Attachment

COUPLE RELATIONSHIPSBrooke c. Feeney, Nancy L. Collins

PARENT-CHILD RELATIONSHIPSKaren S. Rosen, Fred Rothbaum

Normative Attachment Processes in Adulthood

Attachment refers to a specific type of bond that has four defining features:

1. proximity maintenance—the attached individual wishes to be in close proximity to the attachment figure;
2. separation distress—the attached individual experiences an increase in anxiety during unwanted or prolonged separation from the attachment figure;
3. safe haven—the attachment figure serves as a source of comfort and security such that the attached individual experiences diminished anxiety when in the company of the attachment figure; and
4. secure base—the attachment figure serves as a base of security from which the attached individual engages in explorations of the social and physical world.

Bonds of attachment are found in some, but not all, relationships of emotional significance— only those that are critical to an individual's sense of security and emotional stability (Weiss 1982). Adult pair bonds, in which sexual partners mutually provide security to one another, are presumed to be the prototypical attachment relationship in adulthood (see Hazan and Zeifman 1999 for a review).

John Bowlby (1969/1982, 1973, 1980, 1988) proposed that attachment bonds involve two behavioral systems—an attachment system and a caregiving system. First, individuals come into the world equipped with an attachment behavioral system that is prone to activation when they are distressed and that serves a major evolutionary function of protection and survival (Bowlby 1969; Bretherton 1987). The attachment system is, thus, a safety-regulating system that solidifies enduring emotional bonds between individuals that contribute to reproductive success. Although there are normative developmental changes in the expression of the attachment system across the lifespan, the basic function of the attachment system remains constant (Hazan and Zeifman 1999). Adults as well as children benefit from having someone looking out for them—someone who is deeply invested in their welfare, who monitors their whereabouts, and who is reliably available to help if needed. Consistent with this idea, research indicates that intimate relationships play a critical role in promoting health and well-being in adulthood, and that relationship disruption in adulthood is associated with a wide range of adverse health outcomes (see Uchino, Cacioppo, and Kiecolt-Glaser 1996 for a review).

Second, attachment theory stipulates that the caregiving system is another normative, safety-regulating system that is intended to reduce the risk of a close other coming to harm (Bowlby 1969/1982, 1988). Caregiving refers to a broad array of behaviors that complement a partner's attachment behavior, and may include help or assistance, comfort and reassurance, and support of a partner's autonomous activities and personal growth (Collins and Feeney, B., 2000; Kunce and Shaver 1994). Responsive caregiving in situations of distress restores feelings of security and gives the attached individual confidence to explore the environment and productively engage in social and achievement activities. Unlike parent-child relationships, which have clearly defined caregiving and care-seeking roles, adult intimate relationships are reciprocal and mutual. Therefore, in well-functioning attachment bonds, adult partners should be able to comfortably rely on one another in times of need, sometimes as care-seekers and sometimes as caregivers (Collins and Feeney, B., 2000).

Individual Differences in Adult Attachment Styles

Although the need for security is believed to be universal, adults differ systematically in their beliefs regarding attachment relationships and in the way they maintain and regulate feelings of security. Differences in attachment style are thought to be rooted in underlying differences in internal working models of self (as worthy or unworthy of love) and others (as responsive or unresponsive). Working models are thought to develop, at least in part, from interactions with important attachment figures and, once formed, are presumed to guide social interaction and emotion regulation in childhood and adulthood (Ainsworth et al. 1978; Bowlby 1973; Collins and Read 1994; Main, Kaplan, and Cassidy 1985).

Although the basic tenets of attachment theory argue for the existence of attachment bonds throughout the lifespan, the systematic investigation of attachment processes in adult couple relationships did not begin until Cindy Hazan and Philip Shaver (1987) identified styles of attachment in adulthood that parallel those observed among infants. Subsequent advances in the conceptualization and measurement of these styles have led adult attachment researchers to recognize four prototypic attachment styles, which are derived from two underlying dimensions. These dimensions are referred to as anxiety and avoidance, and they are most often assessed through self-report questionnaires (for reviews, see Brennan, Clark, and Shaver 1998; Crowell, Fraley and Shaver 1999). The anxiety dimension refers to the degree to which an individual is worried about being rejected or unloved; the avoidance dimension refers to the degree to which an individual avoids (versus approaches) intimacy and interdependence with others. The four attachment styles derived from these two dimensions are:

1. Secure adults are low in both attachmentrelated anxiety and avoidance; they are comfortable with intimacy, willing to rely on others for support, and are confident that they are loved and valued by others.
2. Preoccupied (anxious-ambivalent) adults are high in anxiety and low in avoidance; they have an exaggerated desire for closeness and dependence, coupled with a heightened concern about being rejected.
3. Dismissing avoidant adults are low in attachment-related anxiety but high in avoidance; they view close relationships as relatively unimportant, and they value independence and self-reliance.
4. Finally, fearful avoidant adults are high in both anxiety and avoidance; although they desire close relationships and the approval of others, they avoid intimacy because they fear being rejected.

Consistent with the major tenets of attachment theory, adult attachment researchers have argued that these different styles of attachment can be understood in terms of rules that guide individuals' responses to emotionally distressing situations (Fraley and Shaver 2000), which have evolved, at least in part, in the context of parental responsiveness to signals of distress (Kobak and Sceery 1988). For example, secure attachment is organized by rules that allow acknowledgment of distress and turning to others for support. In contrast, avoidant attachment is organized by rules that restrict acknowledgment of distress, as well as any attempts to seek comfort and support from others, whereas preoccupied attachment is organized by rules that direct attention toward distress and attachment figures in a hypervigilant manner that inhibits autonomy and self-confidence.

Although most of the empirical work on adult couple relationships (summarized below) utilizes self-report measures of adult attachment style, several interview measures have also been developed (Bartholomew and Horowitz 1991; Crowell and Owens 1996; George, Kaplan, and Main 1985) and are increasingly used to study adult intimate relationships (e.g., Cohn et al. 1992; Crowell et al., in press). However, these measures are not yet widely used in couples research, in part because they are time-consuming to administer and difficult to code (all require specialized training). Moreover, several studies have found relatively weak convergence between some self-report and interview measures of adult attachment (e.g., Shaver, Belsky, and Brennan 2000). The reasons for these modest effects are not well understood, and researchers continue to debate a variety of unresolved measurement and conceptual issues regarding the assessment of attachment adult style (see Crowell, Fraley, and Shaver 1999, for an overview).

Stability and Change in Adult Attachment Styles

Attachment theory argues that individual differences in attachment style will be relatively stable over time in part because working models tend to function automatically and unconsciously, and because they serve to direct attention, as well as organize and filter new information (Bowlby 1988; Bretherton 1985, 1987; Collins and Read 1994; Shaver, Collins, and Clark 1996). However, it cannot be assumed that the attachment styles observed in adulthood (between romantic partners) are identical to those formed in infancy (between children and parents). Longitudinal studies have obtained mixed results regarding the stability of attachment styles from infancy to early adulthood (for reviews, see Allen and Land 1999; Crowell, Fraley, and Shaver 1999). Although there is some evidence for the importance of family experiences in the development of adult attachment processes, there is little evidence of a simple or direct relationship between childhood attachment style and adult romantic attachment style.

Although there is little evidence of direct continuity from childhood to adulthood, there is evidence for stability across adulthood (see Feeney J., 1999 for a review). Studies of adult romantic attachment have shown moderate to high stability of attachment style over intervals ranging from one week to four years (e.g., Baldwin and Fehr 1995; Collins and Read 1990; Davila, Burge, and Hammen 1997; Fuller and Fincham 1995; Scharfe and Bartholomew 1994). Of course, some observed instability may reflect problems in measurement. Nonetheless, it is also the case that some instability reflects actual change in working models over time and appears to be shaped by changing inter-personal circumstances (e.g., Davila, Karney, and Bradbury 1999; Fuller and Fincham 1995). Attachment researchers are continuing to investigate the continuity and the lawful discontinuity of attachment patterns over time. Adult attachment style is best considered a relatively stable personal characteristic that is sensitive to current relationship experiences and open to change over time.

Studies of Adult Romantic Attachment

Since Hazan and Shaver's (1987) seminal study of adult romantic attachment, there has been a burgeoning of research on this topic within social, personality, and clinical psychology. Studies of adult romantic attachment have generally focused on the examination of attachment style differences in overall relationship quality and in specific relationship processes involving emotion, behavior, cognition, and psychophysiology. Although it is not possible to review all of these studies in this entry, some important findings to emerge from the adult romantic attachment literature are highlighted.

Relationship quality and stability. With regard to overall relationship quality, a large body of research indicates that secure adults develop relationships that are happier and better functioning than their insecure counterparts (e.g., Bartholomew and Horowitz, 1991; Collins and Read 1990; Feeney , J., and Noller, 1990; Hazan and Shaver 1987; Simpson 1990). Secure adults tend to be involved in relationships characterized by frequent positive emotion and high levels of interdependence, commitment, trust, and satisfaction. These individuals have high self-esteem, are generally positive and self-assured in their interactions with others, and report an absence of serious interpersonal problems. Anxious/preoccupied adults, on the other hand, tend to be involved in relationships characterized by jealousy, frequent negative affect, and low levels of trust and satisfaction. They report a strong desire for commitment in relationships and exhibit a controlling (over-dominating) interpersonal style. Avoidant adults tend to be involved in relationships characterized by low levels of interdependence, commitment, trust, and satisfaction. They also report low levels of distress following relationship breakup. Similar to anxious/preoccupied individuals, their relationships tend to involve more frequent negative emotions and less frequent positive emotions; however, the negative nature of their relationships stems from discomfort with intimacy rather than obsessive preoccupation with partners.

Although insecure adults tend to have less satisfying relationships, their relationships are not always less stable. For example, in a four-year prospective study, Lee Kirkpatrick and Cindy Hazan (1994) found that the relationships of anxious/ambivalent (preoccupied) respondents were quite stable over time despite their initial, negative ratings of relationship quality (see also Kirkpatrick and Davis 1994). Likewise, in a four-year prospective study of newlyweds, Joanne Davila and Thomas Bradbury (2001) found that insecure individuals were more likely to be involved in unhappy but stable marriages over time. These studies suggest that insecure adults may be more willing than secure adults to tolerate unhappy relationships, perhaps because they are less confident about their available alternatives.

Interpersonal behavior. In addition to studying attachment style differences in relationship quality, a growing body of research examines how secure and insecure adults differ in their interpersonal behavior in a variety of relationship contexts. Although some of this research relies on self-reported behavior, many studies utilize observational methods to examine behavior in laboratory and field settings. These studies have revealed that (a) secure individuals tend to be more effective support-providers and support-seekers than insecure adults (e.g. Carnelley, Pietromonaco, and Jaffe 1996; Collins and Feeney, B., 2000; Feeney, J., 1996; Feeney, B., and Collins, 2001; Kobak and Hazan 1991; Kunce and Shaver 1994; Simpson, Rholes, and Nelligan 1992); (b) secure adults tend to use more constructive strategies for dealing with conflict than insecure adults (e.g. Pistole 1989; Simpson, Rholes, and Phillips 1996); (c) secure adults exhibit more effective communication styles (Feeney, J., Noller, and Callan 1994) and more adaptive patterns of self-disclosure (Mikulincer and Nachshon 1991) than insecure adults; (d) secure individuals tend to respond more adaptively than insecure adults to separations from their partner (Cafferty et al. 1994; Feeney, J., 1998; Fraley and Shaver 1998); and (e) relative to secure and anxious/ambivalent individuals, avoidant individuals experience lower levels of intimacy, enjoyment, and positive emotion, and higher levels of negative emotion, in their daily interactions with others (Tidwell, Reis, and Shaver 1996).

Attachment style differences in adult sexual behaviors have also been documented. For example, avoidant individuals are more likely than secure individuals to engage in "one-night stands" (Brennan and Shaver 1995; Hazan, Zeifman, and Middleton, 1994 as cited in Feeney, J., 1999) and have more accepting attitudes toward casual sex (Feeney, J., Noller, and Patty 1993). Relative to secure and avoidant individuals, anxious/ambivalent individuals (especially women) tend to engage in intercourse at a younger age and to report a larger number of lifetime sexual partners (Bogaert and Sadava 2002); they are also more likely to experience unwanted pregnancy (Cooper, Shaver, and Collinsand 1998). Relative to their insecure counterparts, secure adults are less likely to have sex outside their primary relationship, more likely to be involved in mutually initiated sex, and more likely to enjoy physical contact that is both intimate and sexual (Hazan, Zeifman, and Middleton 1994, as cited in Feeney, J., 1999).

Cognition and perception. Research on interpersonal perception in couples indicates that secure and insecure adults differ in the way that they construe their relationship experiences (see Collins and Allard 2001 for a review). For example, secure adults are more likely than insecure adults to make benign (relationship-protective) attributions for their partners' transgressions (Collins 1996) and to change their perceptions of relationship partners after receiving information that disconfirmed their expectations (Mikulincer and Arad 1999). Attachment models also appear to shape memories of daily social interactions (Pietromonaco and Barrett 1997). Other research shows that avoidant adults tend to suppress their attachment systems by restricting the encoding and accessibility of attachment-related thoughts and memories (Fraley, Garner, and Shaver 2000; Fraley and Shaver 1997; Mikulincer and Orbach 1995). However, psychophysiological studies reveal that although avoidant individuals may report that relationships are unimportant to them, they exhibit elevated physiological responses when separated from their partner in stressful situations (Feeney, B., and Kirkpatrick 1996), and they are just as physiologically stressed as other individuals when they discuss losing their partners (Fraley and Shaver 1997).

Adult Attachment Processes Across Cultures

Research on adult attachment processes has been conducted all over the world and measures of adult attachment style have been translated into many different languages. Nevertheless, most of the empirical work reviewed above comes from industrialized countries, with predominantly Western cultures including Australia, Canada, Germany, Israel, Italy, Portugal, The Netherlands, the United Kingdom, and the United States. There is a growing interest in attachment processes in countries with predominantly Eastern cultures (including China, Japan, and Korea), but this early research has not yet been published in English language journals. In addition, to our knowledge, there is no published research on adult romantic attachment in nonindustrialized societies. Thus it is not possible in this entry to draw conclusions regarding similarities or differences in adult attachment processes between Western and non-Western cultures, or between industrialized and nonindustrialized societies. However, across a variety of Western, industrialized countries, there appears to be a great deal of convergence in normative attachment processes and in the consequences of secure and insecure attachment styles for relationship outcomes.

Conclusion

In conclusion, theoretical and empirical work in the study of attachment indicates that feelings of security are maintained and regulated, at least in part, through the development of intimate relationships with significant others who can serve as a reliable safe haven in times of need. Thus, understanding adult relationships requires an understanding of attachment dynamics, which have been shown to have important implications for personal health and well-being, as well as for relationship functioning.

Attachment Theory

John Bowlby was a psychoanalytically trained clinician who integrated several theoretical perspectives, including ethology (Lorenz 1935; Tinbergen 1951), psychoanalysis (especially object relations theory [Fairbairn 1952; Klein 1932; Winnicott 1958]), general systems theory (Bertalanfly 1968), and cognitive psychology (Erdelyi 1985), into his theory of attachment (Bowlby 1969). Bowlby originally described attachment as a dynamic behavioral system and delineated the set goals and functions of the system within a context of natural selection and survival. He highlighted the ways in which the attachment system is related to the exploratory, fear, and affiliative behavioral systems. Because these systems are organized and in balance, the activation of one is related to activation of the others (Bowlby 1969).

Bowlby delineated several stages in the development of attachment to the mother. During the stage of indiscriminate sociability (birth to six weeks), infants respond to a variety of social and nonsocial cues without showing a preference for a particular person. During the phase of discriminating sociability (six weeks to six or seven months), infants begin to show a preference for the mother, smiling and vocalizing more readily in her company. They learn the contingencies of this relationship, developing expectations about the mothers' response to particular signals and cues. During the stage of attachment (seven months to two years), infants are able to use the mother as a secure base for exploration and to return to her for comfort when distressed. Infants prefer to be in the company of their mother and seek proximity to her, but are able to venture away to explore their environment. Once an attachment has developed, infants are more likely to protest when with an unfamiliar person (stranger anxiety) or when separated from the mother (separation anxiety). Finally, after two years of age, children move into the stage of goal-corrected partnership. At this point, children are able to recognize that the mother may have needs or goals that are different from their own. The developing capacity for tolerating frustration while delaying the gratification of needs marks this shift in the attachment relationship. There is a new understanding of reciprocity and turn-taking, thereby allowing each partner to modify his or her goals in the service of strengthening the attachment relationship.

Though Bowlby described the goal-corrected partnership as the last phase in the development of attachments, he also acknowledged that attachments remain important throughout the life span and continue to undergo profound changes. Significant organizational shifts may occur within the attachment system, and between the attachment, exploratory, fear, and affiliative systems, and new individuals (in addition to the mother) may serve as attachment figures. As attachments become more abstract and sophisticated, and less dependent on behavioral indices of contact maintenance and proximity seeking, they are also more difficult to measure (Bowlby 1969). Still, attachment behaviors will be evident even during childhood and adolescence, particularly when individuals are afraid, sick, distressed, or reunited with an attachment figure following a long absence (Ainsworth 1990).

Other theorists built on Bowlby's writings in important ways. Mary Ainsworth, a developmental psychologist, identified individual differences in patterns of attachment and studied maternal caregiving behaviors during the first year that contribute to these different attachment patterns at one year of age. Ainsworth's contributions to the development of attachment theory are so significant that the theory is often referred to as the Bowlby-Ainsworth theory of attachment (see, for example, Vaughn and Bost 1999). L. Alan Sroufe and Everett Waters (1977) incorporated motivational and affective components into attachment theory, describing attachment within an organizational perspective. Still others expanded Bowlby's description of multiple attachments (Cassidy 1999) and of developmental changes in attachments beyond the infancy period (Greenberg, Cicchetti, and Cummings 1990).

The Assessment of Attachment in Infancy, Childhood, Adolescence, and Adulthood

When the construct of attachment was originally introduced, attachment relationships were conceptualized as being critical throughout the life span (Bowlby 1969). However, the research that followed Bowlby's original ideas focused initially on the infancy period. This was because of the theoretical framework out of which attachment theory emerged, the developmental perspective within which attachment research evolved, and the underlying assumptions made regarding the situations that activate attachment behaviors and enable the classification of attachment patterns (see Schneider-Rosen 1990 for an elaboration of these ideas). Since 1980, conceptual models and new methodologies have been introduced that have expanded the field of attachment (Bretherton 1985; Cassidy and Shaver 1999; Greenberg, Cicchetti, and Cummings 1990). The result of these efforts is that there are now several classification schemes available to assess individual differences in attachment relationships in infancy, childhood, adolescence, and adulthood.

The most popular and commonly used measure to assess patterns of attachment is Mary Ainsworth and Barbara Wittig's Strange Situation (1969). Indeed, it was the introduction of this standardized procedure that led to the explosion of research on individual differences in attachment patterns and enabled questions regarding the precursors to, and consequences of, these different patterns to be explored. The Strange Situation relies on the use of a series of increasingly stressful situations during which infant behaviors towards the caregiver are observed and coded. Infant-caregiver dyads are then assigned into one of three attachment patterns (Ainsworth et al. 1978) based on the organization of specific infant behaviors throughout the Strange Situation.

Securely attached infants (representing approximately 65% of those classified by the Strange Situation) seek interaction with their caregiver, although not always in close proximity. If they are upset by their caregiver's departure, they are easily calmed and well able to return to exploration upon their caregiver's return to the playroom. Anxious-avoidant infants (20% of those classified) show little or no tendency to interact with or maintain contact to their caregiver in the Strange Situation. They show little or no distress upon separation, avoid the caregiver upon reunion by ignoring, looking away, or moving past the caregiver rather than approaching, and are more inclined to interact with the stranger. Anxious-resistant infants (10% of those classified) show little exploratory behavior and are wary of the stranger. They demonstrate a strong desire to maintain proximity to the caregiver following separation combined with an angry resistance to the caregiver upon reunion. They are unable to be comforted or calmed by their caregiver. Their ambivalence toward the caregiver is reflected in both seeking contact and then angrily resisting it once it is achieved. The percentages of infants classified in each of the attachment categories vary across groups and (in particular) cultures.

A couple cuddles with their infant son. Infants as young as six weeks old show attachment to their caregivers and develop expectations about their caregivers' response and affection. ARIEL SKELLEY/CORBIS

Many researchers found that there were some infants who did not fit into any of these three attachment categories. The introduction of the disorganized/disoriented (Main and Solomon 1990) category (5% of those classified) was based on the observation of contradictory, misdirected, stereotypical, frozen, dazed, or rapidly changing affective behavior in the Strange Situation (Lyons-Ruth and Jacobvitz 1999). Infants classified as disorganized/disoriented show a combination of both avoidant and resistant behaviors, reflecting an apparent confusion about whether to avoid or approach the caregiver. They fail to exhibit a clear or consistent strategy for coping with separation. These infants appear to be most stressed by the Strange Situation and may be the most insecure (Hertsgaard et al. 1995).

Although the Strange Situation has been used extensively in attachment research and a clear majority of infants can be classified into one of the four attachment categories, there are some who remain critical of this laboratory-based procedure. Michael Lamb and Alison Nash (1989) argue that the Strange Situation lacks ecological validity; in other words, it does not occur in natural surroundings. Ross Thompson (1988) claims that independent behavior in the Strange Situation is often mistakenly interpreted as reflective of the insecure/avoidant attachment pattern. And temperament researchers (e.g., Kagan 1995) challenge the use of the Strange Situation by arguing that individual differences in behavioral inhibition can explain the behaviors characteristic of children assigned to the attachment categories.

A second widely used measure is the Attachment Q-set (Waters and Deane 1985), which is appropriate for use with one- to five-year-olds. The Q-set involves either a parent or a trained rater observing the child-caregiver dyad in and around the home and sorting ninety-one cards containing attachment-related statements into nine piles ranging from most to least descriptive of the child. The score derived from the Q-set reflects the degree to which the attachment relationship is secure. The Q-set measure was designed as an ecologically valid alternative to the Strange Situation in that the behaviors that are rated are those that occur in more natural settings. However, critics of the Q-set methodology argue that the instrument may not be measuring attachment behaviors (those that are elicited in response to stressful circumstances) but rather correlates of those behaviors. Moreover, attachment theory pertains to the quality of attachment, whereas the Q-set method provides a quantitative, continuous measurement of attachment security (Schneider, Atkinson, and Tardif 2001). Only modest convergence has been found in a recent meta-analysis between the Strange Situation and the Attachment Q-set (r = 0.26; IJzendoorn, Vereijken, and Ridsen-Walraven in press).

There are several other techniques that have been developed to assess attachment security for preschoolers, children, adolescents, and adults (see Solomon and George 1999). The proliferation of new instruments suggests the many directions in which attachment theory has been applied, as well as the need for integrative approaches to assessment in the future.

Parental Caregiving, Infant Temperament, and the Development of Attachment Relationships

One of the assumptions pervading attachment theory and research is that variations in maternal responsiveness to the child's needs lead to individual differences in attachment security (Ainsworth et al. 1978). Early work, which obtained the strongest associations between maternal responsiveness and child security, focused on maternal sensitivity, availability, acceptance, and cooperation (Ainsworth et al. 1978). Since then, research on the association between maternal responsiveness and quality of attachment has yielded mixed results (see Rosen and Rothbaum 1993 for a review). Although many studies have found higher quality caregiving in dyads that are classified as secure, the magnitude of the effects in most of these studies is small (DeWolff and IJzendoorn 1997; Rosen and Rothbaum 1993). The failure to account for a larger portion of the variance in attachment security has led some to conclude that a move to the contextual level is essential in future studies of the caregiving antecedents of attachment security (IJzendoorn and De Wolff 1997). Researchers could consider, for example, the conditions under which caregiving influences attachment (Belsky 1997) or a more complex family systems analysis of the dynamics involved in attachment patterns (Cowan 1997).

The modest associations between caregiving and attachment security have led investigators to look beyond caregivers' influence on attachment patterns (Sroufe 1985). Many researchers have studied temperamental characteristics as potential determinants of individual differences in attachment. Complex and interesting associations have been found for certain temperamental qualities, for specific age groups, and for particular high-risk populations (Vaughn and Bost 1999). The link between temperament and attachment security may not be direct (Belsky and Rovine 1987; Seifer et al. 1996). Rather, although some of the behaviors seen in the Strange Situation may be related to temperament, the preponderance of evidence indicates that the attachment relationship and the confidence of the infant in the caregiver's responsiveness are not determined by temperament alone but by a complex interactional history (Vaughn and Bost 1999). It is most likely that a secure attachment will evolve in relationships where there is a "good fit" between the infants' temperament and the caregiving they are provided, whereas insecure attachments are more likely to develop when highly stressed or insensitive caregivers fail to accommodate to their infants' particular temperamental qualities (Boom 1994).

Consequences of Attachment for Children's Emotional Development and Social Relationships beyond the Family

John Bowlby (1973) and Mary Ainsworth and her colleagues (1978) maintained that the assessment of individual differences in infant-caregiver attachment would be critical not only to better understand the antecedents of attachment relationships but also to identify the consequences of variations in attachment security for the child's later development. To date, there are dozens of studies that have explored the longitudinal associations between early mother-child attachment and later functioning. For example, securely attached infants are more curious and persistent in toddlerhood, more empathic with peers, and show higher levels of self-esteem than children with insecure attachments. Securely attached infants are also more likely to be curious, self-directed, sensitive to others, and eager to learn in preschool at three-and-a-half years. Significant associations have been found between attachment security and children's interactions with unfamiliar age-mates and adults (see reviews by Thompson 1998, 1999; Weinfield et al. 1999).

At six years of age, securely attached infants engage in more positive interactions with peers in school. In middle childhood and adolescence, children with a history of secure attachment have been found to be more ego resilient and socially competent and to display better cognitive functioning. A follow-up in a camp setting at eleven and twelve years found that those who were securely attached as infants displayed better social skills and had closer friends than their age-matched peers who were insecurely attached as infants (reviewed in Thompson 1999). Children with insecure attachments during infancy are more likely than those with secure attachments to have poor peer relations (see Schneider, Atkinson, and Tardif 2001 for a meta-analysis) and to display deviant behavior in adolescence (Allen et al. 1998; Carlson 1998). Moreover, infant attachment classifications predict later adult attachment categories on the Adult Attachment Interview (Hesse 1999).

There has been considerable controversy as to what factors contribute to the predictive power of attachments. Some theorists believe that children develop internal working models of their early relationships and that these models mediate between early attachment experiences and later social competence. Based on the early relationship with their attachment figures, infants begin to develop expectations for their caregivers' behavior in response to their signals and cues. Infants create representations or models of what to expect from their world and of how they can expect to be treated by others. If infants are treated in a responsive and consistently sensitive manner, then they develop models of the world as good and of the self as deserving and valued. If, on the other hand, infants are responded to inconsistently or in a rejecting manner, or if infants are ignored, the world is seen as insensitive and unpredictable and the self is viewed as unworthy. These "internal working models" (Bowlby 1969, 1973) of self and relationships are carried forward into new experiences with new interactional partners, influencing children's subsequent behavior and their expectations regarding the sensitivity and contingent responsiveness of others (Waters et al. 1995).

Internal models become more sophisticated and stable with age (Bowlby 1969; DeWolff and IJzendoorn 1997). They are amenable to change (with consistent or life altering changes in the environment) but cannot be modified easily. The developmental processes involved in the elaboration and consolidation of working models are far from understood (Thompson 1999). Understanding these processes is important for comprehending the role of internal models in the continuity between early attachment and later functioning.

Not all theorists agree that internal working models are adequate for explaining the link between early attachment security and subsequent child adjustment. Several other mechanisms have been implicated, such as emotional security, the continuity of caregiving experiences, and the mediating effect of basic features of the child's affective functioning (Kochanska 2001). Jerome Kagan (1995) suggests that other nonattachment constructs such as temperament might account for this association. Michael Lewis and Candice Feiring (1989) maintain that there are many important socialization agents (other than parents) that influence children's social relationships and may account for the associations between attachment and later social functioning.

Attachment and Culture

Attachment theory is often assumed to have universal applicability. To test the universality of four critical hypotheses of the theory, Marinus van IJzendoorn and Abraham Sagi (1999) reviewed studies from a variety of non-Western cultures— including Africa, China, Israel, and Japan. Given the diversity of cultures and the complexity of the attachment behaviors examined, there was impressive support for the universality of the first hypothesis examined. Specifically, there are similar patterns of proximity seeking, proximity maintaining, and separation protest by infants in relation to their primary caregivers in stressful situations. The second hypothesis, that most children are securely attached, received "rather strong" support as well. In the eleven non-Western cultures (the African societies of Dogon, Efe, Ganda, Gusil, Hausa, and !Kung San; China; Israel [Kibbutz and city]; and Japan [Tokyo and Sappora]) for which data are available, between 56 percent and 80 percent of children are securely attached. Although there were fewer direct tests of the third hypothesis (i.e., the sensitivity hypothesis: that security is fostered by sensitive responsiveness to infants' signals) and the fourth hypothesis (i.e., the competence hypothesis: that security in infancy is associated with later social competence), IJzendoorn and Sagi (1999) conclude that "the universal validity of attachment theory appears to be confirmed in the cross-cultural research" (p. 730).

A somewhat different portrait of cultural differences is provided by Robin Harwood and her colleagues (Harwood, Miller, and Irizarry 1995). These authors suggest that Euro-American, as compared to Puerto Rican, mothers were more likely to evaluate toddler behavior in terms of the development of independence and self-confidence, whereas the Puerto Rican mothers placed more emphasis on the development of respectfulness. These findings highlight the existence of cultural variation in the meaning of social competence, as well as in the meaning of behaviors characterized as secure (at least in the eyes of their caregivers). In a recent study, Vivian Carlson and Robin Harwood (in press) found differences between Puerto Rican and Euro-American mothers that "call into question a single universal definition of maternal sensitivity, instead providing evidence that sensitive caregiving behaviors may be culturally constructed . . ." (p. 17).

Fred Rothbaum and his colleagues (Rothbaum et al. in press; Rothbaum et al. 2000; Rothbaum et al. 2001) maintain that extant notions of attachment are infused with Western ideals and preconceptions because attachment theory has been championed by Western thinkers and the studies have overwhelmingly involved Western samples.

Although most attachment theorists acknowledge that culture influences specific attachment behaviors, they tend to view culture as an overlay on biologically determined human behavior. By contrast, Jerome Bruner (1990) views culture and biology as inseparable aspects of the attachment system. Rothbaum and his colleagues (2000, 2001) call into question the universality of the sensitivity and competence hypotheses for the same reasons as Harwood—what constitutes sensitive caregiving and social competence are culturally constructed. Because Rothbaum and his colleagues focus on findings from Japan rather than Puerto Rico, their concerns add to those raised by Harwood. The evidence from Japan indicates that behavior that is highly valued in the United States, such as autonomy and self-assertion, is seen as immature in Japan.

Beneath the debate over the universality of attachment lie important points of agreement. First, all of the investigators would agree that: (a) there are propensities for attachment behaviors by caregivers and children that are common to all humans; (b) there are important cultural differences in how these propensities are manifested; (c) the final verdict has not yet been reached as to whether there are fundamental cultural differences in attachment because much more cross-cultural evidence is needed. The disagreement revolves around what constitutes a "fundamental" difference in attachment. We should probably avoid such debatable labels and focus instead on the ways in which key attachment constructs are conceptualized and manifested in different cultures. This would lead to research that does not simply rely on Western based measures of attachment (as did most of the studies reviewed by IJzendoorn and Sagi 1999) but focuses as well on widely accepted concepts and beliefs from the cultures being examined and devises measures to explore them. This process would lead to a much more inclusive theory of attachment that embraces cultural differences.

Conclusion

The research generated by attachment theory has yielded an impressive array of studies providing considerable support for many of the theory's underlying premises. It is clear that early attachments have a profound impact on young children's developmental trajectories and on the intergenerational transmission of attachment patterns. Researchers have increasingly highlighted assumptions and biases of attachment theory that pose difficulties when applying the theory to non-Western cultures. The recent focus on context (including, for example, inter- and intracultural differences) and the study of multiple attachments across the life span reflect new directions that are important for the theory's development.

It is undeniable that attachment theory has had a profound impact on the field of developmental psychology. Its continued growth speaks, in part, to the intellectual breadth of its founders ( John Bowlby and Mary Ainsworth), to the talented group of investigators who have continued in their tradition, to the enormous wealth of data generated by questions evolving from attachment theory, and to the theory's flexibility in accommodating new and unanticipated research findings while remaining clear about, and committed to, the central tenets of the theory.