Early Childhood Antecedents of Aggression and Passive-Withdrawal in Early Elementary School

Early Childhood Antecedents of Aggression

and Passive-Withdrawal in Early Elementary

School

Bruce Renken, Byron Egeland, Denice Marvinney,Sarah Mangelsdorf, and L. Alan Sroufe

University of Minnesota

ABSTRACT Early childhood antecedents of elementary school aggressionand passive-withdrawal were evaluated in a large longitudinal study {N = 191)With teacher ratings as outcome measures, a vanety of predictor vanables wereselected from a comprehensive data base Predictors were selected to representseveral major factors (a) a developmental history of insecure attachment andpoor adjustment, {b) inadequate or hostile parental care, and (c) stressful or cha-otic hfe circumstances Support was found for the prediction of these behaviorproblems from early childhood measures (up to one third of the vanance), withresults varying with sex and outcome measure Results were stronger for boysthan girls, and stronger for aggression than for passive-withdrawal Boys' out-come was strongly related to attachment classification at 18 months It is sug-gested that these findings support Bowlby's view that early acquu-ed "workingmodels" of self and other affect later interpersonal functioning

An important aspect of child psychopathology which remains poorlyunderstood is that of its antecedents in early childhood While much canbe leamed from cross-sectional studies and from the retrospective re-ports of parents of disturbed children, the best method for identifyingthese antecedents is comprehensive longitudinal study of unselectedsamples The present study examines the antecedents of childhood ag-

This research has been supported by a grant from the National Institute of MentalHealth (MH 40864 01) Send repnnt requests to L Alan Sroufe, Institute of ChildDevelopment, University of Mmnesota, 51 East River Road, Minneapolis, MN55455 Sarah Mangelsdorf is now at the University of Michigan Denice MarvinneyIS now at Indiana University

Journal of Personality 57 2, June 1989 Copynght © 1989 by Duke UniversityPress CCC 0022-3506/89/$! 50

258 Kenken et al

gressive/antisocial behavior and passive-withdrawn behavior using therich data base of the Minnesota Mother-Child Project (Egeland &Sroufe, 1981, Enckson, Sroufe, & Egeland, 1985, Sroufe, 1983, Waters,Vaughn, & Egeland, 1980)

For several reasons, the Mother-Child Project offers a uniquely valu-able opportunity to study the development of behavioral disorders Be-gun in 1975, the project has earned out frequent and vaned assessmentsof both mothers and children, separately and m interaction with eachother, from the prenatal penod through elementary school Assessmentshave included well-validated observational techniques in addition toquestionnaire and self-report measures The sample is large (N = 191for the present study) and at high nsk for the development of problems ofall types, making it likely that children with particular types of behaviorproblems will be adequately represented

Aggressive/antisocial behavior is a suitable target for prospectivestudies for two reasons First, it is a childhood behavior problem withhigh frequency, notable stability, and negative sequelae Kagan andMoss (1962), Richman and colleagues (1982), Olweus (1979), and Rut-ter and Garmezy (1983) have all reported aggression to be stable acrossvanous phases of childhood Outcome studies (e g , Robins, 1978,West, 1982) have shown that aggressive, antisocial behavior m child-hood IS a strong predictor of a vanety of problems m adulthood, mclud-mg cnminality, alcoholism, and senous emotional disturbance

Second, several bodies of literature point to three sets of clear corre-lates of aggression/antisocial behavior, all of which may be accessed inour data The first of these is hostile, abusive, or punitive parental treat-ment (Becker, 1964, McCord, McCord, & Zola, 1959, Quinton & Rutter,1985) The second is mantal disharmony, chaotic home life, and high lifestress (Emery, 1982, Hethenngton, Cox, & Cox, 1982, Patterson, 1982,Rutter, 1971)

The third set of correlates suggested by the literature is a develop-mental history of avoidant attachment in mfancy and negativistic behav-ior in the toddler and preschool penod This developmental pattem hasbeen linked with matemal rejection/hostility and unavailability m earlychildhood (especially chronic rebuff when the infant seeks physical con-tact), and avoidant attachment in infancy has been specifically rdated toaggression in the preschool years (Egeland & Sroufe, 1981, Enckson etal , 1985, Londerville & Main, 1981, Sroufe, 1983, Troy & Sroufe,

Antecedents of Aggression 259

1987) This link IS supported theoretically as well Such pattemso^ careinterfere with and distort the child's normal tendency to seek emotionalcloseness with others dunng times of stress or arousal and instead acti-vate the normative angry response to literal abandonment (Bowlby,1973)

For the most part these bodies of evidence consist of contemporaneousassessments or studies confined to early childhood Nevertheless, onewould expect these three domains to form the basis of a predictive studyof aggression m middle childhood, especially if considered in combi-nation Assessments available m our data, beginning in infancy, ade-quately tap each of these domains

Passive-withdrawal, while of interest m its own nght, serves as a use-ful contrast for this study of the ongins of aggression There is less guid-ance from the literature as to potential antecedents of this behavior pat-tem Still, certain hypothesized factors may be suggested First, onewould expect general effects of poverty and neglect on the child's senseof agency and resourcefulness The inadequate resources for stimula-tion, disorganization, instability in the living situation, and physical ne-glect or inconsistency in parenting would seem to be important factorsIn other words, the child might not be physically abused or rejected, butthere may simply be insufficient support for autonomous functioning andthe growth of effectance Such care has been associated with an anxious/resistant pattem of attachment (Ainsworth, Blehar, Waters, & Wall,1978, Egeland & Sroufe, 1981, Sroufe, 1985), and may be expected topredict passive-withdrawal as well Thus, two domains of correlates ofmiddle childhood passive-withdrawal are suggested first, a pattem ofinadequate parental care, and second, a developmental history of anx-lous/resistant attachment and poor coping and mastery efforts Measuresof such factors are again well represented in our longitudinal data base

Our developmental conceptualization of aggression and passive-with-drawal IS guided by Bowlby's (1973) concept of "intemal workingmodels " In Bowlby's theory one denves core expectations about the en-vironment, and especially conceming the availability and responsive-ness of others and the potency and worthiness of the self, from early in-teractive experience Thus, an infant who has experienced reliable,sensitive care (responsiveness to its signals, overtures, moods, andstates) will come to expect not only that the caregiver is available but thathe or she (the infant) is effective in eliciting care In short, the infant will

260 Renken et ca

be secure in the attachment relationship Very different models of self,other, and relationships would derive from expenences of chronic rebuffor mconsistent, haphazard care

These internal working models provide guiding frameworks for fur-ther transactions with the environment, including the selection and inter-pretation of experience and the way one approaches others A child thatexpects to be rebuffed by others may not seek them out when needy, mayinterpret benign behavior as hostile, and may strike out aggressivelyThus, continuity m individual adaptation is not due so much to the pres-ence of static traits in the child as to a transactional process guided byexpenence

This developmental perspective leads to more complex predictionsthan the expectations of literal behavioral stabihty Infants with anxious/avoidant attachment (who tum away from or ignore their mothers uponreunion following a bnef separation) are predicted to show later aggres-sion, not passive-withdrawal (although they are considered "emotionallyisolated" from others, Sroufe, 1983) This occurs for two reasons First,later social encounters will be guided by the expectations that others arenot available, that social encounters are not rewarding, and that the selfIS unworthy of care These children have internalized a rejecting rela-tionship and carry that forward to new social relationships (Sroufe &Fleeson, 1986)

Second, the early avoidance itself is seen as a displacement behaviorresulting from conflict between the strong, biologically based tendencyto approach the caregiver when distressed and the anger and pain asso-ciated with the expenence of the caregiver as unavailable Direct expres-sion of anger might further alienate this rejecting caregiver, leaving theinfant completely vulnerable Avoidance in this context has the functionof checking such expression (Mam, 1977)

Likewise, those with anxious/resistant attachment (who are difficultto settle upon reunion, with much crying and, in fact, often explicit an-ger) are predicted to be quite passive later, rather than to show hostileaggression This again denves from a consideration of the workingmodels of self and other being constructed m the early years Because ofmconsistent, chaotic, and haphazard care these infants have come todoubt the certainty of care They have not given up on the possibility ofnurtunng adult care, but they doubt the effectiveness of their efforts toachieve it or to function autonomously without it They remain adult-onented and emotionally dependent Their agency is compromised and

Antecedents of Aggression 261

they are inept with peers, resulting in the appearance of apathy and lackof social salience A quite different denvative from this attachment pat-tem IS a tense, anxious, hyperactive pattem of adaptation Researchersare just beginning to explore the concept of branching developmentalpathways (Sroufe, in press)

The goals, then, underlying the present article are twofold First, wewill undertake a multivanate investigation of the antecedents of aggres-sion and passive-withdrawal By considenng several factors thought tobe linked to these behavior patterns, separately and m combinationthrough a senes of regression analyses, it may be possible to predict ag-gressive behavior and passive-withdrawal in the elementary years evenstarting in the first years of life Such data is important for prevention andearly intervention efforts Second, we are interested in examining the dif-ferential prediction of Ainsworth's two pattems of anxious attachment minfancy We hypothesize that the anxious/avoidant pattem will predictaggression, the anxious/resistant pattem will predict passive-with-drawal, and that secure attachment will not be related to either form ofbehavioral problem

METHOD

Subjects

Subjects are part of the Minnesota Mother-Child Project, an ongoing longi-tudinal study of a group of children considered to be at high nsk for caretak-lng problems Two hundred sixty-seven mothers receiving prenatal carethrough the Minneapolis Public Health Clinic m 1975-1977 were recruitedfor the study Of these mothers, 41% had not completed high school at thetime of the baby's birth, 62% were unmamed, and 86% of the pregnancieswere unplanned In general the mothers were of low socioeconomic statusand their lives were marked by a high degree of stress and instability As-sessments of mothers and children have included objective psychologicaltests, interviews, questionnaires, and observations of child behavior andmother-child interactions Because some subjects have dropped out over theyears, 191 children (106 boys and 85 girls) constitute the sample for thisstudy

Outcome Measures

For the present study, teacher ratings of aggression and passive-withdrawalm first through third grades were the outcome variables Teacher ratings

262 Kenken et a l

were used rather than parent ratings to avoid a matemal response bias ThisIS particularly important since such biases would also affect many of the pre-dictor vanables Each year the child's teacher completed two standardizedmeasures of child adjustment the Teacher Version of the Child Behavior Pro-file (Achenbach & Edelbrock, 1983) and the Devereux Elementary SchoolBehavior Rating Scale II (1982) An aggression scale denved by Achenbachand Edelbrock from factor analyses of the Behavior Checklist (Achenbach& Edelbrock, 1983) was used as a measure of aggression This scale con-sists of 34 Items showing the highest loading on the aggressive factor' Thepassive-withdrawal scale was formed from nine Child Behavior ProfileItems and four items from the Devereux Rating Scale Table 1 contains a hstof the Items of the two scales The aggression items tap behaviors such asdefiance, destructive behavior, disobedience, disturbmg others, lying, andfighting Passive-withdrawal items are related to apathy, lack of initiative orengagement m activities, and shyness and avoidance of social activityScores were averaged over the 3 years

The validity of each of these measures was evaluated m several waysFirst, Item-total correlations were examined as a measure of the homogene-ity of each scale Second, year-to-year stabilities were determined Third,convergent validity was assessed by companng each measure with an obser-vational rating of the same construct Finally, discnminant validity was as-sessed by companng each measure to a measure of intellectual competence(a subset of the Wechsler Intellectual Scales for Children [WISC-R]) ^

Ratings of aggression and passive-withdrawal were obtained from class-room observations of a subset of the children (n = 101) Graduate studentobservers spent between 4 and 8 hours observing each child m a vanety ofclassroom activities They then rated each child on the Califomia Q-Sort(Block & Block, 1980), a l(X)-item forced ranking of child charactensticsInterobserver reliability for the entire Q-sort averaged between 50 and 60,slightly less than the average interobserver reliability of 65 (lntraclass cor-relation) for children aged 3 and 4 in the Block's ongmal study (Block &Block, 1980) From the larger set of Q-sort items, a set of 8 items was cho-

1 Achenbach and Edelbrock denved separate aggression factors for their male andfemale samples, resulting in a 35-item scale for boys and a 38-item scale for girlsThirty-four of these items were the same for both sexes, and the nonoverlappmgItems showed quite low loadings on the factors We decided to use only the 34 over-lapping ltenis, since this alternation produced a minuscule change in children's rel-ative scores and resulted m a substantial savings m computer time and complexityof analyses2 When the subjects were m third grade, the Similanties, Vocabulary, and BlockDesign subtests of the WISC-R were given, and subtest scores were used to denvean estimate of full-scale IQ

Antecedents ol Aggression

Table 1Items ol the Aggression and Passive-Withdrawal Scales

263

367

1619202123242526273739414353

Child4

4260657580

100102111

ArguesDefiantBraggingCrueltyDemands attentionDestroys own thmgsTeasesDisobedientDisturbs othersPoor peer relationsLacks guiltJealousFightsBad fnendsImpulsiveLying, cheatingTalks out of tum

Aggression scale

57676873747677868788899093949597

104

Attacks peopleDisrupts classScreamsActs irresponsiblyShows offExplosiveEasily frustratedStubbornMoodySulksSuspiciousSwearingTalks too muchTeasesTemper tantrumsThreatensLoud

Passive-withdrawal scale

Behavior Profile itemsFails to finish thingsLikes to be aloneApatheticWon't talkShyStares blanklyFails to carry out tasksUnderactiveWithdrawn

Devereux items11

21

4348

Avoids involvement orcommunication"Just sits and doesn'tparticipate^Needs precise directionsRelies upon teacher fordirections

Note All Items m the aggression scale were taken from the Teacher Versionof the Child Behavior Profile (Achenbach & Edelbrock, 1983) The first nineItems in the passive-withdrawal scale were also taken from the TeacherVersion of the Child Behavior Profile, while the last four items were takenfrom the Devereux Elementary School Behavior Rating Scale II (Devereux,1982)a These items were not used in the first grade because they were not includedin the earlier version of the Devereux scale being used at that time

264 Renken et a l

sen to form an aggression scale, and a set of 11 items was chosen to form apassive-withdrawal scale Items were chosen on the basis of content, withthe intent of matching as closely as possible the content of the teacher ratingscales of aggression and passive-withdrawal ^

Predictor VanablesThe Mother-Child Project has an extensive body of data on mothers andchildren from birth through early elementary school A limited number ofpredictor vanables were selected to represent as efficiently as possible themajor domains of matemal and child functioning These domains, and thevanables selected to represent them, are descnbed m this section

Child history cf social-emotional competence At 12 months and again at18 months, the quality of the mother-child attachment relationship was as-sessed m the Strange Situation (Ainsworth & Wittig, 1969), a standardizedsequence of eight episodes involving bnef separations from the mother andcontact with a stranger Ratings are made of the infant's response to thestress of separation, response to reunion, and response to the stranger, withand without the mother present Each relationship was categonzed, at eachof the two assessments, as secure, lnsecure-avoidant, or msecure-resistantClassifying an infant as secure indicates that the infant is able to rely on thecaregiver's supportive presence as a means of reducing distress Classifi-cation as insecure-avoidant indicates that instead of tuming to the mother attimes of distress, the infant actively avoids interaction with the mother Clas-sification as insecure-resistant indicates that the infant seeks contact withthe mother when distressed, but signs of anger and continued distress aremanifested while seeking contact Attachment classification has been exten-

3 Q-sort aggressive scale 2 (Reflected) Is considerate and thoughtful of other chil-dren. 6 (Reflected) Is helpful and cooperative, 13 Charactenstically pushes andtnes to stretch limits, sees what s/he can get way with, 62 (Reflected) Is obedientand compliant, 79 Tends to be suspicious and distrustful of others, 80 Teases otherchildren (including siblings). 85 is aggressive (physically or verbally), 95 Over-reacts to minor frustrations, is easily lmtated and/or angered Q-sort passive-with-drawal scale 8 Tendstokeepthoughts, feehngs, or products to self, 21 (Reflected)Tnes to be the center of attention (e g by showing off, demonstrating accomplish-ments, volunteenng, etc ), 35 Is inhibited and constncted, 36 (Reflected) Is re-souiteful m mitiatmg activities, 44 When in conflict or disagreement with others,tends to yield and give m, 45 Tends to withdraw and disengage when under stress,52 Is physically cautious, 82 (Reflected) Is self-assertjve, 85 (Reflected) Is ag-gressive (physically or verbally), 93 (Reflected) Behaves in a dominating mannerwith others, 98 Is shy and reserved, makes social contacts slowly

Antecedents of Aggression 265

sively validated as a measure of important individual differences in early life(Ainsworth et al , 1978, Bretherton & Waters, 1985, Matas, Arend, &Sroufe, 1978)

Attachment classifications were compared du^ectly with outcome cate-gones in the second phase of the analyses, with the use of contingency ta-bles For the first, multivanate prediction-of-outcome phase, however, itwas desirable to denve continuous measures of attachment from the cate-goncal ratings This was accomplished in the following manner A new var-iable was created to represent each attachment type, with values "0," " 1 , "and "2 " For example, a child who was avoidantly attached at both the 12-and the 18-month assessments received a score of 2 on the avoidant attach-ment vanable, a child who was avoidantly attached at only one of these as-sessments received a score of 1, and a child who was avoidantly attached atneither assessment received a score of 0 Thus, each subject was given ascore on each of three quasi-continuous scales, one for each type of attach-ment ''

In addition to attachment classification, several other vanables were in-cluded under the heading of the child's social-emotional competence At 24and 42 months, children were assessed in a senes of problem-solving tasks,sometimes with the mother present and sometimes in her absence Ratmgswere made of a vanety of the child's affective responses and coping behav-iors Four- to seven-point scales were developed by research staff to assesseach vanable, and ratings were earned out on a subset of subjects until sat-isfactory reliability was attained (Matas et al , 1978) Final ratings were ob-tained by averaging across two observers Where differences of more thantwo scale points occurred, ratmgs were determined by conference of the twoobservers

Selected for this study were ratings of negative affect, enthusiasm, andcoping efforts Negative affect was rated on a 4-point scale, with the lowend indicating "neutral or positive affect, with no instances of negative af-fect" and the high end indicating "predominant negative affect, half the timeor more spent either trying to leave the task or crying, hitting, etc " Enthu-siasm was rated on a 7-point scale, with the low end indicating no enthusi-asm or an active effort to avoid the task, and the high end marked by clearenthusiasm, effort, involvement, and enjoyment of the activity Coping wasalso rated on a 7-pomt scale, marked at the low end by "extremely low cop-ing ability Even with support this child becomes disorganized in the

4 This procedure may appear to violate the assumpuon of discrete categones ofattachment However, we believe that given the nature of group data and the exis-tence of measurement error, it is fair to assume that the "2" group contains moreactual examples of the designated attachment type than the " 1" group, which m tumcontains more examples than the "0" group

266 Kenken et al

face of challenge " and at the high end "The child stays organized andinvolved in the face of great challenge or frustration "

Maternal care variables Vanables relating to the quality of matemal care-takmg were drawn from several sources Observation-based ratings of moth-ers' behavior toward their children were made at 42 months Mothers andchildren were engaged m a set of videotaped teaching tasks, m which moth-ers W2re required to explain the tasks and guide the child's performanceRating scales were developed in a manner similar to that descnbed above forthe child vanables

Of special relevance for the present study were ratings of matemal hostil-ity toward the child and a factonal score related to the mother's sensitivityand emotional supportiveness toward the child The 7-point scale of mater-nal hostility was designed to reflect the mother's expression of anger or re-jection, with high scores indicating a tendency to overtly reject the child,blame him or her for mistakes, and otherwise make explicit the message thatshe does not support the child emotionally A factor analysis was earned outon 14 ratings of both mother and child behavior m the teaching task, of threefactors extracted, the second factor seemed to represent a general factor ofmatemal competence, with its largest loadings on "mother's supportivepresence" (83), "mother's respect for the child's autonomy" ( 79), and"mother's quality of instruction" ( 77) This factor was included in the pres-ent study as a measure of matemal caretaking skill

Supplementing the observational measures, the Home Observation forMeasurement of the Environment (HOME, Caldwell, 1979), a matemal in-terview measure designed to assess the degree to which the mother providesan appropriate environment and facilitates emotional and intellectualgrowth, was administered when the child was in first grade The total scorewas taken as another global measure of matemal skill and supportiveness

In 1984, when most of the children were 7 and 8 years old, each child'scomplete record was examined for evidence of abusive or grossly negligentcaretaking practices Simple dichotomous classifications were made to in-dicate the presence or absence of each of several types of abuse These di-chotomous vanables (with " 1" indicating the presence of abuse and "0" m-dicating Its absence) were treated as continuous vanables in the multipleregression analyses Most cases of abuse found ongmated m the infancy andtoddler penods The two types of abuse considered relevant for this studywere physical abuse and gross negligence Of the 191 subjects in the sample,16 were found to have suffered physical abuse and 17 were found to haveexpenenced gross neglect

Finally, an assessment of mothers' intelligence was included as an addi-tional indicator of matemal competence This consisted of an IQ estimate

Antecedents of Aggression 267

based on three subtests (Similanties, Comprehension, and Block Design) ofthe Wechsler Adult Intelligence Scale

Life stress Several measures of stressful life circumstances were avail-able Part of this information comes from questionnaires and interviews con-ducted regularly over the entire period of the study, in which mothers pro-vided information about current life events and relationships At eachassessment mothers completed a 44-item hfe events questionnaire, withItems relating to a vanety of social and economic stressors A total score wascomputed by summing the number of stressors, with weights assigned ac-cording to the seventy of each stressor (Egeland, Breitenbucher, & Rosen-berg, 1980) Based on interview responses, researchers assigned eachmother a l-to-7 rating at each assessment indicating the degree of emotionalsupport available to the mother from her partner, fnends, extended family,and social service agencies As with other researcher rating scales in thisproject, ratmgs were earned out on a subset of subjects until adequate reli-ability was obtained Analyses of the relation of life stress vanables withoutcome measures were initially broken down by developmental periodSince no difference was found across penods, the life events and emotionalsupport scores were averaged across assessments from (child's) age 1 to 6years

As an additional measure of the degree of life stress, a measure of socio-economic status was denved by combining mformation about parents' edu-cational and income levels when the child was m the third grade Mothers'and fathers' educational levels (averaged if two parents were present) wereplaced on a 7-point scale and family income level was indicated on a 6-pointscale These two scores were then averaged to form an estimate of socioec-onomic status (SES)

Selecting sets of predictors for each outcome Predictor vanables were se-lected separately for each of the two outcomes, aggression and passive-with-drawal, in accordance with the hypotheses stated above Predictors ofaggression were organized into three domains, or "sets" (a) A develop-mental history of avoidant attachment and negative affect The continuousvanable representing a history of avoidant attachment (descnbed above) wasincluded here Also included were observational ratings of negative affecttaken from each of three problem-solving tasks, one at 24 months (motherpresent), and two at 42 months, one with mother present (the "teachingtask") and one in her absence {b) A history of hostile and abusive matemaltreatment Matemal hostility observed in the 42-month teaching task and apositive record of physical abuse were taken as measures of this type of care(c) Stressful life circumstances This domain was represented by threemeasures the life events score, emotional support score, and SES

268 Renken et a l

Zero-order correlations were computed between each predictor vanableand aggression, with the expectation of a significant relation in each caseNext, to assess the full predictive power of each domain or set of predictors,a multiple correlation was computed between the aggression vanable andeach set of predictor vanables Finally, to estimate the overall degree of pre-diction and to compare the power of different predictive domains, hierarchi-cal multiple regressions were earned out in which the aggression vanablewas regressed successively on the three sets of predictors The child's de-velopmental history vanables were entered first, followed by the matemaleare vanables, followed by the life stress vanables '

Predietor vanables were seleeted for passive-withdrawal m a similar man-ner Two sets of predietors of passive-withdrawal were selected in accord-ance with the hypotheses stated above (a) Resistant attachment and a de-velopmental history of diminished enthusiasm and weak coping attemptsThis included the continuous vanable representmg a history of resistant at-tachment Also included were ratings of coping attempts and enthusiasm mthe 24-month session, and ratings of agentic behaviors and enthusiasm in the42-month sessions (b) Incompetent care The four vanables mcluded herewere the total score on the HOME, the 42-month matemal care factor score,a record of grossly negligent care, and low estimated matemal IQ Multiplecorrelations and a hierarchical multiple regression were earned out withthese predietors m the same manner as with the aggression predictors In thehierarchical multiple regression, the developmental history predietor set wasentered first, followed by the matemal eare vanables

To address the issue of the relation of attachment classification and out-come, a strategy consistent with the qualitative nature of the attachment datawas chosen Scores on the aggression and passive-withdrawal scales wereused to define three outcome groups an aggressive, a passive-withdrawn,and a control group Those children seonng in the top 20% of their sex onaggression constituted an aggressive group * An attempt was made to definea passive-withdrawn group in an analogous manner (the top 20% of eachsex on the passive-withdrawal scale) but the groups so defined were found

5 Other ordenngs would be justifiable and interesting, but to mvestigate multipleordenngs would go beyond the appropnate boundanes of this article We felt that thechild's own history would be expected to be the strongest predictor of outcome, andthat It would be interesting to see if maternal care and hfe circumstances added sig-nificant predictive power6 A percentage cutoff was used instead of the usual standard deviation because thedistnbution of scores on the outcome measures was not normal The distnbution washeavily skewed toward the low (nonpathological) end TWenty percent was chosen asa cutoff in order to obtain a clearly ckviant group that was sizeable enough to affordmeaningful statistical analyses

Antecedents of Aggression 269

to include a substantial number ot children with high aggression scores Inthe interests of defining a group which met our a pnon conception of pas-sive-withdrawal as distinctiy different from aggression, a final group wasobtained by selecting those children who fell within the top 30% of their sexon passive-withdrawal and in the bottom 50% of their sex on aggression ^The remaining subjects constituted the control group Childrens' attachmentstatus was then compared with their outcome status, with the expectationthat avoidant attachment would be associated with aggression and resistantattachment with passive-withdrawal

All correlational and regression analyses were performed separately foreach sex I*revious work with this sample has shown marked sex differences,particularly in regard to the relation of maternal and environmental vanablesto child outcomes (Pianta, Egeland. & Sroufe, m press)

RESULTS

Descriptive and Validity Analyses

The mean aggression score was 11 66 (SD = 11 27) and the mean pas-sive-withdrawal score was 6 01 {SD = 4 27) Boys' mean scores weresignificantly higher on both scales (14 44 vs 8 28 for aggression, / =3 7S,p< 01, 6 77 vs 5 09 for passive-withdrawal, t = 265, p <

01)Substantial homogeneity was found for each measure Average item-

total correlations were 68 for the aggression scale and 64 for the pas-sive-withdrawal scale (A = 191), with all correlations greater than 45(all ps< 01) Both scales showed substantial stability over the 3-yearpenod studied With 170 to 185 observations at each year, the aggressionscores correlated 59 between Grades 1 and 2, 59 between Grades 2 and3, and 50 between Grades 1 and 3 Passive-withdrawal scores showedslightly less stability, with correlations of 43, 61, and 39 for the samecompanson (allps < 01)

7 Because only those children scoring in the bottom half of the distnbution of ag-gressive scores were considered eligible for membership in the passive-withdrawalgroup, the pool of potential passive-withdrawn children shrank considerably In or-der to maintain sufficient group size it then became necessary to reiax the passive-withdrawal cutoff from the top 20% to the top 30% of each sex We recognize thatthese alternations m group definitional criteria may be criticized as inelegant Itshould be noted, however, that they were earned out independently of any exami-nation of outcomes of the analyses, and that they were done in the interest of obtain-ing groups of theoretical interest and of sufficient size

270 Renken et al

The correlation between the teacher-rating scales (averaged across the3 years) and scales constructed from the observation Q-sorts was 57, df— 99, p < 01, for aggression and 46, df = 99, p < 01, for passive-withdrawal, suggesting a substantial level of consistency across mea-surement methods There was no correlation between aggression and IQfor either boys or girls, but passive-withdrawal was negatively correlatedwith IQ for both boys, r(92) = - 30, p = 002, and girls, r(76) =- 26,p = Oil This suggests that there is a relationship, although nota strong one, between low intellectual and motivational competence inschool and passive-withdrawal

Multivanate Prediction of Outcomes

Table 2 shows the zero-order correlations among the predictor and out-come vanables Tables 3 and 4 show the results of the multiple regres-sion analyses In Table 2, the correlations of predictor vanables withboth outcome vanables are shown, although each vanable was only hy-pothesized to predict one of the outcome measures For boys, eight ofthe nine predictors were significantly related to aggression Several ofthese predictors were also related to passive-withdrawal in boys Forgirls, negative affect at 42 months, matemal hostility at 42 months,stressful life events, and mother's emotional support were significantlyrelated to aggression Several of the predictors of aggression were sig-nificantly related to passive-withdrawal in girls

Prediction of passive-withdrawal was less successful for both boysand girls Five of nine hypothesized predictors were significantly relatedto passive-withdrawal in boys agentic behavior at 42 months, enthusi-asm at 42 months, matemal neglect, the total HOME score, and the fac-tonal score of matemal caretaking skill Two of these predictors werealso significantly related to aggression in boys Only two of the hypoth-esized predictors related significantly to passive withdrawal in girls en-thusiasm at 42 months and matemal IQ One of the predictors was sig-nificantly related to aggression m girls

Table 3 shows the results of multiple regression analyses performedwith each of the three sets of predictors of aggression That is, a set ofpredictors representing the child's developmental history (Set I in Table3), another set representing matemal hostility and abuse (Set II), and athird set representing stressful life circumstances (Set III) were eachused to predict the aggression cntenon Table 3 also shows the results of

Antecedents of Aggression 271

table 2Zero-Order Conelcrtions of Predictor Vctnables With Aggression and

Passive-Withdrawal, for Boys and Guls

Predictor vanable

Predictors of aggressionAvoidant attachmentNegative affect

24 monthsNegative affect 1

42 monthsNegative affect 2

42 monthsPhysical abuseMaternal hostility

42 monthsStressful life eventsEmotional support for

motherSES

Predictors of passivityResistant attachmentCoping attempts

24 monthsEnthusiasm 24 monthsAgentic behavior

42 monthsEnthusiasm 42 monthsMaternal neglectTotal HOME scoreMaternal caretaking skillMatemal IQ

BoysAggressive

23**

- 03

19*

26**39**

23**27**

- 31**- 30**

— 12

1601

- 12- 12

15- 30**- 26**

00

Passive

29**

18*

- 11

1427**

18*04

- 20*- 16

05

- 16- 15

- 30**- 19*

20*- 39**- 31**- 10

GirlsAggressive

03

- 01

27**

1214

32**32**

- 26**- 10

- 08

- 0913

10- 07

0203

- 19*- 17

Passive

05

- 05

23*

1325**

1627**

- 18- 30**

02

- 07- 12

02- 31**

15- 18- 08- 27**

Note <i/vanes between 78 and 104 for boys and between 70 and 84 for girls* Ol<p< 05** 001 <p< 01

hierarchical regression analyses in which the predictor sets were enteredsequentially into the prediction equation (see rows in Table 3 labeled"Sets I and II," and "Sets I, II, and III") For boys, each of the threesets, taken singly, performed equally well Each produced a significant

272 Renken et al

Tald«3Piediction of Aggression With Multiple-Predictor Sets

Vanable sets

BoysI

II

UI

GirlsI

II

III

DevelopmentalhistoryMaternalhostilityLife stressI and 11I. II, and III

DevelopmentalhistoryMatemalhostilityLife stressIandUI, II, and III

R

41

44425258

29

33354250

R'

17

19182733

08

11121825

F

4 69

11 735 284 133 62

1 60

4 833 012092 05

4

23

69

4

2369

df

,91

,97,74,68,65

,70

.77,66,57,54

P(F)

< 0 1

< 0 1< 0 1

<01< 0 1

18

01030705

Note The developmental history predictor set contains four variables avoidantattachment, negative affect (24 months), negative affect I (42 months), and negativeaffect II (42 months) The matemal hostility predictor set contains two variablesf)hysical abuse and matemal hostility (42 months) The hfe stress predictor set containsthree vanables stressful life events, emotional support for mother, and socioeconomicstatus (SES)

R^ in the range of 17 to 19 In the hierarchical regression analysis, en-tenng the matemal hostility predictor set following the developmentalhistory set produced a significant mcrease m prediction (probability ofchange in F = 02), with i?^ increasing from 17 to 27 Adding the hfestress predictor set mto the equation brought R^to 33, a nonsignificantincrease Inclusion of all three predictor sets in the equation produced asignificant/?^of 33

For aggression m girls, the developmental history predictor set pro-duced a nonsignificant R^ of 08, while the matemal hostility and thestressful life events sets each produced small but significant levels of pre-diction (/?^ = 11 and 12, respectively) In the hierarchical multipleregression for girls, adding the matemal hostility predictor set to the de-velopmental history set produced a significant increase in R^ from 08 to18 (probability of change in F = 04), while the addition of the life

Antecedents of Aggression 273

Table 4Prediction ol Passive-Withdrawal With Multiple-Predictor Sets

Vanable sets

BoysI

II

GirlsI

n

DevelopmentalhistoryIncompetentcareI and II

DevelopmentalhistoryIncompetentcareI and II

R

36

4252

34

3148

R'

13

1827

11

1023

F

2 65

4 183 05

1 74

1 931 90

df

5,90

4,789,73

5.68

4,709,58

p(.F)

03

< 0 1< 01

14

1207

Note The developmental history predictor set contains five vanables resistantattachment, coping behavior (24 months), enthusiasm (24 months), agentic behavior(42 months), and enthusiasm (42 months) The incompetent care predictor set containsfour vanables matemal neglect, matemal caretaking skill, total Home Observation forMeasurement of the Environment (HOME) score, and matemal IQ

Stress predictors produced a further (nonsignificant) mcrease to 25 Theoverall level of prediction afforded by the inclusion of all three sets in theprediction equation (R^ = 25) was significant

Table 4 shows the multiple regression results for the prediction of pas-sive-withdrawal For boys, both the developmental history and the ma-temal care predictor sets produced significant levels of prediction inde-pendently, with R^ values of 13 and 18, respectively In the hierarchicalregression analysis, addition of the matemal care predictors to the devel-opmental history predictors produced a significant mcrease in R^ from13 to 27 (probability of change m F < 01) Overall prediction (/?' =27) was highly significant

For girls, the prediction of passive-withdrawal was less successfulNeither of the predictor sets produced a significant level of predictionindependently, and the overall level of prediction (R^ = 23) only ap-proached significance

To evaluate the specificity of these predictor sets, each set of predic-tors was used to predict both outcomes In the case of aggression, thepassive-withdrawal predictors did not produce a significant level of pre-

274 Kenken et al

diction for either boys or girls The aggression predictors, however, per-formed rather well as predictors of passive-withdrawal The combinedaggression predictors produced a significant level of prediction, R^ =22, F(9, 65) = 2 04, /J = 05, for boys, though this was not as strong

as the prediction afforded by the passive-withdrawal predictors Forgirls, the aggression predictors did a slightly better (though still nonsig-nificant) job of predicting passive-withdrawal, R^ = 50, F(9, 54) =2 02, p = 06, than did the passive-withdrawal predictors On inspec-tion of the zero-order correlations (see Table 2), it can be seen that sev-eral of the aggression predictors, most notably physical abuse, stressfullife events, and SES, are related to passive-withdrawal in girls

Relations Among Attachment a n dOutcome Measures

Table 5 shows the relations between the outcome groups and attachmentgroups for boys, with separate data provided for the 12-month and 18-month attachment data No relationship was found between attachmentand outcome for girls Results for boys were highly significant using the18-month data, but failed to reach significance with the 12-month dataThe pattem in the data, however, is the same in both cases There is aclear tendency for avoidantly attached boys to be overrepresented in theaggressive group, while resistantly attached boys were highly overrepre-sented in the passive group The relationship between resistant attach-ment at 18 months and outcome was particularly striking Only 1 of 21resistantly attached boys became aggressive, while 7 became passive-withdrawn, accounting for more than half of the passive-withdrawngroup

DISCUSSION

Our analysis supports all three predictive factors suggested by the liter-ature as antecedents of aggression for boys, and provides moderate sup-port for two of them for gn-ls Thus, attachment history and early socialadaptation were significantly related to an aggressive outcome in boys,and the factors of harsh parental treatment and stressful life circumstan-ces were related to aggression in both boys and girls It is noteworthy thatthe attachment relationship is predictive of aggression (in boys) even

Antecedents of Aggression 275

Table 5Relation of Attachment and Outcome Classifications in Boys

Outcome group J

12-month attachmentControl groupAggressive

groupPassive groupTotal

18-month attachmentControl groupAggressive

groupPassive groupTotal

'Vvoidant

groups'16

72

25groups'"

9

72

18

Secure

39

94

52

44

123

59

Resistant

13

56

24

13

17

21

Total

68

2112

101

66

201298

a n = lOl,x^ = 6 60,df= 4,p = 16b « = 98,x' = I100,df= 4,p< 01

though It IS assessed before aggression is even a part of the child's be-havioral repertoire In combination, these predictors accounted for athird of the aggression vanance in boys Such power denves from com-bining predictors and also combining outcome assessment across teach-ers It IS expected that they would be considerably stronger with morestable outcome vanables (for example, multiple judges at a given age)

Prediction of passive-withdrawal, while significant, was not quite asstrong There are three possible reasons for this outcome First, there areless substantial theoretical or empincal bases for developing predictivehypotheses Thus, it is difficult to know whether better prediction mightnot have been provided by a different set of predictors Second, passive-withdrawn behavior may form a less coherent and detectable pattem ofbehavior in this age group than does aggression This possibility is sug-gested by the lower stabihty of passive-withdrawal as well as the loweragreement between teachers and observers ( 46 vs 57 for aggression)It should also be noted that fewer items went into this scale Third, teach-ers often may not view such lntemahzing behaviors as especially note-worthy or problematic

The specificity of our predictor sets was supported to some degree forboys That is, aggression predictors were more accurate in predictingaggression than passivity and vice versa This was not the case for girls.

27« Kenken etal

the aggression predictors predicted passive-withdrawal and aggressionequally well Thus, both specificity and overlap were found to a consid-erable degree m our multivanate prediction equations

For boys, considerable predictive specificity was found for infant at-tachment status, particularly as assessed at 18 months Aggressive/anti-social behavior was predicted by avoidant attachment and passive-with-drawal by resistant attachment This finding is important because itsuggests not only that behavioral and emotional problems may be pre-dicted from infancy, especially for boys (see also Lewis, Femng, Mc-Guffog, & Jaskir, 1984), but also that the avoidant and resistant pattemsof anxious attachment may have different sequelae There has previouslybeen little evidence in the literature to support this contention (Sroufe,1983, m press)

There were several striking differences between the girls' and boys'results First, the overall level of prediction was better for boys than forgirls This may be related to the fact that girls' scores clustered nearerthe low, nonpathological end of both the aggression and passive-with-drawal scales A restncted degree of vanability among girls would beexpected to produce weaker correlations with predictor vanables Sec-ond, the two sets of predictors failed to discriminate between aggressionand passive-withdrawal m gnrls to the same degree as m boys This maybe due, m part, to cultural influences on girls in the direction of inter-nalizing behaviors (Block, 1979) Thus, any negative environmentalevents may press girls toward passive-withdrawal In an earlier evalua-tion we found that passivity in girls in kindergarten was predicted by an-ger ratmgs m our toddler tool problem assessment (Marvinney, 1985)This was not true for boys Third, there were differences in the type ofvanables that related to pathology m girls and in boys Most notably, at-tachment classification was not related to either aggression or passive-withdrawal m girls, although it is related to vanables such as social com-petence (Sroufe, m press) On the other hand, attachment status relatedto pathology in expected ways among boys

Processes Underlying Continuity

It would be possible to explain the several predictors of aggression (orpassive-withdrawal) m several ways First, there is a genetic explana-tion One might argue that avoidant attachment, harsh treatment, and hfestress are all markers of a certain genetic makeup, passed from the parentto the child, and expressed also in aggressive behavior Second, a mod-

Antecedents of Aggression 277

elmg explanation would mate sense of the relations found among harshtreatment, physical abuse, and aggression Third, it might be argued thatthese several factors are indicators of ongoing negative environmental in-fluences Thus, aggression in third grade is due to contemporaneous en-vironmental input rather than being "caused" by history

While each of these explanations has ment, we prefer a more lntegra-tive viewpoint for a number of reasons First, certain aspects of the datasuggests greater complexity than is implied by a genetically based traitinterpretation For example, evidence suggests that parental emotionalunavailability is highly predictive of child aggression (Egeland &Sroufe, 1981, Sroufe, 1983, Troy & Sroufe, 1987) It is not obvious howsuch disparate behaviors can be seen as expressions of a single genotypeThe same can be said for the link between avoidant attachment and lateraggressive behavior Moreover, research has shown that harsh parentaltreatment predicts child aggression more strongly than do measures ofparental aggressiveness per se (Huesman, Eron, Lefkowitz, & Walder,1984, Patterson & Dishion, 1988) These same examples of complexityargue against a simple modeling explanation Finally, accounting for thefindings by parents' contemporaneous behavior ignores the child's in-creasingly active role in influencing his or her environment as develop-ment proceeds

Therefore, we prefer to explain the link between antecedents and out-come by refemng to Bowlby's concept of the lntemal working modelThis can be illustrated in the case of aggression, where the workingmodel has the advantage of unifying diverse antecedents, each of whichcan be seen as contnbutmg to the child's hfe perspective The world isseen as threatening, alien, and punitive, and other people are seen as un-available or even as a source of pain In contrast to the usual expectationthat relationships are sources of mutual support and enhancement, sucha child has leamed that in relationships the vulnerable are exploited orpushed aside With this perspective on relationships, and with consid-erable anger denving from unmet emotional needs of the most basic sort,this child enters the world of peers with a very different onentation fromthat of the well-nurtured child A tendency to be hostile and deliberatelyhurtful toward others (those who are weak) leads the child to be activelyrejected by peers (LaFrenier & Sroufe, 1985) and disliked and disci-phned by teachers (Sroufe & Fleeson, 1988)

Links among avoidant attachment, lntemal working models of othersas alien, and aggression are beginning to be explored (Sroufe, 1988)One interesting example is the work of Dodge (Dodge & Frame, 1982)

278 Benken et al

and subsequent work by Suess (1987) in West Germany Dodge finds thataggressive children more frequently interpret ambiguous social situa-tions as entailing hostile mtent, which may be seen as implying a cog-nitive deficit In a prospective longitudinal study, however, Suess re-cently found that it is children with histories of anxious/avoidantattachment (who have indeed expenenced rejection) who show this attn-butional style Therefore we would attnbute these mdividual differencesm attnbution not to an inherent cognitive deficit, but to expenence whichmade reasonable the child's interpretation, however mappropnate it mayactually be in a given instance Viewing others, then, as hostile the ag-gressive child strikes out, in fact leadmg others to dislike him/her, andthe process continues

This view of the developing child, whose intemal working modelsguide active transactions with the social environment, gives coherenceto the strong links between early expenence and later aggressive behav-ior we found in this study It also makes understandable the strong tiesbetween aggression m childhood and later adult problems

REFERENCES

Achenbach T M . & Edelbrock, C (1983) Manual for the teacher's report form andteacher version of the child behavior profile University of Vermont, Department ofPsychiatry, Burlington, VT

Ainsworth, M , Blehar, M , Waters, E , & Wall, S (1978) Patterns of attachmentHiUsdale, NJ Lawrence Erlbaum

Ainsworth, M , & Wittig, D (1969) Attachment and exploratory behavior of one yearolds m a strange situation InB M Foss{Ed), Determinants cf infant behavior {Vol4. pp 111-136) London Methuen

Becker, W C (1964) Consequences of different kinds of parental discipline In MHoffman & L Hoffrtan (Eds), Review of child deveU^ment research {Vol 1, pp169-208) New York Russell Sage Foundation

Block, J H (1979 September) Pbrsonality development in males and females The in-fluence of differential socialization Master Lecture Senes of the American Psycho-logical Association New York, NY

Block, J H , & Block, J (1980) The role of ego-control and ego-resihency in the or-ganization of behavior In W A Collins (Ed ), University of Minnesota symposiumon child psychology {Vol 13, pp 39-101) HiUsdale, NJ Lawrence Erlbaum

Bowlby,J (1973) Attachment and loss Vol II Separation Anxiety and anger LondonHogarth Press

Bretherton, I , & Waters, E (Eds ) (1985) Growing points of attachment theory andresearch Monographs of the Society for Research in Child Development, 50 (1, 2,Serial No 209)

Antecedents of Aggression 279

Caldwell B (1979) Home observation for measurement of the environment LittleRock University of Arkansas Center for Early Development and Education

Devereux Elementary School Behavior Rating Scale II (1982) Devon PA The Dever-eux Foundation

Dodge K , & Frame. C (1982) Social cogmtive biases and deficits m aggressive boysChtld Development, S3 620-635

Egeland B Breitenbucher, M & Rosenberg D (1980) Prospective study of the eti-ology of child abuse Journal of Consulting and Chnical Psycholog\ 48 195-205

Egeland, B ,& Sroufe, L A (1981) Attachment and early maltreatment Child Devel-opment. 52 44-52

Emery, R E (1982) Interparental conflict and the children of discord and divorce Psy-chological Bulletin 92 310-330

Enckson, M F Sroufe, L A & Egeland B (1985) The relationship between qualityof attachment and behavior problems in pre-school in a high-risk sample In I Breth-erton & E \^aX&K{EAs ) , Growing points of attachment theory and research Mono-graphs (^ the Society for Research in Child Development 50 (Serial No 209) 147-165

Hethenngton, E M . Cox, M , & Cox R (1982) Effects of divorce on parents andchildren In M E Lamb (Ed ), Nontraditional families Parenting and child devel-opment {pp 233-238) Hillsdale NJ Lawrence Erlhaum

Huesman, L R , Eron, L D , Lcfkowitz, M N , & Walder. L (1984) The stability ofaggression overtime and generations DevelopmentalPsvcholog\. 20, 1120-1134

Kagan, J , & Moss, H (1962) From birth to maturm A studv in psychological devel-opment New York Wiley

LaFrenier P , & Sroufe, L A (1985) Profiles of peer competence in the preschool In-terrelations between measures, influence of social ecology, and relation to attachmenthistory Developmental Psychology, 21 56-69

Lewis, M , ftinng, C , McGuffog C & Jaskir J (1984) Predicting psychopathologyin six-year-oIds from early social relations Child Development, 55 123-136

Londerville, S ,& Main, M (1981) Secunty of attachment, compliance, and matemaltraining methods in the second year of life Developmental Psvchologv, Yl, 289-299

Main, M (1977) Analysis of a peculiar form of reunion behavior seen in some daycarechildren Its history and sequelae in children who are home-reared InR Webb (Ed),Social development in daycare (pp 33-78) Baltimore Johns Hopkins University

PressMarvinney, D (1985, Apnl) Passive withdrawal in high-risk children A study cf con-

tinuity and change in adaptation Paper presented at the Biennial Meeting of the So-ciety for Research in Child Development, Toronto

Matas, L , Arend, R , & Sroufe, L A (1978) Continuity of adaptation in the secondyear The relationship between quality of attachment and later competence Child De-velopment, 49,547-556

McCord, W, McCord, J ,&Zola, I K (1959) Origins of crime New York ColumbiaUniversity Press

Olweus, D (1979) Stabihty of aggressive reaction patterns in males kreyisw Psycho-

logical Bulletin, 16, 644-666Patterson, G R (1982) Coercive family process Eugene, C» Castalia

280 Renlcen et cd

Patterson G R , &Dishion,T (1988) A mechanism for transmitting the antisocial traitacross generations InR Hinde&J Ste\cnsoR-Hmde {Eds), Relations between re-lationships within families (pp 283-310) Oxford Oxford University Press

Pianta, R C Egeland B & Sroufe. L A (in press) Matemal stress and children'sdevelopment Prediction of school outcomes and identification of protective factorsIn J Rolf, A Masten D Cicchetti, K Nuechterlem, & S Weintraub (Eds ), Riskand protective factors in the development ofpsvchopatholog\ Cambndge, MA Cam-bndge University Press

Quinton, D & Rutter, M (1985) Family pathology and child psychiatnc disorder Afour-year prospective study In A R Nicol (Ed ), Longitudinal studies tn child psy-chology and psychiatry Practical lessons from research experience (pp 91-134)New York Wiley

Richman N , Stevenson, J ,& Graham, PJ (1982) Pre-school to school A behavioralstud\ New York Academic Press

Robins, L (1978) Sturdy childhood predictors of adult anti-social behavior Replica-tions from longitudinal studies Psychological Medicine, 8, 611-622

Rutter, M (1971) Parent-child separation Psychological effects on the children Journalof Child Psychology and Psychiatry, 12, 233-260

Rutter, M , & Garmezy G (1983) Developmental psychopathology In E M Hether-ington (Ed) P H Mussen{SenesEd), Handbook cf child psychology Vol 4 Socialand personality development {pp 775-911) New York Wiley

Sroufe, L A (1983) Infant-caregiver attachment and patterns of adaptation m pre-school The roots of maladaptation and competence InM Itrlmutter (Ed ), A/i/ine-sota sympostum in child psvchology (Vol 16, pp 41-81) Minneapolis University ofMinnesota Press

Sroufe, L A (1985) Attachment classification from the perspective of mfant-caregiverrelationships and infant temperament Chtld Development, 56, 1-14

Sroufe, L A (1988) The role of infant-caregiver attachment m development InJ Bel-sky & T Nezworski (Eds ), Clinical implications cf attachment (pp 18-38) Hills-dale, NJ Lawrence Erlbaum

Sroufe, L A (m press) Psychopathology and developmental deviation InD Cicchetti(Ed ), Rochester symposium in developmental psychopathology

Sroufe, L A , & Fleeson, J (1986) Attachment and the construction of relationshipsIn W W Hartup &Z Rubin {Eds ) , Relationships and devel(^ment{pp 51-72) NewYork Cambndge University Press

Suess, G (1987) AuswirkungenfruhkindlicherBindungserfahrungen aufKompetenz imKindergarten [Consequences of early attachment orgamzation for com{>etence in pre-school] Unpublished doctoral dissertation. University of Regensburg, West Ger-many

Troy, M ,& Sroufe, L A (1987) Victimization among preschoolers Role of attachmentrelationship history Journal cf the American Academy of Child and Adolescent Psy-chiatry 26, 166-172

Waters, E , Vaughn, B & Egeland, B (1980) Individual differences in mfant-motherattachment relationships at age one Antecedents in neonatal behavior m an urban,economically disadvantaged sample Child Development, 51, 203-216

Antecedents of Aggression 281

West, D J (1982) Delinquency Its roots, careers and prospects London HeinemannEducational

Manuscript received February 16 1988, revised September J2 1988