earlier, developmentalresearch has increasingly relied on both approachesto classify individuals and social contexts, a trendclosely related to sociological debates on newmethods (e.g., sequencing) for understating life-course dynamics (Abbott, 1995). Our experiencesuggests that the integration of multiple methodsmight very well be the best way to pursue theseimportant goals.

    This content downloaded on Sun, 13 Jan 2013 16:07:49 PMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    14/17

    276 Crosnoe and NeedhamSecond, we viewed the adolescent friendshipgroup on the mesosystem level, as a context thatinteracts with other ecological settings to directdevelopment. Specifically, we drew on recent socio-logical research to examine how the peer networkand the school moderate the associations between

    friends' characteristics (and friendship group pro-files) and adolescent behavioral outcomes.Beginning with the peer network, we found thatnetwork centrality, but not density, moderated thedevelopmental significance of friendship factors. Forexample, centrality strengthened the associations ofboth friends' drinking and membership in themaladjusted friendship profile with adolescentbehavior. More centrally located adolescents aremore exposed to behavioral models in the network.As a result, drinking in the network as well as theconfiguration of characteristics captured in the mal-

    adjusted profile are likely to have a greater impacton those at the center of the network. These findingsare similar to those of Haynie (2001), although forbehavioral problems (as opposed to delinquency)centrality was the primary network moderatorrather than density. This discrepancy suggests adomain-specific conditioning effect for differentaspects of social networks. These findings are alsorelevant to new fields of inquiry in social networkresearch that center on the structural aspects ofsocial cohesion, including their role in the flow ofnorms and information (Moody & White, 2003).Turning to school context, the role of friends, bothpositive and negative, was greater in academicallyrigorous schools, which suggests parallel, but re-lated, phenomena. Echoing earlier research on re-bellion (Stinchcombe, 1964), adolescent associationsthat are marginalized in a school (e.g., drinkingfriends in academically oriented schools) can fostermore problematic behavior, but social redundancy(e.g., friends' values reinforcing institutional values)is especially protective. In these same schools,adolescents in engaged friendship groups, whosefriends had both positive and negative character-istics, typically had more behavioral problems thanin other schools. Such adolescents might respond tomixed messages, among friends and between theirfriends and the school, in problematic ways. Anotherschool context moderated the significance of adoles-cent friendships in a different way. With oneexception, our results suggest that the negativeaspects of adolescent friendships were buffered inschools high in intergenerational bonding or, alter-natively, that adolescents in more problematic friend-ship groups benefited more from attendance at suchschools. This general pattern suggests that an ethos

    of intergenerational caring, closeness, and mentoringin an institution might be a form of protection amongcertain groups of students. The different patternsdemonstrated by these two school characteristicsmay indicate that they represent qualitatively differ-ent dimensions of the social context of schools-onenormative (e.g., standards) and one socioemotional(e.g., connections among people).For these mesosystem-level analyses, we expectedthat the network characteristics would strengthenthe developmental significance of all friendshipfactors and that the school characteristics wouldstrengthen the more positive factors and weaken themore negative factors, but the observed pattern ofinteractions was more complicated. Thus, our resultsindicate that contextual variability in the develop-mental significance of friendship factors is highlyspecific to the different dimensions of social contextand the friendship group being examined. Each ofthe examples of contextual variability that we pre-sented tells an interesting story and should be ex-plored more deeply. Doing so, and possibly lookingfor more general patterns of contextual variability,could be valuable. Coupled with the extensive pastresearch demonstrating the effects of social contextson friendship associations, such research more fullycaptures the complex ecology of adolescent devel-opment.Future research on adolescent friendships or otherrelationships can build on the approaches taken inthis study-variable centered, person oriented,interactive and correct some of its limitations. Forexample, this study focused on the developmentalsignificance of friendship group membership in arelatively short time frame, but a long-term perspec-tive would be valuable. Such a perspective mightlead to the examination of the developmentaltrajectories of adolescents who have different typesof friends or the comparison of friendship groupprofiles in adolescence, young adulthood, and otherstages of the life course. Such inquiries will be madepossible with the new wave of Add Health, whichwill follow respondents across the transition toyoung adulthood.Another potential improvement concerns themeasurement of the friendship group. This studyfocused on only four characteristics of friends, butother multidimensional treatments of the friendshipgroup can move beyond behavior and mental healthto examine the demographic characteristics offriends, the affective aspects of friendships, or somecombination of these. Likewise, this study focusedon a limited number of social contexts and a limitednumber of indicators of these social contexts, but a

    This content downloaded on Sun, 13 Jan 2013 16:07:49 PMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    15/17

    Adolescent Friendship 277more thorough investigation of contextual variabil-ity of friendship dynamics could move beyond thismore narrow focus. The family context is certainly alogical area in which to extend this research. Suchresearch can be done with Add Health or with otherdata sources that contain more extensive familyinformation. The upcoming educational supplementto Add Health, which will collect high schooltranscripts and data on course material and contentfor all respondents, will allow a much more detailedrendering of institutional context (at the school andcurricular levels) that will greatly benefit this line ofresearch.

    Pursuing these lines of inquiry is an importantgoal. By taking an ecological perspective on thedevelopmental significance of these groups, thisstudy extends the rich literature on adolescentfriendship. At the same time, the lessons of thisstudy and of the literature in which it is groun-ded-that friendships are multidimensional andcontext specific-can extend beyond the specifictype of personal relationship and the specific stageof the life course examined here. Drawing parallelsacross relationships with spouses, romantic partners,parents, offspring, coworkers, friends, and others atmultiple life stages can advance our more generalunderstanding of the ecology of human develop-ment. Understanding similarities in the multidimen-sional, contextualized nature of various relationshipswould be a good start.

    ReferencesAbbott,A. (1995).Sequenceanalysis:New methodsfor oldideas. Annual Reviewof Sociology,21, 93-113.Agnew, R. (1991).The interactiveeffects of peer variableson delinquency. Criminology,29, 47-72.Antonucci, T., & Akiyama, H. (1995). Convoys ofsocial relations:Family and friendships within a lifespan context. In R. Blieszner & V. Hilkevitch (Eds.),Handbookof aging and thefamily (pp. 355 -371). Westport,CT:Greenwood.Aseltine,R. (1995).A reconsideration f parentaland peerinfluences on adolescentdeviance.Journal fHealthandSocial Behavior,3, 103-121.Astone, N. M., Nathanson, C., Schoen, R., & Kim, Y. J.(1999).Family demography,social theory,and invest-ment in social capital. Populationand DevelopmentReview,25, 1-31.Bearman,P., Jones, J., & Udry,J. R. (1997).TheNationalLongitudinal Study of AdolescentHealth: Researchdesign.RetrievedNovember1, 2002,fromhttp://www.cpc.un-c.edu/projects/ addhealth/design.html.Billy,J.,Rodgers,J.,&Udry,J.R. (1984).Adolescentsexualbehavior and friendship choice. Social Forces, 62,653-678.

    Bronfenbrenner,U., & Morris,P. (1998).The ecology ofdevelopmentalprocesses.In W. Damon (Ed.),Handbookof child psychology, Vol. 1 (pp. 993-1028). New York:Wiley.Brown, B. B. (1990). Peer groups and peer cultures. InS. Feldman& G. Elliott(Eds.),At the thresholdpp. 171-196).Cambridge,MA: HarvardUniversityPress.Brown,B.B.,Clasen,D., &Eicher,S. (1986).Perceptionsofpeer pressure,peer conformitydispositions, and self-reported behavior among adolescents. DevelopmentalPsychology,22, 521-530.Chantala, K., & Tabor,J. (1999). Strategies o performdesign-basedanalysis using the Add Health data. RetrievedNovember 1, 2002, from http://www.cpc.unc.edu/projects/addhealth/ strategies.html.Cleveland,H. H., & Wiebe,R. P. (2003).The moderationof adolescent-to-peer similarity in tobacco use byschool levels of substance use. Child Development,74,279-291.Coleman,J. S. (1988).Social capital and the creation ofhuman capital. AmericanJournalof Sociology,94, 95-120.Cook,T.D., Herman,M.R.,Phillips,M.,&Settersten,R.A.Jr. (2002).How neighborhoods,families, peer groups,and schools jointly affect changes in early adolescentdevelopment. Child Development,73, 1283-1309.Crosnoe,R. (2000).Friendships n childhood and adoles-cence: The life course and new directions. SocialPsychologyQuarterly,63, 377-391.Crosnoe,R.(2002a).Academicandhealth-relatedrajectoriesin high school:The intersectionof genderand athletics.Journalof Health and Social Behavior,43, 317-335.Crosnoe, R. (2002b).High school curriculum track andadolescent associationwith delinquent friends.Journalof AdolescentResearch,17, 144-168.Crosnoe, R., Cavanagh, S., & Elder, G. H. Jr. (2003).Adolescent friendships as academic resources: Theintersection of social relationships, social structure,and institutional context. Sociological Perspectives, 46,331-352.Eccles,J., & Barber,B. (1999).Studentcouncil,volunteer-ing, basketball, or marching band: What kind ofextracurricularnvolvement matters?Journal f Adoles-cent Research,14, 10-43.Eder,D., & Parker,S. (1987).The culturalproductionandreproductionof gender: The effect of extracurricularactivities on peer-group culture. Sociology of Education,60, 200-213.Ennett,S., & Bauman,K. (1996).Adolescent social net-works:School, demographic,and longitudinal consid-erations. Journalof AdolescentResearch,11, 194-215.Fletcher,A., Darling, N., Steinberg,L., & Dornbusch,S.(1995).Thecompanythey keep:Relationof adolescents'adjustmentand behaviorand their friends'perceptionsof authoritativeparenting n the social network. Devel-opmentalPsychology,31, 300-310.Furstenberg,F, Cook, T., Eccles, J., Elder, G. H. Jr.,&Sameroff, A. (1999). Managing to make it: Urbanfamiliesand adolescent success. Chicago: University of ChicagoPress.

    This content downloaded on Sun, 13 Jan 2013 16:07:49 PMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    16/17

    278 Crosnoe and NeedhamGiordano, P., Cernkovich, S., & Pugh, M. (1986). Friend-ships and delinquency. AmericanJournalof Sociology,91,1170-1201.Gold, M. (1970). Delinquent behaviorin an American city.Belmont, CA: Brooks-Cole.Harris, K. M., Duncan, G. J., & Boisjoly, J. (2002).

    Evaluating the role of "nothing to lose" attitudes onrisky behavior in adolescence. Social Forces, 80,1005-1039.Hartup, W., & Stevens, N. (1997). Friendships andadaptation in the life course. PsychologicalBulletin, 121,355-370.Haynie, D. L. (2001). Delinquent peers revisited: Doesnetwork structure matter? AmericanJournalof Sociology,106, 1013-1057.Hirschi, T. (1998). Social bond theory. In F T. Cullen &R. Agnew (Eds.), Criminological theory: Past to present(pp. 167-174). Los Angeles: Roxbury.Holland, A., & Andre, T. (1987). Participation in extra-curricular activities in secondary school: What isknown, what needs to be known? Review of EducationalResearch,57, 437-446.Hussong, A. M., Hicks, R. E., Levy, S. A., & Curran, P. J.(2001). Specifying the relations between affect andheavy alcohol use among young adults. Journal ofAbnormalPsychology,110, 449-461.Kandel, D. (1996). The parental and peer contexts ofadolescent deviance: An algebra of interpersonal influ-ences. Journalof Drug Issues, 26, 289-315.Kubitschek, W., & Hallinan, M. (1998). Tracking andstudents' friendships. Social Psychology Quarterly, 61,1-15.Magnusson, D., & Cairns, R. B. (1996). Developmental

    science: Toward a unified framework. In R. B. Cairns,G. H. Elder, & E. J. Costello (Eds.), Developmental cience(pp. 7-30). New York:Cambridge University Press.

    Matsueda, R., & Anderson, K. (1998). The dynamics ofdelinquent peers and delinquent behavior. Criminology,36, 269-308.

    Moody, J. (2001). Race, school integration, and friendshipsegregation in America. American Journal of Sociology,107, 679-716.Moody, J., & White, D. R. (2003). Social cohesion andembeddedness: A hierarchical conception of social

    groups. AmericanSociologicalReview,68, 103-127.Mounts, N., & Steinberg, L. (1995). An ecological analysesof peer influence on adolescent grade point average anddrug use. DevelopmentalPsychology,31, 915-922.Quiroz, P. A., Gonzalez, N. E, & Frank, K. A. (1996).Carving a niche in the high school social structure:Formal and informal constraints on participation in theextracurriculum. Research in Sociology of Education, 11,93- 120.

    Sanders, M. G., & Jordan, W. J. (2000). Student-teacherrelations and academic achievement in high school.In M. G. Sanders (Ed.), Schooling students placed atrisk:Research,policy, and practicein the educationof poorand minority adolescents (pp. 65-82). Mahwah, NJ:Erlbaum.Shouse, R. C. (1996). Academic press and sense of

    community: Conflict and congruence in American highschools. Research n Sociology of Educationand Socializa-tion, 11, 173-202.Stinchcombe, A. (1964). Rebellion n a high school. Chicago:Quadrangle.Wasserman, S., & Faust, K. (1994). Social networkanalysis:Methods and applications. Cambridge, England: Cam-bridge University Press.Xie, H., Cairns, B. D., & Cairns, R. B. (2001). Predicting teenmotherhood and teen fatherhood: Individual character-istics and peer affiliations. Social Development, 10,488-511.

    AppendixBreakdownof Four FriendshipGroup Profiles on the Characteristicsof Friendsand the Demographic,Network, andSchoolCharacteristicsof Adolescents

    MeanMaladjusted Disengaged Engaged High functioningCharacteristics of friendsFriends' alcohol useFriends' emotional distressFriends' academic achievementFriends' extracurricular participationCharacteristics of adolescentsGender (female)

    Age (years)Parent educationFamily structure (two-parent)Non-Hispanic White

    2.32a1.43a2.52c1.18c

    .56b17.01a4.63b

    .48cd.65a

    1.03b1.41a3.05b2.33a

    .76c.89b2.40d1.05d

    .53bc16.55c4.49c

    .47cd.37c

    .51d.80c3.29a1.90b.62a

    16.80b5.38a.60b.64a

    .49c16.22d5.40a.63a.56b

    This content downloaded on Sun, 13 Jan 2013 16:07:49 PMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 Holism, Contextual Variability, And the Study of Friendships in Adolescent Development

    17/17