You are on page 1of 18

Developmental Psychology 2008, Vol. 44, No.

4, 939 –956

Copyright 2008 by the American Psychological Association 0012-1649/08/$12.00 DOI: 10.1037/0012-1649.44.4.939

The Child Attachment Interview: A Psychometric Study of Reliability and Discriminant Validity
Yael Shmueli-Goetz, Mary Target, Peter Fonagy, and Adrian Datta
University College London
While well-established attachment measures have been developed for infancy, early childhood, and adulthood, a “measurement gap” has been identified in middle childhood, where behavioral or representational measures are not yet sufficiently robust. This article documents the development of a new measure—the Child Attachment Interview (CAI)—which seeks to bridge this gap. The CAI is a semistructured interview, in which children are invited to describe their relationships with their primary caregivers. The coding system is informed by the Adult Attachment Interview and the Strange Situation Procedure, and produces 4 attachment categories along with a continuous measure of attachment security based on ratings of attachment-related dimensions. The main psychometric properties are presented, including interrater reliability, test–retest reliability, and concurrent and discriminant validities, both for normally developing children and for those referred for mental health treatment. The CAI correlates as expected with other attachment measures and predicts independently collected ratings of social functioning. The findings suggest that the CAI is a reliable, valid, and promising measure of child–parent attachment in middle childhood. Directions for improvements to the coding system are discussed. Keywords: attachment, middle childhood, psychometric properties

The nature of attachment relationships has been the subject of empirical investigation for half a century and has given rise to an impressive body of literature. The quality of parent– child relationships during infancy and early childhood has been considered to constitute a significant factor in later personality and the development of psychopathology (Dozier, Stovall, & Albus, 1999; Green & Goldwyn, 2002; Greenberg, 1999; Weinfield, Whaley, & Egeland, 2004). Several studies have found associations between insecure attachment with the primary caregiver(s) in infancy and poor social competence and peer relations, increased hostility and aggression, and lower ego resilience in the preschool and preadolescent years (Lyons-Ruth, Connell, Zoll, & Stahl, 1987; Shaw & Vondra, 1995; Sroufe, 1983; Sroufe, Egeland, & Kreutzer, 1990; Stams, Juffer, & van IJzendoorn, 2002; Urban, Carlson, Egeland,

Yael Shmueli-Goetz, Mary Target, Peter Fonagy, and Adrian Datta, Subdepartment of Clinical Health Psychology, University College London. Yael Shmueli-Goetz is now at the Anna Freud Centre, London. Mary Target and Peter Fonagy are now affiliated with both the Subdepartment of Clinical Health Psychology of the University College London and the Anna Freud Centre. Adrian Datta is now affiliated with Child and Adolescent Mental Health Services, Guernsey, United Kingdom, and the Anna Freud Centre. We would like to acknowledge the support of United Kingdom Mental Health Foundation Grant SCYP 97/07 and the Trustees of the Anna Freud Centre. We would also like most gratefully to acknowledge the roles of Karin Ensink, who is project manager for the wider project based at the Anna Freud Centre, and Pasco Fearon for help with the latent class statistical analysis. Correspondence concerning this article should be addressed to Mary Target, Subdepartment of Clinical Health Psychology, University College London, 1-19 Torrington Place, London WC1E 7HB, United Kingdom. E-mail: 939

& Sroufe, 1991). Recent research has clarified that social context critically moderates the strength of such predictions and interacts with specific classes of attachment. Belsky and Fearon (2002) have analyzed the National Institute of Child Health and Human Development (NICHD) data set to clarify the relationships between attachment insecurity and poor developmental outcomes. Social contextual risk factors helped to predict some preschool outcomes (e.g., cognitive development indexed by readiness for school) and moderate the direct influence of attachment for others (e.g., social competence and expressive language). While all children were affected by high contextual risk, those who had been avoidant as infants showed impaired social– emotional and language development at intermediate levels of environmental risk. This helps us to understand why the relationship between early attachment and adult attachment may be more readily observed in high-risk samples relative to low-risk ones (Hamilton, 2000; Waters, Merrick, Treboux, Crowell, & Albersheim, 2000; Weinfield, Sroufe, & Egeland, 2000). Moreover, attachment insecurity in infancy and early childhood has been shown to predict various forms of psychopathology in adolescence and adulthood (Ogawa, Sroufe, Weinfield, Carlson, & Egeland, 1997; Warren, Huston, Egeland, & Sroufe, 1997). What has been somewhat lacking is a better understanding of the way in which the attachment system interfaces with personality development (e.g., Allen et al., 2003; Larose & Bernier, 2001; Marsh, McFarland, Allen, McElhaney, & Land, 2003; Ziv, Oppenheim, & Sagi-Schwartz, 2004) and the emergence of psychopathology in middle childhood (e.g., Bar-Haim et al., 2002; Easterbrooks, Biesecker, & Lyons-Ruth, 2000). One reason for this lack may be the absence of appropriate measurement instruments for assessing attachment in the school years. Two measurement approaches have dominated attachment research. Prototypical of the first approach, the Strange Situation Procedure (SSP; Ainsworth, Blehar, Waters, & Wall, 1978) was

and a number of attachment-focused doll-play procedures (e. Kaplan. Pederson & Moran. Goldberg. 1996. in the Trowell et al. George. none of the girls was classified as showing disorganization of attachment (J. slipping into talking about fantasy events. 1999. Wright. 2003. van IJzendoorn. Trowell. 1996. family drawings (e. Madigan. the available assessments. Additionally. Seifer. Numerous studies have established the reliability and validity of both the SSP (Bar-Haim. 66). Scharf. these expectations are captured in his/her projective response to an attachment-related drawing or story. becoming suddenly inappropriately intimate with the interviewer. 2002. 1990). separation–reunion procedures akin to the SSP have yielded analogous attachment classifications for 2.g. Many researchers have developed semiprojective procedures for early and middle childhood. & Marvin. Smith. some investigators have used a version of the AAI (Ammaniti et al. 2004). 2000. driven by the supposition that inferred mental representations would reflect or represent children’s attachment organization in the same way that behavior patterns do. representational measures rest on the assumption that a child applies knowledge of his/her own family to the standard drawings. NICHD Early Child Care Research Network. on occasion. Koenig. Bureau..940 SHMUELI-GOETZ. Smith. 1988. 2000. Slough & Greenberg. 1994. Treboux. 1989) and the AAI (Bakermans-Kranenburg & van IJzendoorn.. . Crowell. The strongest predictor of infant security of attachment was the coherence of the transcript.. Semantic knowledge of cultural stereotypes. & Vetter. Stanley. Goldwyn. 2000. Kaplan. & Goldberg. van IJzendoorn & Bakermans-Kranenburg. the study of attachment in early and middle childhood continues to prove a challenge. & Smith. longitudinal prediction from parent–infant attachment or assessment of maternal attachment status in an attempt to validate these measures.5. adopting a stylized manner copied from an action hero. & Pederson. it is worth considering whether there were nonverbal signs not captured by the adult coding system.. Lefever. 1988). many new measures have appeared. However. these interviews underrepresent disorganized attachment representations and overrepresent dismissing attachment strategy (Ammaniti. and manner. poses difficulties in developing reliable behaviorally derived classification systems for these age groups. and Cassidy’s (1985) seminal article describing the development of an interview-based method. & Green. Children may show attachment disorganization through many aspects of behavior not tapped by the AAI coding (e. Steele. 2004). It seems likely that. quantity.. Marvin. 1992. 2000) because of developmental differences. Main & Cassidy. 1991. Described by Main et al. others have found only weak associations and moderate-to-low stability over time (e. the absence of a sequence of critical moments in the behavioral assessment of separations and reunions comparable to those provided in the SSP.g. including modifications of the behavioral approach of the SSP and representational instruments. Main & Cassidy. The predominance of indirect projective approaches to the study of attachment in the preschool and middle childhood years is due to the premise that attachment organization cannot be captured through direct questioning. & Kivenson-Baron. 1993.. marked an important turning point in the study of attachment beyond infancy and represents the second dominant measurement approach. 1985). together with. The publication of Main. Mental representations have been elicited through a variety of procedures: the Separation Anxiety Test (SAT. Binney. 1997). & Barglow. (2002) study of known sexual abuse.g. This article explores the possibilities of more direct personal assessment. TARGET. Cyr. Slough & Greenberg. and there is concern about low test–retest reliability (e. AND DATTA developed to examine infants’ and toddlers’ behavioral strategies for maintaining proximity to their attachment figures. Ladd. 1994. 2004. Speranza. cultural) ideal representations... 1992.. & Steele. Moran. 1988. Mongeau.. 2001). & Tambelli. coupled with children’s growing verbal abilities. & O’Connor. but such an assumption fails to consider that elicited representations may reflect other (e. (1985) as the “move to the level of representations” (p. 1997) and have led to an exponential growth in the study of attachment relationships. Stanley. An infant or child showing behaviors considered to reflect an insecure attachment is thus assumed to hold a set of expectations that he/she is unloved and will not receive comfort and support. Fonagy. Gomille. at least in the younger age range. 1989. While it might be that the girls had “rehearsed” their abuse histories in earlier interviews such that disorganization of abuse narratives was not seen. including quality. Vaughn. SSP studies examining the stability of attachment patterns from infancy to preschool have produced mixed results. Green. & Goldwyn. Crittenden.. 1985). 1988). However. While measures designed to assess attachment organization in infancy and adulthood are well established. 1999). displaying silly or regressive behavior with the interviewer. mainly the SAT. Coherence was conceptualized in part as the consistency between semantic and episodic memory and was illustrated in terms of adherence to—and violations of—Grice’s (1975) maxims of discourse. Slough & Greenberg. Shouldice & StevensonHinde. 1978) and for 6-year-olds (Main & Cassidy. family photos (Main & Cassidy. Benoit & Parker. have not attained the rigorous validation of the SSP and the AAI.g. van IJzendoorn. Fox. the Adult Attachment Interview (AAI. for school-age children. Cassidy. & Main. Moss. wishful thinking. 16-year-olds directly about attachment relationships and experiences. & 4. 1995. & the MacArthur Working Group on Attachment.g. personal communication. & St-Laurent. Resnick. Waters. or episodes witnessed secondhand do not necessarily tell us anything helpful about the child’s expectations of his/her own attachment figures’ emotional availability. relation. October 2001). 1990) to ask 10. or representations that are not veridical or that contain distortions of the child’s own subjective experiences.5-year-olds (Ainsworth et al.g. Mayseless. While some studies employing separation–reunion procedures have reported clear associations between children’s responses and attachment representations (George et al. the AAI was conceived as a way of predicting infants’ Strange Situation attachment patterns and was conceptualized as requiring individuals to recount and reflect on early attachment-related relationships and experiences while maintaining coherent and collaborative discourse. 1995) and validity (e. For assessment during the preschool years. For example. and making grimaces while continuing to speak fairly coherently). In recent years. with stability ranging from 38% to 72% (Bar-Haim et al. Smith. Bowers. Nevertheless. Shouldice & Stevenson-Hinde. 1990). FONAGY. Hamilton. These have shown the expected associations between classifications derived behaviorally and representationally and to a certain extent between the attachment representations of child and mother (Gloger-Tippelt. Sutton. Madigan. Most studies have used behavioral and representational instruments. 1990.

Interrater reliabilities with expert as well as naı ¨ve coders are reported. and socioeconomic status. While purely behavioral measures of attachment (separation and reunion) are no longer appropriate in the school years. & Abraham. gender. The construct validity of the CAI is examined in three domains: the overlap with other narrative-based measures of attachment and the SAT (Wright et al. this may still be the case for the school-age child. 2005). In addition. These scales show adequate internal consistency and 2-week test–retest reliability.. We evaluated the discriminant validity of the CAI by testing the relation between attachment classifications and age. This article presents some of the findings relating to the CAI’s psychometric properties for both clinical and nonclinical samples of adequate size. asking children to rate statements that refer to their expectations. should the need for help arise (Kerns et al. Fourth. 1987. includes developmentally appropriate scaffolding. we predicted that some of these domains would be related to attachment (e. Aspelmeier. followed by test–retest reliability across a 3-month and a 12-month interval. First. p. Recently. In infancy. including clinical as well as normative samples. The coding and classification system of both the CAI and AAI represents a compromise between the two. 1923. We felt that the CAI should be developed using a sufficiently large group. 1996). as in the AAI. although this integration is expected to have occurred by adulthood.. in press) and that some of the categories such as “earned secure” are difficult to identify in such investigations. as cited in Ainsworth. therefore. 1994. Furthermore. the stability of attachment classification across time is a vital aspect of validity.g. a caregiver interview coded on 14 dimensions.. 2004). we hoped to help overcome what has been called the “measurement roadblock” in attachment (Greenberg. and there is evidence that it does happen in two thirds of children by college age (Furman & Simon. highlighting the need for a developmentally appropriate interviewer stance. 1982) and that this tendency is also shown in children’s memory and recall capacity (with younger children showing a memory bias toward recent events. Fraley. Morgan. and behavior in relation to a specific attachment figure. several questionnaires have been developed (Kerns. The traditional categorical approach and the identification of fairly broad categories do not allow for a fine-tuned analysis of distinct attachment dimensions and their possible interaction. the correspondence between attachment classification of parent and child on the basis of the AAI and previous robust findings of intergenerational consistency of attachment classifications (van IJzendoorn & Bakermans-Kranenburg. 1985). 56).g. While the Security and Avoidant Scales have proven to be correlated with other measures of attachment. p. Schneider. & Contreras. any measure should attempt to assess the degree to which children perceive their parents as available and accessible. 2000).CHILD ATTACHMENT INTERVIEW 941 More recently. 2001). the lack of “narrative diachronicity” (Bruner. Linear scales have often been constructed as complementary to the classification systems. Fonagy. Tomich. conclusive evidence has emerged from a taxometric study that secure versus dismissing discourse on the AAI is a continuously distributed dimension (Roisman. Such an approach would allow us to distinguish a developmental limitation (expressed through an impoverished narrative) from a dismissing strategy. and is piloted and validated on both normative and clinical samples. in light of evidence that young children tend to describe themselves in terms of the immediate present (Damon & Hart. affect. 2000). the “set goal” of the attachment system during the middle childhood years is no longer considered to be the physical proximity but rather the availability of the attachment figure (Bowlby. May 2003). 1999. Second. and yet it has rarely been considered. children seem to have independent working models of attachment in relation to their caregivers. & Belsky. children cannot yet fully mask or control behaviors that indicate anxiety or other emotions. The CAI is a narrativebased assessment that relies on a level of linguistic competence. Several central conceptual and methodological considerations guided the development of the Child Attachment Interview (CAI). special consideration should be given to the way in which narratives are elicited. In developing the CAI. A general problem in the development of attachment instruments has been that they have been standardized on relatively small samples of the normal population and then applied to clinical and high-risk groups (e. Coding of nonverbal behavior is therefore available as an additional helpful indicator of attachment strategies. p. we cannot assume that an integrated state of mind with respect to attachment has been achieved in middle childhood (Bretherton. the CAI seeks to assess children’s perceptions of their attachment figures’ current availability. relies on nonverbal as well as verbal communication in the analysis of narratives. There is a controversy among attachment researchers over whether attachment behaviors or representations should be described using continuous or categorical variables (see Developmental Psychology’s special section. 1990). As such. In this way. 1995). These considerations indicate the need for an age-specific interview and coding system for assessment of attachment status through narratives. it might simply reflect the child’s linguistic egocentricity (Piaget. The interviewer needs to provide “scaffolding” (appropriate cues) to help children to remember attachment experiences (see Nelson. Wallis & Steele. 19) of child and adolescent narratives has to be distinguished from a preoccupied or disorganized attachment strategy. the Preoccupied Scale has not shown a consistent pattern of associations with other attachment measures (Kerns. which may have predictive value in this still-formative period of development. we have pursued both strategies and anticipate that further research will clarify whether a continuous measure of insecurity or a categorical system will provide greater powers of prediction. 2005). 486) by developing a new interview to assess children’s internal models of attachment relationships based on questioning them directly about their experiences with and perceptions of their primary caregivers. and a multidomain measure of adaptation. but relatively few studies have adopted a dimensional approach in establishing attachment organization. 1993). 1981). Schlegelmilch. For these reasons. Ensink. Fitzgerald. it may be more appropriate to ask children about current rather than past relationships. & Janes. We have attempted to find a better compromise between using indirect assessment of representations and simply using an adult interview and coding system. impulse control and parent– child . elicits separate representation of attachment figures. Like Belsky and Fearon (2002). Statements are rated along a security continuum and for avoidant and preoccupied coping. Nor does lack of coherence necessarily reflect attachment insecurity. the relation between attachment and expressive language competence as well as IQ is explored. the Hampstead Child Adaptation Measure (Target. focusing on aspects of reliability and discriminant validity. Third.

␹2(1. 1985) and the AAI (George et al. typically.9 (18. additional probes are used to elicit relevant instances or episodic detail. added partly to help the child get used to talking with a stranger about personal matters but also to investigate possible meaningful links between self-descriptions and attachment representations. often described in the pilot interviews. and loss.20. N ϭ 226) ϭ 4. but it also incorporated elements of behavioral coding based on videotapes and segmented narratives into relationship episodes (Luborsky & Crits-Christoph. The mean age at interview was significantly higher in the nonreferred group. nonreferred samples Characteristic Mean age in years (SD) Mean verbal IQ (SD) % Boys % Middle class % White % Living with two parents Nonreferred (n ϭ 161) 10. p Ͻ . Guided by the above 12-year-olds.50. we adapted them for use with 7. 11 met exclusion criteria. The current version of the CAI (see Appendix) comprises 15 questions. 1994). “Is that what usually happens?” “Did you?” “Is there anything else you remember?” “That is a good example. A further important difference was that the CAI focused on recent attachment-related events and current attachment relationships rather than the memory of relationships in earlier childhood.40. and where possible.2 (18. ␹2(1. N ϭ 226) ϭ 0. the style was considerably different: The interview needed to be consistent enough to reveal structural variations in response and flexible enough to help children with its demands without compromising validity. 1990). representations of his/her primary caregivers. No other differences approached statistical significance. TARGET. we reviewed the questions on the Berkeley Autobiographical Interview (Main et al.2) 102. or severe mental health problems).9 82. separation. The remaining 89 were approached to participate in the study. Throughout the interview. would not (e. The demographics of these children. The representative sample consisted of 301 children whose parents were approached for permission for their children to participate in a study of social and emotional development in childhood. 1990). as did 73% (65) of the clinical sample. owing to the restricted attentional capacity of younger children. The interview needed on average to be about half the length of the AAI. A number of pertinent issues concerning the child’s experience of conflict with caregivers. ␹2(1. An independent subsample of an additional 33 children was identified to take part in the 1-year follow-up. However. ns ns ns. allowing the coder to include Table 1 Demographic Characteristics of the Two Sample Groups (Nonreferred and Referred) Psychiatric referred vs. interested comments such as.5 33. N ϭ 226) ϭ 0. and the division of narrative about relationships into relationship episodes (Luborsky & Crits-Christoph.9 (1.4 Statistic t(224) ϭ 2.3) 58. as the CAI also needed to activate the attachment system to elicit attachmentrelated information. AND DATTA relationships). and the interview was scored from videotape rather than from transcript. Of these children.10. The interview opens with a warm-up question eliciting information relating to family composition. The protocol of questions was closest in content to the AAI.21. The interviewer also provides scaffolding to assist the child in telling the story. 10 met one of a small number of exclusion criteria (the child or main caregiver not speaking English at the level of an average 7-year-old..05 t(156) ϭ 1.0 44. represented 55% of those eligible. Approximately 55% (161) of the school sample approached agreed to participate. required new questions about arguments with and between parents.. can you tell me more about it?” “Was it after school?” and “Who was there?” The CAI Coding and Classification System. hurt. ns ␹2(1. theoretically unrelated to attachment. times of conflict. 1985). ns . A clinical sample was drawn from 100 children who were 7–12 years old and had been consecutively referred for psychiatric assessment to three London specialist child mental health clinics. psychosexual development and physical self-care). so we could evaluate its psychometric properties.1 Referred (n ϭ 65) 10. are shown in Table 1. A subsample of 46 was drawn for a 3-month test– retest. Measures The CAI protocol. The interview protocol was developed according to three assessment models: the AAI (in which narrative is analyzed). the SSP (focusing on behavior in current attachment relationships). the child being the subject of current family court proceedings. or the child having a previously established learning disability. without known mental health problems) were recruited from urban and rural schools.9) 99. whereas others.0 47.2 70. FONAGY. with both parents and child giving informed consent. Of these children.4 (1.8) 50. A further important difference from the AAI was the inclusion of a set of questions about the child’s perception of himself/herself as a person at the beginning of the interview. this means giving nonspecific. only those with psychosis and autistic spectrum disorders were excluded. distress.. illness.3 40. separate classifications were derived for the child’s relationship with each caregiver. Method Participants Two hundred twenty-seven children took part in at least one CAI. N ϭ 226) ϭ 1.942 SHMUELI-GOETZ. The final sample. divided into referred and not referred. As for the SSP.65. The coding and classification system was also partly modeled on the AAI (Main & Goldwyn. 161 of these children (ages 7–12 years. the main probes are indicated in the protocol (see Appendix). This is followed by a series of questions tapping the child’s self-representation. The exclusion criteria for the clinical sample were the same except that among children with mental health problems.g. significant sensory loss.

Alongside the linguistic analysis. scales were adapted from the AAI coding system. this dimension was found frequently to include not only anger but also denigration and contempt. Disorganization of attachment was coded categorically for presence versus absence of certain markers of disorganization informed by two major sources: behavioral manifestations. At the low end of the scale. Three State of Mind scales were designed to capture aspects of attachment narratives: Coherence. with the anchor points at odd values illustrated with examples. some children displaying Involving Anger were also derogating toward the relevant parent(s). In our first coding scheme. including those identified as characteristic of disorganized attachment in the . Idealization. together with the Disorganized category of infant attachment. they were also absorbed and entangled by their memories of episodes with the attachment figure. or Very Insecure. there were many nonverbal markers in the videotaped interviews that could not have been noted in AAI transcripts. For instance. marked anxiety. recalibrated. Three scales (Involving Anger. with the help of the interviewer. Dismissal. together with positive qualities of emotional openness. 1990. We therefore gave them the same names as used for the three main categories of adult attachment: Dismissing. We determined attachment classifications independently for each parent using an algorithm for combining the scale ratings. 1995). 1997. However. and Involving Anger. The Use of Examples scale was partially based on the AAI’s Insistence on Lack of Recall scale. children provided either no examples or very impoverished descriptions. Use of Examples. children were heavily biased toward either positive or negative. Balance of Positive and Negative References to Attachment Figures. all aiming to assess the child’s overall current state of mind with respect to attachment. together with a consideration of the overall consistency. balance of representations. it was vital to describe the ways in which “involving anger” was shown by children. Using an extensive and systematic qualitative analysis of a large number of pilot interviews not included in the present report. 2001). A global interview score was then assigned for each scale. and Balance.CHILD ATTACHMENT INTERVIEW 943 nonverbal and paralinguistic behavior. and Emotional Openness. Nine scales were coded. Dismissal. We classified Involving Anger and Conflict Resolution as “Active Conflict” scales. the child presented a picture containing both positive and negative descriptions. and Overall Coherence were based on AAI scales but adapted to reflect developmentally appropriate responses. the emerging categories showed considerable overlap with those prototypically observed in infancy and adulthood. Where relevant. Children’s ability to describe constructive resolutions to conflict has been closely linked to attachment security (Oppenheim. At the high end. A low score would be given to children showing marked idealization.. poor use of examples. The Emotional Openness scale was developed to assess children’s ability to express and label emotions and to ground them in descriptions of interactions with attachment figures. clear examples that vividly illustrated the adjectives or general description given in answer to a question. changes of posture in relation to the interviewer. the scales of Involving Anger. At the low end of the scale. we further assigned a level of security with regard to each parent: Very Secure. Again. Secure. As in AAI narratives. and Resolution of Conflict. The procedure differed from AAI and SSP coding in that it rested on coding of the videotaped narrative in terms of relationship episodes. Furthermore. At the high end of the scale. the child described situations in which he/she actively sought to resolve a conflict. and illustrated specifically for children’s videotaped narratives. this would result in a Cannot Classify category. Overall Coherence was rated similarly to the AAI’s Coherence scale: on the basis of scores on the “State of Mind” scales. A high score would indicate an absence of these distortions. Thus. Although we did not start clustering interviews with the intention of replicating the adult attachment classification prototypes. Secure. At the low end of this scale. Slough & Greenberg. These indicated the degree of realism and integration of the representations of relationships with each parent. and conflict resolution. The Balance of Positive and Negative References to Attachment Figures scale was based on the assumption that secure children would more readily recognize and integrate positive and negative aspects of parental figures. use of examples. changes in tone of voice. 1997) and was conceptualized in the CAI as Resolution of Conflict. criteria were developed to enable the coders to assign interviews to three-way and four-way attachment classifications. In the AAI. children showed a very limited range of emotional expression and made few references to emotional states even when encouraged to do so. All scales were coded from 1 to 9. At the high end. thus presenting more balanced descriptions. but the CAI scale also reflected children’s ability to provide relevant and elaborated examples. The main difference from the AAI classification concerned the Preoccupied category. To obtain a Secure classification. and contradictions between verbal and nonverbal expressions were considered during assessment. and Dismissal as “Avoidance” scales. and Dismissal) were rated separately for mother and father. Wright et al. so that the overall impression was of a balanced view of the relationships being described. We were influenced by Sroufe’s (1996) affectregulation model and studies that have identified the importance of emotional openness as an element in children’s attachment-related narratives and as a marker of security of attachment (Oppenheim. Use of Examples. as reflected in both his or her narrative and nonverbal behavior. children gave detailed. the child must have been assigned a rating of approximately 5 or above on all scales except Idealization. children described situations that seemed to have no resolution. and strong involving anger. which. Based on this qualitative analysis. a simple behavioral analysis of children’s responses to the interview situation and questions was included. where a score of 3 or less was expected. and Preoccupied. The remaining four scales were Emotional Openness. At the high end. Insecure. we specified algorithms for making this judgment. children used a range of appropriate emotional terms and showed different emotional states in relation to different subjects and situations as well as an understanding that different people have different feelings that may change over time. in contrast to the AAI. were elaborated into as full a narrative as the child could manage. Idealization. At the low end of this scale. we further developed the coding and classification scheme by categorizing interviews into clusters of predominant attachment themes (see Shmueli-Goetz. development. because the speaker would be regarded as showing strong features of both preoccupation and dismissal of attachment. these child narratives made it clear that while the children were denigrating of the attachment relationship. Maintenance of eye contact. and reflection. Idealization.

Good reliability and validity data have been reported by Schneider (1997). 1987).g. two randomly selected subsamples of families were contacted and asked to participate in the second phase of the study. and Sentence Assembly. The WISC–III UK is a well-established and validated measure of the intellectual abilities of children aged 6 –16. Seven of the HCAM dimensions code the quality of interpersonal relationship (with mother. only subtests of the CELF–R (UK version) developed to assess expressive language in children 8 years and older were employed. Slough & Greenberg. a shortened WISC-III UK was used. Prorated scores were derived for Verbal. F5 ϭ some preoccupation with attachment figures) and four types of insecurity (DS1 ϭ dismissing of attachment. One interviewer administered the AAI and HCAM interviews to the main caregiver. Ratings of the SAT were undertaken while judges were blind to CAI classifications. these three dimensions were eliminated. AND DATTA SSP (Main & Solomon. SATs were rated by Adrian Datta. Wiig. 1992). Procedure Administration. which were labeled as “mild” or “severe” on the basis of existing scoring systems (Shouldice & Stevenson-Hinde. FONAGY. Coding and interrater reliability of the CAI. ␬ ϭ . 1995) is a semiprojective test designed to assess children’s narrative responses to representations of separations from parents. TARGET. 1992.81. exploration/play. which gave rise to an overall classification (Secure or Insecure) and subclassifications based on five types of security (F1 ϭ some setting aside of attachment. 1995). (c) Picture Arrangement. peers.70 for 15 reliability transcripts). For the purposes of the current study. psychosexual development. 1995). 1983) and thus cover the full range of possible childhood adaptation. Adult Attachment Interview (George et al. The anchor points correspond as closely as possible to those of the Child Global Assessment Scale (Shaffer et al. Performance and Full-Scale IQ scores from the subtests. stepfather. Wright et al. adolescence. emotion and its regulation. Wright et al. Coders of the HCAM interviews were blind to CAI codes assigned by a separate group (five master’s-level graduates in psychology and Mary Target).. and adulthood. Two interviewers with experience in the administration of the CAI conducted the battery of assessment measures. For this study. DS2 ϭ devaluing of attachment. although these are also considered when coders assign a classification. and had trained the other two coders to good reliability (82% agreement. E2 ϭ angry/conflicted). F3 ϭ secure with free valuing of attachment. and confidence and self-esteem. & Secord. The CAI was conducted in a private room. including the CAI. had achieved satisfactory reliability (86% agreement. and adults). each of whom A copy of the CAI protocol and details of training in the Coding and Classification Manual can be obtained from the corresponding author. Recalling Sentences. and disruptions of narrative used in classifying certain AAI transcripts as Unresolved. 2000). ␬ ϭ .. Slough & Greenberg. appropriate to each of four age bands (Target et al. along with other markers such as inappropriately familiar behavior toward the interviewer. 1990). including the following four subtests: (a) Similarities. but with some additional markers appropriate for the behavior of older children. 1976. Adrian Datta had received formal training in coding the SAT from Dr. The AAI was developed as a way of predicting infants’ Strange Situation attachment classifications.. the duration of the CAI ranged from 20 min to 1 hr and 20 min. and a clinical psychologist colleague. The CELF–R is an established and widely used standardized language measure designed to assess receptive and expressive language skills in children aged 5–16. the sessions were videotaped. Clinical Evaluation of Language Fundamentals—Revised (CELF–R. Wechsler. The transcripts were coded using Resnick’s (1993) revised rating scales. while the second interviewer carried out the child assessments. It requires individuals to recount and reflect on early attachment-related experiences while maintaining coherent and collaborative discourse (Main.67 for 15 reliability transcripts). with a score of 70 being the boundary between normal and clinically impaired adaptation. Performance. with separate sets of anchor definitions for each dimension. effortful control. (1995). During the CAI. In order to reduce the chances of Type I error. the interviewer initially explained the study and ensured that the child felt at ease and was willing to take part.944 SHMUELI-GOETZ. Yael Shmueli-Goetz. 2000). and Nonattachment-Related Processes. we used nine SAT photographs. stepmother. Four dimensions would not be predicted to be linked to attachment: physical self-care. Resnick. 1 . clear contradictions between verbal and nonverbal behavior (e. There were three independent coders of CAI interview videotapes. An expressive language standard score was then derived by adding the standard scores for each subtest and converting the sum. siblings. Semel. Attachment-Related Processes. The classification system yields categories that are parallel to the Strange Situation patterns. father. Target et al.. developmentally sensitive 100-point scale. Cronbach’s ␣ ϭ . Cronbach’s ␣ ϭ . resilience in the face of adversity. 1985). E1 ϭ passive.1 Wechsler Intelligence Test for School Children III (WIS–III UK. F2 ϭ secure but restricted. 1990).. Raw scores for each subtest were converted into norm-referenced standard scores. and FullScale IQ. Three subtests were used: Formulated Sentences. Since in this sample some children did not have siblings and many had no stepparents. we combined the three sets of dimensions into three scales: Quality of Relationship. F4 ϭ some preoccupation with attachment.. giggling about the death of an attachment figure) were considered. Following an interval of 3 or 12 months. In addition. Cronbach’s ␣ ϭ . Hampstead Child Adaptation Measure (HCAM. and (d) Block Design. The caregiver interviews for the HCAM are coded on a global scale and 16 dimensions.79. Following Wright et al.. A number of dimensions concerned psychological processes believed by some to be related to the quality of attachment relationships: impulse control. The child’s responses to the SAT were audiotaped and transcribed verbatim. and conformity/compliance. Each of these is a carefully anchored. The SAT (Klagsbrun & Bowlby. rather than his/her actual or probable experiences. as suggesting disorganization of the attachment system. A second CAI was then administered by the same interviewer for the evaluation of test–retest reliability. The AAI is a semistructured interview comprising 25 questions covering early childhood. The AAI attempts to capture the participant’s current state of mind with respect to attachment. Separation Anxiety Test (SAT.88. The WISC-III UK comprises 13 subtests from which three composite scores can be derived: Verbal. (b) Vocabulary. 1990.

which are standard measures of agreement between independent coders on a categorical judgment. correlations between CAI scales.77 .25–.68 .89–. Results This section first reports on the nonreferred sample.78 . only one scale (Idealization of Father) yielded unacceptable ICCs.85 . with the confidence interval including a negative correlation.82 95% CI .82 .83–.79 .. Secure. were familiar with current attachment assessment methodologies. Yael Shmueli-Goetz coded the total sample.66–.70–.53–. Insecure) and levels of security (i.47–. each judge independently coded an additional 30 interviews.66 .CHILD ATTACHMENT INTERVIEW 945 had postgraduate degrees.46–.79 .94 . test–retest reliability.74–. The second and third coders then received training from the first coder.15–. Interclass correlations (ICCs) for the three coders were computed and are shown in Table 2.94 Naı ¨ve coders & expert codersc r .85 Ϫ.94 .81–. the three judges jointly coded 30 cases.86e .87 with no unacceptably low agreements. c N ϭ 68 cases.88 .92 .65–.69 . The median ICC for all scales was .91d .91 . Table 2 presents the ICCs and Pearson correlations for the three phases of the reliability test.84 .94 . .89e Note.90 .71 . and naı ¨ve coders are shown in Table 3 and were found overall to be consistently high.70 .39–.73–.83 .86 .94 .80 .90–. Following training. and discriminant validity.70–.88 .e. randomly drawn from 226.88 .38 . and were involved in developing the coding system described earlier.85 . In Phase 1 (training phase).89–. we repeated these tests on the referred sample and made statistical comparisons between the two samples. a N ϭ 30 cases. the second and third judges each coded one half of the sample.97 .47 . In order to establish the psychometric properties of the instrument for a clinical population.92 .83–. which indicated very strong agreement among the three coders. Interrater reliabilities were computed in three phases.94 .77–.78–. also showed a high correlation between two raters. In the first attempt to establish interrater reliability.68 .89 .78–.85–. which led to further refinement of the CAI Coding and Classification System (subsequently renamed Version II).91 . the number of disorganized classifications was too small to estimate agreement. three.87 .63–.52–.76–. the median r being .89 .89 .83 .81–. The second assessment of interrater reliability.90 95% CI .81–. For the three coders and naı ¨ve coders. and their agreement with the experienced coders was computed. We assessed interrater agreement for the main classifications (i.74 .53–. e N ϭ 57 cases.88 .72d .95 .91 .84 . The relations of classifications given by the two.89 .91 .64d .81–.. b N ϭ 50 cases. CI ϭ confidence interval.66–. across 50 cases. In Phase 2.95 r .95 .91 .79 . Very Secure.91 .79 . The low ICC observed in relation to the Idealization of Father scale may reflect the absence of information about fathers that typified many of the CAI narratives (Idealization is difficult to assess if there is little information against which to measure the child’s descriptions).58 .85–.81 with the exception of two scales: Involving Anger and Idealization of Father. another 50 randomly selected interviews were coded by two other master’s-level coders with limited knowledge of attachment theory. the internal consistency of the CAI. Coding was carried out blind in regard to whether children came from the referred or nonreferred sample. respectively.e. The CAI reliability training entails 3 days of coding relatively difficult cases and discussion when discrepancies and ambiguities arise.97 .75 .86 .65–.94 .81–.66–. d N ϭ 62 cases.94 .94 .88.90 .90 .81 .79 . Secure vs.95 .90 . Insecure and Very Insecure) using the kappa statistic and Spearman rho.87 .82–.82 . who had not been involved in the development of the measure but who had received training as described earlier. median r being .79–.90 .93 . ICC ϭ interclass correlation. A subsample of the referred and nonreferred Table 2 Interrater Reliability of Scores on Child Assessment Interview Scales Pearson r ICCs for three expert codersa Scale Emotional Openness Balance Use of Examples Anger With Mother Anger With Father Idealization of Mother Idealization of Father Dismissal of Mother Dismissal of Father Conflict Resolution Coherence Level of Security with Mother Level of Security with Father Combined Security rating ICC .81–. As a further test of the robustness of the system.97 . in Phase 3 undergraduate (“naı ¨ve”) students with no previous knowledge of attachment theory were given the 3 days’ training and subsequently asked to code an additional 68 interviews.89 Two expert coders 95% CI .72–.72–. who was familiar with many more child cases and had been certified as reliable in coding the AAI. Assessment of interrater reliability with naı ¨ve coders also produced high agreement.55–.96 .95 .80 .88 . including the distribution of attachment classifications.58–.89 .78–.

92 .66 .81 .89 to estimate Combined Median . a b Father Median Range .84).93 .84 n too small to estimate Combined .83 .89 . a high proportion of children were classified as Secure with respect to both mother and father (66% and 64%.81–.78 .80 .69 .83 Range . The discrepancies were mainly accounted for by the eight interviews of children who were coded as Secure with mother but not father (5%). Children classified as Disorganized were not included in later calculations because of the small number of cases in this sample. TARGET.78Ϫ. As shown in Table 5. % Father No.92 .81 .87 . there was a predominance of the Dismissing classification within the Insecure group (30% for both mother and father).88 . % a Insecure Overall Separatea Dismissing Overall Separatea Preoccupied Overall Separatea Disorganized Overall Separatea Nonreferred sample 107 66 98 64 98 61 106 66 97 63 97 60 54 33 56 36 47 29 49 30 51 33 42 26 48 30 46 30 52 32 45 28 43 28 49 30 7 4 10 6 11 7 5 3 8 5 9 6 6 4 6 4 6 4 Referred sample 20 30 14 23 17 26 20 30 14 23 17 26 46 70 46 77 49 74 40 61 41 68 43 65 37 56 37 62 39 59 33 50 33 55 35 53 9 14 9 15 10 15 7 11 8 13 8 12 6 9 5 8 6 9 Separate ϭ values when Disorganized interviews were classified separately. observed means were fairly high on the scales that are considered to reflect a Secure strategy (ranging from 5.77Ϫ. and HCAM.946 SHMUELI-GOETZ.80 .83 .58 . % Mother No. The frequency of Preoccupied attachment was low at 4% for mother.25 to 2.84 . Scales expected to be characteristic of particular attachment classes are marked after the name of each scale in square brackets.86 .85Ϫ.92 . 6% for father. Thus.90 .79 .78–. Only 4 children were coded as Secure with father but not mother (3%).90).74Ϫ. % Combined No.80 Two coders kappab Mother . respectively.78 . using the SAT.84 – .60 .86 n too small samples was then assessed to examine construct validity. All children coded as Disorganized with one parent were also Disorganized with the other.80 Range .86 .58 . was broadly in line with distributions reported in other studies.67 .87 . Nonreferred Sample The distribution of attachment classifications. 61% when combined).76 .78 Father . and 7% combined.84 . AND DATTA Table 3 Interrater Reliability of Classifications and Levels of Security Three coders kappaa Mother Classification Secure/Insecure Three-way Four-way Disorganization N ϭ 30 cases.78 Mother . FONAGY.92 . ␬ ϭ .85 Naı ¨ve coders kappac Father .84).87 Median . N ϭ 50 cases. AAI.84Ϫ. .58 . shown in Table 4.78 Combined .85 .77–. Separate analyses of variance were carried out on each of the Table 4 Distribution of Attachment Classifications for the Nonreferred and Referred Samplesa Secure Attachment Mother No. % Combined No.17 to 5. with low means shown for those scales considered as possible indices of Insecure attachment (ranging from 1. The concordance for three-way classifications of interviews based on attachment to mother and father was very high (92%. c N ϭ 68 cases.87 .66 . % Father No.

72 41. and Dismissal of Father (but not Idealization of Father) were higher in the Dismissing group than in the Preoccupied group.04 1.65 1.38 0.12a 1.5 years.77a 2.32a 4.36 1.09a 3.41 1. There is a strikingly low correlation between the Involving Anger scores for mother and father.7. which showed significant differences in mean ratings for interviews falling within the three classifications.27b 2. 31 children (13 girls.50 1.00b 2.27b 3.71 49. with all retest interviews coded by one of the two other experienced coders.54b 3.06 1. Conflict Resolution.55 1. From the larger sample.b 4.14 1.69 145. P ϭ Preoccupied).c 3. with those below the diagonal representing the nonreferred sample and those above it the referred sample. We assessed the internal consistency of the three groups of CAI scales using Cronbach’s alpha. interviews classified as Insecure rated significantly lower than those classified as Secure.16 M 5.17 1.08 0.72b 6.11 17.24 2. Involving Anger With Mother and Father were specific to Preoccupied children.82).0 years. range 76 –128).84 22.60 1. there was low consistency (␣ ϭ .77 0. with children whose interviews were coded as Dismissing being least secure and as Preoccupied being significantly more insecure than those in the Secure group.5. and their mean full IQ was 102. and. range 78 –154).69 1. Emotional Openness.07 0. representing 70% of those asked to participate. Most stability coefficients for CAI scales were quite high over 3 months: the median was . range 77–146). Scales that were expected to differentiate Dismissing children by and large separated them from those with Preoccupied or Secure classifications: Dismissal of Mother. Table 6 presents the correlations between CAI scales.98 0.81 1. it only emerged in relation to one parent.12 0. Their mean verbal IQ was 108. It seems that most children did not show this form of involving anger with either of their parents.94 1. The set of Avoidance scales also showed high internal consistency (␣ ϭ .69 (range from .15a SD 1.4 (SD ϭ 17.91a 4.28 1.72a 4.14 36. Means with same subscript are not significantly different to Tukey’s test contrasting three groups.86b 2.24a 4.64a 1. with a range of 6.78 108.90).18 1. 24 boys).15a 1.18 57.43 0.24a 1.45b.10 0.99 1. Of these 67 children.70 51. with those classified as Preoccupied rated highest on Involving Anger scales. and these are displayed in Table 5.25 1. Involving Anger With Father . On the State of Mind scales. and overall ratings of level of security.51 1.25 120. Our expectations concerning characteristics of narratives by children from the three attachment groups were broadly confirmed by these analyses. range. The mean age was 9.36b 2. For Active Conflict scales.09 0.32).12a.05 1. 18 boys) agreed to take part in the 1-year follow-up phase.91 9. There was considerable variability in the stability of the scales.32a 4.40b 4.26c 3. The highest correlation with Combined Security was with Coherence (as is found in the AAI).02b 2.49 0. Tukey’s post hoc tests were performed to identify significant group differences.91b Preoccupied (N ϭ 11) SD 1. which was not surprising.50 0.58 1.8. 73–155). the interviews classified as Dismissing are also rated highest on Avoidance scales. First test interviews were all coded by Yael Shmueli-Goetz.90 34.51 76. Most correlations were expected.46 0. Low Use of Examples was also characteristic of the Dismissing children.88c 2. although Idealization and Dismissal appear to intercorrelate highly.29 to .0 70.38a 4.87.42 2. Their mean verbal IQ was 106.76 ANOVA (df ϭ 2.30 0. 158) 45. ANOVA ϭ analysis of variance.0 –12.03 1.76a 4.19 13.09b 6.8 years.91b 5.0 (SD ϭ 13.85 Secure (N ϭ 98) M 6.5 (SD ϭ 18. and their mean expressive language score was 101.22a 4.6 years.08c 2. D ϭ Dismissing.54a 5.80b 2. and those who did tended to be angry with one parent while viewing the other positively. Pearson product moment correlations between the two test periods were computed for each of the CAI scales along with stability for main attachment classifications as assessed by the kappa statistic.85 2.47b 5.4 (SD ϭ 18. Idealization of Mother.25b 2.82b 1.28c SD 1.7–11. The mean scores on these scales are displayed in Table 5. For example. The mean age was 9. while Involving Anger With Mother appeared to be highly stable across 3 months. To evaluate the test–retest reliability of the CAI.51 31. Coherence. as expected.67c 6.26 2. Secure children were rated significantly higher on the following scales: Balance (between positive and negative). with a range of 7.08 0. The Combined Insecurity rating distinguished all three groups.36b 2. because on the rare occasions that Involving Anger was manifest. Attachment category predicted to be linked to each scale is given in square brackets (S ϭ Secure.45b 2. The set of State of Mind scales yielded a Cronbach’s alpha of . these are reported in Tables 7 and 8.26a 3.CHILD ATTACHMENT INTERVIEW 947 Table 5 Child Attachment Scales and Levels of Security For Nonreferred Sample Dismissing (N ϭ 52) Scale/security level Emotional Openness (S) Balance (S) Use of Examples (D) Involving Anger With Mother (P) Involving Anger With Father (P) Idealization of Mother (D) Idealization of Father (D) Dismissal of Mother (D) Dismissal of Father (D) Conflict Resolution (S) Coherence (S) Level of Security: Mother (S) Level of Security: Father (S) Combined Insecurity rating (S) State of Mind (S) Active Conflict (P) Avoidance (D) M 4.07b 6.39a 4.28a 1.06a 3.58 0. This represented 50% of those invited. scales.87 0. 46 agreed to take part (22 girls.41a 4.41a 3.31 Note.37a 2.92a 1.07 1.76 1.81 1.46 0.9.60 1.09 1. we contacted 67 nonreferred children after a 3-month interval on the basis of geographical proximity to a research center.56b 1.53 0.

69 and .80 .16 Ϫ. 2.49‫ءء‬ Ϫ.78 .60‫ءء‬ . 11. Of the parent-specific scales.78 .46 .38‫ء‬ Ϫ.83–. IA-M/F ϭ Involving Anger With Respect to Mother/Father.40‫ء‬ .33‫ءء‬ Ϫ.73‫ءء‬ . 10.35 .44 .27–.53‫ءء‬ 8 Ϫ.69 .37‫ء‬ .67‫ءء‬ .59‫ءء‬ Ϫ.90 .68‫ءء‬ .66 .55‫ءء‬ .56‫ءء‬ Ϫ.87 .78 over 3 months.68 . by contrast.63 .00–.33‫ءء‬ Ϫ.07 Ϫ.41‫ءء‬ Ϫ.25 — .77 .51–. However.35‫ء‬ Ϫ.01 Ϫ.75‫ءء‬ 11 .81‫ءء‬ Ϫ.32 — .04 Ϫ. and Use of Examples were relatively stable.00 Ϫ. RES ϭ Resolution of Conflicts.83 .22‫ء‬ Ϫ.82 .09 Ϫ.35–.58 .83 .90 .75 Ϫ. ID-M/F ϭ Idealization With Respect to Mother/Father. and 69%.31–.72 .93‫ءء‬ 12 Ϫ.03 .70‫ءء‬ Ϫ.60‫ءء‬ .22‫ء‬ 5 Ϫ. as it was rare for a child to score above 1 (the lowest rating) on this scale with respect to either parent.66‫ءء‬ .23‫ء‬ Ϫ. particularly Idealization and Involving Anger With Father.60‫ءء‬ . UoE ϭ Use of Examples.04 .80.06 Ϫ.02 Ϫ.01 Ϫ. In addition.59‫ءء‬ Ϫ.55–.84‫ءء‬ Ϫ.81 . EO BAL UoE IA-M IA-F ID-M ID-F DS-M DS-F RES COH MSEC FSEC CSEC 1 — .42 .09 Ϫ.80 . The Level of Security ratings were relatively stable. 6.84‫ءء‬ — Note.49–.57‫ءء‬ . Idealization of both parents was somewhat unstable.82‫ءء‬ .02 . 12.51‫ءء‬ — .24–.80‫ءء‬ . somewhat more so for mother than for father.75 .32‫ءء‬ .55‫ءء‬ — .76‫ءء‬ Ϫ. it was 83%.75‫ءء‬ Ϫ. and for the four-way.00–.53‫ءء‬ Ϫ. was far less so.86‫ءء‬ .52‫ءء‬ Ϫ.52‫ءء‬ Ϫ.79‫ءء‬ 9 Ϫ.75‫ءء‬ Ϫ. but.25 .70 .81‫ءء‬ .73‫ءء‬ .21‫ء‬ .89‫ءء‬ .62 .73‫ءء‬ .75).49‫ءء‬ Ϫ.58‫ءء‬ .06 6 Ϫ.69‫ءء‬ Ϫ. The aggregate scales of State of Mind.35–. Emotional Openness.11–. The kappas compared favorably with the interrater reliability of .71–.08 .25 .69‫ءء‬ Ϫ.92‫ءء‬ .67‫ءء‬ 3 . which were both poor.63 .42‫ءء‬ .50–.16 Ϫ.42‫ء‬ Ϫ. and Coherence seemed highly consistent at the two testings. the figures were 82%.18 Ϫ.71–.67 .79 .75‫ءء‬ Ϫ.29 . ‫ ءء‬p Ͻ .48–.68‫ءء‬ . TARGET.77‫ءء‬ .24 .62 .75‫ءء‬ . MSEC ϭ Mother Level of Security. 5. For classifications of the relationship to father. 9.37‫ءء‬ Ϫ.24–.57 and . between .65‫ءء‬ Ϫ.11 .44‫ءء‬ Ϫ.94 .12 . Table 7 Test–Retest Reliability of Child Attachment Interview Scales Over 3 Months and 1 Year Test–retest at 3 months (n ϭ 46) Scale Emotional Openness Balance Use of Examples Involving Anger With Mother Involving Anger With Father Idealization of Mother Idealization of Father Dismissal of Mother Dismissal of Father Conflict Resolution Coherence Level of Security: Mother Level of Security: Father Combined Insecurity rating State of Mind Active Conflict Avoidance Reliability .08 .56‫ءء‬ Ϫ. it was 89%. Of the aggregate scales.42–. the percentage of agreement was 85%.56‫ءء‬ . the Use of Examples.948 SHMUELI-GOETZ. 7. for the three-way.87‫ءء‬ .60‫ءء‬ .87‫ءء‬ 2 .42‫ء‬ Ϫ.60‫ءء‬ .12 .54 .74 .46–.81‫ءء‬ Ϫ.34 .10 .64‫ءء‬ Ϫ.05 .80 .16 Ϫ. Dismissal and Level of Security were highly stable for both mother and father.63 . 14.67‫ءء‬ .68 .61‫ءء‬ Ϫ.75‫ءء‬ Ϫ.61‫ءء‬ Ϫ.70‫ءء‬ Ϫ.001.91‫ءء‬ 13 Ϫ.55‫ءء‬ Ϫ.43‫ءء‬ Ϫ.59‫ءء‬ .08 to .13–.33‫ءء‬ .68 .61‫ءء‬ Ϫ.70‫ءء‬ .52‫ءء‬ Ϫ.20 Ϫ.37–.69‫ءء‬ Ϫ.15–.54 95% confidence interval . classifications over 3 months were stable. COH ϭ Overall Coherence.46‫ءء‬ Ϫ.65‫ءء‬ — .08 Ϫ.91‫ءء‬ — . As shown in Table 8.06 .21‫ء‬ Ϫ.69‫ءء‬ .05 .66–.46‫ءء‬ Ϫ.60‫ءء‬ Ϫ.49‫ءء‬ .03 .56 Ϫ.85 .01 .29–.63 .10 .54 .73‫ءء‬ 10 .11–.29‫ءء‬ .25 .05 Ϫ. DS-M/F ϭ Dismissal With Respect to Mother/Father.05 .36‫ء‬ Ϫ.85‫ءء‬ .11 . with a median correlation of . Table 7 suggests that the stability of the scale scores across a 1-year interval was more moderate.82 .64‫ءء‬ .73 95% confidence interval .76 .48‫ءء‬ Ϫ.57 .63–.64‫ءء‬ .65‫ءء‬ Ϫ.56 Ϫ.42‫ءء‬ Ϫ.57‫ءء‬ .64‫ءء‬ — .54‫ءء‬ Ϫ.00–.94‫ءء‬ — Ϫ.77‫ءء‬ Ϫ. For the two- way Secure–Insecure split with respect to mother.82 .00–.25‫ء‬ Ϫ.52 . Emotional Openness. with correlations between .53‫ءء‬ .57‫ءء‬ — Ϫ.00–.01.24 .83 .83‫ءء‬ — .89 . 8. whereas the parent-specific scales had low stability.17 Ϫ.19 Ϫ.43‫ءء‬ Ϫ. FSEC ϭ Father Level of Security.11 Ϫ.70 .57 .35 .90 .55 .53–. 13.09 Ϫ.03 Ϫ.39 .92‫ءء‬ . 3.43‫ءء‬ Ϫ.42–. CSEC ϭ Combined Security rating. it is difficult to attribute too much significance to low stability.63‫ءء‬ Ϫ.64‫ءء‬ Ϫ.54 (ranging from .84‫ءء‬ Ϫ.29‫ءء‬ Ϫ.44‫ءء‬ Ϫ.71 .87 .62‫ءء‬ Ϫ.09 . AND DATTA Table 6 Correlation Matrix For Child Attachment Interview Scales For Nonreferred Sample Below the Diagonal (N ϭ 161) and Referred Sample Above the Diagonal (N ϭ 66) Variable 1.62 .17 .75 . respectively. and Avoidance were all reasonably stable.18 — .82‫ءء‬ . EO ϭ Emotional Openness.28–. Again.06 .90‫ءء‬ — . 4. Bal ϭ Balance of Positive/Negative References to Attachment Figures.56‫ءء‬ Ϫ.65‫ءء‬ . there was considerable spread: Coherence.86‫ءء‬ 4 . 75%.86‫ءء‬ . FONAGY. Active Conflict.63‫ءء‬ Ϫ.71 .60‫ءء‬ 7 Ϫ.26 . ‫ء‬ p Ͻ .35‫ءء‬ Ϫ.45‫ءء‬ . State of Mind had the highest stability and Active Conflict the lowest.24–.56–.84 Test–retest at 1 year (n ϭ 33) Reliability .89‫ءء‬ 14 Ϫ.67.81‫ءء‬ Ϫ.07 .

01.70 (18. the referred sample scored significantly lower on the scales associated with attachment security: Emotional Openness.42 Referred Sample As shown in Table 4. In addition. N ϭ 214) ϭ 28. The table also lists the effect size of each difference.70) 11 (30. for the three-way.67 . and for the four-way. The distribution for both mother and father and for the combined coding for the Secure– Insecure split was found to differ significantly from the distribution of attachment patterns observed for the nonreferred sample: for mother.00) 23 (21. ns Secure 11. N ϭ 157) ϭ 1. 76%. and for the four-way. ␹2(1. Avoidance. and Coherence. ␹2(2. with a predominance of the Dismissing strategy (56% and 62%. and medium effect sizes were seen for the aggregated scales for State of Mind.20) 76 (71. and one. of Boys (%) No.20.001.41) 29 (52. ns ␹2(2.64 . and Active Conflict.5.00 1 year (n ϭ 33) Mother .4. This result was replicated when we considered three-way attachment classifications: for mother.38) 99.33. Verbal IQ.CHILD ATTACHMENT INTERVIEW 949 Table 8 Test–Retest Reliability (Cohen’s kappa).52 .20) 5 (9. As shown in Table 8.64 . The referred sample scored higher on scales expected to be associated with insecurity of attachment—strikingly so for Dismissal and somewhat so for Idealization and Involving Anger. ns F(1.62) 99.74 .60) 70 (70.00 Father . The findings suggest however that the assignment of overall attachment classifications was not related to these variables.85 (15. p Ͻ .83) 98.36) 49 (49.36. The most substantial effect sizes were associated with Coherence and the overall rating of Insecurity.37) 99.02.76 (1.40) 37 (68. N ϭ 227) ϭ 25. the distribution of attachment patterns for the referred sample revealed a high prevalence of Insecure classifications with respect to both mother and father (77% for each). p Ͻ . ns Statistic F(1.76.50) 8 (7.50) 98.60) 39 (70.50) 48 (48. N ϭ 227) ϭ 23. it was 85%. 152) ϭ 2.53 (15.60) 101. N ϭ227) ϭ 22.4. ns Statistic F(1. ns ␹2(1.001. N ϭ 227) ϭ 23. A series of one-way analyses of variance (ANOVAs) and chi-square tests were computed. 110) ϭ 0. N ϭ 161) Ͻ 1. 56) ϭ 0. N ϭ 157) Ͻ 1. Consistently. (%) White Black Asian No.70) 22 (22. for father. socioeconomic status (SES).16 (1. While observed correlations were in the expected directions. The correlations between CAI scales for the referred sample are presented in Table 6.71 1. and for both parents combined.001.20) Attachment—Father Insecure 10. ns ␹2(1. and as shown in Table 9. ␹2(2.30) 26 (45. none of the variables—including age.67 Father . gender. the scales of Involving Anger with respect to both parents correlated very weakly with the remaining CAI scales.77) 30 (55. Just over a quarter of the sample were rated as Secure with both parents. ns ␹2(1.00) 5 (9. There was a nonsignificant tendency for those rated as Insecure With Mother to be slightly younger.60) 11 (31. N ϭ 161) Ͻ 1. expressive language.48. 56) ϭ 0. N ϭ 153) ϭ 2. 69%. For the two-way Secure–Insecure split with respect to mother. it was 76%.28 (19. respectively).13 (1.0.50) 27 (46. attachment classifications were relatively stable. at 3 Months and 1 Year of Attachment Classifications With Mother and Father 3 months (n ϭ 46) Classification Secure/Insecure Three-way Four-way Disorganization Mother . N ϭ 154) Ͻ 1. p Ͻ . it was 79%. In examining the discriminant validity of the CAI.08 F(1. and just under 10% were coded as Disorganized with at least one parent. one-way ANOVAS did not reveal statistically significant differences between secure and insecure children (see Table 9).50 (18. ␹2(1.73 (1. although somewhat below the coefficients obtained with an interval of 3 months between test periods. Use of Examples.2. p Ͻ . we explored the relation between demographic variables and attachment classifications with respect to both parents. these test–retest reliability figures are encouraging and generally suggest that children’s security classifications can expect to be reasonably stable in a low-risk sample. ns ␹2(1. calculated as the difference in the means divided by the pooled standard deviation.001. of middle class (%) Race: No. Table 9 Relations Between Attachment Classification With Mother and Father and Demographic Variables.30) Insecure 10. ethnicity.10) ␹2(1. Balance.10) 32 (47. and attachment classifications. for the three-way. ␹2(1. living with two parents (%) Secure 11. 159) ϭ 3.52 . and for both parents combined.70) 28 (45.90) 11 (20. N ϭ 214) ϭ 27. 104) ϭ 0.56 .001. ns ␹2(2.97) 101. N ϭ 151) Ͻ 1. N ϭ 153) Ͻ 1. Overall. for father. ns F(1.12 (18.69 .90 (12.66 (12. p Ͻ .60) ␹2(1.3. ns F(1.001.81 .20) 7 (7.74. the scales of Dismissal and Levels of Security with respect to both parents emerged as highly correlated.67 1.61) 98.53 . ns . p Ͻ . p ϭ . In establishing the relation of verbal IQ.50) 12 (22. ␹2(2. The stability of the classification of the interviews in relation to father was noticeably below that in relation to mother: For the two-way Secure–Insecure split. Table 10 contrasts the mean CAI scale scores for the nonreferred and the referred samples.29) 51 (47.or two-parent household—approached statistical significance. the percentage of agreement was 85%.10) 46 (53. and Expressive Language in Nonreferred Sample Attachment—Mother Variable Mean age (SD) Mean verbal IQ (SD) Mean expressive language (SD) No.

83 0.21 4.01.92‫ءءء‬ 18.98 2.06 1.11 ANCOVA (df ϭ 1.12 3. As predicted.78 M 4.03‫ءءء‬ 6. 2 . The kappa coefficient computed for this contingency table was 0.35 1.49 0.70 1. 83) ϭ 3.11 4.84 5. ␹2(9.00.00‫ءءء‬ 30.05 2. only a three-way analysis was possible.49 0. as is customary (van IJzendoorn & Bakermans-Kranenburg.81 0.26‫ءءء‬ 28.77‫ء‬ 5. or Preoccupied characterization.11‫ءءء‬ 16.20‫ءء‬ 10.03 1. F(2. and F subclassifications.1.51 1. 2003).9.27 2. As expected.59 4. approximate t ϭ 3.14 1.00‫ءءء‬ 17. N ϭ 88) ϭ 23.66 2. p Ͻ .58 5.92 Nonreferred (n ϭ 161) M 5. F(2.83. Eighty-eight mothers were administered the AAI.65 1.60 0.27‫ءءء‬ 38.40 0. respectively.93.52 1.03.91‫ء‬ 1. AND DATTA Table 10 Means and Standard Deviations for Child Attachment Interview Scales and Levels of Security in Nonreferred and Referred samples Total (N ϭ 227) Scale/security level Emotional Openness Balance Use of Examples Involving Anger With Mother Involving Anger With Father Idealization of Mother Idealization of Father Dismissal of Mother Dismissal of Father Conflict Resolution Coherence Level of Insecurity: Mother Level of Insecurity: Father Combined Insecurity rating State of Mind Avoidance Active Conflict M 5.43 for Active Conflict. Coefficient kappa was calculated as an estimate of agreement (␬ ϭ .59 SD 1.19 1. and .03. p Ͻ . children did not significantly differ on verbal IQ and expressive language as a function of their attachment security.03 2. As regards the relation of verbal IQ.82 1.55 0.25 0. with the Secure children having the highest scores.94‫ءءء‬ 6.002. and Attachment-Related Processes. While the majority of the 222 participants fitted into one of three classes with a profile of scores that was consistent with a Secure.65‫ءءء‬ 16.87 2. Figure 1 displays the means and standard errors of the three groups on the global score and the three scales of the HCAM.34 2.25 1.78 1. p Ͻ .33 2.63 2. ethnicity. 225) 14. TARGET.or two-parent household—approached statistical significance.80 0.36 1. This reflects a 64% agreement of these independently coded measures of attachment.68 1.04 1. one-way analysis of variance yielded significant F ratios for two of the three scales: Quality of Relationships. p Ͻ . The association between CAI and AAI on the four-way categorization was highly significant.66 1. so the association between CAI and SAT involved a three-way categorization of DS.84 2. It is noteworthy that 83% of the 12 children classified as Disorganized in this sub- sample had mothers whose interviews were coded Unresolved or CC. the AAI CC code was combined with the Unresolved code.56 1. Effect size is calculated as the difference between means divided by the pooled standard deviation. expressive language.36 0.56 2.38 5.90.36 0. Because of the restricted sample.35 0. E.005).98 Referred (n ϭ 66) SD 1. ‫ء‬ p Ͻ . ‫ ءء‬p Ͻ . Discussion The aim of this study has been to develop a new assessment of child–parent attachment in middle childhood to address the measurement gap in the study of attachment relationships in this age group.04.67 3.80 1.44 1. SES.2 As for further exploring the discriminant validity of the CAI (Table 11).14 5.10 1.36 0. There were 50 Dismissing.66 1. and attachment classifications for the referred sample. and 26 Secure caregivers.76 2.05.48 0. Active Conflict yielded lower consistency (Cronbach’s ␣ of .18 2.30 10.48 4.99 3.76‫ءءء‬ 11.63 3.75 0.43 SD 1.73 1.16. with the exception of gender. t Ϸ 3.73‫ءءء‬ 48.88 0. internal consistency for State of Mind and Avoidance was .84 4.15 5. and these were harder to interpret. p Ͻ .36 1.72.49).55 0.57 1. ns.98‫ءءء‬ Effect size (d)a 0.39 2.43 0. 1996). The SAT does not have a Disorganized category.48 1.42 3.02 0.87 and .81 1.14 1. Nonattachment-Related Processes was not significantly related to security: F(2.83 0.83 1.87 1.70 1.63 1.50 3.88 1. F(2.33 5. ‫ ءءء‬p Ͻ . we conceived and adapted the We attempted to confirm that these hypothesized categories of attachment could be identified as latent classes by trying to fit a latent class model to the scale scores (Vermunt & Magidson. For the sample as a whole. 83) ϭ 1.75 5.55 1.54 5.24 2. Similar to the findings reported for the nonreferred sample. three additional classes (n Ͻ 10) were all required for a nonsignificant fit. FONAGY. which was coded independently of the CAI coding. and one. there was a difference among the three groups on the Global scores.87 1.97 5.950 SHMUELI-GOETZ. 83) ϭ 3.79 0. State of Mind and Avoidance yielded high level of consistency with a Cronbach’s alpha of .11 4.45 2.62 0.91 1. Dismissing.76 2.62 Note.81 1.70 1. p Ͻ .80. 10 Preoccupied. SAT protocols were obtained from 67 (40%) of the sample.36.08‫ء‬ 5.43 2.65 1. The primary caregivers of 86 children provided interviews for HCAM ratings.45 2.81 4.26 3.54 1. The relevant data are available on request from the corresponding author.34 2.17 0.55 2. As the CAI does not have a Cannot Classify (CC) code. respectively. Informed by the AAI.34 1.30 3.84. The findings suggest that while boys were more likely to be assigned an Insecure classification.86 and . attachment classifications were assigned independently of the remaining variables.94 2. 83) ϭ 6. Probably a larger sample size is required before a readily interpretable model can be satisfactorily fitted to these scale scores.004.001. none of the remaining demographic variables—including age.25 1.58 0.19 5. For internal consistency with respect to the referred sample.

55.27 (13.68) F(1. (%) 7 (35.7) 29 (64.84.15 (1.9) ␹2(1. ns 110. N ϭ 66) ϭ 6.2) 22 (47. 57) ϭ 1. The CAI elicits both representations of attachment with respect to the parents and nonverbal evidence of emotional and cognitive responses to attachment issues. N ϭ 58) Ͻ 1.0) 5 (11. we found compelling evidence that children aged 8 –12 can respond to direct questions about their attachment relationships and that observed structural and behavioral variations appear to reflect individual differences in their attachment organization. it is both representational and behavioral.00 (13.80 (17.2) ␹2(1. We assume that the specific questions of the CAI—asked by an unfamiliar adult—activate the child’s attachment system.CHILD ATTACHMENT INTERVIEW 951 Table 11 Relations Between Attachment Classification With Mother and Father. (%) ␹2(2. N ϭ 62) Ͻ 1.86) 100.25) F(1. Manchester Child Attachment Story Task.0) 3 (7.02. and SAT and as a potentially useful clinical instrument. the CAI calls on the child to relate memories of interactions with his/her parents.9) 0 (0) 4 (9.95. showing that naı ¨ve raters could (with 3 days’ training) reliably Figure 1.61. 23) ϭ 0. N ϭ 60) ϭ 3. ns 10.4) ␹2(1. N ϭ 51) Ͻ 1. N ϭ 64) ϭ 1.80( F(1. of middle class (%) 9 (45. and Expressive Language Among Children Who Had Been Referred For Treatment Mother Variable Secure Insecure Statistic Secure Father Insecure Statistic Mean age (SD) 10.1) 2 (14.1) 12 (85.63. N ϭ 56) Ͻ 1.7) 36 (83.3) 3 (7. and Demographic Variables. In this sense.81.3) Asian 2 (10. We assume that the child’s style of discourse and quality of responses may be seen as enacting their attachment strategy.6) 21 (45.34) F(1.0) 34 (81. Is the CAI reliable? We began by examining interrater reliability. of boys.10) 84. ns 89. Error bars correspond to 95% confidence intervals.1) ␹2(1. ns ␹2(2. ns White 17 (85. Mean adaptation scores on the Hampstead Child Adaptation Measure of 86 children grouped according to attachment classifications obtained with Child Attachment Interview. No.6) ␹2 (1. N ϭ 57) ϭ 1. 63) ϭ 2. ns Mean verbal IQ (SD) 109. ns 7 (53.59. Verbal IQ. 27) ϭ 0.64 (17. p Ͻ .43 (9. ns No. ns 5 (35.7) Black 1 (5.22) F(1. the latter being potentially important in the assessment of disorganization of attachment. ns 6 (42. ns CAI to offer a developmentally appropriate interview protocol and coding system.9) 13 (30.24) F(1. This study initially investigated whether children in middle childhood could comply with the demands of an interview about attachment relationships. 44) ϭ 2. ns Mean expressive language (SD) 89.55) 100.0) 31 (68.10 (1. We sought systematically to test the new measure and to evaluate it as a viable alternative to semiprojective techniques such as the Attachment Story Completion Task.0) No.0) 12 (28.61.55 (1.20) 85. which will be the subject of a future communication. 39) ϭ 1.6) ␹2(1.75 (23. ns Race: No.42) 10.40) 10. . As with the AAI.54 (1.77 (20. Following extensive interviews with both nonreferred and referred children. which potentially causes stress in a way that triggers the attachment system. living with both parents (%) 9 (46.

but in children currently experiencing.952 SHMUELI-GOETZ. we would be surprised not to find a higher rate of insecurity in a clinical group. (b) children are probably affected quite considerably by their day–to-day experiences. Goldberg. Coherence. to parental separation or emerging domestic conflict. Coherence was in fact the best predictor of overall security. Finding about a quarter of the referred sample to be coded as Secure could reflect either insensitivity in the instrument or an indication of heterogeneity in the clinical sample.e. such interviews were coded as Secure. AND DATTA code the interview. (95% CI. van IJzendoorn.35– 0. test–retest reliability was reduced to 0. the lower end of the 95% confidence interval (CI) range for the stability coefficient remained above 0. ethnicity) nor cognitive variables (IQ and expressive language) appeared to predict attachment classification in either the referred or the nonreferred sample. 2004). 1988) than in high-risk ones (Cicchetti & Barnett. unambiguously secure responses that reflect upon experience after the event). TARGET. The discovery of lower stability rates for the parent-specific scales. The construct validity of the instrument is partly supported by the predictable pattern of associations between scales and attachment categories and by the internal consistency of two out of three of the theoretically derived groups of scales (State of Mind and Avoidance). From a developmental perspective. Of current measures of middle childhood attachment. say. Over 1 year. Closer scrutiny of the Secure clinical interviews suggests that most of these were coherent accounts of disturbed parent– child relationships. The CAI seems to have sound discriminant validity. Distinguishing State of Mind and Current Experience aspects of narratives in the coding system might improve reliability and the capacity for CAI codes to predict other aspects of the child’s functioning. in press). Ammaniti et al. irrespective of specific experience. the classification of security could be qualified in a way that permits future study of the correlates of childhood coping strategies. Kerns et al.7. 1992. we might argue that attachment security may be most effectively coded on two separate bases: (a) on coherence of narrative and (b) on the representation of the parents’ responses to the child’s expressed attachment needs. particularly with respect to father. which may account for the fact that greater stability was found in lower risk samples (Howes & Hamilton. was of some concern. (2005) reported 50% across a 3-month period using the same procedure.79). as in preschool assessments of attachment. had been referred by their school). and (c) some apparent changes in attachment status may reflect actual changes in attachment security linked.. Seifer et al. SES. The distribution of attachment classifications broadly conformed to the distributions reported in other studies of different age groups on other attachment measures. Perhaps children found it more difficult to recount interactions with their fathers in detail and therefore tended to provide fewer and less elaborated descriptions (often characterized by descriptions of activities or outings/“doing stuff”). The second might reflect something closer to the observed interactions. There is evidence to suggest that. (2000) reported 71% in children 10 –14 years old using the Attachment Interview for Childhood and Adolescence. Within the classification currently based on coherent State of Mind. In addition to the stability of attachment categories.. These figures suggest reasonable stability of attachment classifications over 1 year. Narratives that have high narrative coherence but describe negative interactions might be a source of difficulty in achieving interrater reliability. Main & Cassidy. Without suggesting that attachment and psychological disorder are coterminous. Kroonenberg. 0. Promisingly. corresponds to the level of security as assessed by the State of Mind. The child’s coherence might reflect an effort to maintain a good relationship with his or her parents despite current negative interactions. The stability of our combined security rating is therefore encouraging. even though different interviewers and coders were involved. Encouraging findings emerged with regard to stability of attachment classification over a 3-month and over a 1-year period: This stood at 85% for the Secure–Insecure distinction with respect to mother at both 3 months and 1 year and at 83 % for four-way classifications at 3 months and 76% at 1 year. Studies using diverse methodologies to examine stability in middle childhood have generally reported variable rates of agreement over somewhat shorter periods. and Dismissal showing high stability while others showed moderate to low stability after 1 year. They resemble the earned-secure adult attachment narrative pattern (i. rejection. for both the nonreferred and referred samples (DeKlyen. Over a 3-month period. 1991. since neither demographic variables (age. Recent findings suggest that the AAI taps into a continuous dimension of security rather than ideal types (Roisman et al.. observed in a . at least in relation to mothers (figures with respect to father were slightly lower). van IJzendoorn & Bakermans-Kranenburg. as in the AAI. raising questions about whether the continuous measure of security is stable only for these scales. The first. 1996. for example. The level of agreement between the SAT and the CAI three-way classifications suggests convergent or concurrent validity. changes in attachment status are linked to changes in the quality of caregiving and in life circumstances. which would potentially affect specific scales such as Involving Anger or Dismissal toward a parent and the emotional tone of particular episodes recounted in relation to particular questions. at least in preschool children. Our results fall at the higher end of this range. Criterion validity was strongly supported by the predominance of Insecure classifications in the referred sample. This may be achieved at the cost of impairment in other domains of functioning—as suggested by the clinical status of many of these children (who. although future studies are needed to clarify whether their differences in approach tap different aspects of the construct. Granot and Mayseless (2001) reported 94% across 3 months using a doll–story completion task. The highly significant transgenerational association. FONAGY. The observed reliability was limited by the considerable variability in the stability of scale scores contributing to this coefficient. Main & Cassidy (1988) reported 62% across 1 month using their separation–reunion procedure. we wanted to establish the stability of the CAI as a continuous measure of attachment security. & Frenkel. and it is possible that different adults (in the absence of considerable training) elicit different responses. 1992). 1996.62. the CAI seems a valid measure of the attachment construct. in many cases. Several factors may explain the lower stability observed with respect to some of the other CAI scales: (a) a strong test of stability was applied through the use of different interviewers and coders on different occasions. the SAT is closest to the CAI. Internal consistency for Active Conflict was lower probably due to the restricted range of Involving Anger scores in our sample. with several important scales such as Emotional Openness.

CHILD ATTACHMENT INTERVIEW 953 substantial subsample. between the AAI classifications of the mothers and their children’s CAI codes also indicates convergent validity. let alone producing relatively coherent responses to unfamiliar personal questions. The 4% rate of disorganization in this study’s nonclinical sample corresponds with the approximate 5% rate in previous studies. However. our referred children were coded as Disorganized in less than 10% of cases— compared with 50% in other high-risk. which will allow us further to refine the coding of Disorganization. Admit- tedly. We intend to add a new scale to our coding system.. In the light of the psychometric analysis of Roisman and colleagues (in press). and gross motor behaviors. We found high concordance between classifications of attachment to mother and father. the low representation of the Preoccupied classification in both the nonclinical and clinical samples reflects the difficulties of previous studies in identifying Preoccupation. 2003).g. Given the very high concordance between attachment classifications with mother and father. Only 8 children showed discordant classifications. (2005) also showed an inconsistent pattern of associations with other attachment measures. 2005. and relatively coherent descriptions. Thus. The coding criteria were also intentionally conservative to avoid overcoding of Disorganization. The current coding system has only broad guidelines for capturing these nonverbal indicators. Hillman. p. but the Insecure child might struggle to perform the task and manage the interpersonal stress. Preoccupation was expressed in negative. 1999). a small group of children were neither angry. we and our collaborators have recruited several high-risk samples from various communities. Main & Cassidy (1988) and Wartner. In terms of future work. which raises the issue of whether children in middle childhood hold separate internal working models with respect to mother and father. and measures of psychological capacities that are often attachment related (e. We could also refine the coding of the child’s nonverbal behavior. Hopper. overarching current state of mind. and thus to address. drumming of the fingers. at least in part. In collaboration with the Romanian Follow-Up Team. although this is currently being addressed. The interview does not replace parental and teacher reports. there are also causes for concern about the CAI’s validity. Kaniuk. First. This is in itself a stressful task.. such as flattening of affect.g. The CAI is potentially a clinically relevant research tool and a systematic and consistent assessment of the child’s experience of the family situation. The extent to which the child is physiologically or behaviorally stressed by the interview situation should therefore reveal stable attachment strategies. In our sample. eye contact. or have come to integrate these representations into a unitary. although the coefficient of agreement was modest compared with infant transgenerational studies (van IJzendoorn. nor fearfully absorbed in intrusive traumatic memories (as some adult participants in the AAI). Predictive validity of the CAI was supported by the association between the measures of the child’s current functioning and his or her attachment classification. A Secure child should cope better with this task. self-soothing gestures. Using behaviorally derived classifications. the difficulties in identifying more clearly those who show a Preoccupied strategy. 1998. The coding of Disorganization of attachment was partly validated by the difference in rates of disorganization and organization between clinical and nonclinical groups. as conceptualized within the AAI. there now exist extensive studies of children’s emotional expressions.. The Preoccupied scale of Kerns et al. Children with secure attachment relationships seemingly have superior social adaptation (e. which appear strongly linked to the content of the narrative. 2001) and preschool children (Goldwyn et al. . In the CAI. Sroufe. absorbing. & Henderson. This could suggest a lack of sensitivity in the CAI’s current coding of Disorganization. As such. and Suess (1994) were. Verschueren & Marcoen. as shown in infant behavioral assessments. due to inadequate numbers. These children might currently be miscoded as Secure because of their extensive examples. & Fonagy. FremmerBombrik. 2005). but it gives a window onto the child’s own experience of family relationships and parental availability. mainly in the United States. emotion regulation. looking away from the camera. clinical samples. It was particularly encouraging. emotional openness. we developed a Q-sort coding for Disorganization. and often depressing memories (similar to the “inchoate negativity” [Main & Goldwyn. but as an ordinal scale. in terms of the CAI’s discriminant validity. rather. The relationship with the interviewer is also underexplored. Few studies have assessed this concordance. while Disorganized children almost invariably had mothers with unresolved loss or trauma... the CAI could be applied in the evaluation of the outcome of therapy or as part of an assessment for placement of children in foster-care or being considered for adoption. such as ratings of the quality of the child’s relationships. effortful control). that attachment security was associated theoretically with attachment-related aspects of functioning. This finding may be explained by the relatively mild reasons for clinical referrals and the fact that the referred children often had no background of significant trauma. confused. for which the child is unlikely to have been prepared. 1999. unable to include children considered Ambivalent-Dependent. McGauley. repetitive. this index may currently be preferable to the overall classification for each caregiver. and findings suggest a moderate-to-high concordance rate (Kerns et al. the measure has a nonsignificant positive skew in the nonreferred sample and a slightly negative skew in the referred sample. the preliminary findings of which suggest that the interview is sensitive to unusual phenomena commonly observed in neglected Romanian adoptees some years after adoption. 168] of participants’ interviews coded as Preoccupied in the AAI). which captures this excessively absorbed style of preoccupation. Hodges. Turton. The CAI requires the child to talk to a stranger about his/her relationship with each parent. it currently seems most efficient to use a single index of security based on the continuous measure derived from the rating scales. and so on. Grossman. 1992). his/her narrative would thus possibly be impoverished or confused. The correspondence between adult Unresolved or Cannot Classify codes and the Disorganized classification of CAI narratives deserves further exploration—as also suggested by measures of attachment among infants (Hughes. it brings together stimulation of attachment feelings and memories and the presence of a sympathetic stranger (analogous to the Strange Situation). Green et al. 2000) and by studies concerning adopted children and their adoptive families (Steele. Although many of the findings are encouraging. it would provide a reliable and valid indicator of the degree of individual security. most mothers with Unresolved classifications had children who were not coded as Disorganized.

& Hart. 11. New York: Guilford. victims and bully/victims in middle childhood.. et al. (2002). Developmental Psychology. 3. Smith. Luborsky. Goldwyn. Treboux. Kuperminc. The Manchester Child Attachment Story Task: Relationship with parental AAI. Cassidy & P.. Ainsworth.. Child Development. B. S. Chicago: University of Chicago Press. Bruner.. (1999). University of Rome “La Sapienza. Ammaniti & D. & Binney. Ortu. H. (1992). & Barnett. M... Autobiographical memory: Reports in adolescence. S.. & Goldwyn. H. Kerns. Discriminant validity of the Adult Attachment Interview. (2002). Speranza. S.. Grice. P. J. Attachment based assessments of parent-child relationships in middle childhood. Attachment theory: Retrospect and prospect. (2000). Fox.. J. 614 – 626. (2000).. M. V. In K. Understanding transference: The CCRT method. J. D. 292–307. W. & Goldwyn. Attachment and . (1976). George. Journal of Child Psychology and Psychiatry and Allied Disciplines.. Concordance in attachment states of mind and styles with respect to fathers and mothers. 67. N. & Main. Continuity and discontinuity of attachment from infancy through adolescence. 71– 84. Child Development. T. In M. Perceived family relationships of bullies. 62. A. K.. et al.). Disorganised attachment behaviour among infants born subsequent to stillbirth. FL. D. M. Journal of Consulting and Clinical Psychology.. R. New York: Guilford.). Social support processes: Mediators of attachment state of mind and adjustment in late adolescence. K. D. Cassidy & P. J. 25. C. 835– 846. & Wall. Aspelmeier. Bar-Haim. 41.. T. Attachment representations in 6-year-olds: Related longitudinally to the quality of attachment in infancy and mother’s attachment representations. Children’s relationships with caregivers: Mothers and child care teachers. M.. J. Annotation: Attachment disorganisation and psychopathology. V. Dazzi. 35. K. A. Richardson (Eds.. 2. 215–232. C. & Fearon. (2000).” Ammaniti. A psychometric study of the Adult Attachment Interview: Reliability and discriminant validity.... 7–26. M.. & Lyons-Ruth. 1444 –1457. New York: New York University Press. D. Crittenden. Moore. SAT. (1991). C.. Child Development. Internal working models of attachment during late childhood and early adolescence: An exploration of stability and change. (2001). W.. Stern (Eds. 328 –346. 23. 463– 488).. Shaver (Eds. 53. J. R. representation. Stovall. Cicchetti. McElhaney. Sutton. Bar-Haim. A.. Attachment and Human Development. Koenig.). Bakermans-Kranenburg. M. A.954 References SHMUELI-GOETZ. V. Cummings (Eds. The development of self-understanding from infancy through adolescence. C. British Journal of Projective Psychology and Personality Study. Infant–mother attachment security. M. contextual risk.. (2000). Ainsworth. & Simon. C. (1982). 14.. (1999).. 2.. A. 65. M. Journal of Social and Personal Relationships. Attachment and psychopathology in childhood. Y. IAL: Intervista sull’attaccamento nella latenza [AICA: Attachment Interview for Childhood and Adolescence]. Damon. 3–35. D. New York: Academic Press. van IJzendoorn. & Green. M. Maternal representations of attachment during pregnancy predict the organization of infant–mother attachment at one year of age. J. 357–365. Attachment in the pre-school years: Theory. Journal of Child Psychology and Psychiatry and Allied Disciplines. (1991). New York: Basic Books. 2. 96 –120.. Benoit. Tomich. B. M. (1990). K. (1996). J. 469 – 496). Gloger-Tippelt. Attachment and Human Development. 2. G. N. Granot. AND DATTA psychopathology in adulthood. Waters. 2584 –2599. P. Klagsbrun. Logic and conversation. C. 530 –541.. & Albus. Miami. Unpublished protocol. Child Development. & Tambelli. W. 36. Biesecker. (2001). Belsky. R. D.. & Hamilton. The narrative construction of reality. West. Child Development. Shaver (Eds. 841– 864. E. Unpublished scoring manual.. M. & Bernier. Schlegelmilch. M. 41–58). M. Stability and change of attachment at 14. Bhangoo. 21. 791– 801. 64. Attachment and Human Development.. K. M. M. C. (2003). A.. Journal of Child Psychology and Psychiatry. Attachment organization in preschool aged maltreated children. Stanley. R.. Cassidy. Larose. (1985). Greenberg. Morgan (Eds. R. B.. 24. A. (1993). Assessing attachment in middle childhood. 859 – 866. Green. Smith. E. Psychoanalysis and development: Representations and narratives. Attachment organization in three. 497–519)... Bretherton. 71. A. Attachment in middle childhood (pp. A. & Bowlby. A. Kerns & R. K. Marvin.. S. Easterbrooks. Journal of Applied Developmental Psychology. Child Development. 40.. (1985). Turton. Steele. and 58 months of age: Behavior.. 170 –187. & Vetter. & the MacArthur Working Group on Attachment. P. (1994). L. R. M. 48 –70. (1992). University of Illinois at Chicago.. Development and Psychopathology. Candelori. C.. H. 431– 450. Muscetta. M.. 50(1–2. J. Patterns of attachment: A psychological study of the Strange Situation. New findings in attachment research and their potential implications for developmental psychopathology in childhood. and early development: A moderational analysis.. (1994).. 3. N. 381–388. E.. Handbook of attachment: Theory. Adult Attachment Interview. 43. E. & Fonagy. G. University of California at Berkeley. Cicchetti. Morgan. P. D.. 1239 –1247. Y. & E. Gomille. In M. J. D.and four-year-olds: Coding guidelines.. Smith. Green. & Steele. (1989). (2002). De Coro. (1978). FONAGY.. Dozier. NJ: Erlbaum. Attachment and Human Development. E. 891–905. Unpublished manuscript. (1990). 690 – 694.. J. and child behaviour. (2000). TARGET. M. N. M. K. Blehar. New York: Guilford. A new method of evaluating attachment representations in young school-age children: The Manchester Child Attachment Story Task. V..). Allen. 46 –70). research and intervention (pp. and life events. J.. P.. A. (2000).). S. & van IJzendoorn. Hillsdale. (1999). Preschool assessment of attachment. Waters. M. P. & Mayseless. G. A. (2002). Family Relations Institute. A. D. Monographs of the Society for Research in Child Development. L. A. Developmental Psychology. E. 42. & Crits-Christoph. J.. I. P. 4. K. Bowers. A secure base in adolescence: Markers of attachment security in the mother–adolescent relationship. Greenberg. Ammaniti. 870 – 879. In J. & O’Connor. L.. Handbook of attachment: Theory. Beck. 397– 411. Development and Psychopathology. Stability and transmission of attachment across three generations. O. D. J. Furman. 318 –339... Syntax and semantics: Speech acts (pp. M. (1996). Crowell. (1975). 63. E. The emergence of childhood bipolar disorder: A prospective study from 4 months to 7 years of age.. G.. In R. Howes. DeKlyen. & Parker... Hamilton. Hopper. P. A. research.. Canadian Journal of Psychology. D. R. Response to separation from parents: A clinical test for young children. Infancy predictors of emotional availability in middle childhood: The roles of attachment security and maternal depressive symptomatology. C. M. Unpublished manuscript. L... Fonagy. R. H. Kerns. Epilogue: Some considerations regarding theory and assessment relevant to attachment beyond infancy. 69 –73. and clinical applications (pp. C. N. Pe ´ rez-Edgar. & Contreras. P. Attachment and Human Development. R. In J. Child Development. M. research and clinical applications (pp.. J.. M. J. R.). Kaplan. (1981). Fox. Hughes. (2001). Attachment security and adjustment to school in middle childhood. Series 209). M. (1994). S. O’Beirne-Kelly. K. Stanley. 29. T.. S. E. Disruptive behavior disorders and intergenerational attachment patterns: A comparison of normal and clinic-referred preschoolers and their mothers.. Fitzgerald. International Journal of Behavioral Development. & Abraham.. et al. Cole & J. P. & Marvin.. Land. Developmental Psychology. (1990).. 74. Attachment and Human Development.. F. G. P. 293–310. (2004). McGauley. (2005). K.

. (2002). H.. M. 445–560. Carlson. Attachment and Human Development. DC. (1990). and interpretation of the Separation Anxiety Test for 11. (1993). Pederson. & Secord.. van IJzendoorn. T. 407– 474). H. S. Patterns of individual adaptation across childhood.. Ladd. 38. National Institute of Child Health and Human Development (NICHD) Early Child Care Research Network. (1990). 76 –99. 430 – 444. L.. J. Wiig. I. Goldberg. R.. Adult attachment scoring and classification system. Journal of Child Psychotherapy. J. J.. M. M. Greenberg. childhood. research and intervention (pp. Egeland... Intergenerational transmission of parenting: A review of studies in nonclinical populations. Attachment theory: Social. Fraley. Coping with security distress: The separation anxiety test and attachment classification at 4. K. Hillman.CHILD ATTACHMENT INTERVIEW Lyons-Ruth. 50. Hodges.. Developmental Review. M. 33. Development and the fragmented self: Longitudinal study of dissociative symptomatology in a nonclinical sample. Shaw. Ensink. Development and Psychopathology. Psychological Science. A. coding. 7. J. (1990). and clinical perspectives. Nelson. Switzerland: Delachaux & Niestle. Main. Child Development. longitudinal study from birth to adulthood. Chicago: Chicago University Press. M. Development and Psychopathology.). 234 –247. Royal College of Psychiatrists. O. Cummings (Eds. Weeramanthri. (1991). Glasser. 121–160). Moss. 14-year-olds.. University of London. 415– 426. TX: Psychological Corp. M.. (2002). J. A. Developmental Psychology. Categories of response to reunion with the parent at age 6: Predictable from infant attachment classifications and stable over a 1-month period. Slough. Unpublished doctoral dissertation. LaGasse. Main. & Pederson. (2000). Paper presented at the biennial meeting of the Society for Research in Child Development. 349 –367. Rockville. Muir. Egeland. Maternal sensitivity. E. In I. et al.... 223–232. Madigan. & Cassidy. 23. (2004). Schneider. (1996). G.. (pp. L. Neuchatel. London: University College London and the Anna Freud Centre. & E. 187–205. (1996). Urban.. E. 3. Unpublished manual. 847– 862. S. & Goldberg. and adulthood: A move to the level of representation. Attachment representations and adoption: Associations between maternal states of mind and emotion narratives in previously maltreated children. .. parental responsiveness. Measuring attachment in early adolescence: A manual for the administration. Sroufe. Trowell. Kerr (Eds. S. MD. M. R. Attachment and development: A prospective. Minnesota Symposia on Child Psychology (Vol. 75. & Leitch. M. A. Mongeau.. Stams.. & Kivenson-Baron. T. K.). with selected implications for clinical work. Shankaran. M.. T. Moran. (1992). M. S. L. & Vondra. Psychotherapy for sexually abused girls: Psychopathological outcome findings and patterns of change. H. I.. 915–927. (2004). 16. Ogawa. P.. Raters’ manual for the Hampstead Child Adaptation Measure (HCAM).. 61. & Egeland. (2003). N. developmental. Emotional development: The organization of emotional life in the early years. 12. 1228 –1231. (1987). Oppenheim. N. (2005). & Goldwyn. M. The Attachment Doll-Play Interview for preschoolers. (1987). Attachment and Human Development. N. 451– 467. P. (1998). Madigan. F. (1994). G. New Directions for Child Development. W.. Security in infancy. K. B. Goldberg. Clinical Evaluation of Language Fundamentals—Revised (CELF–R). 24... D.. G. McFarland.. 67. C. G. NJ: Erlbaum.. Sroufe. K. J.. Hillsdale. S. J. (1997). Barry. E. T. Fonagy. 681– 697. C. Attachment. Sroufe. Adolescents’ attachment representations and developmental tasks in emerging adulthood. & Janes. 48. C. A children’s global assessment scale (CGAS).. & Moran. and infant attachment behavior. Infant– caregiver attachment and patterns of adaptation in preschool: The roots of maladaption and competence. & Solomon. In M.. Schneider. Resnick. & Stahl. B.. (1983). McElhaney. Archives of General Psychiatry. van IJzendoorn. D. Sadowski. 387– 403. R.. Shaffer. In M. Kolvin. Developmental Psychology. D. Child Development.. J. 41– 83). Connell. (1997.. A. Developmental Psychology. 1363–1373. I. (2003). Resnick.). April). Development and Psychopathology. Five-year-olds’ representations of separations from parents: Responses from the perspective of self and other. San Antonio. S.. Juffer.. 15. D.. Main. R. A. R. C. Attachment status in children prenatally exposed to cocaine and other substances. 40. & Henderson. S. J. B. J. E. A.. K.. Developmental Psychology. (1983). 40.. D. Child care and family predictors of MacArthur preschool attachment and stability from infancy. & Cassidy. Psychological Bulletin. Child Attachment Interview: Development and validation.. P. Adult attachment rating and classification system: Manual in draft (Version 6. 7–14. & Sroufe. & Kreutzer.). 67– 84. Target. Carlson. Piaget. Naı ¨ve observers’ perceptions of family drawings by 7-year-olds with disorganized attachment histories. and temperament in early childhood predict adjustment in middle childhood: The case of adopted children and their biologically unrelated parents. 806 – 821. Unpublished manuscript. Y. M. P. England. Child Development. The psychological and social origins of autobiographical memory. D. University of California at Berkeley. (1985). (1923). M. infant attachment. Mayseless. S. F. Semel. (1995). H. T. M. A picture is worth a thousand words: Children’s representations of family as indicators of early attachment. 323–334. Kaniuk. C. & St-Laurent. (2001). Fisher. 223–241. Attachment and Human Development.. S. Recent studies in attachment: Overview. Bird. 117. (1988). 180.. Bureau.. (1997). Zoll. E. (2001). J. M. 331–348. Main. Developmental Psychology. J. G. Westat. (1992). G. & Land.. M. H. Sroufe.. M. The correspondence between the Strange Situation at 12 months and the Separation Anxiety test at 11 years in an Israeli kibbutz sample. Chicago: University of Chicago Press. NJ: Analytic Press. B. M. L. R. F. L. 850 – 868. Gould. and infant attachment: A meta-analysis on the predictive validity of the Adult Attachment Interview. K. H. D.. & Goldwyn. Berelowitz. Developmental Psychology. R. Brasic.. J. L. Bauer.. J. Bretherton & E. 29. L. Cyr. (2003). H. Adult attachment representations. pp. 955 Scharf.0). 855– 879. R. & Stevenson-Hinde. Infant attachment security and maternal predictors of early behavior problems: A longitudinal study of low-income families. 6. Shmueli-Goetz. L. In S. S. autonomy. D. University College London. pp... 335–357.. & Greenberg. Infants at social risk: Relationships among infant maltreatment.. L. University of California at Berkeley. maternal behavior. A. Unpublished manuscript.. P. (1995). Allen. M. Journal of Child Psychology and Psychiatry. Inc. Procedures for identifying infants as disorganized/disoriented during the Ainsworth Strange Situation. Unpublished doctoral dissertation. (1995). 40. England.. Perlmutter (Ed. & van IJzendoorn. J.. L. Le langage et la pensee chez l’enfant [Language and thought in children]. Washington... and multifinality in adolescent internalizing and risky behavioral symptoms. Roisman. 5.5 years. Attachment during the preschool years: Theory. The fate of early experience following developmental change: Longitudinal approaches to individual adaptation in childhood. Growing points of attachment theory and research: Monographs of the Society for Research in Child Development (Vol. Weinfield.. Correlates of attachment at age 3: Construct validity of the preschool attachment classification system. (1997). Expressions of the attachment relationship outside of the Strange Situation. (2004). A. 9.. 37. J. A taxometric study of the adult attachment interview.. Steele. D.. (2004).. 23. & Bada. M. Kaplan. & Belsky.. Developmental Psychology. Sroufe. Journal of Abnormal Child Psychology.. Waters (Eds.. International Journal of Behavioral Development. 19 –37. D. Seifer. (in press). 20. New York: Cambridge University Press. Measuring adaptation in middle childhood: The development of the Hampstead Child Adaptation Measure. Marsh. 4. R. I. Hillsdale. (1993). Main. Main. & J. Shouldice. J. D. 66 –104). Ambrosini.

(1996). Y.. H. (1995). Weinfield. K. 36. 1. van IJzendoorn. Journal of Child Psychology and Psychiatry. TARGET. Latent Gold (Version 3. Attachment and Human Development. 695–702. 327–348. 259 –268. Wallis. J. (2004). Can you tell me about the people in your family? —The people living together in your house? (then ask about extended family. Weinfield. (2001). van IJzendoorn. J. Attachment representations in mothers. (1992). J. D. Security of attachment in 8. 3. Wright. What happens when your mum gets cross with you or tells you off? —Story. Fremmer-Bombrik. J. P. In L. Have you been away from your parents for longer than a day? (If child is not living with parents [e. & Albersheim. If you could make three wishes when you are older. and coherence in attachment from infancy to late adolescence: Sequelae of organization and disorganization.. M. Child Development. V. What happens when you’re ill? —Example. 70. A. Can you tell me three words to describe your relationship with your mum. Egeland.. P. Attachment and psychopathology (pp. G. Child Development. Attachment behavior. G. M. 4. Warren. FONAGY. Is there anyone that you cared about who isn’t around anymore? 14. J. 8 –21. Representation of self and socioemotional competence in kindergartners: Differential and combined effects of attachment to mother and father. 60. 6. H. (2000). J. Huston.. what it’s like to be with your Mum? —Examples for each. L.. (1999). Attachment security from infancy to early adulthood: A 20-year longitudinal study. A. B.. and clinical groups: A meta-analytic search for normative data. H. P. K. Kroonenberg.. San Antonio. In what ways would you like/not like to be like your mum/dad? 17. (1989). The relative effects of maternal and child problems on the quality of attachment: A meta-analysis of attachment in clinical samples. 3. Can you tell me about a time when you were really upset and wanted help? —Story. Atkinson & K. 64. fathers. E. Zucker (Eds. L. B. ask about time when he/she left parents. Wechsler Intelligence Scale for Children (3rd UK Ed.). 9. Waters. & Barglow. Merrick... Journal of Consulting and Clinical Psychology. O. 757–774. discontinuity. 36. Intergenerational transmission of attachment: A move to the contextual level.. Tell me three words that describe yourself. (1992). Social information processing in middle childhood: Relations to infant-mother attachment. B. Child Development. that is. what would they be? Received July 18. adolescents. New York: Guilford Press. & Magidson. N. J. Wartner. B. D. & 12-year-olds: A revised version of the Separation Anxiety Test. Whaley.. 59. 135–170). 684 – 689. [Questions 3 & 4 repeated for Dad or other main caregivers] 5. N. (2000). Wechsler. your point of view... Child Development.. its psychometric properties. want to know what things are like in your family. and clinical interpretation. M. G.. (2003). & Frenkel. (Elementary-school-aged children:) Have you ever been touched in the private parts of your body by someone much older than you? (For older children:) Have you ever been touched sexually by someone when you didn’t want him/her to do it? —Story. H. 183–201. U. Have you ever been hit or hurt by an older child or a grown-up in your family? —Story. 65. Belmont.g... 12. MA: Statistical Innovations.).956 SHMUELI-GOETZ. 637– 644. & Steele. Vermunt. Sroufe.. & Bakermans-Kranenburg. Has anyone important to you ever died? Has a pet you cared about died? —Story. D. What did you feel and others feel? 13. or confused you? —Story. Crowell. 73–97. discontinuity and their correlates. S. 2006 Revision received August 13. that is. 728 –737. A. How do you feel? 16. B. 2007 Ⅲ . 11. Child and Appendix Child Attachment Interview (CAI) Protocol Introduction—interview not a test. van IJzendoorn. R. M. 71.. TX: Psychological Corp. Lefever. & Smith.0. and temperament during infancy. 71.) —Story. AND DATTA adolescent anxiety disorders and early attachment. S.1) [Computer software]. Ziv. Attachment and Human Development. & Sagi-Schwartz. Goldberg. & Egeland. (1997). Continuity. S. Attachment patterns at age six in South Germany: Predictability from infancy and implications for preschool behavior. Vaughn. & Sroufe. L. K. attachment security. (1994). C. Has anything [else] really big happened to you that upset. J. What happens when you get hurt? —Example. is in foster care]. How did you and parents feel? What was it like when you saw them again? 15. K. G. P. (2004). Child Development. 6. (1997).. 6. M. J. Do your parents sometimes argue? —Story. Attachment from infancy to early adulthood in a high risk sample: Continuity. M. & Bakermans-Kranenburg.. Grossman.. 147–156... & Egeland. 8. —Examples. Oppenheim... Journal of the American Academy of Child and Adolescent Psychiatry. Verschueren. Treboux. & Suess. E. Attachment representations in adolescence: Further evidence from psychiatric residential settings.. Binney. Child Development. 2007 Accepted November 27. 1014 –1027. L. E. what sort of person you are. Do you ever feel that your parents don’t really love you? —When? Do they know you feel that? 7. Attachment and Human Development. L. S. —Have you been badly hurt by someone outside your family? 10. Seifer..) 2. A. scared..