12
Correlates of Attachment at Age 3: Construct Validity of the Preschool Attachment Classification System Ellen Moss, Jean-Franc ¸ois Bureau, Chantal Cyr, and Chantal Mongeau Universite ´ du Que ´bec a ` Montre ´al Diane St-Laurent Universite ´ du Que ´bec a ` Trois-Rivie `res This study examined correlates of attachment at age 3 to further validate preschool separation–reunion measures. Three-year-olds (N 150) and their mothers participated in a separation–reunion protocol, the Preschool Attachment Classification System (PACS; J. Cassidy & R. S. Marvin with the MacArthur Working Group on Attachment, 1992), and a mother– child interaction session during a laboratory visit. Mothers also completed psychosocial measures and, along with teachers, evaluated child behavior problems. The secure and disorganized groups received, respectively, the highest and lowest interaction scores. Disorganized children showed a higher level of teacher-reported externalizing and internalizing problems than did secure children. Mothers of insecure children reported higher child externalizing (all insecure groups) and internalizing (avoidant group) scores, more personal distress related to emotional bonding (disorganized group), childrearing control (ambivalent group), and child hyperactivity (avoidant group). Results strongly support the validity of the PACS as a measure of attachment in 3-year-olds. At preschool age, the quality of attachment to the caregiver is presumed to influence several processes that impact directly on the child’s adaptive behavior in both home and school settings (Erick- son, Sroufe, & Egeland, 1985; Thompson, 1999). Compared with the vast infant literature, relatively few studies have examined associations among preschool attachment, family variables, and child social behavior. The Preschool Attachment Classification System (PACS; Cassidy & Marvin with the MacArthur Working Group on Attachment, 1992) is a laboratory-based categorical system, the objective of which is to capture the particular secure or insecure attachment strategy of the 3- to 5-year-old child primarily on the basis of separation–reunion behavior. An accumulating list of studies has demonstrated associations between the Cassidy– Marvin system and measures of mother– child interaction, child representations of self and other, social competence, and socio- emotional adaptation (Achermann, Dinneen, & Stevenson-Hinde, 1991; Bar-Haim, Sutton, Fox, & Marvin, 2000; Barnett, Kidwell, & Leung, 1998; Bretherton, Ridgeway, & Cassidy, 1990; Cicchetti & Barnett, 1991; Easterbrooks, Davidson, & Chazan, 1993; Green- berg, Speltz, DeKlyen, & Endriga, 1991; Howes & Hamilton, 1992; Manassis, Bradley, Goldberg, Hood, & Swinson, 1994; Moss, Rousseau, Parent, St-Laurent, & Saintonge, 1998; Moss & St-Laurent, 2001; NICHD Early Child Care Research Network, 2001; Stevenson-Hinde & Shouldice, 1995; Turner, 1991). Most of these studies have focused on children of older preschool age, with few examining attachment patterns and their correlates in 3-year- olds. Accordingly, our aim in this study was to examine concurrent associations between PACS attachment classifications and exter- nal criteria representing core predictions of attachment theory in order to further validate this measure for this age group. The two domains that have been most strongly associated with attachment in childhood both theoretically and empirically are the quality of the caregiver– child relationship and child socioemotional adapta- tion. Attachment groups were also compared on aspects of family ecology and maternal psychosocial state. Attachment Relations at Early Preschool Age When Bowlby wrote his definitive works on attachment theory (Bowlby, 1969/1982, 1973, 1980), he conceptualized attachment relationships as serving a broad-based adaptive function relevant to the life span rather than to only restricted periods of develop- ment. The child’s use of the caregiver as a secure base for explo- ration, a haven of safety, and a source of comfort is a continual challenge requiring renegotiation at every developmental stage. With the growth of language and perspective-taking skills at early preschool age, the child becomes able to communicate his or her own intentions and plans, understand those of the caregiver, and engage in negotiations aimed at jointly achieving a common goal (Marvin, 1977). The child’s internal working model, or internal- ized representation, of caregiver responsiveness should influence expectations concerning interactive exchanges with both caregiv- ers and other adults or peers in extrafamilial settings. Studies with samples of 3-year-olds support these theoretical ideas in establishing associations between attachment security and emotional openness in children’s narratives (Bretherton et al., Ellen Moss, Jean-Franc ¸ois Bureau, Chantal Cyr, and Chantal Mongeau, Department of Psychology, Universite ´ du Que ´bec a ` Montre ´al, Montreal, Quebec, Canada; Diane St-Laurent, Department of Psychology, Universite ´ du Que ´bec a ` Trois-Rivie `res, Trois Rivieres, Quebec, Canada. This research was supported by grants received from the Social Sciences and Humanities Research Council of Canada (SSHRC) and the Conseil Que ´becois de la Recherche Sociale (CQRS) to Ellen Moss and by doctoral and postdoctoral fellowships from the SSHRC and the CQRS to Jean- Franc ¸ois Bureau and Chantal Cyr. Correspondence concerning this article should be addressed to Ellen Moss, De ´partement de Psychologie, Universite ´ du Que ´bec a ` Montre ´al, C.P.8888 Succ Centre-Ville, Montre ´al, Que ´bec H3C 3P8, Canada. E-mail: [email protected] Developmental Psychology Copyright 2004 by the American Psychological Association 2004, Vol. 40, No. 3, 323–334 0012-1649/04/$12.00 DOI: 10.1037/0012-1649.40.3.323 323

Correlates of Attachment at Age 3. Construct Validity of the Preschool Attachment Classification System

Embed Size (px)

DESCRIPTION

Moss Et Al (2004)

Citation preview

  • Correlates of Attachment at Age 3: Construct Validity of the PreschoolAttachment Classification System

    Ellen Moss, Jean-Francois Bureau, Chantal Cyr, andChantal Mongeau

    Universite du Quebec a` Montreal

    Diane St-LaurentUniversite du Quebec a` Trois-Rivie`res

    This study examined correlates of attachment at age 3 to further validate preschool separationreunionmeasures. Three-year-olds (N 150) and their mothers participated in a separationreunion protocol, thePreschool Attachment Classification System (PACS; J. Cassidy & R. S. Marvin with the MacArthurWorking Group on Attachment, 1992), and a motherchild interaction session during a laboratory visit.Mothers also completed psychosocial measures and, along with teachers, evaluated child behaviorproblems. The secure and disorganized groups received, respectively, the highest and lowest interactionscores. Disorganized children showed a higher level of teacher-reported externalizing and internalizingproblems than did secure children. Mothers of insecure children reported higher child externalizing (allinsecure groups) and internalizing (avoidant group) scores, more personal distress related to emotionalbonding (disorganized group), childrearing control (ambivalent group), and child hyperactivity (avoidantgroup). Results strongly support the validity of the PACS as a measure of attachment in 3-year-olds.

    At preschool age, the quality of attachment to the caregiver ispresumed to influence several processes that impact directly on thechilds adaptive behavior in both home and school settings (Erick-son, Sroufe, & Egeland, 1985; Thompson, 1999). Compared withthe vast infant literature, relatively few studies have examinedassociations among preschool attachment, family variables, andchild social behavior. The Preschool Attachment ClassificationSystem (PACS; Cassidy & Marvin with the MacArthur WorkingGroup on Attachment, 1992) is a laboratory-based categoricalsystem, the objective of which is to capture the particular secure orinsecure attachment strategy of the 3- to 5-year-old child primarilyon the basis of separationreunion behavior. An accumulating listof studies has demonstrated associations between the CassidyMarvin system and measures of motherchild interaction, childrepresentations of self and other, social competence, and socio-emotional adaptation (Achermann, Dinneen, & Stevenson-Hinde,1991; Bar-Haim, Sutton, Fox, & Marvin, 2000; Barnett, Kidwell,& Leung, 1998; Bretherton, Ridgeway, & Cassidy, 1990; Cicchetti& Barnett, 1991; Easterbrooks, Davidson, & Chazan, 1993; Green-berg, Speltz, DeKlyen, & Endriga, 1991; Howes & Hamilton,1992; Manassis, Bradley, Goldberg, Hood, & Swinson, 1994;

    Moss, Rousseau, Parent, St-Laurent, & Saintonge, 1998; Moss &St-Laurent, 2001; NICHD Early Child Care Research Network,2001; Stevenson-Hinde & Shouldice, 1995; Turner, 1991). Most ofthese studies have focused on children of older preschool age, withfew examining attachment patterns and their correlates in 3-year-olds. Accordingly, our aim in this study was to examine concurrentassociations between PACS attachment classifications and exter-nal criteria representing core predictions of attachment theory inorder to further validate this measure for this age group. The twodomains that have been most strongly associated with attachmentin childhood both theoretically and empirically are the quality ofthe caregiverchild relationship and child socioemotional adapta-tion. Attachment groups were also compared on aspects of familyecology and maternal psychosocial state.

    Attachment Relations at Early Preschool Age

    When Bowlby wrote his definitive works on attachment theory(Bowlby, 1969/1982, 1973, 1980), he conceptualized attachmentrelationships as serving a broad-based adaptive function relevantto the life span rather than to only restricted periods of develop-ment. The childs use of the caregiver as a secure base for explo-ration, a haven of safety, and a source of comfort is a continualchallenge requiring renegotiation at every developmental stage.With the growth of language and perspective-taking skills at earlypreschool age, the child becomes able to communicate his or herown intentions and plans, understand those of the caregiver, andengage in negotiations aimed at jointly achieving a common goal(Marvin, 1977). The childs internal working model, or internal-ized representation, of caregiver responsiveness should influenceexpectations concerning interactive exchanges with both caregiv-ers and other adults or peers in extrafamilial settings.

    Studies with samples of 3-year-olds support these theoreticalideas in establishing associations between attachment security andemotional openness in childrens narratives (Bretherton et al.,

    Ellen Moss, Jean-Francois Bureau, Chantal Cyr, and Chantal Mongeau,Department of Psychology, Universite du Quebec a` Montreal, Montreal,Quebec, Canada; Diane St-Laurent, Department of Psychology, Universitedu Quebec a` Trois-Rivie`res, Trois Rivieres, Quebec, Canada.

    This research was supported by grants received from the Social Sciencesand Humanities Research Council of Canada (SSHRC) and the ConseilQuebecois de la Recherche Sociale (CQRS) to Ellen Moss and by doctoraland postdoctoral fellowships from the SSHRC and the CQRS to Jean-Francois Bureau and Chantal Cyr.

    Correspondence concerning this article should be addressed to EllenMoss, Departement de Psychologie, Universite du Quebec a` Montreal,C.P.8888 Succ Centre-Ville, Montreal, Quebec H3C 3P8, Canada. E-mail:[email protected]

    Developmental Psychology Copyright 2004 by the American Psychological Association2004, Vol. 40, No. 3, 323334 0012-1649/04/$12.00 DOI: 10.1037/0012-1649.40.3.323

    323

  • 1990) and peer representations (Cassidy, Kirsh, Scolton, & Parke,1996), mothers enhanced psychosocial and interactive function-ing, and childrens more positive interactive style (Laible &Thompson, 1998; Moss, Gosselin, Parent, Rousseau, & Dumont,1997; NICHD Early Child Care Research Network, 2001; Teti &Gelfand, 1997; Teti, Gelfand, Messinger, & Isabella, 1995) andlower levels of externalizing and internalizing behavior problems(Erickson et al., 1985; Moss, Parent, Gosselin, Rousseau, & St-Laurent, 1996; Shaw, Keenan, Vondra, Delliquadri, & Giovanelli,1997; Shaw, Owens, Vondra, & Keenan, 1996). Studies alsosupport the idea that disorganized 3-year-olds are at higher riskthan secure and other types of insecure 3-year-olds on a variety ofmeasures related to child behavior, maternal psychosocial func-tioning, and motherchild interaction (De Mulder & Radke-Yarrow, 1991; Manassis et al., 1994; Moss et al., 1996; NICHDEarly Child Care Research Network, 2001; Radke-Yarrow, Cum-mings, Kuczynski, & Chapman, 1985; Shaw et al., 1996, 1997;Teti et al., 1995).

    The preschool-age attachment patterns of secure, avoidant, andambivalent children are conceptually similar to those displayed bysuch children in infancy. However, two profiles are associatedwith disorganization at preschool age. Children who display abehaviorally disorganized profile fail to show a coherent strategyfor dealing with reunion and display sequences of behavior thatseemingly lack a goal or that evidence a collapse of strategy (Main& Solomon, 1990). Children classified as insecure controllingshow role-reversed behaviors with the caregiver of a punitive orcaregiving type. Although at least two studies have shown that themajority of infants with a disorganized attachment profile manifesta controlling strategy at age 6 (Main & Cassidy, 1988; Wartner,Grossman, Fremmer-Bombik, & Suess, 1994), when and how thistransition occurs is unknown. It has been common practice tocombine both controlling and disorganized children in one groupfor analyses. Therefore, there are relatively few studies providingdata either on the proportion of children manifesting controlling ordisorganized behaviors at different points in the preschool periodor on the differential developmental correlates of these subgroups.Two studies that included normative samples suggested that at age4, the majority of disorganized/controlling children still displayedthe disorganized behaviors shown in infancy (8%14% of thesamples), with fewer (4%8%) showing controlling behavior(Howes & Hamilton, 1992; Stevenson-Hinde & Shouldice, 1995).However, other studies have suggested that these proportions mayvary according to sample characteristics (Cichetti & Barnett; 1991;Speltz, Greenberg, & DeKlyen, 1990). A few studies that sepa-rately analyzed developmental correlates of these subgroups sug-gested some greater resilience in family and child risk profiles forthe controlling group (Moss & St-Laurent, 2001; Teti, 1999).Clearly, further studies that separately examine the development ofcontrolling and behaviorally disorganized children are needed tofurther clarify these issues.

    Preschool Attachment and the ChildCaregiverRelationship

    One of the core elements of attachment theory is the relationbetween security and caregiver sensitivity and responsiveness. Bythe 3rd year of life, the attachment relationship is transformed intoa goal-corrected partnership (Bowlby, 1973). Secure-base behav-

    ior is more likely to be manifested in verbal exchanges with thecaregiver that emphasize co-constructive experiences rather thanbeing restricted to nonverbal proximity-seeking behavior (Waters& Cummings, 2000). A number of studies using measures ofpreschool attachment based on separationreunion have shownthat security is associated with more open emotional expression, ahigher level of child participation in joint problem-solving activ-ities, and more reciprocal control during motherchild interactionin both home and laboratory contexts (Barnett et al., 1998; Moss etal., 1997, 1998; NICHD Early Child Care Research Network,2001; Stevenson-Hinde & Shouldice, 1995). Studies have furtherindicated that the quality of communication between mothers andchildren classified as disorganized or controlling is the most dis-rupted compared with that of other insecure groups (Main, Kaplan,& Cassidy, 1985; Moss et al., 1998; Moss & St-Laurent, 2001;Stevenson-Hinde & Shouldice, 1990).

    Because the caregiver also fulfills a role in scaffolding emergingsocioemotional competencies such as accurate perspective takingand the ability to take responsibility in regulating the socioemo-tional content of relationships, the caregivers perceptions regard-ing his or her own mental state and personal relationships may berelated to preschool attachment patterns (Teti & Gelfand, 1997).Studies have found that maternal depression, dysfunctional maritalrelations, high parenting stress, and difficulties with the caregiverrole are linked to attachment insecurity (Davies & Cummings,1994; Egeland & Farber, 1984; Moss et al., 1998; Teti et al., 1995).Among insecure attachment groups, some studies suggest that it isthe mothers of ambivalent/dependent and disorganized childrenwho differ most from the mothers of children in other attachmentgroups on psychosocial measures, with the former showing greaterparenting stress and difficulty in managing the caregiving role(Moss et al., 1998; Scher & Mayseless, 2000; Stevenson-Hinde &Shouldice, 1995; Teti et al., 1995) and the latter showing higherlevels of depression or altered states of mind, which may interferewith responsivity (Hesse & Van IJzendoorn, 1998; Lyons-Ruth,Connell, Grunebaum, & Botein; 1990; Radke-Yarrow et al., 1985).However, because these profiles are largely based on infant sam-ples, more studies of family correlates of early preschool attach-ment are needed.

    Preschool Attachment and Behavior Problems

    Past research supports the view that the secure childs capacityto be appropriately self-reliant and to negotiate interpersonal dif-ficulties should engender a lower level of behavior problems in thepreschool setting than that shown by insecure peers (Easterbrookset al., 1993; Greenberg et al., 1991; Moss et al., 1998; NICHDEarly Child Care Research Network, 2001; Parke & Waters, 1989;Teti & Ablard, 1990; Wartner et al., 1994). The most consistentassociations between attachment and behavior problems have beenfound for disorganized or controlling children. In several longitu-dinal studies with high-risk and normative samples, children clas-sified as disorganized in infancy showed a higher rate of external-izing problems at preschool and school age than any otherattachment group (Lyons-Ruth, Alpern, & Repacholi, 1993;Lyons-Ruth, Easterbrooks, & Cibelli, 1997; Shaw et al., 1996).Similarly, children in the controlling attachment group at pre-school and school age are significantly more likely to be rated ashighly externalizing or aggressive (Moss et al., 1996, 1998; So-

    324 MOSS, BUREAU, CYR, MONGEAU, AND ST-LAURENT

  • lomon, George, & De Jong, 1995). A few studies have establishedassociations between disorganized or controlling attachment andinternalizing symptoms (Carlson, 1998; Moss et al., 1998; Shaw etal., 1997). Recent studies that included all four attachment groupshave generally revealed that, compared with the disorganizedgroup, avoidant and ambivalent children showed lower levels ofbehavior problem risk (Bates, Maslin, & Frankel, 1985; Fagot,1995; Fagot & Pears, 1996; Lyons-Ruth et al., 1993, 1997; Mosset al., 1996, 1998; Rothbaum, Schneider-Rosen, Pott, & Beatty,1995; Shaw et al., 1996; Solomon et al., 1995). However, a fewstudies have found that ambivalent children have higher rates ofexternalizing problems than the secure group (Cohn, 1990; Moss etal., 1996). Similar associations have been found between theavoidant pattern and internalizing problems (Goldberg, Gotowiec,& Simmons, 1995; Lyons-Ruth et al., 1997; Moss et al., 1998).Further studies that include all classification groups and target theearly preschool period are needed. The objective of the presentstudy was to examine correlates of attachment security at age 3 inorder to further validate the PACS separationreunion measure.Because studies with both infants and early school-age childrenhave shown that attachment security is associated with moresynchronous patterns of caregiverchild interaction, better mater-nal psychosocial functioning, and fewer child behavior problems,we expected to demonstrate these associations at early preschoolage. In line with previous studies, we expected that insecurechildren would be at increased risk for behavior problems (mother-and teacher-reported) compared with secure children and thatdisorganized children would be at greater risk than those mani-festing the avoidant or ambivalent pattern. A further objective ofthis study was to increase our understanding of disorganization bydocumenting the proportions of behaviorally disorganized versuscontrolling children at this age and the degree and type of devel-opmental risk associated with each of these subgroups. We ex-pected that measures of sociodemographic level, mono- or bipa-rental status, and child sex would be largely independent ofattachment, thus demonstrating the divergent validity of the PACS.

    Method

    Participants

    Study participants were 153 French-speaking motherchild dyads whowere recruited for an ongoing longitudinal project focusing on develop-mental adaptation as a function of parentchild relationships. Participantswere recruited through preschools in diverse socioeconomic areas of Mon-treal with the intention of forming a sample representative of the generalpopulation. Analysis of background variables for the sample indicated amean age of 44 months (SD 4.31, range 34 years). The sample wasquite heterogeneous with respect to income level, with 18% of familiesearning under $20,000, 40% earning between $20,000 and $50,000, and42% earning $50,000 or more. Average maternal education was 16 years(SD 2.93 years), with 19% of mothers having 11 or fewer years ofschooling and the rest having some college or university-level training.Twenty-seven percent of the sample children were living in mother-headedsingle-parent families, and the rest were living with both parents.

    Measures and ProceduresData for the study came from the PACS separationreunion measure,

    maternal reports (the Child Behavior Checklist [CBCL], the Beck Depres-sion Inventory, the Parental Stress Index, the Life Experiences Survey, the

    Parental Childrearing Inventory, information on social networks), obser-vation of motherchild interaction, and teacher reports of child behaviorproblems (the Preschool Socio-Affective Profile). Mothers and their chil-dren also participated in a laboratory visit. The maternal report measureswere sent to the home 2 weeks in advance of the laboratory visit. Duringthe laboratory visit, questionnaires were gone over and checked with themothers for any missing or misunderstood items. The laboratory protocolwas as follows: (a) greeting and motherchild free play (10 min); (b)separationreunion procedure (20 min); (c) child play with research assis-tant while mother completed questionnaires in a separate room (40 min);and (d) snack time (10 min).

    Separationreunion procedure. The modified separationreunion pro-cedure described by Cassidy and Marvin with the MacArthur WorkingGroup on Attachment (1992) for preschool-age children and validated inprevious studies (cited above) was followed. It consists of four episodeslasting 35 min each (in this study, 5 min): (a) separation between motherand child; (b) reunion; (c) second separation; and (d) second reunion.During both separations, the child was left alone. Following the separa-tions, the mother was told to rejoin the child but received no specificinstructions concerning the reunions. The separationreunion sequencetook place in a room in which age-appropriate toys were scattered. Thechilds attachment classification was based on behavior observed duringboth reunions.

    Attachment classification. Coders classified reunion behavior usingcriteria from the PACS (Cassidy & Marvin with the MacArthur WorkingGroup on Attachment, 1992). Ratings were based on observer evaluationsof the childs physical proximity to the mother, affective expression, andverbal exchanges. Conversational patterns assume increasing importanceas a function of child age. The secure (B) pattern is categorized by relaxed,mutually enjoyable parentchild interaction. The secure child uses thecaregiver as a base that facilitates exploration of the environment. Theinsecure avoidant (A) pattern is characterized by the childs physical andaffective avoidance of the parent. The child will typically ignore parentalverbal initiatives, and parentchild discussions are often short, with littleelaboration by one partner of topics initiated by the other. Or the conver-sations that take place are focused on the childs play activities rather thanon any reference to the child and the parent or to their relationship.Generally, children classified as avoidant avoid intimacy rather than avoidtheir attachment figures in a more global manner. Children classified asinsecure ambivalent/dependent (C) alternatively show resistance and con-flictual behavior patterns or excessive immaturity evidenced by passivityor dependent behaviors such as following the parent around the room ortrying to be held by him or her. Interactions between parent and child oftenseem to interfere with child exploration. Children are classified as insecuredisorganized (D) or insecure other (IO) if they seem unable to use thecaregiver as a secure base for exploration but do not clearly show the A orC pattern. In the case of the disorganized group, children may displaydisordered, incomplete, or undirected sequencing of movements or someconfusion or apprehension; children classified as insecure other may dis-play other anomalous behavior or a combination of other insecure patterns.The child classified as insecure controlling attempts to control the parentsbehavior, often in a caregiving or punitive manner. Caregiving behavior ismanifest when the child is focused on helpfully guiding, orienting, orcheering up the parent. A punitive child uses hostile, directive behaviorwith the caregiver that may include verbal threats or harsh commands.Certain children include both caregiving and punitive elements or a generalcontrolling style characterized by a pattern of role reversal, with the childdirecting the parents activities and conversational exchanges.

    The videotaped reunions were coded by two individuals who wereunaware of participants scores on any other study measures. Both coderswere trained by Robert S. Marvin and achieved reliability with him on aseparate sample of tapes. All discrepancies were resolved by a reviewingof the tapes until consensus was reached. Overall agreement for the majorclassifications calculated on 30% of sample cases was 90%.

    325PRESCHOOL ATTACHMENT CORRELATES

  • Motherchild interaction. Two coders rated the quality of motherchild interaction from the videotaped snack time interaction. An overallrating and eight 7-point subscales were used to capture the following globalaspects of parentchild behaviors, with higher scores considered moreoptimal: coordination (from interaction that flows smoothly toward mutu-ally defined goals to little or unproductive interaction); communication(from clear verbal and nonverbal exchanges to inconsistent, incongruentpatterns); partner roles (from appropriate parentchild role assumption toa pattern of role reversal); emotional expression (from balanced and sharedexpression of both positive and negative affective states to restricted orexaggerated expression); responsivity/sensitivity (from attunement betweenmother and child to an intrusive or ignoring response style); tension/relaxation (from calm, comfortable interaction to a tense, anxious climate);mood (from generally positive to negative); enjoyment (from sustainedwarmth and obvious pleasure to displeasure); overall (from generally highquality, i.e., responsive and harmonious, to poor quality, i.e., indifferent orconflictual).

    Coders were unaware of all other information concerning the families,including attachment classifications. Interrater reliability (Pearson intra-class correlations) calculated on 25% of cases ranged from .75 to .85. Thescales have been shown to distinguish the motherchild interactive patternsof 3- to 7-year-old children with different attachment classifications, withthe disorganized and controlling groups showing the most dysfunctionalpatterns, and have shown concurrent and longitudinal associations withbehavior problem ratings and school performance (Cyr & Moss, 2001;Moss et al., 1998; Moss & St-Laurent, 2001; Moss, St-Laurent, Cyr, &Humber, 2000). In addition, snack time ratings of motherpreschool-childinteractions during a laboratory visit using a similar procedure were foundto correlate .71 with behavior observed in the home setting within a periodof 3 months (Dubois-Comtois & Moss, in press).

    Teacher-reported behavior problems. Teachers evaluated child socialadaptation using the Preschool Socio-Affective Profile (PSP; LaFreniere,Dumas, Capuano, & Dubeau, 1992). Teachers had known the participantsfor at least 3 months before completing the questionnaire and completedtheir ratings within 3 months of the attachment evaluation. The PSP is an80-item rating scale that measures behavior problems using scores on threefactors: Social Competence (SC), AngerAggression (AA), and AnxietyWithdrawal (AW). The instrument permits classification of children intoeither externalizing or internalizing problem groups using standardized zscores calculated separately for boys and girls. In validating the instrument,approximately 10% of the nonclinical population was defined as havingproblem behavior (similar to the proportion used for the CBCL). Behaviorproblem groups are formed using cutoff points of 1.0 SD (score) for eachof the AA or AW scores. Continuous scores for externalizing and inter-nalizing behavior and overall social competence can also be computed.LaFreniere et al. reported very good interreliability between independentteacher reports on the same child (correlations ranging from .72 to .89),internal consistency (correlations ranging from .79 to .91), and testretestreliability of the three PSP factors as well as orthogonality of the AA andAW factors (correlations ranging from .74 to .93). The validity of the PSPhas been supported by substantial correlations with CBCL Teacher ReportForm ratings (Achenbach & Edelbrock, 1981), observations of behaviorwith teachers and peers, and motherchild interactions (LaFreniere et al.,1992; Moss et al., 1998).

    Mother-reported behavior problems. Mothers evaluated child behav-ior problems using the CBCL (Achenbach & Edelbrock, 1981). Thisinstrument consists of 118 items scored on a 3-point scale (not true,somewhat or sometimes true, very true or often true). The CBCL generatesscores on eight subscales, five of which are grouped into an Internalizingdimension, including Withdrawn, Somatic Complaints, and Anxious/De-pressed subscales, and an Externalizing dimension, including Delinquentand Aggressive subscales. The instrument has excellent psychometricqualities. Testretest reliability was established for the short term (7 days;average of .89 for all scales) as well as for the long term (average of .75

    after 1 year and .71 after 2 years for all scales). Analyses reveal that almostall the CBCL items allow one to discriminate between clinical and non-referred children, indicating excellent content validity. Construct validityhas been demonstrated by significant correlations between the CBCLscales and the analogous scales of the Revised Behavior Problem Checklist(Achenbach, 1991).

    Family background. A family background questionnaire containingitems regarding demographic information and the family context wascompleted by mothers. Information related to family income, parentaleducation and occupation, single-parent status, and numbers of persons inthe household was included. The measure also included questions concern-ing the extent of the social network. Respondents were asked to rate thefrequency of social contacts with parents, colleagues, friends, and profes-sionals on a 0 (never) to 3 (frequently) scale, which yielded a total score.This total was used in analyses concerning the extent of social contact.

    Beck Depression Inventory (BDI). This questionnaire is a well-known21-item self-report measure that taps affective, cognitive, and physiologi-cal symptoms of depression (Beck, Ward, Mendelson, Mock, & Erbaugh,1961). The total score represents both the prevalence and severity ofsymptoms; a cutoff score of 10 is suggested for mild depression, 19 formoderate depression, and 30 for severe depressive illness. The intensity ofeach symptom is self-evaluated on a 03 scale. The validity of the BDI hasbeen demonstrated for clinical and nonclinical populations (Beck, Steer, &Garbin, 1988; Bumberry, Oliver, & McClure, 1978).

    Parental Stress Inventory (PSI). Mothers completed the 101-item PSI,a maternal self-report measure focusing on sources of perceived stressrelated to the parental role (Abidin, 1992). In addition to an overall stressscore, the instrument generates subscale scores in the maternal and childdomains. The Maternal domain taps seven dimensions: depression, feelingsof competence, attachment to child (i.e. investment in the parenting role),couple relations, social isolation, health, and sense of role restriction. TheChild domain taps maternal perceptions of six child characteristics: adapt-ability, demandingness, mood, hyperactivity, acceptability (conformitywith parental expectations), and reinforcing (to parent). The PSI has beenwidely used, and acceptable concurrent, construct, discriminant, and fac-torial validity and reliability have been reported (Abidin, 1992; Abidin,Jenkins, & McGaughey, 1992).

    Life Experiences Survey (LES). This 57-item inventory of positive andnegative life experiences includes possible life events likely to have had anegative impact on the family, for example, illness, death of familymembers, divorce, job loss, and severe financial problems (Sarason, John-son, & Siegel, 1978). Mothers were asked to indicate which of these eventswere experienced in the family during the previous year. Because studieshave indicated that negative life events are more likely to be correlatedwith stress-related dependent variables than are either positive life eventsor total scores, only the negative scale was used in this study. Psychometricstudies using this instrument have demonstrated good testretest reliabilityover a 5-week interval and significant associations between negative lifechange scores and stress-related dependent variables (Sarason et al., 1978).

    The Parental Childrearing Inventory. This questionnaire, developed toevaluate parental practices and attitudes concerning the rearing ofpreschool-age children, has two distinct parts (Rouzier, 1986). The firstincludes 54 items related to parental teaching activities, which reflectsensitivity to the childs zone of proximal development. More specifically,higher scores on this subscale reflect the parents awareness of his or herchilds learning potential as well as the parents engagement in activitiespromoting learning (e.g., When my child has difficulty doing something,I ask questions to help him find solutions). Parents evaluate the extent towhich they engage in different practices on a scale ranging from 0 (never)to 3 (always).The second subscale measuring parental control attitudes iscomposed of 17 affirmations that differentiate attitudes on a continuum ofpermissivenessrigidity, with higher scores indicating more authoritarianattitudes (e.g., A child will later be grateful that he was raised withseverity) and lower scores indicating a more laissez-faire approach. Par-

    326 MOSS, BUREAU, CYR, MONGEAU, AND ST-LAURENT

  • ents evaluate the extent to which they agree with each expressed attitude ona scale ranging from 1 (agree) to 5 (disagree). Studies have shown that theinstrument has good testretest reliability (r .70) over an 8-week intervaland internal consistency ( .83). Predictive validity for the inventory hasalso been demonstrated in distinguishing childrearing strategies of parentsof a normative sample (n 120) of 3- to 5-year-old children (Rouzier,1986). Mothers and fathers elevated scores on the Autonomous Learningsubscale were positively correlated with childrens lower Behavior Prob-lem scores and higher scores on the Language, Cognitive, Motor andOverall developmental scales of the Denver Developmental Screening Test(rs .26 and .48). Parents higher scores on the Control scale werepositively associated with higher problem scores and negatively predictedthe same Denver subscale and overall scores (rs .24 and .42).

    Results

    Breakdown of PACS Attachment ClassificationsTable 1 shows the distribution of participants according to

    separationreunion classification and gender. Overall, 69% of thesample was classified as secure (B), 10% as avoidant (A), 10% asambivalent/dependent (C), 7% as disorganized (D), 3% as control-ling (Cont.), and 1% as insecure other (IO). In the Cont. group, 2children were subclassified as caregiving, 2 as punitive, and 1 ascontrollinggeneral. Because the number of IO subjects was in-sufficient for data-analytic purposes, they were dropped fromfurther analyses of the sample. In line with past research, whichhas combined the disorganized and controlling groups on theassumption that the latter is a developmental transformation of theformer, and recent work suggesting that separate analysis of thedevelopmental trajectory of each group is warranted at this stage,we included both a combined D group (D/Cont.) and the separatedisorganized group in analyses. Because the size of the controllinggroup (n 5) was insufficient for statistical analyses, only de-scriptive cell means for this group are presented. Chi-square anal-yses revealed no differences in attachment classification distribu-tion as a function of gender, 2(3, N 146) 1.09, ns.

    Preliminary AnalysesPreliminary analyses assessed associations between family de-

    mographic variables and attachment classification groups. Corre-lational analyses between maternal education, family income, mar-ital status, and age of the child revealed significant associationsbetween maternal education and family income (r .30, p .01),maternal education and marital status (r .22, p .01), andfamily income and marital status (r .56, p .01). Consequently,we calculated a composite variable representing socioeconomicstatus (SES) based on the mean of the standardized scores of these

    three variables. There were no significant differences betweenattachment classification groups (one-way analyses of variance[ANOVAs]) on SES or child age. However, because half of thedependent variables used in subsequent analyses showed signifi-cant associations with the SES composite variable (rs ranging from.16 to .43), SES was used as a covariate in all analyses.

    Attachment and MotherChild Interaction

    Our first hypothesis concerned differences between secure andinsecure attachment groups on the motherchild interaction mea-sure. Table 2 displays means for the overall rating and the eightsubscales of the motherchild interaction measure for the fourPACS groups. Because principal-components factor analysis con-ducted on all subscales yielded a single factor explaining 85.3% ofthe variance, only the overall scale representing reciprocated,balanced and open communication (at the optimal end) was used inthe analyses. However, means for all subscales are presented inTable 2 for descriptive purposes. Analyses of covariance(ANCOVAs; SES included as a covariate) with planned compar-isons that specified differences between the B group and each ofthe insecure groups, and between the combined avoidant andambivalent groups (A/C) and the D/Cont. and D groups, wereconducted on the overall interactive quality scores. Analyses re-vealed a significant main effect with contrasts showing that allinsecure groups had significantly lower scores than the B group,indicating poorer coordination, problems in role structuring, lessopen emotional expression, greater tension, and less enjoyment ofthe interaction: A versus B, F(1, 120) 16.95, p .01; C versusB, F(1, 120) 9.25, p .01; D/Cont. versus B, F(1, 120) 33.74, p .01; D versus B, F(1, 115) 38.13, p .01. A seconda priori contrast indicated that both the D/Cont. and D groups hadsignificantly poorer interactive quality than the combined A/Cgroup: F(1, 44) 4.87, p .05, and F(1, 39) 8.59, p .01,respectively. Examination of the overall and subscale meansshowed that the Cont. group was similar to the A and C groups onoverall interactive quality and on the Coordination, Communica-tion, Roles, and Enjoyment subscales and was closer to the Bratings for emotional expression, lack of tension, and overallmood.

    Attachment and Maternal Psychosocial Measures

    Table 3 shows the summary statistics for maternal psychosocialmeasures. Using similar ANCOVA analyses and planned compar-isons (described above), we analyzed differences between theattachment groups on maternal depression, parenting stress, stress-ful life events, social networks, and childrearing practices. Nosignificant differences were found on depression or on the overall,maternal, or child domain scores of the PSI. However, the indi-vidual subscales of the PSI were also analyzed separately becauseof potential theoretical links with attachment variables. Analysesof individual subscales revealed that mothers of the D/Cont. groupreported more problems related to couple relations than did moth-ers of either the B group, F(1, 120) 5.89, p .05, or the A/Cgroup, F(1, 44) 4.78, p .05. Mothers of D children, comparedwith mothers of secure children, reported higher stress related tocouple relations, F(1, 115) 2.7, p .05, and attachment to thechild, F(1, 115) 3.7, p .05. High scores on the Attachment to

    Table 1Distribution of Attachment Classifications by GenderAttachment classification Girls Boys Total

    Secure 50 56 106Avoidant 7 8 15Dependent 5 10 15Disorganized 5 5 10Controlling 3 2 5Insecure other 1 1 2

    327PRESCHOOL ATTACHMENT CORRELATES

  • Child subscale reflect problems in emotional bonding with thechild and a parents perceived inability to accurately read andunderstand the childs feelings or needs. Higher stress levels onother maternal psychosocial measures were also evident for the Aand C groups. Mothers of A and C children reported higher stressrelated to child adaptability than did mothers of D children, F(1,39) 4.0, p .05. Mothers of the A group also reported higherstress related to child hyperactivity/distractibility than did mothersof the B group, F(1, 120) 3.3, p .05. Concerning socialnetworks, mothers of A children reported more frequent contactswith friends than did mothers of secure children, F(1, 120) 8.5,p .01. Analyses of parental childrearing practices showed thatmothers of C children reported a more rigid attitude concerningchildrearing, F(1, 120) 4.1, p .05. No group differencesemerged on the Teaching Practices (autonomy-promoting) scale orconcerning the frequency of stressful life events. Examination ofcell means for Cont. children showed that, in comparison with themothers of the D group, mothers of the Cont. group reportedslightly higher levels of stress on overall scores and almost allsubscales. They also had lower scores on parental rigidity than didthe mothers of any other group.

    Attachment and Child Behavior ProblemsTable 4 displays means and standard deviations for teacher-

    reported (PSP) and mother-reported (CBCL) behavior problems asa function of attachment classification. Results of the ANCOVAanalyses and planned comparisons (described above) revealed thaton teacher reports, both the D/Cont. and D groups had significantlyhigher levels of externalizing problems than either the B group orthe A/C group: B versus D/Cont., F(1, 120) 15.5, p .01; B

    versus D, F(1, 115) 20.9, p .01; D versus A/C, F(1, 39)11.2, p .01; D/Cont. versus A/C, F(1, 44) 6.42, p .05. Forinternalizing problems, both D/Cont. and D children also hadhigher levels than B children: F(1, 120) 2.6, p .05, and F(1,115) 2.6, p .05, respectively. The mean for Cont. children waslower than that for the D group and comparable to those of the Aand C groups on the externalizing scale. On the internalizing scale,Cont. children resembled the D group in being higher than the Cgroup and comparable to the A group.

    Analyses of the mother-reported externalizing scales showedthat the A, C, and D groups had higher levels of aggressivebehavior than their secure peers, with no differences between theD and A/C groups: F(1, 120) 3.8, p .05; F(1, 120) 4.7, p.05; and F(1, 115) 2.8, p .05, respectively. However, contrastsinvolving the D/Cont. group were not significant. Examination ofthe mean externalizing score for Cont. children showed it to belower than those of all the other insecure groups and higher thanthat of secure children. Mothers of the A group also rated theirchildren higher on internalizing symptoms than did mothers ofsecure children, F(1, 120) 4.6, p .05 (see Table 4).

    DiscussionThe overall objective of this study was to examine the construct

    validity of the PACS (Cassidy & Marvin with the MacArthurWorking Group on Attachment, 1992) at early preschool age byevaluating concurrent associations with key domains related boththeoretically and empirically to attachment behavior. These do-mains included motherchild interactive patterns and behaviorproblem ratings. In addition, we examined whether family andmaternal psychosocial variables would distinguish between attach-

    Table 2MotherChild Interactions as a Function of Attachment Classifications: Means and Significant Between-Groups Contrasts

    Motherchild interactiona

    Attachment classification

    Significant planned contrastsB

    (n 106)A

    (n 15)C

    (n 15)D/Cont.(n 15)

    D(n 10)

    Cont.(n 5)

    Overall 4.55 3.39 3.68 2.85 2.44 3.67 B A**, C**(1.05) (1.05) (1.06) (1.06) (1.05) (1.04) B D**, D/Cont.**

    A/C D**, D/Cont.*Coordination 4.67 3.47 3.47 3.17 2.80 3.66

    (1.08) (1.13) (0.64) (1.03) (1.14) (1.02)Communication 4.59 3.73 3.80 2.99 2.40 3.88

    (1.26) (1.39) (0.94) (1.28) (1.35) (1.22)Partner Roles 4.67 3.67 3.60 3.07 2.70 3.49

    (1.14) (1.18) (0.83) (1.10) (1.34) (1.07)Emotional Expression 4.52 3.60 3.47 2.76 2.00 4.05

    (1.03) (0.83) (0.64) (0.99) (1.05) (0.94)Responsivity 4.69 3.60 3.53 3.19 2.60 4.08

    (1.12) (0.99) (0.64) (1.07) (1.08) (1.05)Tension/Relaxation 4.91 4.07 3.87 3.43 2.80 4.46

    (1.22) (1.33) (1.13) (1.26) (1.23) (1.24)Mood 4.77 3.73 3.53 3.26 2.60 4.28

    (1.12) (1.39) (0.83) (1.12) (0.97) (1.09)Enjoyment 4.46 3.53 3.40 2.80 2.20 3.68

    (1.05) (1.25) (0.63) (1.04) (1.14) (1.01)

    Note. Means are adjusted for socioeconomic status. Values in parentheses are standard deviations. A insecure avoidant; B secure; C insecuredependent; D insecure disorganized; D/Cont. insecure disorganized and controlling; Cont. insecure controlling.a Range 17.* p .05. ** p .01.

    328 MOSS, BUREAU, CYR, MONGEAU, AND ST-LAURENT

  • ment classification groups given the paucity of information avail-able for this age group. The associations found here betweenattachment and the childs ability to regulate socioemotional be-havior are in keeping with the core theoretical notion of thecoherence of development across tasks (Sroufe, 1979). Our resultsstrongly support the construct validity of the PACS with childrenas young as 3 and provide important descriptive data on correlates

    of attachment for this understudied age period. In addition, theresults of this study further our understanding of disorganization.In keeping with recent theoretical and empirical work suggestingthat behaviorally disorganized and controlling children differ notonly with respect to the degree of organization of their attachmentbehavior but also in family and child risk profiles (Teti, 1999), wecoded and analyzed data separately for these subgroups. The

    Table 3Summary Statistics for Maternal Psychosocial Measures in Relation to Attachment Classification

    Measure

    Attachment classification

    Significantplanned contrasts

    B(n 106)

    A(n 15)

    C(n 15)

    D/Cont.(n 15)

    D(n 10)

    Cont.(n 5)

    Maternal depression (BDI)a 6.68 6.79 6.13 8.84 9.05 8.68(6.07) (5.96) (6.08) (6.22) (6.04) (2.62)

    PSI overallb 224.89 231.02 231.79 237.01 231.78 248.59(39.02) (38.84) (39.38) (39.85) (39.17) (40.36)

    PSI Mother domain 126.28 126.33 128.10 134.47 131.32 141.30(23.46) (23.33) (23.68) (23.90) (23.54) (23.85)

    Depression 20.52 20.04 20.65 21.83 21.51 22.57(5.70) (5.67) (5.75) (5.72) (5.72) (5.76)

    Attachment to Child 13.10 12.73 13.52 14.35 14.41 14.23 B D*(3.05) (3.03) (3.08) (3.03) (3.07) (3.00)

    Role Restriction 18.96 20.59 19.03 19.29 19.41 19.21(5.03) (5.00) (5.07) (5.15) (5.04) (5.09)

    Competence 29.17 29.90 28.70 30.73 28.96 34.40(6.25) (6.22) (6.31) (6.48) (6.27) (6.41)

    Social Isolation 13.58 13.72 14.32 13.35 12.41 15.27(3.97) (3.95) (3.97) (4.10) (3.99) (4.11)

    Couple Relations 18.59 17.87 19.12 21.86 21.32 22.93 B D*, D/Cont.*(4.89) (4.86) (4.94) (4.84) (4.91) (4.61) A/C D/Cont.*

    Health 13.52 13.75 14.12 14.35 14.07 15.26(3.01) (3.00) (3.04) (3.44) (3.02) (3.98)

    PSI Child domain 98.97 104.60 103.30 102.19 100.08 106.99(19.34) (19.22) (19.53) (19.58) (19.40) (19.84)

    Adaptability 26.50 27.61 28.20 25.20 23.93 27.91 A/C D*(5.47) (5.45) (5.52) (5.66) (5.49) (5.57)

    Acceptability 11.17 11.88 11.85 12.41 11.59 14.11(3.04) (3.02) (3.09) (3.08) (3.05) (3.00)

    Demanding 19.32 19.65 19.00 21.12 21.22 20.97(5.20) (5.17) (5.24) (5.18) (5.22) (5.32)

    Mood 10.28 11.29 11.04 10.52 10.02 11.02(3.31) (3.29) (3.34) (4.01) (3.29) (3.32)

    Distractibility 22.93 25.38 24.55 23.73 24.11 23.14 B A*(4.87) (4.85) (4.92) (4.91) (4.89) (4.96)

    Reinforcing to Parent 9.35 9.24 9.56 10.13 10.06 10.31(3.01) (2.99) (3.02) (3.34) (3.01) (3.30)

    Stressful life eventsc 2.87 2.56 2.76 2.73 2.25 3.62(1.47) (1.46) (1.48) (1.49) (1.47) (1.50)

    Social networkdFamily 2.71 2.79 2.79 2.57 2.65 2.42

    (0.61) (0.60) (0.61) (0.62) (0.61) (0.58)Friends 2.46 2.99 2.45 2.43 2.64 2.01 B A**

    (0.67) (0.67) (0.68) (0.70) (0.67) (0.68)Parental childrearing practicese

    Autonomy-promotion 106.17 104.90 104.17 108.10 107.35 109.27(10.60) (10.57) (10.73) (10.66) (10.66) (10.56)

    Parental rigidity 48.21 46.24 52.50 48.07 51.13 42.28 B C*(7.62) (7.55) (7.71) (8.01) (7.62) (8.83)

    Note. Means are adjusted for socioeconomic status. Values in parentheses are standard deviations. A insecure avoidant; B secure; C insecuredependent; D insecure disorganized; D/Cont. insecure disorganized and controlling; Cont. insecure controlling.a BDI Beck Depression Inventory (range 063). b PSI Parental Stress Index (overall: 1st percentile 131; 99th percentile 320). c Numberof events. d Number of people. e Autonomy-promoting teaching practices, range 0162; parental rigidity, range 085.* p .05. ** p .01.

    329PRESCHOOL ATTACHMENT CORRELATES

  • results indicated that the common practice of combining Cont. andD children in one group for data analyses may be masking thegreater resilience of the former and the risk status of the latter incertain domains. These contributions are discussed in greater detailin the following paragraphs.

    This study is one of the few to describe the prevalence ofdifferent disorganized profiles in normative samples at age 3. Inline with other comparable samples, approximately 11% of oursample displayed a D/Cont. attachment classification. Two thirdsof this group (7% of the total sample) showed the D classificationfirst identified in infants by Main and Solomon (1990), whereasonly a third (3% of the total sample) showed a Cont. profile. Datafrom the only other study to separately report Cont. versus Dproportions at age 3 (Cichetti & Barnett, 1991) also indicated thatthe majority of the D/Cont. group are disorganized rather thancontrolling at this age. A review of other studies that have exam-ined the composition of the D group at different points between 3and 6 years of age suggests that the proportion of controllingchildren increases over the preschool period (Cicchetti & Barnett,1991; Cyr, Bureau, & Moss, 2003; Speltz et al., 1990). It has alsobeen shown that the proportions of D versus Cont. children in-crease with risk (e.g., 25% vs. 11% for a maltreated samplecompared with 17% vs. 11% for a normative comparison sample;Cichetti & Barnett, 1991). Hence, the transformation from a dis-organized to a controlling attachment classification by age 6,which has been demonstrated by Main and Cassidy (1988) andWartner et al. (1994), may not occur for all D children. Longitu-dinal studies tracking the evolution of the D subclassificationsthrough the preschool period are needed.

    One of the key attachment-related domains in the literature isthat of motherchild interaction. Our results indicated differencesbetween secure and insecure attachment groups on a measure ofthe socioaffective quality of motherchild interaction at preschoolage. Consistent with the theoretical idea of goal-corrected partner-ship, our interactive rating system examined the nature of motherchild communications including role balance, emotional openness,and mutual responsiveness. The secure group was rated higher

    than each of the insecure groups. These results support the theo-retical idea that securely attached preschoolers should be capableof maintaining a discourse pattern with the caregiver involvingopen emotional expression, negotiation, and reciprocal control ofcommunication patterns (Cicchetti & Schneider-Rosen, 1984;Marvin, 1977). Our findings are supported by previous postinfancystudies indicating poorer coordination and emotional attunement inthese insecure groups when compared with secure dyads (Main etal., 1985; Moss et al., 1998; NICHD Early Child Care ResearchNetwork, 2001; Stevenson-Hinde & Shouldice, 1995) and demon-strate continuity with infant studies (Egeland & Farber, 1984;Isabella & Belsky, 1991; Lyons-Ruth, Bronfman, & Parsons,1999; Smith & Pederson, 1988). Our findings are also in line withthose of studies conducted both at infancy and early school age(Jacobvitz & Hazen, 1999; Lyons-Ruth, Bronfman, & Parsons,1999; Moss et al., 1998; Moss & St-Laurent, 2001; Teti, 1999) inshowing that the D/Cont. group, and the D group in particular,manifests the most dysfunctional interactive pattern of all attach-ment groups.

    Maternal psychosocial reports provided further evidence of dys-functions that might constitute inherent risk factors to the dyadicfunctioning of mothers and their disorganized children. Mothers ofD/Cont. and D preschoolers reported higher stress on the Attach-ment subscale of the PSI and more difficulty in couple relations. Itis important to point out that the PSI Attachment subscale wasdesigned to assess a particular component of the caregivers at-tachment system, that is, his or her intrinsic investment in theparenting role, rather than broader aspects of the parentchildattachment relationship (Abidin, 1992). Consistent with this theo-retical emphasis, in this study, the scale did not discriminatebetween mothers of secure children and mothers of children withorganized insecure attachment patterns, further supporting the ideathat this subscale should not be seen as a more general measure ofthe attachment relationship. However, the differences demon-strated here between the secure and disorganized groups on theAttachment subscale are supported by other studies, indicating thatthe caregiving system of mothers of disorganized children may be

    Table 4Summary Statistics for Child Behavior Problems in Relation to Attachment Classification

    Measure

    Attachment classification

    Significant planned contrastsB

    (n 106)A

    (n 15)C

    (n 15)D/Cont.(n 15)

    D(n 10)

    Cont.(n 5)

    Teacher-reported problems (PSP)aExternalizing 8.43 9.10 9.48 11.46 12.67 9.42 B D/Cont.**, D**

    (2.63) (2.62) (2.65) (2.77) (2.63) (2.59) A/C D/Cont.*, D**Internalizing 7.60 8.12 7.22 8.56 8.76 8.19 B D/Cont.*, D*

    (2.08) (2.07) (2.10) (2.13) (2.08) (2.06)Mother-reported problems (CBCL)b

    Externalizing 12.67 16.35 16.85 15.83 16.53 14.80 B A*, C*, D*(6.94) (6.91) (7.00) (7.18) (6.96) (7.12)

    Internalizing 6.00 8.55 6.75 6.44 7.26 4.89 B A*(4.34) (4.32) (4.38) (4.35) (4.34) (4.34)

    Note. Means are adjusted for socioeconomic status. Values in parentheses are standard deviations. A insecure avoidant; B secure; C insecuredependent; D insecure disorganized; D/Cont. insecure disorganized and controlling; Cont. insecure controlling; PSP Preschool Socio-AffectiveProfile; CBCL Child Behavior Checklist.a Range for Externalizing and Internalizing 020. b Range for Externalizing 033; range for Internalizing 031.* p .05. ** p .01.

    330 MOSS, BUREAU, CYR, MONGEAU, AND ST-LAURENT

  • disrupted by unresolved trauma or loss, resulting in severe diffi-culties related to decoding and responding to the affective needs oftheir children (George & Solomon, 1999). Several studies havedescribed the pattern of helplessness or abdication of the parentalrole that characterizes the caregiving model and behavior of moth-ers of D infants (George & Solomon, 1996; Lyons-Ruth & Block,1996). According to Lyons-Ruth and Jacobvitz (1999), it is likelythat disorganized preschoolers, like disorganized infants, reactwith conflict and apprehension to this maternal pattern, leading todysregulation in the parentchild relationship. Such dysregulationis likely to provoke child compensation for maternal insufficiencythrough exaggeration of controlling, coercive behavior or extremeinhibition.

    The second domain most strongly associated with attachmenttheoretically and empirically is socioemotional adaptation, mea-sured in this study by both teacher-reported and mother-reportedbehavior problems. At preschool age, a high level of aggressive,disruptive behavior is the most likely indicator of problematicsocioemotional adaptation. The findings of this study are conver-gent with those of studies conducted using infant (Carlson, 1998;Lyons-Ruth et al., 1993, 1997; Shaw et al., 1996; Wartner et al.,1994) and early school-age attachment measures (Greenberg et al.,1991; Moss et al., 1996, 1998; Solomon et al., 1995) in showingthat, according to teacher reports, children who manifest a disor-ganized (D or D/Cont.) attachment with their mothers are at greaterrisk for behavior problems of an externalizing and internalizingnature than are any of the other attachment groups. The peerbehavior of disorganized children has been described as shiftingbetween social withdrawal and extremely aggressive episodes (Ja-cobvitz & Hazen, 1999). As interpreted by these authors, becausedisorganized children may not believe that they can master thechallenges of engaging competently with peers and may see peersas a potential threat, they may demonstrate fight-or-flight behav-iors, that is, aggression accompanied by fearful affect in whichthey aggress against peers in order to ward off perceived threats.

    A further objective of the present study was to examine theextent to which the organized insecure categories, A and C, areassociated with behavior problem profiles. Neither of these groupswas rated by teachers as at risk for either externalizing or inter-nalizing problems. However, our results also indicated that, alongwith mothers of the disorganized group, mothers of these orga-nized insecure groups reported higher externalizing symptomsthan did mothers of secure children. In addition, according tomaternal reports, avoidant children were the only group ratedhigher on internalizing symptoms. In interpreting the significanceof mother and teacher reports in determining childrens adaptation,it is important to point out that although some of the discrepancyin mother and teacher reports may be due to contextual differencesin childrens behavior across settings, maternal reports are alsosubject to bias. Recent research has shown that mothers overreportsymptoms relative to teachers and that maternal reports are aslikely to be influenced by the mothers own adjustment (i.e.,anxiety and stress) as by the childs actual level of problembehavior (Seiffge-Krenke & Kollmar, 1998; Treutler & Epkins,2003; Youngstrom, Loeber, & Stouthamer-Loeber, 2000). Resultsfrom other maternal report measures in this study corroborate thisidea.

    Apart from the parental stress reported by mothers of D children(previously described), mothers of the C group reported greater

    difficulty in controlling their childrens behavior and mothers of Achildren reported their preschoolers to be higher on hyperactivity/distractibility than did mothers of secure children. Thus, althoughit is likely that these maternal reports of organized insecure (A andC) children also reflect a certain level of child interactive diffi-culty, the extent of deviance from the larger peer group is unlikelyto be as great as that of D children. As illustrated in case studies,preschool children who were disorganized were the least compe-tent with peers, becoming aggressive, fearful, or displaying odd,contradictory behavior when initiating play (Jacobvitz & Hazen,1999). Our results are therefore consistent with the existing evi-dence of behavior problem profiles for these organized insecuregroups (Goldberg et al., 1995; Lyons-Ruth et al., 1997; Moss et al.,1998) as well as with a general pattern of lower risk in comparisonwith the disorganized group.

    The final objective of this study was to explore the degree ofdevelopmental risk associated with the controlling and disorga-nized subclassifications of the D/Cont. group. Although the size ofthe controlling subsample in this study did not permit us to conductseparate statistical comparisons between this group and otherattachment classification groups, examination of mean scores onoutcome measures for this subgroup suggested a mixed pattern.Controlling children had scores on dyadic coordination and com-munication that were higher than those of the D group and com-parable to those of the A and C groups. On interactive subscales,which measured the emotional tone of motherchild interactions,means for the controlling group were higher than those of all otherinsecure groups, approaching those of secure children. A parallelpattern was also evident on behavior problem measures. Control-ling childrens mean behavior problem scores were generallylower than those of their D peers and similar to those of the A andC groups. Together these findings support the theoretical idea thatthose preschoolers who lack a coherent strategy for accessing thecaregiver may be at greater risk than those who have developedsome form of organized attachment, albeit an insecure one (George& Solomon, 1999; Teti, 1999).

    However, this more positive profile for the controlling group incomparison with the disorganized group was not evident on ma-ternal psychosocial measures. Means for mothers of controllingchildren were generally the highest of all attachment groups onparental stress and the lowest on perceived parental control. Theseresults are consistent with the interview data of Solomon andGeorge (1999) indicating that mothers of controlling children feelhelpless and out of control in their caregiving relationship. Theseresults suggest that despite the somewhat positive child behavioralprofile shown by controlling children in the early preschool period,these children face serious challenges within the caregiverchildrelationship. The controlling childs strategy of reorienting awayfrom seeking comfort and protection and the meeting of his or herown needs and toward maintaining engagement with the parent onthe parents terms is likely to increase the likelihood of childpsychopathology (Bowlby, 1969/1982). As parentchild role re-versal becomes more manifest with development, behavior prob-lem profiles for this group may also become more evident (Moss,Thibaudeau, Cyr, & Rousseau, 2001). However, both the type(internalizing or externalizing) and severity of risk profiles maydiffer depending on the punitive or caregiving nature of the childscontrolling strategy (Lyons-Ruth, Bronfman, & Atwood, 1999;Solomon & George, 1999; Teti, 1999).

    331PRESCHOOL ATTACHMENT CORRELATES

  • In conclusion, the findings of this study provide strong supportfor the construct validity of the CassidyMarvin classificationsystem as a measure of the quality of parentchild attachmentduring the early preschool years. Attachment groups, coded withthe PACS, differed in predicted directions on outcome variablesassociated with motherchild interactive patterns and child socio-emotional adaptation, both core elements of attachment theory. Inaddition, insecure groups differed from secure children onattachment-related aspects of maternal psychosocial functioning.There was also considerable evidence indicating that the preschoolmeasures of disorganization differentiated the D group from the Aand C groups on a risk continuum. The fact that attachment groupsdid not differ as a function of gender or background measures suchas parental revenue, educational level, and monoparental statusfurther supports the divergent validity of the PACS.

    This study was limited by its focus on concurrent relations andshould be followed by research relating 3-year-old attachment toprior attachment processes in infancy and to later developmentalprocesses and outcomes. It is particularly important to furtherclarify the risk status of the D subgroups. Although our data onmotherchild interaction patterns and behavior problems suggestthat controlling children may be more resilient than their behav-iorally disorganized peers, the sample size was insufficient toconfirm these differences. On the other hand, because the majorityof contrasts indicated that both the D/Cont. and D groups showedpoorer functioning than children classified as B, a risk pattern forboth groups is still evident. Future studies involving the separateanalysis of trajectories of children with either a controlling or adisorganized classification are greatly needed to resolve theseambiguities and to advance our understanding of the family andchild psychosocial correlates of preschool disorganization.

    References

    Abidin, R. (1992). The determinants of parenting behavior. Journal ofClinical Child Psychology, 21, 407412.

    Abidin, R., Jenkins, C. L., & McGaughey, M. C. (1992). The relationshipof early family variables to childrens subsequent behavioral adjustment.Journal of Clinical Child Psychology, 21, 6069.

    Achenbach, T. M. (1991). Manual for the Child Behavior Checklist/418and 1991 profile. Burlington: University of Vermont, Department ofPsychiatry.

    Achenbach, T. M., & Edelbrock, C. S. (1981). Behavioral problems andcompetencies reported by parents of normal and disturbed children agedfour through sixteen. Monographs of the Society for Research in ChildDevelopment, 46(1, Serial No. 188).

    Achermann, J., Dinneen, E., & Stevenson-Hinde, J. (1991). Clearing up at2.5 years. British Journal of Developmental Psychology, 9, 365376.

    Bar-Haim, Y., Sutton, B., Fox, N., & Marvin, R. (2000). Stability andchange of attachment at 14, 24, and 58 months of age: Behavior,representation, and life events. Journal of Child Psychology and Psy-chiatry and Allied Disciplines, 41, 381388.

    Barnett, D., Kidwell, S. L., & Leung K. H. (1998). Parenting and pre-schooler attachment among low-income urban African-American fami-lies. Child Development, 69, 16571671.

    Bates, J. E., Maslin, C. A., & Frankel, K. A. (1985). Attachment security,motherchild interaction, and temperament as predictors of behaviorproblem ratings at age three years. In I. Bretherton & E. Waters (Eds.),Growing points of attachment theory and research. Monographs of theSociety for Research in Child Development, 50(12, Serial No. 209),167193.

    Beck, A. T., Steer, R. A., & Garbin, M. G. (1988). Psychometric propertiesof the Beck Depression Inventory: Twenty-five years of evaluation.Clinical Psychology Review, 8, 77100.

    Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961).An inventory for measuring depression. Archives of General Psychiatry,4, 561571.

    Bowlby, J. (1969/1982). Attachment and loss: Vol. 1. Attachment. NewYork: Basic Books.

    Bowlby, J. (1973). Attachment and loss: Vol. 2. Separation. New York:Basic Books.

    Bowlby, J. (1980). Attachment and loss: Vol. 3. Loss, sadness and depres-sion. New York: Basic Books.

    Bretherton, I., Ridgeway, D., & Cassidy, J. (1990). Assessing internalworking models of the attachment relationship: An attachment storycompletion task for 3-year-olds. In M. T. Greenberg, D. Cichetti, &E. M. Cummings (Eds.), Attachment in the preschool years (pp. 349).Chicago: University of Chicago Press.

    Bumberry, W., Oliver, J. M., & McClure, J. N. (1978). Validation of theBeck Depression Inventory in a university population using psychiatricestimate as the criterion. Journal of Consulting and Clinical Psychology,46, 150155.

    Carlson, E. A. (1998). A prospective longitudinal study of disorganized/disoriented attachment. Child Development, 69, 19701979.

    Cassidy, J., Kirsh, S. J., Scolton, K. L., & Parke, R. D. (1996). Attachmentand representations of peer relationships. Developmental Psychology,32, 892904.

    Cassidy, J., & Marvin, R. S., with the MacArthur Working Group onAttachment. (1992). Attachment organization in three- and four-year-olds: Procedures and coding manual. Unpublished manuscript, Univer-sity of Virginia.

    Cicchetti, D., & Barnett, D. (1991). Attachment organization in maltreatedpreschoolers. Development and Psychopathology, 3, 397411.

    Cicchetti, D., & Schneider-Rosen, K. (1984). Theoretical and empiricalconsiderations in the investigation of the relationship between affect andcognition in atypical populations of infants. In C. Izard, J. Kagan, & R.Zajonc (Eds.), Emotions, cognition and behavior (pp. 306406). NewYork: Cambridge University Press.

    Cohn, D. A. (1990). Childmother attachment of six-year-olds and socialcompetence at school. Child Development, 61, 152162.

    Cyr, C., Bureau, J.-F., & Moss, E. (2003, April). Stability of attachmentfrom preschool to school age: The impact of quality of motherchildinteraction and family risk. Paper presented at the meeting of the Societyfor Research in Child Development, Tampa, FL.

    Cyr, C., & Moss, E. (2001). The role of motherchild interaction anddepression at preschool age in the prediction of child attachment atschool-age. Canadian Journal of Behavioral Sciences, 33, 7787.

    Davies, P., & Cummings, E. (1994). Marital conflict and child adjustment:An emotional security hypothesis. Psychological Bulletin, 116, 387411.

    De Mulder, E. K., & Radke-Yarrow, M. (1991). Attachment with affec-tively ill and well mothers: Concurrent behavioral correlates. Develop-ment and Psychopathology, 3, 227242.

    Dubois-Comtois, K., & Moss, E. (in press). The relation between attach-ment at preschool age and motherchild interaction in the lab and homesettings. Canadian Journal of Behavioral Science.

    Easterbrooks, M. A., Davidson, C. E., & Chazan, R. (1993). Psychosocialrisk, attachment, and behavior problems among school-aged children.Development and Psychopathology, 5, 389402.

    Egeland, B., & Farber, E. (1984). Infantmother attachment: Factors re-lated to its development and changes over time. Child Development, 55,753771.

    Erickson, M. F., Sroufe, L. A., & Egeland, B. (1985). The relationship ofquality of attachment and behavior problems in preschool in a high risksample. In I. Bretherton & E. Waters (Eds.), Growing points in attach-

    332 MOSS, BUREAU, CYR, MONGEAU, AND ST-LAURENT

  • ment theory and research. Monographs of the Society for Research inChild Development, 50(12, Serial No. 209), 147186.

    Fagot, B. I. (1995). Classification of problem behaviors in young children:A comparison of four systems. Journal of Applied Developmental Psy-chology, 16, 95106.

    Fagot, B. I., & Pears, K. C. (1996). Changes in attachment during the thirdyear: Consequences and predictions. Development and Psychopathol-ogy, 8, 325344.

    George, C., & Solomon, J. (1996). Representational models of relation-ships: Links between caregiving and attachment. Infant Mental HealthJournal, 17, 198216.

    George, C., & Solomon, J. (1999). Attachment and caregiving: The care-giving behavioral system. In J. Cassidy & P. R. Shaver (Eds.), Handbookof attachment: Theory, research and clinical applications (pp. 649670). New York: Guilford Press.

    Goldberg, S., Gotowiec, A., & Simmons, R. (1995). Infantmother attach-ment and behavior problems in healthy and chronically ill preschoolers.Development and Psychopathology, 7, 267282.

    Greenberg, M. A., Speltz, M. L., DeKlyen, M., & Endriga, M. C. (1991).Attachment security in preschoolers with and without externalizingproblems: A replication. Development and Psychopathology, 3, 413430.

    Hesse, E., & van IJzendoorn, M. (1998). Parental loss of close familymembers and propensities towards absorption in offspring. Developmen-tal Science, 1, 299305.

    Howes, C., & Hamilton, C. (1992). Childrens relationships with child careteachers: Stability and concordance with parental attachments. ChildDevelopment, 63, 867878.

    Isabella, R. A., & Belsky, J. (1991). Interactional synchrony and the originsof infantmother attachment: A replication study. Child Development,62, 373384.

    Jacobvitz, D., & Hazen, N. (1999). Developmental pathways from infantdisorganization to childhood peer relationships. In J. Solomon & C.George (Eds.), Attachment disorganization (pp. 127159). New York:Guilford Press.

    LaFreniere, P. J., Dumas, J. E., Capuano, F., & Dubeau, D. (1992).Development and validation of the Preschool Socioaffective Profile.Psychological Assessment, 4, 442450.

    Laible, D. J., & Thompson, R. A. (1998). Attachment and emotionalunderstanding in preschool children. Developmental Psychology, 34,10381045.

    Lyons-Ruth, K., Alpern, L., & Repacholi, B. (1993). Disorganized infantattachment classification and maternal psychosocial problems as predic-tors of hostile-aggressive behavior in the preschool classroom. ChildDevelopment, 64, 572585.

    Lyons-Ruth, K., & Block, D. (1996). The disturbed caregiving system:Relations among childhood trauma, maternal caregiving, and infantaffect and attachment. Infant Mental Health Journal, 17, 257275.

    Lyons-Ruth, K., Bronfman, E., & Atwood, G. (1999). A relational diathesismodel of hostilehelpless states of mind: Expressions in motherinfantinteraction. In J. Solomon & C. George (Eds.), Attachment disorgani-zation (pp. 3370). New York: Guilford Press.

    Lyons-Ruth, K., Bronfman, E., & Parsons, E. (1999). Maternal frightened,frightening, or atypical behavior and disorganized infant attachmentpatterns. In J. Vondra & D. Barnett (Eds.), Atypical attachment ininfancy and early childhood among children at developmental risk:Monographs of the Society for Research in Child Development, 64(3),6796.

    Lyons-Ruth, K., Connell, D., Grunebaum, H., & Botein, S. (1990). Infantsat social risk: Maternal depression and family support services as me-diators of infant development and security of attachment. Child Devel-opment, 61, 8598.

    Lyons-Ruth, K., Easterbrooks, A., & Cibelli, C. (1997). Infant attachmentstrategies, infant mental lag, and maternal depressive symptoms: Pre-

    dictors of internalizing and externalizing problems at age 7. Develop-mental Psychology, 33, 681692.

    Lyons-Ruth, K., & Jacobvitz, D. (1999). Attachment disorganization:Unresolved loss, relational violence, and lapses in behavioral and atten-tional strategies. In J. Cassidy & P. R. Shaver (Eds.), Handbook ofattachment: Theory, research and clinical applications (pp. 520554).New York: Guilford Press.

    Main, M., & Cassidy, J. (1988). Categories of response to reunion with theparent at age six: Predictable from infant attachment classifications andstable over a 1-month period. Developmental Psychology, 24, 415426.

    Main, M., Kaplan, N., & Cassidy, J. (1985). Security in infancy, childhood,and adulthood: A move to the level of representation. In I. Bretherton &E. Waters (Eds.), Growing points of attachment theory and research:Monographs of the Society for Research in Child Development, 50(12,Serial No. 209), 66104.

    Main, M., & Solomon, J. (1990). Procedures for identifying infants asdisorganized/disoriented during the Ainsworth Strange Situation. InM. T. Greenberg, D. Cicchetti, & M. Cummings (Eds.), Attachment inthe preschool years: Theory, research, and intervention (pp. 121160).Chicago: University of Chicago Press.

    Manassis, K., Bradley, S., Goldberg, S., Hood, J., & Swinson, R. P. (1994).Attachment in mothers with anxiety disorders and their children. Journalof the American Academy of Child and Adolescent Psychiatry, 33,11061113.

    Marvin, R. S. (1977). An ethologicalcognitive model of the attenuation ofmotherchild attachment behavior. In T. Alloway, L. Krames, & P.Pilner (Eds.), Advances in the study of communication and affect: Vol. 3.Attachment behavior (pp. 2560). New York: Plenum Press.

    Moss, E., Gosselin, C., Parent, S., Rousseau, D., & Dumont, M. (1997).Attachment and joint problem-solving experiences during the preschoolperiod. Social Development, 6, 117.

    Moss, E., Parent, S., Gosselin, C., Rousseau, D., & St-Laurent, D. (1996).Attachment and teacher-reported behavior problems during the pre-school and early school-age period. Development and Psychopathology,8, 511525.

    Moss, E., Rousseau, D., Parent, S., St-Laurent, D., & Saintonge, J. (1998).Correlates of attachment at school age: Maternal reported stress,motherchild interaction, and behavior problems. Child Development,69, 13901405.

    Moss, E., & St-Laurent, D. (2001). Attachment at school age and academicperformance. Developmental Psychology, 37, 863874.

    Moss, E., St-Laurent, D., Cyr, C., & Humber, N. (2000). Lattachementaux periodes prescolaires et scolaires et les patrons dinteractionsparentenfant. In G. M. Tarabulsy, S. Larose, D. R. Pederson, & G.Moran (Eds.), Attachement et developpement : Le role des premie`resrelations dans le developpement humain. Sainte-Foy, Quebec, Canada:Les Presses de lUniversite du Quebec.

    Moss, E., Thibaudeau, P., Cyr, C., & Rousseau, D. (2001, April). Control-ling attachment and child management of parental emotion. In D. R.Pederson & C. A. DeOliviera (Chairs), Attachment and the socializationof emotions. Symposium conducted at the biennal meeting of the Societyfor Research in Child Development, Minneapolis, MN.

    NICHD Early Child Care Research Network. (2001). Child-care and fam-ily predictors of preschool attachment and stability from infancy. De-velopmental Psychology, 37, 847862.

    Parke, K., & Waters, E. (1989). Security of attachment and preschoolfriendships. Child Development, 60, 10761081.

    Radke-Yarrow, M., Cummings, E., Kuczynski, L., & Chapman, M. (1985).Patterns of attachment in two- and three-year-olds in normal familiesand families with parental depression. Child Development, 56, 884893.

    Rothbaum, F., Schneider-Rosen, K., Pott, M., & Beatty, M. (1995). Earlyparentchild relations and later problem behaviour: A longitudinalstudy. Merrill-Palmer Quarterly, 41, 133151.

    Rouzier, F. (1986). E laboration dun questionnaire devaluation des atti-

    333PRESCHOOL ATTACHMENT CORRELATES

  • tudes et pratiques educatives des parents denfants dage prescolaire(Q. E. A. P. E. P.). Unpublished manuscript, Departement des sciencesde leducation, Universite du Quebec a` Montreal, Montreal, Quebec,Canada.

    Sarason, I. G., Johnson, J. H., & Siegel, J. M. (1978). Assessing the impactof life changes: Development of the Life Experiences Survey. Journal ofConsulting and Clinical Psychology, 46, 932946.

    Scher, A., & Mayseless, O. (2000). Mothers of anxious/ambivalent infants:Maternal characteristics and child care context. Child Development, 71,16291639.

    Seiffge-Krenke, I., & Kollmar, F. (1998). Discrepancies between mothersand fathers perceptions of sons and daughters problem behavior: Alongitudinal analysis of parentadolescent agreement on internalizingand externalizing problem behavior. Journal of Child Psychology andPsychiatry and Allied Disciplines, 39, 687697.

    Shaw, D. S., Keenan, K., Vondra, J. I., Delliquadri, E., & Giovanelli, J.(1997). Antecedents of preschool childrens internalizing problems: Alongitudinal study of low-income families. Journal of the AmericanAcademy of Child and Adolescent Psychiatry, 36, 17601767.

    Shaw, D., Owens, E., Vondra, J., & Keenan, K. (1996). Early risk factorsand pathways in the development of early disruptive behavior problems.Development and Psychopathology, 8, 679699.

    Smith, P. B., & Pederson, D. R. (1988). Maternal sensitivity and patternsof infantmother attachment. Child Development, 59, 10971101.

    Solomon, J., & George, C. (1999). The place of disorganization in attach-ment theory: Linking classic observations with contemporary findings.In J. Solomon & C. George (Eds.), Attachment disorganization (pp.332). New York: Guilford Press.

    Solomon, J., George, C., & De Jong, A. (1995). Children classified ascontrolling at age six: Evidence of disorganized representational strate-gies and aggression at home and at school. Development and Psycho-pathology, 7, 447463.

    Speltz, M. L., Greenberg, M. T., & De Klyen, M. (1990). Attachment inpreschoolers with disruptive behavior: A comparison of clinic-referredand nonproblem children. Development and Psychopathology, 2, 3146.

    Sroufe L. A. (1979). The coherence of individual development. AmericanPsychologist, 34, 834841.

    Stevenson-Hinde, J., & Shouldice, A. (1990). Fear and attachment in2.5-year-olds. British Journal of Developmental Psychology, 8, 319333.

    Stevenson-Hinde, J., & Shouldice, A. (1995). Maternal interactions andself-reports related to attachment classifications at 4.5 years. ChildDevelopment, 66, 583596.

    Teti, D. M. (1999). Conceptualizations of disorganization in the preschoolyears: An integration. In J. Solomon & C. George (Eds.), Attachmentdisorganization (pp. 213242). New York: Guilford Press.

    Teti, D. M., & Ablard, K. E. (1990). Security of attachment and infantsibling relationships: A laboratory study. Child Development, 60, 15191528.

    Teti, D. M., & Gelfand, D. M. (1997). The preschool assessment ofattachment: Construct validity in a sample of depressed and nonde-pressed families. Development and Psychopathology, 9, 517536.

    Teti, D. M., Gelfand, D. M., Messinger, D. S., & Isabella, R. (1995).Maternal depression and the quality of early attachment: An examinationof infants, preschoolers, and their mothers. Developmental Psychology,31, 364376.

    Thompson, R. A. (1999). Early attachment and later development. In J.Cassidy & P. R. Shaver (Eds.), Handbook of attachment: Theory, re-search, and clinical applications (pp. 265286). New York: GuilfordPress.

    Treutler, C. M., & Epkins, C. C. (2003). Are discrepancies among child,mother, and father reports on childrens behavior related to parentspsychological symptoms and aspects of parentchild relationships?Journal of Abnormal Child Psychology, 31, 1327.

    Turner, P. J. (1991). Relations between attachment, gender, and behaviorwith peers in preschool. Child Development, 62, 14751488.

    Wartner, U. G., Grossman, K., Fremmer-Bombik, E., & Suess, G. (1994).Attachment patterns at age six in south Germany: Predictability frominfancy and implications for preschool behavior. Child Development, 65,10141027.

    Waters, E., & Cummings, M. E. (2000). A secure base from which toexplore close relationships. Child Development, 71, 164172.

    Youngstrom, E., Loeber, R., & Stouthamer-Loeber, M. (2000). Patternsand correlates of agreement between parent, teacher, and male adoles-cent ratings of externalizing and internalizing problems. Journal ofConsulting and Clinical Psychology, 68, 10381050.

    Received January 15, 2003Revision received November 3, 2003

    Accepted December 31, 2003

    334 MOSS, BUREAU, CYR, MONGEAU, AND ST-LAURENT