MATERNAL INTERACTION STYLE, THE FAMILY AND ATTACHMENT OUTCOMES

Understanding of early relationship processes has been moving away from simplified linear relationships in favor of multifaceted approaches to child development (Moran & Pederson, 1992). Incorporation of these more dynamic models into the field of attachment research has been advocated (Mangelsdorf, et al., 1990), but not duly accomplished to date. Relations between maternal sensitivity and attachment outcomes are still expected to be linear, although empirical support for this notion is limited. Studies vary in methodology, findings and interpretation. Building on previous research (Schiller & Seifer, 1992), this study addressed methodological issues as related to assessment of maternal sensitivity and attachment constructs as well as placed their relationship within the context of family interaction. Using videotaped data of 51 mother-infant dyads, we identified the relevant components of maternal sensitivity as related to attachment outcomes. Data consisted of (1) six weekly naturalistic observations of free-play interaction in the home at 6 months (2) Ainsworth Strange Situation at 12 months during a laboratory visit; (3) home-based Q-sort measures of attachment security derived from both maternal and observer reports; and (4) self-report and interview measures of family functioning. Scoring systems appropriate to each of these assessments were used. Multiple home assessments were used so that a series of observations could be aggregated to form reliable measures of the maternal sensitivity scales. Results indicated that (1) aggregation of multiple home observations produces highly reliable and consistent measures of sensitivity (2) maternal sensitivity was related to both home (Q-sort) and laboratory (Ainsworth classifications) measures of attachment, although the Q-sort method produced more robust findings, (3) measures of family functioning were positively related to Q-sort attachment, but not to Ainsworth classifications of security, ( 4) Interview-based measures of family functioning were related to both sensitivity and observer-reported Q-sort security, while (5) Self-report measures of family functioning were related to mother-reported Q-sort security. Current findings are discussed in the context of previous attachment research. Methodological as well as theoretical explanations are considered.


Statement of the problem
Understanding of early relationship processes is moving away from simplified linear relationships in favor of multifaceted approaches to child development . Human behavior and development are increasingly conceptualized as a dynamic process affected by, and interacting with, many variables. Integrative approaches have been progressively replacing theoretically polarized positions; transactional models (Sameroff and Chandler, 1975) are talcing the place of more linear conceptualizations as more accurate explanations of complex phenomena. This evolution is also becoming evident in research on early attachment.
The long-standing dichotomy, for example, between temperament and attachment is slowly giving way to an appreciation of the mutual dependency of such constructs (Seifer & Schiller, in press;Susman, Waldman, Kalkose and Egeland, 1992;Calkins and Fox, 1992;Vaughn, Stevenson-Hinde, Lefever, Shouldice, Trudel, Belsky, Waters and Kotsaftis, 1992). This in turn opens up more complex models of development and at the same time challenges researchers to develop methods of measuring and describing interactive behaviors and qualities of relationships.
Incorporation of these compelling approaches specifically into the field of attachment research has been advocated by researchers from several perspectives. Drawing on systemic concepts, Mangelsdorf et al., (1990), and Pederson and Moran, (in press) have argued that relationships are necessarily seen as an "organized whole" (Sroufe and Fleeson, 1988) in which the individual characteristics of each participant play an active role. Additional support for this notion of adaptive pathways, rather than predetermined positive and negative traits or relationships, comes from the field of developmental psychopathology . Similarly, the role of context in shaping behavior has been strongly implicated in behavioral research (Seifer and Sameroff, 1986;Sameroff and Emde, 1989) but not as yet systematically applied to the study of the marital and family context in shaping the developing relationship between a caregiver and child.
In spite of these theoretical advances and initial efforts at implementation, the process of change has been slow and inconsistent (Pederson and Moran, in press).
Specifically, the subject of this study, relations between maternal sensitivity and attachment outcomes, are still often expected to be linear, although both theoretical and methodological support for this notion is limited. Studies to date have varied in methodology, findings, and interpretations. There have been inconsistencies ranging from operationalizing constructs (especially sensitivity) to the type, number and context of assessment. Mixed findings or findings with small effect size have been open to diverse interpretations. While some researchers maintain the position that maternal sensitivity is the single most effective predictor of attachment (Isabella, 1993), others emphasize the weak or inconsistent relations (Rosen and Rothbaum, 1993).
The purposes of this study were: (1) to address some of the methodological inconsistencies revealed in the literature through careful selection and implementation of assessment measures. Aggregated measures of sensitivity and multiple methods of assessing attachment were used; and (2) to place the attachment relationship within a broader context of development and family life by conducting multiple assessments in the home environment and by gathering information to help explore the role of family and marital relationships through both self-report and clinical interview methods.

.Justification for and Simificance of the Study
Importance of Attachment The attachment system, as first conceptualized by Bowlby (1969Bowlby ( , 1982, is a species-typical set of adaptive responses whose evolutionary value apparently is to protect its vulnerable members. Bowlby proposed that the "attachment system" was comprised of several patterned behavioral responses including attachment behaviors (such as crying, calling, reaching, and following) that can be "activated" when the young find themselves in risky or threatening situations. The adaptive function of these behavioral patterns consists of returning the infant to closer, proximity to its attachment figure, who provides caretaking and protection. Secure attachment, in Bowlby's terms, is a well-defined, efficiently activated system, that will function as needed to provide proper protection for the vulnerable child, as well as a secure base for exploration of the environment. Operationally defined, securely attached children are able to use their mothers effectively and without anxiety as a "secure base" from which to gain comfort when needed, and then return to exploratory tasks.
The quality of attachment, often conceived as the degree to which an infant's balance of exploratory vs. safety needs are met by the caregiver, has been seen as an important contributor to lifelong development. Theoretically, the child's capacity to seek comfort, have his needs met by caregivers and return to the business of growth and learning is crucial for the development of "secure" expectations about himself and the world. As attachment theorists point out, a caretaker's pattern of contingent, sensitive behavior over time provides a foundation for the development of trust and positive self-worth (Sroufe and Waters, 1977;Bretherton, 1985). Such positive primary relationships provide the basis for approaching relationships with others in an open, productive attitude which is more likely to result in fulfilling close relationships.
Attachment theorists invoke a theoretical mechanism, the "internal working model" to describe the control system that regulates attachment behavior (Bowlby, 1980;Stern, 1989). An internal working model is a motivational system that regulates behavior on a symbolic, largely non-conscious level and guides one's interpretations and interactions with the environment. According to Stern, (1989) these models, which begin as flexible representations, based on early caregiving experiences, develop into less malleable schemas that determine how a person approaches interactions with others, and in turn how others react thereby confirming the original model. In this way, early attachment relationships, through ongoing interplay with the environment help dictate a pattern of interpersonal relatedness.
Recent empirical work has substantiated the claim that secure attachment patterns are an important marker of positive and adaptive socio-emotional development. Security of attachment has been associated with positive oµtcomes in cognitive development in both normative and delayed populations (Donovan & Leavitt,1978;Mahoney et. al., 1985;Bakeman & Brown,1980;Bornstein & Tarnis-LeMonda, 1988). Other beneficial effects of secure attachment include social behavior and communication skills in the developing child (Clarke & Seifer, 1985;Lutkenhaus, Grossmann & Grossmann, 1985)) as reflected in more stable and rewarding relationships (Clarke-Stewart & Hevey, 1981;Hubbs-Tait, 1987;Park & Waters, 1989;Main, 1983;Patterson, Cohn & Kao, 1989). As noted above, security of attachment has been seen as a foundation for later emotional regulation and personality development (Main, Kaplan and Cassidy, 1985;Sroufe and Fleeson, 1986). Further, Sroufe and Fleeson argue that the way in which early behavior is organized vis-a-vis another important person, will shape the course of other significant relationships in later life. It is hypothesized that infants who receive sensitive, timely responsive care acquire a set of positive expectations about close relationships with others, which form a foundation for satisfying and productive adult relationships.
Sroufe , on the other hand (1988) provides a theoretical framework for how less adequate attachment behaviors may lead to maladaptive peer relationships. In the case of an insecure child, who presumably has internalized a working model based on unfulfilling response to his needs, one possible scenario might be to avoid others, rather than to approach them openly. When faced with rejection, he is likely to find confirmation for his internal working model and to retreat further into an avoidant strategy. Another child faced with variable, or inconsistently available parental response, may develop an ambivalent working model and in anticipation of unpredictable behavior on the part of others, rely on angry behavior as a coping strategy further contributing to his own rejection. Such cycles, if perpetuated, may result in social isolation, withdrawal and depression. Crittenden (1992) outlines a related model. She sees insecure children as those who fail to learn adaptive coping mechanisms for managing affect. These children, who theoretically fail early on to receive appropriate responsiveness from their caregivers, do not learn to regulate their affective states, but rather to intensify their signals. These children are at risk for exhibiting aggressive, maladaptive behaviors associated with conduct disorders. Matas et al., (1978), Sroufe, (1988) conducted an extensive longitudinal project that lends support for these hypotheses.
These researchers found that insecure preschoolers had a higher incidence of selfesteem problems, depression and generally less adaptive peer relationships. Lyons- Ruth and colleagues (1993) also reported that children judged as disorganized with regard to attachment strategy at eighteen months were more likely to engage in maladaptive, aggressive peer relationships in preschool.
Beyond early childhood, researchers are beginning to examine the effects of early attachment on adolescent development. Insecure attachment is described as a risk factor for adolescent depression (Kobak et al., 1991;Greenberg et al., 1991), with continued significant effects on peer relationships (Urban, et al., 1991). In line with this interest, Main and Goldwyn (1984) developed the Adult Attachment Interview.
Administered to adults, this semi-structured interview is hypothesized to reflect adult internal working models of relationships, which presumably continue to influence and guide their relationships with others, including their offspring, thus contributing to intergenerational transmission of relational processes. Significant relations between adult and infant security have been reported (van IJzendoorn, 1993;Fonagy, Stele, & Steele, 1991).
To summarize, attachment is viewed by theorists as an important regulatory and developmental construct. Originally conceptualized as an evolutionary, speciestypical mechanism, the attachment system, as it develops into internal working models is seen as having lasting effects on personality and interpersonal development throughout the lifetime.

Importance and Definitions of Sensitivity
Attachment theory implies that maternal sensitivity is one of the primary factors determining a secure mother-infant attachment. In their classic study, Ainsworth and colleagues examined the relationship between patterns of maternal sensitivity to infant's cues, over the first year of life, and the quality of attachment observed at twelve months (Ainsworth, Blehar, Waters, & Wall, 1978). Findings showed that mothers judged as more sensitive over the course of the year, were more likely to have securely attached infants at 12 months than less sensitive mothers.
Sensitive mothers, as defined by Ainsworth et al. (Ainsworth, Bell, & Stayton,1971;1974) were able to accurately interpret their babies' cues and respond to them appropriately, promptly and consistently. Ainsworth et al. asserted that babies of such mothers, on the basis of accumulated experience, develop the expectation that their needs will be most adequately addressed, the beginnings of positive working models. They learn that their signals are heard and understood; they develop trust.
Secure attachment, then, is viewed as an outgrowth of this basic trust.
At the other extreme, maltreated infants, described as suffering from "caretaking casualty" (Sameroff & Chandler, 1975), have been found to be more likely than their normative counterparts to demonstrate disturbances in the quality of attachment (Lyons-Ruth et al., 1987;Cicchetti & Barnett, 1991) and with it, some farreaching unfavorable consequences (Schneider- Rosen, et al., 1985).
In light of this emphasis and above noted evidence regarding associations between parenting, attachment and developmental outcomes, a closer look at the construct of sensitivity is in order. Noted researchers in the field have emphasized various aspects of the maternal repertoire in their definitions of maternal sensitivity.
Accordingly, several terms have been used interchangeably in the literature to refer to the underlying construct of optimal parenting. Ainsworth's concept of sensitivity is the appropriate and contingent responsiveness exhibited by the mother to her infant's cues. Ainsworth has relied primarily on naturalistic observations particularly in caretaking situations to arrive at global ratings of relative maternal sensitivity (Ainsworth and Bell, 1969). Stern (1977) has focused on the timing and structure of the mother-infant interaction particularly during moments of social play.
"Attunement" is the desired state of mutual responsiveness which is attained by an infant and his mother , provided that the mother is able to perceive the infant's cues and adjust her behaviors to the appropriate level of stimulation. Such interactions, studied in detail during face-to-face interactions, are characterized by periods of mutual greeting, engagement and breaks. The sensitivity of the mother in this case would be most closely associated with her ability to tune up or down according to her infant's needs. Insensitive interaction is often characterized by intrusive, or overstimulating behaviors at times when the inf ant is sending signals for a break, or lack of interesting action when the infant is engaged and clearly available.
Detailed investigation of the face-to-face interaction has been the focus of study of Tronick, Als, and Brazelton (1980), Cohn et al. (1986), and Kaye and Fogel (1980). Second-by-second analyses have been used to describe the steps comprising a continuum of mutual involvement of young inf ants and their mothers in a laboratory setting. "Synchrony" is the term most often used by these researchers to describe the ideal state where each partner is picking up the cues of the other and interacting accordingly. Imitation of baby's behaviors, appropriate pauses, and mutual gaze are some of the more favorable behaviors observed in synchronous dyads. Fogel and Thelan ( 1987) point out that the challenge facing the mother lies in the need for continuous adjustment to the growing capabilities of her developing infant. As the infant becomes capable of longer attention span and ongoing stimulation, the mother needs to expand and change her repertoire accordingly or risk boring and "tuning out" the infant.
In a similar vein, Belsky (Belsky, Taylor, Rovine 1984--Studyll;Isabella, Belsky & von Eye, 1989;Isabella & Belsky, 1991) operationally define sensitivity as "interactional synchrony", which essentially consists of reciprocity of the dyad and the responsivity of the mother. In accordance with attachment theory, Belsky hypothesizes that maternal responsivity to infant's cues lead to a mutually rewarding interaction which in tum leads to the infant's conceptualization of the mother as "available, responsive and trustworthy." According to this model, responsivity and synchrony are the precursors of secure attachment. In fact, maternal responsiveness to infant's cries were the early focus of study in documenting individual differences. Bell and Ainsworth (1972) reported that in their longitudinal sample, infants whose mothers responded quickly and consistently to crying in early infancy, had children who cried less, and used alternative communication more than children of mothers who systematically did not respond in this fashion.
Narrowing the construct even further some researchers have used responsivity alone as an implicit or explicit measure of sensitivity (e.g., Crockenberg & McCluskey, 1985;Lewis & Feiring, 1989). These researchers either distributed selfreport questionnaires aimed at assessing responsivity patterns or counted frequency of responses on a time sampled basis. In interpreting their results these authors often equate these variables with measures of sensitivity.
To reiterate, due to the emphasis placed on maternal sensitivity as a contributing factor to attachment and developmental outcomes, it has been the focus of extensive study. Operational and methodological variability in the literature, however has contributed to ambiguity in interpretation of findings.

Assessment of Attachment
Ainsworth and her colleagues developed a laboratory procedure designed to classify observed behaviors exhibited by the child in response to a series of separations and reunions with his mother in an unfamiliar setting (Ainsworth & Wittig, 1969). Specified behaviors rated over the course of this "Strange Situation" procedure generate three general categories: Secure (B), Avoidant (A), and Resistant (C) (Ainsworth, Blehar, Waters, & Wall, 1978). As stated above, Securely attached infants are identified as using their mother as a "secure base"; they are able to seek out contact with her when needed, and then return to exploratory tasks. Insecure children have traditionally been described as exhibiting one of two patterns, or strategies.
A voidant children are seen as rde-emphasizing their affective needs for fear of being rejected by their caregiver. Their affective tone is neutral, and they often actively avoid proximity to mother at the moment of reunion. Resistant children show an ambivalent attachment pattern; they alternate between soliciting and rejecting contact with their parent Their affective tone is often negative, with petulance persisting through reunion episodes. The A voidant and Resistant attachment classifications are often referred to collectively as "Insecure attachment." Recently, Main and Solomon (1990) introduced an additional secondary category for classification of attachment behavior in the strange situation. The need for an additional category grew out of reported findings that up to 14% of samples were "unclassifiable" according to the three original categories (Main & Weston, 1981 Therefore, these children may exhibit an array of behaviors characteristic of the other three categories, but without the predicted order, consistency, or affective tone. Alternatively, they may display odd or bizarre behaviors, particularly in reunion situations. When faced with a stress on the relationship system, these inf ants exhibit unorganized or bizarre behavior. Since the availability of these criteria, attachment researchers have included the D category in attachment classification of both at risk and normative samples. It has been suggested that as much as 15% of normative samples will be classified as Disorganized (Cicchetti, 1987).
These classifications were originally conceptualized by Ainsworth as outcome variables, designed to examine the consequences of previously observed patterns of maternal behaviors in the home. It was hypothesized that a novel, stressful situation would trigger the infant's pattern of established responses. These behavioral patterns are presumably reflective of accumulated experiences in the ongoing relationship between the child and mother. Attribution of a classification then implies, but does not directly assess, the components of the current relationship or its precursors.
Recent emphasis on systems-based interpretation has led researchers to suggest that the study of a relationship dictates that both the qualities of the mother (sensitivity) and the infant (security) are manifestations of the same dynamic process (Pederson and Moran, in press).
The classification of attachment on the basis of the Strange Situation paradigm has become an accepted convention. The construct of attachment along with the operationally defined attachment behaviors, as observed in the Strange Situation, have been used almost exclusively to index the quality of attachment of a young child to his mother (Pederson et al., 1990), although this approach is associated with a series of limitations. For example, the constraints of a laboratory procedure with variable arousal effects on different children may result in unrepresentative, atypical activation of the attachment system, and consequently atypical response patterns (Vaughn et al., 1994). Recently, however, an alternative approach has been suggested and adapted by some researchers (Waters & Deane, 1985;Pederson et al., 1990;Vaughn & Waters, 1990;Vaughn et al., 1992;Pederson & Moran, in press). The Attachment Behavior Q-sort was introduced by Waters and Deanne (1985) as a way of addressing some of the limitations of assessment in a structured laboratory-based paradigm.
This Q-sort instrument consists of 90 items. Each item is a description of attachment-relevant behavior derived from theoretical and empirical work on attachment. Many items are qualified by specifying a context. These items are printed on cards to be sorted into nine piles according to similarity with the infant's behavior. The completed sort is then compared with the "criterion sort" (a composite sort of the prototypically secure child, as judged by a series of experts). The resulting correlation is interpreted as a continuous measure of the child's relative security of attachment. Some researchers believe that this instrument not only provides a useful complement to the Strange Situation, but also provides some methodological and theoretical advantages in the assessment of attachment (Seifer & Schiller, in press;Waters & Deane, 1985;Pederson et. al., 1990;Waters, 1990, Krupka, et al., 1992, Pederson &Moran, in press). Waters and Deane (1985) point out that a method of assessment closely related to the child's naturalistic environment is more in keeping with Bowlby's original conceptualization of the attachment system. Not only does it promote a more realistic assessment, but it also provides the opportunity to assess a fuller repertoire of attachment behaviors, observed in their natural context. For example, in the home, the child may normally feel safe to explore, to move away from his mother, only checking with her periodically. If a stressful, situation were to develop, however, such as an accidental fall, a loud noise, or the arrival of a stranger, the "safe" distance might be greatly reduced, and the need for more physical contact, rather than distance interaction more immediate.
In addition, a lab setting affects children differently. Based on previous experiences, or temperamental differences, children react quite differently on first entrance to a strange room. The Strange Situation paradigm may therefore prove intensely stressful to one child and not noticeably disturb another. To the extent that such differences affect the "triggering" of the attachment system, they may confound the behaviors observed and classified in the Strange Situation (Crockenberg, 1981;Kagan, 1982;Campos, Barrett, Lamb, Goldsmith & Stenberg, 1983;Vaughn et al., 1992;Vaughn et al., 1994;Goldsmith & Harman, 1994). It must be noted, however, that the arguments presented by some researchers that all relevant behavior observed and scored during the Strange Situation is can be entirely explained by infant attributes, i.e. temperamental style is probably overstated (Vaughn, Lefever, Seifer & Barglow, 1989;Seifer & Schiller, in press). These researchers found that while temperamental differences in children affected certain behaviors during the Strange Situation, such as amount of crying, such differences alone were not enough to account for security of attachment classifications.
Finally, Lamb et al. (1985), have suggested that too much valuable information collected through observation in the Strange Situation is ultimately lost by reducing outcomes to a three-category code. Some researchers have resorted to "converting" the categorical distinction into a continuous variable for use in correlational analyses (Cox et al., 1992). The Q-sort method, on the other hand, yields a continous measure of Security. Others have reported findings based on attachment behaviors observed during the Strange Situation, rather than classifications (Lyons-Ruth et al., 1987). These issues point to the usefulness of an alternative approaches to assessment of attachment.
Several points are important to consider when evaluating this body of literature. First, studies vary greatly in terms of design and assessment methods.
Procedures vary from lengthy home observations coupled with informal diary-like recordings, to relatively brief observation periods analyzed through time-sampling methods. In addition, behaviors of interest vary from free-play, feeding, caretaking , or "regular activities" chosen by the mother. Second, the operational definitions of sensitivity range from molecular behavioral counts to global four-point scales. Third, most studies (with two exceptions) are based on relatively small sample sizes.
Studies employing large samples (Egeland & Ferber, 1984; were inconsistent in the pattern of reported results in terms of the types of maternal behavior and length of observation periods that proved predictive of attachment classifications. Fourth, some researchers, (Goldsmith & Alansky, 1987) have pointed out that the often cited relationship between maternal sensitivity and attachment classification is not as large as assumed by some theorists (Ainsworth, 1978;Grossmann et al., 1985;Crockenberg & McCluskey, 1985;Benn, 1985). Their metanalysis of studies revealed a weak effect for the studies reviewed. Finally, a recent study  failed to replicate Ainsworth's original findings when repeated, extensive observations, and objective measures of sensitivity with demonstrated high reliability were used.
In the Schiller and Seifer study, repeated measures were used to address two major questions. 1. Can a reliable measure of maternal style be obtained using repeated observations over time? 2. Is there a relatively more stable or representative time period (6 vs. 9 months) in assessing the mother-infant relationship, and is one relatively superior in predicting 12 month attachment classification? It was found that through aggregation of 6 observations, a highly reliable measure of maternal sensitivity with strong stability over time could be obtained. The age of assessment did not affect the relation of sensitivity with attachment outcomes. Interestingly, while maternal sensitivity at 6 and 9 month did not relate to Strange Situation (Ainsworth & Witting, 1969) attachment classification, a positive relationship was observed with the Attachment Q-sort (Waters & Deanne, 1985).
These results raise questions not only about the assumptions regarding linear relations between sensitivity and attachment, but also about the methodological implications of assessing each of these constructs.
The mechanisms by which such factors affect relationship development have not yet been specified, but attention has turned toward examining contextual effects on mothers' ability to effectively parent. The role of social support, for example has been seen as a significant contributor to attachment outcomes (Crockenberg, 1981;Sroufe, 1985). Crockenberg (1981) considered maternal responsiveness, infant irritability and social support as important contributors to the developing motherinfant relationship. She found that social support was the single most significant predictor of attachment status. Of additional interest, was the interaction between infant irritability and social support. It appears that for mothers of highly irritable babies social support was especially important in determining the quality of attachment. These findings point to transactional influences among the child, the mother and the environment, rather than purely linear determinants of attachment based on maternal traits or behavior.
Other researchers have focused more on distinguishing the relative differences in style and behaviors of mothers of Secure versus A voidant versus Resistant babies in helping to understand the process of attachment formation. It has been suggested that maternal style of interaction, falling on a continuum according to level of stimulation, can serve as a differentiating measure .
These authors found mothers of Securely attached infants to demonstrate an "intermediate" level of interaction, as compared to the "overstimulation" and "neglect" that characterized the interaction style of mothers of Avoidant and Resistant infants, respectively. Egeland and Farber (1984) describe stable differences among these three sets of mothers in slightly different terms. Caretaking abilities, including general knowledge, timing, and responsivity, as well as maternal feelings and attributions about motherhood were considered. As expected, mothers of Securely attached babies, were most appropriately responsive, while mothers of A voidant babies were characterized as "indifferent" and "unavailable." Mothers of Resistant babies suffered more from lack of awareness, than lack of interest, but also failed to provide sensitive care.
Other researchers (Mangelsdorf et al., 1990;Tronick, 1989) demonstrated that maternal characteristics, such as affective state influence the interactive style of the mother, the degree to which she can be available, sensitive and responsive.
To briefly summarize, in exploring the pathways to sensitive and insensitive parenting, many aspects of caregiver characteristics have been studied. More recent efforts have focused on identifying the mechanisms by which such individual characteristics are translated into behavioral or interactional patterns. The role of social support, infant attributes and the interplay of these factors has resulted in more comprehensive and useful models.

Family Variables
Considering the role of social support and context further, researchers have begun exploring the effects of marriage on parent-infant relations, and attachment specifically. Positive relations have been found between marital functioning (i.e. marital satisfaction, conflict resolution, and level of emotional closeness) with quality of parenting and attachment (Jacobson & Frye, 1991;Howes and Markman, 1989;Goldberg and Easterbrooks, 1984). Jacobson and Frye (1991) and Valley et al., (1992) found significant positive relationships between maternal social support, especially the presence of an intimate relationship, to sensitivity and attachment outcomes. The link between marital quality and attachment has been studied directly (Cox, Owen, Lewis, and Henderson, 1989;Goldberg and Easterbrooks, 1984;Howes and Markman, 1989) establishing the importance of marital adjustment and satisfaction as a significant variable in terms of infant attachment outcomes.
More globally, researchers concerned with understanding developmental outcomes in children often call attention to the role of the family (Stevenson-Hinde, 1990;Marvin, 1992;Schachere, 1989;Minuchin, 1985). Moran and Pederson (1992) for example describe a model where the global context of the family, including social, financial and marital circumstances are translated and communicated by the primary caregiver to the developing inf ant. Theoretical application and integration of the family systems and attachment approaches has been the focus of recent writings as researchers recognize the compatibility of these frameworks and the utility of adaptation of the family system approach to the study of attachment (Stevenson-Hinde, 1990;Marvin, 1992;Schachere, 1989;P. Minuchin, 1985). Family systems theory is based on the assumption that all dyadic relationships occur and exist in context; they are both created and maintained by the larger family system. Therefore, the study of any given dyadic relationship must take into account the forces responsible for its upkeep. As an example applied to the mother-infant relationship, one possible scenario could be considered: A mother receiving insufficient support from her spouse may feel both anxious, lacking in confidence , and frustrated by unmet emotional needs. This emotional state may become expressed through a variety of undesirable interactive patterns with her infant such as over-, or underinvolvement, anxious, or inconsistent style. This in tum may result in heightened or anxious attachment behaviors on the part of the infant, who may be unsure of what to expect when intensifying his signals for attention. An already frustrated parent may then feel progressively overburdened by these demands, becoming more needy in relation to her spouse who in response to increasing demands may also feel overburdened and withdraw further . Into this already complex cycle, one must add the inherent attributes of all relationship patterns , such as personality and relationship histories of the adults and temperamental attributes of the child. No doubt a "difficult" child would intensify his signals with greater intensity in the above scenario, contributing to a more profound relationship and familial disturbance than an "easy" child.
In a comprehensive review of the literature examining the relation of attachment and maternal employment, Schachere (1989) outlines potential mechanisms for differential outcomes within the framework of family systems. She argues that the vulnerability of a child to insecure attachment is affected not simply by maternal employment status, but by the meaning and structure surrounding this choice which is dictated by overall family dynamics. How does the family respond and cope with mother's schedule and availability? Is there a pattern of resentment in the marriage translated into dyadic interactions with the child, or does mother's career bring pride and structure to the family environment? Is increased paternal involvement s.een by the family as an asset or a burden? In asking these bidirectional, rather than linear questions the strengths of family systems can be effectively applied in increasing our understanding of long-standing developmental and relational issues.
In spite of this growing interest in family influences and systemic approaches to the study of development, direct assessment of overall family functioning has been missing from attachment research. Systemic models of family functioning have been developed and successfully applied to research settings, (Miller, Bishop, Epstein and Keitner, 1985;Keitner, Miller, Epstein, Bishop, Fruzzetti, 1987;Epstein, Baldwin, andBishop, 1983, Sameroff, 1988 There is an interchange of its internal subsystems, (the dyad, the individual) and external systems, (the school, community, and work) (Epstein, Bishop, and Levin, 1978;Epstein and Bishop, 1981). As in other models based on systems theory, (Minuchin, 1974), the emphasis is on the family unit and the transactional patterns that are believed to shape the behavior of all members, rather than on any given individual's characteristics.
The  (Epstein, Baldwin, and Bishop, 1983). This study used both methods of family assessment to gain an index of family functioning for the participating families.
Placing the mother-child attachment in the context of the family led to better understanding of this important and complex phenomenon.

Summazy and Research Questions
Achievement of secure attachment in infancy is considered an important and favorable developmental process. It represents an adaptive response from the ethnological and developmental perspectives. It is associated with numerous positive developmental outcomes in cognitive and socio-emotional development. Maternal sensitivity has been postulated as the most significant contributor to positive attachment outcomes, but the definition and assessment of this construct contributed to methodological limitations in many studies examining this relationship. Recent work  showed that maternal sensitivity can be reliably and consistently measured through aggregation methods.
Assessment of attachment has been accomplished to date almost exclusively through a structured lab paradigm. Recent evidence indicates that a more naturalistic approach yielding a continuous (rather than a categorical), measure of attachment may be preferable in accurately capturing the attachment system.
In addition, the family context is often speculated as an important contributor to the developing relationship between infant and mother. The family system has not been adequately studied in this context. This study addressed these issues by examining the following research questions: 1. How are attachment assessment methods related?
By utilizing the two most widely used approaches to assessment of attachment, we were able to compare the Strange Situation and the Q-sort methods.
Further, both trained observers and mothers completed the Q-sort measure allowing for comparisons and exploration of reporter-bias and associated methodological implications.
2, How is maternal sensitivity related to attachment outcomes?
Earlier work ) demonstrated that (a) reliable and stable measure of sensitivity may be obtained through aggregate methods, and (b) maternal sensitivity as assessed in the home at 6 months is not related to Strange Situation outcomes but is positively related to Q-sort measures of security. This study aimed to replicate these findings on a larger more diverse sample.
3a, How is family functionin~ related to attachment? 3b, How do family and marital functionin~ variables mediate the sensitivityattachment relationship?
To help place the mother-infant relationship in broader context, measures of family and marital functioning were administered in this study. Their relation to both sensitivity and attachment outcomes was explored.

Method Subiects
Fifty-one mother-child dyads participated in this study. These subjects were recruited from those who have already completed their participation in a larger study (N=120) of infant temperament at the Bradley Research Center (Seifer, Sameroff, Barrett, & Krafchuk, in press). The Infant Temperament Study was a short-term longitudinal project, studying infants between 4 and 6 months of age. The project involved a variety of procedures and instruments. Only those components which are directly applicable to this study will be described. This sample was chosen because six home observations were made when the infants were 4 to 6 months of age. These subjects were also recruited for participation into one of two studies of family relationships at The Bradley Research Center (The Providence Family Study; or the current study).

Initial Recruitin~ Procedures:
Recruiting was done for the Temperament Study at Women and Infant's Hospital, Providence, RI. A trained research assistant (often the author) screened medical records to identify families who met the following criteria: subjects were healthy newborns, not requiring stay in the intensive care nursery; and were born to families living within driving distance from Providence. Special efforts were made to recruit a diverse group of subjects, representing a variety of racial, SES, marital status, and birth-order factors. Descriptive characteristics of the sample are summarized in Table 2. As can be seen in the Table, and illustrated in Figure 1, although all socio-economic classes were represented, the sample can be best described as largely middle to upper-middle ·class.
Insert Table 2 and Figure 1 here After initial screening, the research assistant approached mothers during the lying-in period at the hospital. The Infant Temperament Study was explained to them in some detail. Those that expressed interest received a one-page description of the study to review with their families. They also signed a consent form granting the research staff permission to contact them by phone when the infant was two to three months old. Every effort was made at the initial recruiting phase, as well as during follow-up phone contact, to include families who were able and willing to participate in weekly home observations and related procedures as part of an extensive longitudinal study. As a final recruiting step, after the 2-3 month follow-up call, the research assistant made an initial home visit to the participating family. Procedures and questionnaires were further explained and informed consent was signed by those families who eventually agreed to participate.

Post Temperament Study Recruitin~ Procedures:
Phone contact was made by the project coordinator of the family studies with only those families who had indicated a willingness to be re-contacted at completion of the Temperament study. Details of study procedures, compensation, and length of participation were explained. Mothers expressing interest in participation, were then contacted to arrange scheduling of initial visits. Informed consent was reviewed and signed by the parent. 6 month Assessment: Home Assessment of Sensitivity: Procedures Six home visits on consecutive weeks were made by a research assistant to the participating families to videotape naturalistic observations in three situations: child playing alone; child playing with his/her mother; and caretaking activities. The same research assistant visited a family throughout the 6-week study. She brought with her a small videotape camera, and a standard set of five, age-appropriate toys. Mothers were not specifically instructed how to play with their infants or whether to use the toys provided. They were informed, however, that a minimum of ten (10)    high reliability (intraclass correlations ranging from r= .78 tor= .88) was achieved by aggregating across six observations. In addition, in the earlier study, 6 and 9 month data were examined. Results showed that both age levels were equally unrelated to attachment outcomes in the Strange Situation, but were similarly related to attachment security as evaluated by the Q-sort, with the 6-month data slightly more strongly related. For the purposes of this study, therefore, the availability and analysis of only the 6 month age level was deemed optimal.
Scoring of maternal sensitivity was also completed as in our previous research. The first 10 minutes of the mother and child playing together were reviewed and scored by one of the trained and reliable raters. The remaining two situations of behaviors (child playing alone and caretaking) were not scored. There were two main reasons for this decision. First, mothers are often not visible on tape during caretaking and play-alone episodes. Second, the chosen coding system for maternal sensitivity, The Parent /Caregiver Involvement Scale (PCIS), was developed by Farran et al. (1986) specifically for periods of interactive play between a mother and her young child (See Appendix I for sample score sheet). Other reasons for employing this scale were that reliability and validity information was available and satisfactory; the scale has been used in three previous research projects, including a longitudinal study of young children; (Farran et al., 1985;1987) ratings were made on a Lik:ert scale based on specified behaviors observed during the course of the interaction. Such well-defined, yet global ratings (as compared to time-sampled coding of specific behaviors) have been recommended as the superior method for assessing individual differences in patterns of behavior such as maternal sensitivity (Jay & Farran, 1981;Cairns & Green, 1979;Waters, 1978).
Home Assessment of Sensitivity: Reliability and Variables Used: Rater reliability was established before final scoring began. Raters (the author and two additional raters) were reliable and experienced in the use of the Parent/Caregiver Involvement Scale (PCIS, Farran et al., 1986) from our previous research projects. However, due to slight variability of method and sample, as well as passage of time, a new training and reliability verification was undertaken. Minor adjustments in the coding procedure were instituted as a result. For example, as this sample included first-born as well as later-born infants, where as our previous sample consisted exclusively of first-borns, the effects of older siblings on the dyadic motherinfant interaction needed to be considered. Appropriate decision rules were applied and added to the coding scheme. Subsequently, raters completed a set of 10 tapes.
Acceptable reliability was demonstrated on subscales as well as summary variables.
Reliability was calculated using intraclass correlations and values exceeded r=.80. As this level of reliability is sufficiently high, the home observation tapes were scored by a single rater, with any unusual or difficult tapes reviewed by another rater (usually the author) for reliability checks. and Appropriateness was made on a five point scale. The ratings were then summed so that data reduction yielded 3 summary variables for each mother-child dyad observed at 6 months. These variables were: Amount of maternal involvement (AMNT), Quality of maternal involvement (QUAL) and Appropriateness of maternal involvement (APPR). Descriptive information for these variables (means and standard deviations) is provided in Table 3. Correlations among these three scales are presented in Table 4.
Insert Tables 3 and 4 here As in our previous research, the three dimensions of the scale were found to be highly interrelated, with correlations ranging from r=.38; p<.05 to r=.97; p<.001.
As expected from previous findings, correlations of measures of Quality and Appropriateness of maternal involvement were very closely related to each other, and relatively less related to Amount of involvement. This pattern of results is to be expected, because Amount refers to the frequency of maternal response, where as Appropriateness and Quality are meant to assess the relative sensitivity with which these responses are delivered.
As discussed above, there were 6 observation points for each dyad around the 6 month age level. Average week-to-week intraclass correlations were examined for each summary variable . While measures of Quality and Appropriateness of maternal involvement were relatively stable (r=.38) as expected from previous research, measure of Amount of involvement was quite variable from week to week (r= .00).
Therefore, in all further analyses only variables QUAL and APPR were considered.
As can be seen in Table 5, six-week aggregation of these two variables improved their reliability to r=.79 and r=.80, respectively resulting in a highly reliable measures of maternal sensitivity .
Insert Table 5 here 12 month Assessment: Home Assessment of Attachment: Procedures and Reliability Each family was visited by one or two observers for a three hour naturalistic observation period. The observation period was scheduled at the family's convenience when mother and child were together and the child was awake . The Attachment Q-sort was used to assess home-based attachment behavior (Waters & Deanne, 1985). For establishment of reliability, the author visited each family with one of three additional raters. Observers remained silent or minimally interactive with mother and child during the first two hours of the observation period. observers (Krupka, et al., 1992) to 6 hours with 2 observers (Vaughn and Waters, 1990). Jacobson and Frye (1991); Moran et al., (1992) report 2-3 hour observation periods with 1 observer yielding adequate samples of behavior necessary for the reliable completion of the Q-sort ratings.
Each observer sorted the 90 behavioral statements (see Appendix II) into nine piles (10 statements each) according to how closely each statement represented the usual observed behavior of the child. The completed sorts were then correlated with the sort of the prototypically secure child to generate a relative security of attachment rating for each child. The criterion sort (see Appendix II) was developed by aggregating completed prototype sorts generated by eight expert judges (Waters & Deane, 1985).
Participating mothers were also asked to complete the Attachment Q-sort at the time of the home visit After 2 complete hours of observation, the observer explained the instrument to the mother and remained available for questions as the mother completed her sort. She was also instructed to respond to the child in her usual way while completing her sort. The inclusion of this procedure had several benefits. First, maternal report, used in conjunction with objective observer ratings was designed to provide a more complete view of the child. Second, it enabled the comparison between observer and maternal ratings on the same instrument. Finally, the insertion of a time period where maternal attention was split between her child and a specific task (Pederson et al., 1990) provided the observer with ample opportunity to witness behaviors relevant for the Q-sort such as maternal prohibitions, child's bids for contact and attention, and child's independent play .

Home Assessment of Attachment: Data Reduction
The Attachment Q-sort yields 2 continuous summary variables of relative security and dependency (QSEC and QDEP) of the child in relation to his mother. In original development of the Q-sort, a measure of dependency was included so as to distinguish secure-base attachment behaviors from a more global neediness exhibited by a child (Waters & Deanne, 1985). In this project both maternal and observer sorts were examined thus, a total of 4 summary variables of Q-sort attachment were considered. Descriptive information for these variables is presented in Table 6 and intercorrelations in Table 7. As can be seen in Table 7, measures of security and dependency were not related in maternal report, (r= -.19), but were moderately related in observer report (r=.33; p<.05). When considering mother-observer agreement , as expected, moderate relations emerged for security and dependency (r= .28; p<.05 and r= .32; p<.05 respectively).
Insert Tables 6 and 7 here Laboratory Assessment of Attachment: Procedures and Data Reduction The 12 month laboratory visit consisted of several protocols, but only the Strange Situation is relevant for this study. The Strange Situation (Ainsworth & Wittig, 1969) is a structured 23-minute paradigm involving a series of separations and reunions of the mother and child, in addition to episodic interactions with an unfamiliar adult ("stranger"). The procedure consists of eight brief episodes: 1) Experimenter brings mother and infant to playroom -30 sec.; 2) mother and infant alone -10 minute free play; 3) stranger enters and sits quietly, talks to mother, engages child in play -3 min.; 4) mother leaves, stranger is left in the room with infant 5) mother returns, stranger leaves, free-play -3 min., 6) mother leaves infant alone in play room -3 min.; 7) stranger returns, attempts to comfort baby if necessary -3 min.; 8) mother returns, stranger leaves, free play -3 min. Each of the last 6 episodes is 3 minutes. If the infant is distressed during the separation episodes ( 4, 6, 7) the episode is shortened to allow mother's return.
The Strange Situation (Ainsworth et al., 1978) was used to assess laboratorybased attachment security. A trained and certified rater of strange situation attachment was responsible for scoring all 51 tapes. This rater was independent and "blind" to both sensitivity and Q-sort attachment observations.
Specified behaviors are coded from videotapes that subsequently yield a security of attachment classification for each child (Ainsworth, et al., 1978). (Please see Appendix ill for reference for full scoring criteria). The behavior scales scored are: (1) Proximity/Contact seeking of the child which captures the intensity and persistence of the child attempts to be in physical contact with mother, (2) Contact Maintaining, which refers to the child's degree of activity and persistence in remaining in physical contact with mother, (3) Resistance, as expressed through angry behaviors such as pushing away, throwing, kicking, batting, squirming out of contact, etc., and (4) Avoidance which is exemplified through moving away, leaning away, turning away, hiding from the mother, or ignoring of bids for attention. Behavior that the child wishes to be in physical and interactive contact with his mother, especially in reunion episodes. The child may or may not be distressed at separation, but if distressed, he is able to go to his mother and quickly become consoled on reunion without substantial anxiety or avoidance. Group (C) infants are exemplified by resistance, or ambivalence to contact and interaction with the mother. Family Assessment: Self-Report: Of the 51 participating mothers, 48 completed the Family Assessment Device (FAD, see Appendix VII, Epstein, Baldwin, and Bishop , 1983), . This 60 item selfreport instrument closely parallels the MCSIFF Interview yielding scores on each of the six dimensions of family functioning as well as a Global Functioning score. The scores on the FAD items range from 1 (very healthy) to 4 (very dysfunctional) .
While both approaches assess the same dimensions of family functioning, they represent different perspectives. The FAD measures the family's perception while the Clinical Rating Scale (CRS) completed by a trained clinician following the completion of the MCSIFF represents a clinician's view. These instruments have been successfully used in previous research with families. Miller et al., (1985) and Epstein et al., (1983) report high reliability and validity estimates for the FAD.  (Archambault, 1992;Keitner, et al., 1992). The FAD has also been shown to have moderate correlations with other family assessment self-reports, demonstrating adequate concurrent validity.
Marital Satisfaction: Self-Report: Forty-eight mothers completed the Dyadic Adjustment Scale (DAS, Spanier, 1976, see Appendix VIII), to assess marital satisfaction. The DAS is a 32 item questionnaire assessing Dyadic Satisfaction, Cohesion, Consensus, and Affectional Expression. The overall Satisfaction score was used in this study. Moderate to high correlations between the DAS and FAD have been previously reported (Dickstein, et al., 1992).
Measures of Marital and Family Functioning: Data Reduction, Reliability As described above, the two measures of family functioning (MCSIFF and FAD) each yield scores on 6 dimensions of family functioning and a summary variable of General Functioning. Thus, the same dimensions are described through self-report and clinical interview methods. It should be noted, however, that the scales of the two instruments are directionally reversed, that is, higher scores on the MCSIFF represent healthier functioning while higher scores on the FAD indicate relatively more disrupted functioning. For the FAD, "healthy functioning" is described as falling below the cut-off scores which range from 1.9 for Behavior Control to 2.3 for Roles dimensions . For the MCSIFF, "healthy functioning" on all dimensions is characterized by scores equal to or exceeding 5. Marital satisfaction was derived from a single DAS variable which represents Overall Marital Satisfaction as reported by mother. For this variable the cut-off score is equal to 100, with higher scores signifying relatively greater marital satisfaction. See Table 8 for descriptive information (means and standard deviations) of these variables and Table 9 for intercorrelatonal information.
Insert Tables 8 and 9 here Table 9 shows characteristics of these family measures in this study. As in previous research ( Miller, et al., 1985), moderate to high correlations were observed between the MCSIFF and FAD scales, with 5 out of 7 corresponding scales being

Results
This study addressed four major issues .
(1) In exploring the relations between attachment and related or contributing constructs, consideration of assessment methodology of attachment itself seems worthwhile . Administration of the two main methods of attachment assessment, conducted both in the home and in the laboratory, and drawing on both the expertise of maternal report as well as the objectivity of trained observers, enabled such comparisons.
(2) Extension and replication of earlier findings  where maternal style as assessed through reliable, aggregated observations in the home was found to be related to Q-sort Security, but not to strange situation Security classification was undertaken . Replication efforts were necessary as earlier work was conducted with a relatively small and homogeneous sample .
(3) Lacking in earlier work, as well as in the field of attachment research in general, is the empirical evaluation of the contribution of the family system to the developing relationship between mother and child . This study aimed to explore relations between family functioning, marital satisfaction, and attachment. Both self-report and interview measures were used . (4) The relation between sensitivity and family and marital functioning as related to attachment outcomes was explored.
Relations Among Attachment Measures: The four indices of Q-sort attachment were as follows: (1) Q-sort Dependency as rated by mother (QDEPM), (2) Q-sort Dependency as rated by observer (QDEPO), (3) Q-sort Security as rated by mother (QSECM), and (4)  Security. Ratings of Dependency were also not significantly related to Security in the Strange Situation r= -.00 (ns) for maternal report and r= .23 (ns) for observer report.
Insert Table 10 here The means for each strange situation attachment group were examined using a one-way ANOVA procedure; results are summarized in Insert Table 13 here Multiple regression analyses were then conducted to further explore the relation between maternal sensitivity and observer reported security. Quality and Appropriateness variables were entered together in a multiple regression equation, as these predictors of attachment were highly correlated measures of sensitivity.
Together, these variables accounted for 46%; (p<.05) of the variance in the Q-sort security rating as judged by an observer. This finding indicates that there is no additive gain in predicting power when sensitivity variables are considered together, as would be expected when the predictors are highly correlated (r=.97, see Table 4).
Family Functionin~ and Attachment: In analyses of family functioning as related to attachment, global summary variables were of most interest. These were mother's rating of Marital Satisfaction All measures reflect data collected around the child's 12 month age level, concurrent with measures of attachment Strange Situation: Correlational analyses were done to examine the relation between family measures and Q-sort attachment. These findings are summarized in Table 15.
Maternal report of Security was significantly related to self report measures of Marital .11; ns, and r=.10; ns for summary variables of interview Family Functioning, selfreport Family Functioning and self report of Marital Satisfaction, in that order).
Measures of Dependency were unrelated to any of the Family Functioning or Marital Satisfaction variables for either mother or observer generated ratings.
Insert Taken together with above reported results of maternal sensitivity as relating to observer reported Security of attachment, only the variables of interest were combined in a hierarchical multiple regression. These were: Quality and

Appropriateness of maternal involvement and the Global variable of interview-based
Family Functioning. As above, Quality and Appropriateness were entered in a single step to the equation. They were entered in alternating order with the Global measure of Family Functioning to determine amount of independent variance explained in the outcome measure of Q-sort security based on observer report. It must be noted that these analyses were performed on a smaller sample (n=24), as MCSIFF data were available for these families. In equation 1, Quality and Appropriateness were entered at Step 1 and the Family Functioning score entered in Step 2. In equation 2, the Family Functioning score was entered at Step 1, with sensitivity variables added at Step 2. These findings are presented in Table 17. Results show that when Family Functioning was partialed in the second equation, (entered on step 1 of the regression) maternal sensitivity variables were still significantly related to Q-sort Security (change in R squared= .29; p<.05). However, when maternal sensitivity was partialed, in the first equation, Family Functioning no longer related to Security (change in R squared= .00; ns).
Insert Table 17 here As family variables were not significantly related to Strange Situation Security, similar analyses were not undertaken with this outcome variable.

Discussion
This study aimed to address the following major questions:

How are attachment assessment methods related?
Two main methods of assessment of mother-infant attachment, the Strange Situation (Ainsworth et al., 1978) classification system and the Attachment Q-Sort (Waters & Deanne, 1985) were used and compared. Additionally, the Q-sort measure was completed independently by both mothers and trained observers to allow for methodological comparisons.
As expected, mother and observer reports of Q-sort attachment were moderately related. Parental report reflects extensive experience with a child, while observer reports are based on relatively short observation periods, but are aided by training and experience with the instrument as well knowledge of nonnative behavior of children in a given age group. Previous research has shown that higher correspondence between mothers and observers may be achieved , but only through an extensive period of parent "training" (feti & Mgourty, 1994). In the current project standard procedures of explanation, brief practice and availability for mother's questions by a trained observer were used, which approximated the extended "trainng" protocol. Given these methods, it appears that parental and observer ratings coincide in terms of the more robust behaviors of the child, but diverge in describing the more subtle behaviors.
Comparison of the Strange Situation, a structured laboratory measure, and the home-based naturalistic observation of the Q-sort yielded less expected findings. In our previous research,  as well as in the validation and development of the Q-sort, moderate correlations between the two methods were found. In our previous work, the two methods of assessing attachment Security were significantly and positively related. In the current sample, however, no such relations were discovered. Several explanations for these findings may be considered. First, in this study, infants were classified into the traditional A/B/C groups as well as the newly developed D classification. This grouping was not used in the original validation, or in subsequent comparisons of the two methods Waters & Deanne, 1985; Vaughn & Waters, 1990). In fact, in Vaughn & Waters' validation study the Q-sort measure was documented as successfully distinguishing Secure vs. Insecure infants but failed to discriminate among the three attachment classifications. Second, this sample, as mentioned above, contained a significant proportion of D-classified infants (see Figure 2) which may have contributed to less typical sample distribution of attachment-relevant behavior, as described by the Q-sort. Third, the emphasis of the two methods is distinct. The Strange Situation aims to tap the aroused attachment system, focusing on the mechanics of separations and the reparative work the dyad engages in during reunions. The Q-sort, on the other hand, describes the relatively non-stressed, naturalistic, "secure-base" behavior of a child in relation to his mother.
This difference in emphasis may in part account for the lack of correspondence between the two methods. It may also help to address the unexpected finding that Qsort measured dependency, not security, discriminated among the Strange Situation groups, with higher dependency evident in the secure group of infants. Infants who are potentially more reactive and expressive in their attachment-reunion behavior may also be the ones to demonstrate a higher level of general dependency in a less stressful environment. Finally, the possibility of sample-specific aberration must be considered. As described above, efforts were made to recruit a representative sample, but a sample of fifty-one dyads may present somewhat atypical characteristics. Such characteristics may not have been directly controlled in the study, but may have expressed themselves in attachment-related behaviors.
Rather than attributing one method with a preferential status in terms of assessment accuracy, these results suggest that the two ways of approaching the complex construct of infant attachment may provide complementary, if not overlapping points of view.

How is maternal sensitivity related to attachment outcomes?
In addressing this question this study aimed to replicate earlier work . In several ways these findings were comparable to the earlier project. First, the method of assessing maternal style was once again shown to be valid and reliable. The Parental/Caregiver Involvement Scale (PCIS) was again used to assess maternal behavior in repeated weekly observations as in our earlier work.
Inter-rater reliability was established to anticipated levels. Aggregated measures (over 6 weekly sessions) also proved highly reliable for the two qualitative measures of maternal style. Unlike the earlier study the quantitative measure, Amount of maternal involvement , proved to be more variable from week to week, resulting in unacceptable aggregated reliability levels. This variable, which was thus excluded from subsequent analyses, had not been an important predictor in earlier work. The implications of this finding are significant from a methodological perspective, as many studies examining maternal sensitivity and attachment often employ undifferentiated qualitative and quantitative ratings, often in a single observational period (See Tables 1-A and 1-B). The current findings imply that such methodology may lead to unstable observations and conclusions that vary from study to study.
Second, methodological variability in the literature, as reviewed in Tables 1-A and 1-B, prompted both the earlier and current study of these issues. In the present study, as in the earlier project, careful attention was given to observer bias, reliability of measures, and assessment procedures. Attachment classifications were done by a trained and reliable independent rater. Both measures of attachment and sensitivity were carefully selected and appropriate levels of reliability were reached before any independent rating of either construct was undertaken. Maternal sensitivity ratings were based on a theoretically and methodologically sound assessment method combined with repeated assessments in the home to produce a highly stable and reliable instrument. Q-sort ratings were done by independent observers, unfamiliar with both sensitivity and Strange Situation ratings. Sensitivity ratings were made based on videotape review by a separate set of independent raters. (Note: the author, although involved in both Q-sort and sensitivity scoring, was never responsible for both assessments of a given family). This level of methodological consistency and scrutiny allowed for careful replication of earlier findings and more definitive answers to the questions posed.
Third, in this study, as in earlier work, a strong relation was found between maternal sensitivity and Q-sort attachment Security, as rated by an observer. These findings are particularly interesting given that maternal sensitivity is assessed at the 6 month age level and attachment Security around the 12 month age. Many studies reviewed (see Tables 1-A  Several possible explanations, although speculative in nature, may be considered as explanations for these findings. First, the relatively higher incidence of D infants in this sample (20 % as compared to the 10-15% which has been suggested for normative samples, Cichetti, 1987) may contribute theoretically and empirically new and unexplained variance to an otherwise normative sample. As previous research documenting association between D classification and abusive parenting was based on high-risk samples (Lyons-Ruth et al., 1987Cicchetti, et al., 1987), expectations for normative samples remain unclear. It may be worth considering, for example, alternative pathways for arrival at a Disorganized pattern of attachment-relevant behaviors within a structured setting for children from abusive versus unremarkable parenting backgrounds . Individual infant characteristics, as well as the constraints of the laboratory procedure itself need to be considered. It has been suggested, for example, that infants vary in the level of arousal experienced within the Strange Situation paradigm. Some have speculated that it may not be applicable to infants with atypical fear, or anger thresholds (Goldsmith & Alansky, 1987;Campos, Barrett, Lamb Goldsmith & Stenberg, 1983). It may also be possible then, that infants from normative backgrounds demonstrating unusual behaviors in the Strange Situation are especially affected by its demands. These children may in fact be more reactive, more needy, and more accustomed to highly responsive care from their mothers. These children may then appear in some ways similar to those suffering from long-standing histories of abuse only within the specific constraints and 47 expectations of a structured paradigm designed to elicit stress and arousal , but these similarities may be superficial, reflective of distinct precursors.
Second, marginal findings with regard to maternal sensitivity and attachment classification are consistent with our previous work as well as suggestions by several researchers that the often cited relationship is not as robust as may be expected (Goldsmith & Alansky,1987;Rothbaum & Rosen, 1993). Many studies (see Table 1-B) report mixed or partially supportive findings of the theoretically important, but empirically controversial relationship.
Given the methodological rigor of this study and the overall consistent pattern of findings with previous work, we conclude that maternal sensitivity, as measured through repeated reliable observations is related to attachment outcomes. The method of attachment assessment chosen may influence the strength of the observed relationship. In other words, attachment, as assessed in a naturalistic setting over a relatively longer observation period was clearly related to similarly assessed maternal style. The emphasis and demands of a structured laboratory paradigm, however produced less robust associations with home-based assessments of sensitivity.

How is family functioning related to attachment?
In analysis of the relation of family functioning to attachment, a similar pattern emerged. The Q-sort measure of Security was, as predicted, related to directly observed, but not to self-reported measures of Family Functioning. In other words, observer-reported home-based ratings of attachment Security were related to Family Functioning as assessed in a clinical interview. Such relations are consistent with theoretical predictions regarding family environments as contexts and perpetuating mechanisms for developing relationships (Schachere, 1990;Stevenson-Hinde, 1990;Goodman, Brogan, Lynch & Fielding, 1993). No comparable relationships were detected with Strange Situation classification, or when only the dichotomous variable (Secure vs. Insecure) was considered. The families of Securely and Insecurely attached infants did not notably differ from each other.
These latter findings are not only counter-intuitive, but also difficult to reconcile with current theoretical thinking about the nature of relationships in context Several explanations need to be considered. First, a limitation is noteworthy, in the interpretation of these results. Only a smaller subsample of families participated in the family interview, and thus comparison for this variable among attachment groups must be interpreted with caution, and considered exploratory in nature. Second, methodological differences between the attachment methods compared need to be considered. As mentioned above, the Q-sort and the Strange Situation varied in period, type and emphasis of observation. Their ratings may then yield related, but not synonymous representations of the attachment construct.
The specificity of the Strange Situation assessment, tapping the stressed attachment system under structured observation, may be less related to the overall context or ongoing patterns of the developing relationship than the home-based naturalistic focus of the Q-sort. Specifically, the advantage of observing a child in a home setting over several hours is the opportunity to witness the ways in which "secure base behavior" is balanced with exploration, how comfort is sought and addressed in relatively unstressed surroundings, and how normative patterns of interaction unfold. This observation of daily life may be closer in spirit to those constructs such as Affective Involvement, Roles, Communication which are assessed in the Family Functioning interview. As described above, the Q-sort measure was also found to be positively and significantly related to maternal style, as observed in the home. These theoretically consistent links were found only with observergenerated reports, however. Again, direct observation of a child's attachment behavior was found to be related to similarly observed maternal sensitivity six months earlier. This type of observation yielded distinct information from that provided by mother, even when the same instrument (the Q-sort) was used.
Based on these findings, the Q-sort, when conducted by trained observers, as compared with the Strange Situation, appears as a relatively more impressive method in terms of providing theoretically consistent links to related constructs of both sensitivity and family relations.

How do family functioning variables mediate the sensitivity-attachment relationship?
Similar to the pattern of findings outlined above, maternal sensitivity as observed at 6 months in the home, was related exclusively to interview-based measures of Family Functioning, not self-report. Further, self-report measures of Family Functioning were related to self-report measures of attachment while interview-based, or observed measures of Family Functioning were related only to observer-reported attachment Security. The impact of methodological choice seems to have played a significant role in determining the results. It appears that mothers and observers provided distinct, but internally consistent representations of these interrelated constructs-sensitivity family and attachment.
To summarize, (1) As expected, Family Functioning was found to be related to both sensitivity and attachment outcomes, although Family Functioning was not independently predictive of attachment if the effects of sensitivity were statistically controlled.
(2) Method of assessment played an important role both in determining outcomes and dictating interpretations.
(3) Although Family Functioning did not explain unique variance in attachment outcome, its relation to maternal sensitivity suggests a mediating role in the sensitivity-attachment relationship. Researchers have suggested several pathways for this effect. Crockenberg (1981) for example has proposed the impact of social support on a mother's ability to provide sensitive responsiveness to her child. The impact of positive marital relationships, both as a means of addressing mother's own needs and as a source of social support has also been cited (Goldberg & Easterbrooks, 1984;Sroufe, 1985;Jacobson & Frye, 1991). More globally, a family atmosphere where the emotional needs of members are given adequate attention and priority is in all likelihood a conducive environment for promotion of sensitive dyadic interaction.

Conclusions and Future Directions for Research
Five major conclusions from this study can be made: (1) Two currently available and widely used methods for assessing infant attachment (The Q-sort and the Strange Situation) may be tapping related, but not largely overlapping aspects of the attachment construct.
(2) Home-based, repeated, and aggregated observations of maternal sensitivity produce a reliable and theoretically meaningful measure.
(3) Maternal sensitivity and attachment were related. Stronger relations were demonstrated with the home-based Q-sort attachment Security measure than with Strange Situation classifications.
(4) Family variables were found to play an important role in the relation of attachment and sensitivity, particularly when home-based attachment measures were combined with interview-based assessments of Family Functioning. Similarly, in selecting a method for attachment assessment, rather than seeking the "better", or more accurate one, it may be more useful to first decide which component of this complex and dynamic process is under investigation and then select the method accordingly.
Future research will need to also focus on several issues raised, by the findings of the study, but to date inadequately addressed in the literature. Specifically, expectations for classification and interpretation of disorganized (D) attachment behavior in normative samples will need to be both theoretically and empirically explored.
In addition, many important variables worthy of investigation were not examined in this study. Individual variables pertaining to the mother as well as the child need to be included. For example, the contribution of the child to the ongoing relationship, i.e. temperamental style; the contribution of the mother i.e. her emotional and physical health, attachment history, life stress. Finally, present findings lend support to recent interest in incorporating family systems perspectives to the study of attachment. However, given the limitations of small samples, current work will need to be replicated to verify and extend findings.
These efforts may be fruitful not only in helping to more fully explore and contextually place the issue of infant attachment, but also in continuing to provide avenues for integration of clinical and research efforts in this area. While some researchers and clinicians have already started this process with promising results (Lieberman 1992;Ericson et al., 1992;Marvin, 1992) more careful investigation into relevant components and points of intersection is well worth continued pursuit of research questions posed. Further elaboration and clarification, for example, of the mechanisms through which relationships within a family help to shape the motherchild relationship, and how it in turn affects the remaining system, would lend compelling support to family-based treatment components of early intervention programs, largely lacking such focus. The ability to consistently and reliably assess maternal style is an exciting research as well as clinical tool applicable and much needed in the expanding short-term work with mother-infant and mother-toddler dyads. More generally, added appreciation and evidence for the influence of context on children's behavior, and its interpretation can help put in perspective brief officebased assessments often conducted for evaluation purposes. The need to observe children in their milieu, to consult the input of parents as well as outside reporters is applicable as much to research as to every-day clinical practices. These are just some of the examples in which questions posed in this project may lend themselves to elaboration in the field of clinical and developmental psychology.   ., Score 5 (Lowest ) 2% I Table 3 Descriptive Information for Home Observations of Maternal Sensitivity (n = 51)      Correlations Among Measures of Family and Marital Functioning            .29*   (Farran, Kaaari., · COlll(or'4 and Jay, 1986)  2. When child returns to mother after playing , 1.8 5.8 child is sometimes fussy for no clear reason .

Amount of maternal involvement
Low: Child is happy or affectionate when he returns to mother between or after play times.
3. When he is upset or injured, child will accept 4.8 2.0 comforting from adults other than his mother.
Low : Mother is the only one he allows to comfort him .
4. Child is careful and gentle with toys and pets. 6.2 4.8

5.
Child is more interested in people than in 6.3
Low: More interested in things than people .
6. When child is near mother and sees 2.2 7.2 something he wants to play with, he fusses or tries to drag mother over to it.
Low : goes to whatever he wants without fussing or dragging mother along. 9. Child is lighthearted and playful most of the 6.5 3.0 time .
Low : Child tends to be serious , sad, or annoyed a good deal of the time.
10. Child often cries or resists when mother takes 2.3 6.0 him to bed for naps or at night.

Security Dependency
11. Child often hugs or cuddles against mother , 7.5 7.4 without being asked or invited to do so.
Low : Child doesn 't hug or cuddle much , unless mother hugs him first or asks him for a hug .
12. Child quickly gets used to people or things 6.0 2.8 that initially made him shy or frightened him.
**Middle if never shy or afraid. 13. When the child is upset by mother ' s heaving, 2.7 7.4 he continues to cry or even gets angry after mother is gone. 26. Child cries when mother leaves him at home 3.3 7.6 with baby-sitter , father, or grandparent.
Low : Doesn 't cry with any of these.

Security Dependency
27. Child laughs when mother teases him. 6.3 4.0 Low: Annoyed when mother teases him . **Middle if mother never teases child during play or conversation.
28. Child enjoys relaxing in mother ' s lap . 7.5 6.4 Low: Prefers to relax on the floor or on furniture .
Middle if child never sits still. generally finds something else to do without returning to mother between activities .
Low: When finished with an activity or toy, he returns to mother for play, affection or help finding more to do.
60. If mother reassures him by saying "It' s OK" 8.5 3.0 or "it won 't hurt you", child will approach or play with things that initially made him cautious or afraid . 86. Child tries to get mother to imitate him, or 6.5 6.2 quickly notices and enjoys it when mom imitates him on her own. 87. If mother laughs at or approves of something 5.8 6.6 the child has done, he repeats it again and again. G. Adult Sexual Gratification: Ability to initiate sex and respond to each other in a sexually and affectively gratifying manner.

3.
Behavior Control: Patterns for handling behavior of children and adults in physically dangerous situations , in meeting and expressing psychobiological needs or drives and in interpersonal relationships.

CHILDREN:
A. Physically Dan~erous Situations: Rules for such situations as playing or running out in the street, playing with matches, alcohol and drugs, using dangerous objects (e.g ., knives , sharp objects, guns), moving into dangerous surroundings.

B.
Psychobiolo~jcai Needs or Drives: rules for eating, sleeping , eliminating, sex and aggression.

C.
Interpersonal Relationships: Rules for getting along with family members and for people outside the family.
D. Enforcement of Rules: How adults enforce rules .

ADULTS:
Appropriate rules for such situations as driving recklessly, alcohol and drugs, smoking, suicide attempts, taking inappropriate risks. Rules for interpersonal relationships.

4.
Problem Solvin~: Family' s ability to resolve problems to a level that maintains effective family functioning. Problems involve: Instrumental Problem Solvin~: Mechanical problems of everyday family life ( e.g., household repairs, planning a trip, buying an appliance).
A. Identification: How instrumental and affective problems are identified.

B.
Communication: How instrumental and affective problems are communicated to the appropriate person(s).
C. Development of Action Alternatives: How family members discuss and agree upon suitable/appropriate plans of action to solve instrumental and affective problems .

D.
Action: How problems to solve problems are put into action.

E.
Monitorin~ and Evaluatin~ Action: How the family checks to see whether or not action plans were acted upon and carried out.

5.
Communication : Recurrent patterns of how instrumental and affective information and messages are exchanged within the family.
A. Extent of Communication : Amount of time adults talk to one another and which parents and children talk to one another.

B.
Clarity of Communications: The extent to which: discussions of everyday issues and understood ; feelings and moods are discussed straightforwardly and are understood by family members; family members listen to each other; family members let one another know that they understand what they've said and, when they don't, ask to clarify it; sensitive topics can be discussed.

C.
Directness of Communication: The extent to which family members answer for themselves , talk directly to the person for whom the message is intended and do not talk about a person in their presence.
6. Affectjye Responsiveness: The degree to which the family and family members respond with the full range of feelings and whether or not these feelings and whether or not these feelings are appropriate for the particular situation and/or behavior .
Emer~ency Emotions: Anger Sadness Fear 7. Affective Involvement: Degree to which the family as a whole shows interest in and values the activities and interests of individual family members.
A. Adult Relationships: Degree to which adults feel that their spouse or some other adult shows a genuine interest in them and in the things that interest them ( e.g., activities, hobbies, etc.). The degree to which this involvement is authentic and caring but allows the individual enough space to feel that they "can do their won thing" and to "think for themselves." B. Parent-Chj!d Relationships: The degree to which children feel that parents show a genuine interest in their activities and interests but also demonstrate authentic concern about their well being (e.g., don't just show interest in the child because it's important to the parent). The extent to which parents feel that they can get close enough to their children and children feel they can get close enough to their parents. The degree to which children feel that they have enough space to "think for themselves" and to "do their own thing."

8.
Family Functionjn~: Summation of the dimensions, identified transactional patterns.     2. It would be nice ifmy relationship succeeded, but I can't do much more than I am doing now to help it succeed.

FAMILY ASSESSMENT DEVICE
I. It would be nice if it succeeded. but I refuse to do any more than I am doing now to keep the relationship going .
0. My relationship can never succeed. and there is no more that I can do to keep the relationship going.