Jay Belsky


The topic of the infant's emotional tie to mother has been a focus of theorizing for hundreds if not thousands of years. Freud, however, is probably responsible for modern scientific interest in the topic, as he asserted that the relationship between mother and baby served as a "prototype" that would shape the remainder of the developing individual's life, especially his capacity to love and to work. But it is John Bowlby, the eminent British psychiatrist, who broke from the ranks of Freudian psychoanalysts to develop the prevailing theory of attachment that guides most developmental research on the topic today. Rather than considering the full scope of research and thinking about the infant-parent attachment relationship in this chapter, I focus upon results from several longitudinal studies that I have been involved in over the past 25 years in order to illuminate the antecedents and sequelae of individual differences in infant-parent attachment security. Thus, in this chapter, I report on findings pertaining to the origins or determinants of individual differences in infant-parent attachment security. In addition to considering the classical question of how the quality of maternal care affects the development of secure and insecure infant-mother attachment relationships, I address the role of temperament, individual differences in infant-father attachments, and broader contextual influences on infant-parent attachment security, including early child care, social support, marital quality, and work-family relations. Before describing this work and summarizing results, it is appropriate to review some core tenets of attachment theory that guided my research as well as that of many others.


Attachment theory can be viewed, in large part, as a theory of personality development, one which emphasizes the role of early experiences in shaping psychological and behavioral development. In contrast to the psychoanalytically trained psychiatrists whose ideas Bowlby came to reject, Bowlby regarded Freudians as inappropriately emphasizing the role of the individual's inner fantasies in shaping personality, at the expense of actual lived experiences. There were many possible aspects of early experience that could be considered developmentally important. Bowlby's clinical experience alerted him to the adverse effects of early separations from the mother on emotional well being, both in terms of the short-term distress it evoked and the longer-term consequences it appeared to have on children. In pondering such phenomena, Bowlby found himself dissatisfied with existing, secondary-drive explanations which suggested that because the mother satisfied the infant's primary need for nourishment, she became associated with feelings of satisfaction and, thereby, came to be regarded positively by the infant.

The Evolutionary Function of Attachment

Drawing upon a variety of theoretical perspectives, Bowlby fashioned an evolutionary theory of the child's tie to the mother: The child's intense affective tie to his mother, which was dramatically revealed by his behavior when away from her, was not the result of some associational learning process, but rather the direct consequence of a biologically-based desire for proximity and contact with adults that arose as a direct result of Darwinian natural selection. That is, infants and young children who protested separation, sought to maintain proximity to the caregiver, and engaged in other behaviors which are now considered attachment behaviors, were more likely than others not evincing them to be cared for well in ancestral human environments and were less likely to be consumed by predators or become lost. As a result, they were more likely to survive and to reproduce. In consequence, these genetically-determined proclivities of infants to become attached to their caregivers became part of the human behavioral repertoire, thereby promoting "the survival of the species" (Bowlby, 1969/1982, 1973, 1980). In the closing section of this chapter, questions are raised about Bowlby's original evolutionary theorizing and an effort is made to cast attachment theory in a more modern evolutionary perspective (Belsky, 1999).

Measuring Individual Differences in Attachment Security

Bowlby's foremost American collaborator and even co-constructor of attachment theory, Mary Ainsworth, used his original theoretical insights to develop a procedure for measuring individual differences in the security of the infant-mother attachment relationship that I adopted to address many of the same questions that interested Bowlby, Ainsworth, and an entire generation of attachment researchers (for review, see Belsky & Cassidy, 1994; Thompson, 1999). Noting the complementary activation and inhibition of two distinct behavioral systems, one which kept the child close to the mother and thereby promoted safety and survival (i.e., the attachment system) and another which fostered exploration and thus promoted learning (i.e., the exploratory system), Ainsworth (Ainsworth, Bell, & Stayton, 1971) came to refer to an "attachment-exploration balance" and the child's inclination to use the caregiver as a "secure base from which to explore" (Ainsworth, 1963). Most infants balance these two behavioral systems, responding flexibility to specific situation after assessing both the environment's characteristics and the caregiver's availability. For instance, when the attachment system is activated (perhaps by separation from the attachment figure, illness, fatigue, or by unfamiliar people and environments), infant exploration and play decline. Conversely, when the attachment system is not activated (e.g., when a healthy, well-rested infant is in a comfortable setting with an attachment figure nearby), exploration is enhanced.

Appreciating in Bowlby's theory the inherently stressful—and attachment-evoking—nature of being in an unfamiliar place, encountering unfamiliar people, and being separated from mother, Ainsworth purposefully designed all these features into a brief, 20-minute, 8-episode laboratory procedure in order to experimentally elicit infant attachment behavior (see Table 3.1). Just as important as the creative insight which led Ainsworth to invent such an experimental paradigm was the fact that she studied in the Strange Situation a small sample of 26 infants from middle-class homes whom she had been systematically observing throughout their first year of life, in their own homes. It was Ainsworth's goal to account for variation in infant attachment behavior in the Strange Situation by carefully considering the nature and course of mother-infant interaction during the first year of life, as it was her thesis that it was the quality of care that the child received at the hands of his caregiver during this developmental period that was principally responsible for individual differences in attachment security (Ainsworth, 1973).

TABLE 3.1 The Strange Situation

Episode Persons present Time

1 Experimenter, parent, infant 1 minute

2 Parent, infant 3 minutes

3 Parent, infant, stranger 3 minutes

4 Infant, stranger 3 minutes*

5 Parent, infant 3 minutes

6 Infant 3 minutes*

7 Stranger, infant 3 minutes*

8 Parent, infant 3 minutes

*Length of period reduced if infant is very distressed.

In light of the fact that one of the most dramatic differences between babies in how they react to separation from mother, especially when left with an unfamiliar individual in a strange place (even if only for a short period of time), involves the extent to which they become distressed, many believed initially that a focus upon crying would prove to be the window on individual differences in attachment security. As it turned out, Ainsworth's ground-breaking research revealed that it was reunion behavior with mother that was principally reflective of the quality of the infant's emotional tie to her mother, a discovery of no small proportion in the history of developmental psychology (Ainsworth et al., 1978).

Indeed, on the basis of variation in the attachment behavior of the 26 infants whom she studied, Ainsworth identified three distinct patterns of attachment that have guided research ever since (Ainsworth et al., 1978). Infants classified as securely attached (pattern B) use the mother as a secure base from which to explore, reduce their exploration and may be distressed in her absence, but greet her positively on her return, and then return to exploration. Secure infants classified into subcategories B1 and B2 are less distressed by separation than those subcategorized as B3 and B4 and greet mother following separation by vocalizing, smiling or waving across a distance rather than by immediately seeking physical contact and emotional comfort (see Figure 3.1).

Initially, two patterns of insecurity were identified using the Strange Situation. (A third, pattern D, has since been recognized but is not discussed here because it has not played a very significant role in



A1 A2


B1 B2

B3 B4


C1 C2

REUNION Ignore Interact across

BEHAVIOR* Move away distance

Approach - avoid

Approach & seek comfort

Seek & resist contact



Low separation distress Long onset latency Quick recovery

Intense separation distress Brief onset latency Prolonged recovery

Figure 3.1 Patterns of attachment and distress: Reunion and separation behavior. Adapted from Ainsworth, Blehar, Waters, and Wall (1978).

my own research. See Lyons-Ruth & Jacobvitz, 1999, for recent analysis of the disorganized attachment pattern.) Infants classified as insecure-avoidant (pattern A) explore with little reference to the mother, are minimally distressed by her departure and seem to ignore or avoid her on return. Infants classified as insecure-resistant (pattern C) fail to move away from mother and explore minimally. These infants are highly distressed by separations and are often difficult to settle on reunions.



Importantly, Ainsworth's contribution to our understanding of attachment went well beyond the development of an important method for measuring individual differences. Most significantly, it was she who first theorized in detail about why some infants would develop secure and others insecure attachments to their mothers and published data addressing this issue. Using highly detailed ratings of the quality of maternal behavior based upon repeated observations of her sample of 26 middle-class families, Ainsworth (1973) found that it was maternal sensitivity that accounted for why some infants behaved in a secure manner in the Strange Situation when one year of age whereas others behaved insecurely. Central to the notion of sensitivity was the mother's ability to read the infant's behavioral and especially emotional cues and respond in a timely and appropriate manner that served the infant's needs. Thus, mothers who reared secure infants responded in a timely manner to their infants' crying, were ready and able to physically handle the infant in a comforting manner, and were neither intrusive nor ignoring of the infant. In fact, it was Ainsworth's contention that insecure-avoidance developed in response to maternal rejection of the infant, especially when it came to the baby's desire for physical contact. Insecure-resistance, in contrast, was the result of inconsistent caregiving in which the mother was sometimes available and responsive but frequently not. Thus, the different patterns of insecurity were theorized—and found—to derive from different ways in which mothers were insensitive in their care of the infant.

Such findings which buttressed attachment theory led to a first round of investigations by Alan Sroufe and his colleagues at the University of Minnesota examining the sequelae of individual differences in attachment. This was of central importance because what might now be regarded as Bowlby-Ainsworth attachment theory was not only a theory of the nature of the infant's tie to the mother and of the determinants of individual differences in attachment security but, as noted at the outset, of personality development. Thus, there were theoretical grounds for expecting that infants securely and insecurely attached to their mothers would develop differently. Importantly, the work by Sroufe (1988) and his colleagues further buttressed the theory by providing evidence consistent with expectations. Specifically, preschoolers with secure attachment histories proved to be more sociable toward strangers than agemates with insecure histories (Pastor, 1981). These same children also were less emotionally dependent upon their teachers, yet more willing to call upon them for assistance when faced with a challenge they could not manage on their own (Sroufe, Fox, & Pancake, 1983). In addition, children with secure attachment histories evinced more empathy and positive affection toward peers (Sroufe, 1983), perhaps accounting for their greater popularity with classmates (Sroufe, Fox, & Pancake, 1983). These children also were less likely to be victimized by peers, or bully them (Troy & Sroufe, 1987). In fact, whereas insecure-avoidant attachment proved to be predictive of increased aggressiveness during the early-elementary school years, insecure-resistance forecast passive-withdrawn behavior (Renkin et al., 1989).

The developmental benefits of attachment security continued through the middle-childhood years and into adolescence. When 10-year-olds were studied at a summer camp established for research purposes (and enjoyment), ratings by camp counselors, as well as observational data, showed that children with insecure attachment histories of both the resistant and avoidant variety were more dependent on adults than were agemates with secure attachment histories (Urban, Carlson, Egeland, & Sroufe, 1991). Moreover, these differences in dependency continued to manifest themselves when the Minnesota children were studied as 15-year-olds (Sroufe, Carlson, & Shulman, 1993).

Children with secure histories also evinced greater social competence as they grew up. Not only was this evident in global ratings provided by school teachers (Weinfield, Sroufe, Egeland, & Carlson, 1999), but also when reciprocated friendships were measured using sociometric techniques, when camp counselors evaluated children, and when children's interactions with agemates were observed at camp (Elicker et al., 1992). In adolescence, teens with secure attachment histories were rated by camp counselors as more competent in general and more socially effective in mixed-gender crowds in particular (Englund, Levy, & Hyson, 1997). An interview study further revealed the girls with secure histories to be judged as more intimate in their interpersonal relationships (Ostoja, 1996). When considered in their entirety, these results from the groundbreaking Minnesota investigation demonstrate, indisputably, that a foundation of a secure attachment history appeared to be a developmental asset which the child carried at least through adolescence, whereas an insecure history was a risk factor or liability when it came to getting along well with others.


The Minnesota group's early research validating classifications of attachment security based on behavior in the Strange Situation proved to be extremely important in shaping my investigatory endeavors. Having received only minimal training in attachment theory as a graduate student at Cornell University in the mid-1970s, but a great deal on the role of family, community, cultural and historical context in shaping human development (Bronfenbrenner, 1978), I found the prospect of integrating these two distinct research traditions pregnant with opportunity. In fact, because my doctoral research had focused upon parenting and infant development, with a special concern for fathering as well as mothering and husband-wife as well as parent-child relationships (Belsky, 1979a,b), it proved rather easy to extend my research horizon to incorporate ideas from attachment theory into my ecologically-oriented program of research on early human experience in the family. What was principally required was the inclusion of Strange Situation assessments into what developed into a series of four short-term longitudinal studies focused upon the opening years of life that I have come to refer to collectively as the Pennsylvania Child and Family Development Project which I directed over the first two decades of my career while I was at Penn State University.

Inclusion of infant-mother and infant-father attachment assessments in these longitudinal studies enabled me to address a number of theoretically important questions. These concern the determinants of individual differences in infant-parent attachment security, including infant temperament, quality of parenting, early child care, and the social context in which the parent-child dyad is embedded. At the same time that this work was going on in central Pennsylvania, my collaboration with colleagues in a 10-site study of infant day care also enabled me to address issues of child care and attachment with greater precision than was possible in my more local investigations and to extend my basic empirical research on attachment to examine the developmental sequelae of individual differences in infant-mother attachment security. In what follows, I summarize many of the results of these inquiries. I consider first work pertaining to parenting and temperament influences on attachment security, before proceeding to consider the broader ecological context in which attachment relationships are embedded. Finally, before making some closing remarks intended to modernize the evolutionary basis of attachment theory, I review my research on nonmaternal child care and infant-parent attachment security, as well as on the developmental sequelae of infant-mother attachment security. It should be noted that no attempt will be made to extensively review related research on these and other topics pertinent to the study of attachment in infancy. For such reviews, see Belsky and Cassidy (1994), Colin (1996), and Cassidy and Shaver (1999).


Even though Ainsworth (1973) had theorized and found that the sensitivity of maternal care during the first year of life predicted attachment security when the child was one year of age, her work was limited in two important respects. First, her sample of 26 was quite small in size. Second, and more importantly, her evaluations of children's attachment security were informed by what she already knew about the quality of care that the children received at home. Thus, her assessment of maternal sensitivity and attachment security were not independent and this compromised the confidence that could be placed in the results of what, in some respects, had to be regarded as a (remarkable) "pilot" study (Lamb et al., 1984).

Like others, I was fascinated by Ainsworth's (1973) sensitivity hypothesis and used my first longitudinal study of marital change across the transition to parenthood to examine the relation between mothering observed on three occasions during the first year of the infant's life and infant-mother attachment security assessed at 12 months. In this study, 56 Caucasian mothers and their infants from working- and middle-class Caucasian families residing in and around the semi-rural central Pennsylvania community of State College where Penn State University is located were observed at home when infants were 1, 3, and 9 months of age. During each observation period, mothers were directed to go about their everyday household routine, trying as much as possible to disregard the presence of the observer. This naturalistic observational approach is one that I have used in all the research to be described. In order to record maternal and infant behavior, we noted the presence or absence every 15 seconds of an extensive series of maternal and infant behaviors, and one particular kind of dyadic exchange in which the infant or mother emits a behavior, the other responds to it, and the first then contingently responds to the other (i.e., three-step interchange).

Because we did not employ the same rating system as did Ainsworth, we needed a way of conceptualizing and parameterizing the frequency scores of particular behaviors that we generated into indices of sensitivity. Toward this end, we theorized that more was not inherently better and thus hypothesized that infants who established secure relationships with their mothers would have experienced neither the most frequent nor least frequent levels of reciprocal mother-infant interaction. In order to create an index of reciprocal interaction, we factor analyzed a set of 15 mother, child, and dyadic frequency scores. At each of three separate ages, the factor structure proved quite similar, with the principle factor reflecting reciprocal interaction. Loading highly on this factor were measures of maternal attention and care (e.g., undivided attention, vocalize to infant, vocally respond to infant, express positive affection, stimulate/arouse infant), infant behavior (e.g., look at mother, vocalize to mother), and dyadic exchange (i.e., three-step interaction).

We theorized that insecure-avoidance might develop in response to intrusive, overstimulating maternal care (which would force the child to turn away from mother) and that insecure-resistance might be the consequence of insufficiently responsive, unstimulating care. Thus we predicted—and found—that mothers of secure infants would score intermediate on the resulting composite index of reciprocal mother-infant interaction (see Figure 3.2). Moreover, as anticipated, we found that mothers of insecure-avoidant infants scored highest on this index, and that mothers of insecure-resistant infants scored lowest (Belsky, Rovine, & Taylor, 1984). In fact, when the reciprocal interaction composite variable was decomposed into indices of maternal involvement and infant behavior, it was clear that it was the former rather than the latter that distinguished attachment groups.

On the basis of these results, a graduate student at the time, Russ Isabella, extended this work in our second and third longitudinal studies by focusing not solely upon the raw, composited frequencies of maternal and infant behavior on which our original index of reciprocal interaction was based, but rather on the close-in-time co-occurrence of mother and infant behaviors which were theorized to reflect synchronous and asynchronous exchanges in the dyad. The subjects of this work were 153 working- and middle-class Caucasian families rearing firstborns drawn from the same community as our first study. Once again, mother-infant dyads had been observed for 45 minutes, this time at two distinct ages—6 and 9 months—using the same time-sampling methodology already described, and infants had been seen with their mothers in the university laboratory at 12 months to measure attachment security using the Strange Situation. For his dissertation Isabella pursued the hypothesis that dyads that fostered secure attachment would be characterized by interactions that appeared synchronous, whereas those that fostered insecurity would look asynchronous. And, based upon our

Mean Levels of Reciprocal Interaction in the Mother-Infant Dyad and Attachment Security

1 month

3 months

9 months i>



Co Qr

Figure 3.2 Reciprocal interaction at 1, 3, and 9 months as a function of one-year infant-mother attachment classification. (Adapted from Belsky, Rovine, & Taylor, 1984.)

earlier findings, it was predicted that insecure-avoidant dyads would be characterized by intrusive, overstimulating interactions and that insecure-avoidant dyads would be characterized by unresponsive-detached caregiving.

To generate indices of synchrony and asynchrony, it was necessary to inspect the behaviors checked off on our behavior checklist during each and every 15-second observation period, along with those checked off in each of the two adjacent 15-second sampling periods. This enabled Isabella to determine whether the infant and mother behaviors that were recorded within three adjacent periods reflected synchronous and asynchronous interactions. Then, relying upon a sophisticated point-prediction analytic technique, results again substantiated our hypotheses (Isabella, Belsky, & von Eye, 1989; Isabella & Belsky, 1991). Thus, findings from the first inquiry were replicated and extended. In fact, not only had interaction processes reflective of overstimulation been related to insecure-avoidance and those reflective of unresponsive-detachment proven predictive of insecure-resistance in three separate samples which relied on similar approaches to recording mother-infant interaction (i.e., time-sampled behaviors) but dramatically different approaches to parameterizing interaction processes, but in a number of other inquiries similar results obtained (e.g., Lewis & Feiring, 1989; Malatesta et al., 1989; Smith & Pedersen, 1988; Leyendecker et al.,1997). Significantly, all these findings were generally consistent with Ainsworth's (1973) original theorizing linking sensitive and appropriately responsive care with the establishment of a secure attachment to mother by baby (for meta-analysis, see De Wolff & van Ijzendoorn, 1997).

The same was true, it turned out, when the NICHD Early Child Care Research Network (1997), of which I was a part, examined the developmental antecedents of infant-mother attachment security, measured when infants were 15 months of age, as part of its effort to examine the effects of early child care on child development (see below). As expected, it was found that higher levels of observed maternal sensitivity when infants were 6 and 15 months of age predicted increased likelihood of a child establishing a secure attachment to mother—in a sample of over 1,000 children, the largest ever examined with respect to the determinants of attachment security. Especially interesting is that a meta-analysis of the results of 66 studies which did not include the findings from the just-cited NICHD Study of Early Child Care showed, as well, that a variety of indices of maternal sensitivity, interactional synchrony, and dyadic mutuality systematically related to attachment security in just the manner detected in my own work and that of the large child-care study (De Wolff & van IJzendoorn, 1997). Clearly, just as Ainsworth (1973) had theorized and found in her original very small sample investigation, the nature of the child's interactional experiences with mother contributed to determining whether an infant developed a secure or insecure attachment.


An alternative explanation of individual differences in attachment to that proposed by Ainsworth (1973) emphasizing the quality of maternal care draws attention to the infant's temperament, especially the dimension of negative emotionality or difficulty (e.g., Goldsmith & Alansky, 1987). In particular, it was argued that insecurity reflects distress in the Strange Situation, which itself is a function of temperament (Chess & Thomas, 1982; Kagan, 1982). A fundamental problem with this interpretation, of course, was that infants classified as both secure and insecure evince great variation in distress in the Strange Situation. Consideration of Figure 3.1 makes it clear that some secure infants typically evince a great deal of distress in the Strange Situation (i.e., those classified B3 and B4), whereas others do not (i.e., B1, B2), and that some insecure infants typically express a great deal of negativity in the Strange Situation (i.e., C1, C2), whereas others do not (i.e., A1, A2).

A possible way of bringing together competing perspectives on the role of temperament in the measurement of attachment security in the Strange Situation occurred to me upon considering results of findings generated by Thompson and Lamb (1984) and by Frodi and Thompson (1985). When it came to the expression and regulation of negative emotion in the Strange Situation, these investigators observed that secure infants receiving classifications of B1 and B2 looked more like insecure infants receiving classifications of A1 and A2 than like other secure infants (i.e., B3, B4); and that secure infants receiving classifications of B3 and B4 looked more like insecure infants classified C1 and C2 than like other secure infants (i.e., B1, B2). This observation raised the following question in my mind: Might temperament shape the way in which security or insecurity is manifested in the Strange Situation (A1, A2, B1, B2 vs. B3, B4, C1, C2), rather than directly determine whether or not a child was classified as secure? That is, might early temperament account for why some secure infants became highly distressed in the Strange Situation (i.e., those classified B3 or B4), whereas others did not (B1, B2), and why some insecure infants became highly distressed in the Strange Situation (C1, C2), whereas others did not (A1, A2). To address this question, we examined data available on 184 firstborn infants participating in our second and third longitudinal studies.

As theorized, we found that measures of temperament obtained during the newborn period, using the Brazelton Neonatal Behavior Exam, and when infants were 3 months of age, using maternal reports, discriminated infants classified as A1, A2, B1, and B2—that is, the ones who cried little in the Strange Situation—from those classified as B3, B4, C1, and C2 (i.e., those who tend to cry more). These measures did not distinguish infants classified secure (B1, B2, B3, B4) from those classified insecure (A1, A2, C1, C2), however (Belsky & Rovine, 1987). More specifically, although secure and insecure infants did not differ on any of the temperament measures in either of the two longitudinal samples, those most likely to express negative emotion in the Strange Situation at 12 months of age scored lower as newborns on orientation (i.e., visual following, alertness, attention) and evinced less autonomic stability (i.e., regulation of state) and scored higher at age 3 months on a cumulative difficulty index. Thus, consistent with my theorizing and the Thompson, Lamb, and Frodi analyses of emotion expression within the Strange Situation, early temperament appeared to affect the degree to which infants became overtly distressed in the Strange Situation but not how they regulated—with or without the assistance of their mother—their negative affect. In sum, early temperament was systematically related to how much distress infants evinced in the Strange Situation, but not whether they were secure or insecure.

In these first investigations, we, like all other investigators, focused solely upon temperament at a single point in time (i.e., during the newborn period or at 3 months of age). Such an approach was generally consistent with the then-prevailing view of temperament as an inborn, stable, constitutional trait not particularly subject to change. Because we had obtained identical temperament reports from mothers when infants were 3 and 9 months of age, we found ourselves in the enviable position of being able to reconceptualize temperament as a characteristic of the infant that was, in theory at least, subject to change. And when we examined change in temperament, as reported by mother, using data from our first longitudinal study, especially interesting results emerged: Infants who at 1 year of age were classified as secure became more predictable and adaptable from 3 to 6 months, whereas the exact opposite was true of infants who would develop insecure attachments to their mothers (Belsky & Isabella, 1988).

Such results raised the prospect that change in temperament may be what security of attachment, at least in some cases, is all about—a notion that Cassidy (1994) advanced in an effort to interpret attachment from an emotion-regulation perspective. That is, what security of attachment reflects is the child's ability to regulate on his own, or to co-regulate with the expected assistance of his mother, his emotions. Secure children, then, are ones who have developed, in the context of the mother-infant relationship, the ability to manage negative affect and share positive affect. Insecure children, in contrast, are inclined to suppress negativity, or lose control of it, and/or fail to share positive feelings. Theorizing, then, that change in the manifestation of emotion over time might reflect emotion-regulation processes and, thereby, be related to attachment security, we examined, using 148 firstborn infants participating in our second and third longitudinal studies, stability and change in two separate dimensions of temperament, positive and negative emotionality, each based upon composited observational and maternal-report measures obtained when infants were 3 and 6 months of age (Belsky, Fish, & Isabella, 1991). More specifically, at each age, we relied upon maternal reports of how often infants smiled and laughed, as well as cried and fussed, and frequencies of these same behaviors observed during the course of two separate observations, one when mother was home alone with the infant and another when mother and father were at home with the child.

To be noted is that it was consideration of the emotional expressions in the Strange Situation of insecure-avoidant and some secure infants, especially those classified B1 and B2, which led us to think about positive emotions as well as negative emotions with respect to temperament change and attachment security. Central to such thinking was the observation that one thing that distinguished these two groups of children (i.e., A1/A2 vs. B1/B2) was that the secure children classified as B1 or B2 in the Strange Situation openly greeted their mothers upon reunion—with smiles, gestures, and vocalizations—whereas insecure-avoidant infants classified A1 or A2 seemed to suppress any proclivity to greet and express positive sentiment. Also influencing our separate focus upon infant positive and negative emotionality was evidence from the literature on emotions and moods in adulthood highlighting the need to distinguish positive and negative emotionality (Belsky & Pensky, 1988).

Relying upon our repeatedly-measured (at 3 and 6 months) composites of infant positivity and negativity, we created four groups of infants with respect to each emotionality dimension: those scoring high on the dimension in question at both points in time, those scoring low at both points in time, those who changed from high to low, and those who changed from low to high. Figure 3.3 graphically illustrates the stability and change groups in the case of negativity. When we examined attachment security at 1 year of age as a function of these stability and change groups, several interesting findings emerged (Belsky, Fish, & Isabella, 1991). First, changes in negative emotionality were not as strongly related to later attachment as changes in positive emotionality. Consistent with the results of Malatesta et al. (1989), however, we discovered that it was infants who declined in the positivity they manifested between 3 and 9 months who were most likely to be classified as insecure in their attachment at 1 year of age. This suggested to us that change in the temperamental dimension of positive emotionality might reflect processes of emotion regulation and thus that emotion regulation and attachment were

STABILITY AND CHANGE GROUPS: Raw Negative Emotionality Mean Scores

to 0

Infant Age in Months Figure 3.3 Infant negativity change groups. (Adapted from Belsky, Fish, & Isabella, 1991.)

very much related (see Cassidy, 1994). Moreover, these data made intuitive sense in suggesting that children who ended up insecure at the end of the first year of life were the ones whose lives, at least while with their mothers, became less pleasurable over time.

Even though change in positive emotionality by itself, but not change in negative emotionality by itself, predicted attachment security, it was not the case that stability and change in negativity were not at all related to attachment security. This is because we discovered that certain combinations of stability and change in negative and positive emotionality predicted attachment security. Specifically, insecurity was most likely to be observed (i.e., 53% of the time) when (1) infant negativity remained high over time (i.e., high-high group) or increased (i.e., low-to-high group) and (2) infant positivity remained low over time (i.e., low-low group) or decreased (i.e., high-to-low group). In contrast, when none of these conditions obtained, insecurity was quite rare (i.e., 6%). To be noted is that in this work looking at change in positive and negative emotionality, we purposefully did not contrast effects of change against those of temperament or emotionality at a single point in time. And the reason for this was that we were explicitly pursuing a developmental approach which focused upon how specific emotional features of temperament change, not simply where they end up at some later time point or where they begin. Indeed, in our mind, one error often made in much temperament research, whether related to attachment or not, involves the assessment of dimensions of temperament at a single point in time. Such an approach fails to acknowledge that dimensions of temperament develop, or at least change; that is, they are not fixed.

When considered in their entirety, the findings summarized above regarding temperament and attachment dispel the notions that temperament determines attachment security in some simple, straightforward fashion or that there is no relation whatsoever between temperament and insecurity. Rather, they clearly and collectively suggest that the relation between these two constructs is complex. The fact, moreover, that stability and change in infant positive and negative emotionality between 3

and 9 months could be predicted using measures of the parent and family functioning obtained before the child was born and of parenting obtained when infants were 3 months of age strongly—with positive features of parents' personalities, their marriages and their parenting predicting improvements in temperament and the reverse being true of negative features of parents, marriages and parenting—suggests that it is a mistake to presume that emotional features of temperament reflect, exclusively, some inborn characteristic of the infant (Belsky et al., 1991). Because they can change, and because such change appears tied to experiences in the family, understanding of such change may tell us as much about the development of attachment security as it does about presumed constitutional features of the child.

Recent studies which take into account biological inheritance provide further evidence that attachment security is not a direct function of temperament. In perhaps the first test of the heritability of attachment security, Ricciuti (1992) combined data from three samples of twins and, after comparing concordance of attachment security across identical twins (who share 100% of the same genes) and fraternal twins (who share 50%), concluded that attachment security was not demonstrably heritable, at least in the case of 12- to 22-month-olds. More recently, O'Connor and Croft (2001) employed behavior genetic modeling of attachment data collected on 110 identical and fraternal twin pairs seen in the Strange Situation as preschoolers and detected only modest genetic influence, but substantial environmental influence—consistent with findings reviewed above linking maternal sensitivity and attachment security.

The largest and most comprehensive study of the heritability of attachment security conducted to date provides further evidence of the role of environmental factors rather than biological or temperamental ones in shaping attachment security. Indeed, when Bakermans-Kranenburg and Bokhorst (2003) subjected to analysis data on more than 200 pairs of twins, siblings, and unrelated children seen in the Strange Situation as infants, they discovered that biological inheritance did not play a role in determining whether children were classified as secure or insecure in their attachment to mother. More specifically, 48% of the variance in attachment security was explained by shared environmental influences and 52% by unique environmental influence and measurement error. Significantly, similar results highlighting the role of shared environment and yielding very little evidence of genetic effects emerged when the focus of attention was security of attachment to father. In contrast, indices of temperament, reflecting degree of distress in the Strange Situation, proved highly heritable. In sum, while security or insecurity does not appear to be a function of biological inheritance or of temperament per se, variation in emotionality seen in the Strange Situation, even if not the regulation of such emotion in the context of interacting with mother, does seem to be heritable.


Through this point I have considered what might be referred to as "classical" determinants of attachment security, namely those considered in most developmental theorizing about the origins of secure and insecure attachment (Belsky, Rosenberger, & Crnic, 1995a). But an ecological perspective on human development, one that underscores the fact that the parent-child dyad is embedded in a family system (Belsky, 1981), which is itself embedded in a community, cultural, and even historical context (Bronfenbrenner, 1979), suggests that if one wants to account for why some infants develop secure and others insecure attachments to mother, father, or even child-care worker, then there is a need to look beyond the "proximate" determinants of mothering and temperament.

Toward this end, we undertook a series of inquires using data collected as part of our longitudinal studies based upon a contextual model of the determinants of parenting that I advanced more than a decade ago which highlights the role of parent, child, and social-contextual factors in shaping the parent-child relationship (Belsky, 1984; see Figure 3.4). As should be apparent from Figure 3.4, the model presumes that parenting and thus the parent-child relationship is multiply determined and that the contextual factors of work, social support, and marriage can affect parenting both directly and indirectly (through personality). Not explicit in the figure, however, but central to the conceptualization

Jay Belsky
Figure 3.4 Determinants of parenting: A process model. Adapted from Belsky (1984).

on which the model is based, is the notion that parenting, and thus the parent-child relationship, is a well buffered system. This means that threats to its integrity stemming from limitations or vulnerabilities in any single source of influence (e.g., work) are likely to be compensated for by resources that derive from other sources of influence (e.g., marriage). Thus, parenting and the parent-child relationship are most likely to be adversely affected when multiple vulnerabilities exist (e.g., difficult temperament plus conflicted marriage) that accumulate and undermine the effectiveness of other sources of influence in promoting parental functioning. It is just such thinking that led us to examine the cumulative impact of multiple determinants of parenting in affecting attachment security, not just the impact of one or another source of influence.

In the first work of this kind that we carried out using data collected as part of our first longitudinal investigation, linkages were examined between attachment security measured at 1 year and (a) mother's own childrearing history reported during the prenatal period; (b) mother's personality assessed using questionnaires at the same point in time; (c) change in mother-reported infant temperament between 3 and 9 months (and already discussed above); (d) change in marital quality between the last trimester of pregnancy and 9 months postpartum (based upon self-reports obtained at both measurement occasions); and (e) prenatal reports by mothers of the friendliness and helpfulness of neighbors (i.e., social support). Results from univariate analyses revealed that mothers of secure infants scored higher than those of insecure infants on a personality measure of interpersonal affection, whereas mothers of avoidant infants scored lowest on a measure of ego strength; that the (mother-reported) temperaments of secure infants became, as indicated earlier, more predictable and adaptable over time, whereas the reverse was true of insecure infants; that insecure infants were living in families in which marriages were deteriorating in quality more precipitously than were secure infants; and that the neighbors of secure infants were perceived as more friendly and helpful than those of insecure infants (Belsky & Isabella, 1988). More important than these univariate findings, however, was evidence that emerged when maternal, infant, and contextual stressors and supports were considered collectively: The more that the family ecology could be described as well resourced (i.e., positive maternal personality, positive change in infant temperament, less marital deterioration), the more likely the child was to develop a secure attachment to mother.

This work was extended using data from our fourth longitudinal study, this one of a sample consisting exclusively of 125 firstborn sons whose families were enrolled when they were 10 months of age (rather than prenatally as in the first three investigations). These subjects were recruited from the same locale as those participating in the first three longitudinal investigations, though we relied upon birth announcements published in the local newspaper rather than names provided by local obstetricians to identify potential participants. Because of our interest in the multiple determinants of parent-child relations, extensive data were collected on a variety of sources of influence. At enrollment when infants were 10 months of age, mothers and fathers completed questionnaire-based personality assessments and self-reports pertaining to the social support available to them and their satisfaction with it, and whether work and family life interfered with each other or were mutually supportive. At 12 and 13 months, infants were seen in the laboratory to assess infant-mother and infant-father attachment, respectively. Following administration of the Strange Situation, parent and infant engaged in a short period of free play (in a separate lab room) and a number of procedures were implemented to evoke positive and negative emotions. For example, an experimenter tried to make the child laugh and smile using hand puppets; and the parent was directed to frustrate the child by taking a toy away from him while he was in a high chair. Videos of infant behavior following the Strange Situation were rated in terms of the extent to which the child expressed positive and negative emotion every 10 seconds; these ratings were then factor analyzed and combined with parent-report measures of temperament obtained at age 10 months in order to create composite indices of positive and negative emotionality.

Using these data, we again found that it is the cumulative vulnerabilities and resources of infants, parents and families that afford the best prediction of both infant-mother and infant-father attachment security, rather than any single variable or factor (Belsky, Rosenberger, & Crnic, 1995b; Belsky, 1996). Figure 3.5 depicts results pertaining to the cumulative effect of infant, parent and social-contextual factors in shaping infant-father attachment security (Belsky, 1996). In this research, three personality measures were composited (extraversion - neuroticism + agreeableness), as were two measures of infant temperament/emotionality (positivity-negativity), and four social-context measures ([social support satisfaction + number of people to provide support] + [work-family support-interference]). The measures derived from compositing each of these three sets of variables (i.e., personality, infant temperament, social context) were then split at its median; a high score on each composite measure was given a value of 1 to reflect conditions considered favorable for the development of a secure infant-father relationship, and a low score was given a value of 0 to reflect conditions presumed less conducive to the development of a secure relationship. These new values were then summed across sets of variables (i.e., parent, infant, social context), resulting in individual family scores ranging from a value of 0-3. As is apparent in the figure, the greater the family resources (i.e., cumulative resource score of 3), the more likely the infant-father relationship was classified as secure on the basis of the infant's behavior with father in the Strange Situation (Belsky, 1996). Similar results emerged in the case of the infant-mother attachment relationship (Belsky et al., 1995b).

Number of Antecedent Domains Above the Median

Figure 3.5 Probability of son-father attachment security as a function of cumulative resources (0 = low, 3 = high).

Number of Antecedent Domains Above the Median

Figure 3.5 Probability of son-father attachment security as a function of cumulative resources (0 = low, 3 = high).


Because of the role that lengthy child-parent separations played in Bowlby's original formulations of attachment theory, concern has been raised often about the consequences of more routine, short-term separations of the kind experienced on a daily basis by children cared for by someone other than a parent when mother is employed. The initial work addressing this issue focused almost exclusively upon children being cared for in very-high-quality, university-based centers and generally failed to reveal any consistent association between day care and attachment insecurity (for reviews, see Belsky & Steinberg, 1978; Rutter, 1981). However, this first wave of attachment day care research used as an index of security the extent to which the child became upset upon separation from parent, even though it was never clear conceptually whether greater or lesser distress should be considered a marker of security (or insecurity).

When we examined the issue of relations between nonmaternal care and security of infant-parent attachment in our second and third longitudinal studies, using the Ainsworth et al. (1978) reunion-based Strange-Situation scoring system, two particularly interesting findings emerged. First, infants who experienced, on average, more than 20 hours per week of such care during their first year were more likely to develop insecure attachments than were children who experienced less nonmaternal care (Belsky & Rovine, 1988). In fact, when I compiled data from my own work and that from other studies of nonrisk samples that were published in the scientific literature , the same pattern emerged (Belsky, 1988). Subsequently, Clarke-Stewart (1989) and Lamb and colleagues (1990) undertook similar analyses, drawing upon both published and unpublished data. The results of these compilations of findings of research carried out in the United States across a variety of nonmaternal caregiving arrangements (e.g., centers, family day care homes, nanny care) all revealed a reliable association between more than 20 hours per week of such nonmaternal care in the first year of life and attachment insecurity. The magnitudes of association varied, however, with Belsky and Rovine (1988), Clarke-Stewart (1989) and Lamb et al. (1990) indicating, respectively, that the rate of infant-mother attachment insecurity was 65%, 24%, and 83% higher among infants with more than 20 hours per week of nonparental care in their first year relative to infants with less time (including none at all) in such care. Subsequent to the publication of this work, however, Roggmann et al. (1994) addressed the same issue using data from five studies they had carried out and failed to replicate these results.

The second noteworthy finding to emerge from our own work concerned infant-father attachment security. Like Chase-Lansdale and Owen (1987) before us, we found that sons with more than 35 or more hours per week of nonmaternal care (in the United States) were more likely to develop insecure attachments to their fathers and thus have two insecure attachments (one to mother and one to father) than were other boys (Belsky & Rovine, 1988). These findings seemed particularly significant as earlier work in our lab (Belsky, Garduque, & Hrncir, 1984) and by others (Easterbrooks & Goldberg, 1987; Howes et al., 1988; Main, Kaplan, & Cassidy, 1985; Main & Weston, 1981) indicated that children with two insecure attachment relationships functioned more poorly than did children with one or more secure attachments.

A variety of observations about and explanations of these findings linking extensive nonmaternal care in the first year of life with elevated rates of insecure attachment have been offered. Consider first the fact that more than half of the children who experience early and extensive infant day care were classified as secure; such variation in response to early day care suggests that separation per se is probably not the principal cause of the elevated rates of insecurity that have been repeatedly chronicled. Consider next that the quality of child care in the United States is known to be limited; thus, elevated rates of insecurity may have as much, or more, to do with the nature of the care infants receive when cared for by someone other than mother than by the fact that mother is not providing the care. Especially notable in this regard is the fact that toddlers are more likely to develop secure attachments to those who care for them in child care when these caregivers are more sensitive, responsive and available to them (Goosens & van IJzendoorn, 1990; Howes et al., 1988); and that security of attachment to caregiver is itself related to more competent social functioning, especially in concert with secure attachment to mother (Howes et al., 1988). These findings would again seem to implicate the quality of nonmaternal care which the child receives when it comes to understanding the developmental effects of early child care. Also to be considered is the quality of care which the child experiences at home when with his parents, especially since extensive time away from the parent may make it more difficult for the parent to provide the kind of care that fosters attachment security (i.e., sensitive care). Of course, the pressures that working parents experience when at home with the child may also make the provision of such care more difficult.

These varied and complex issues which remained unresolved in discussions of effects of day care led to the establishment of the NICHD Study of Early Child Care (NICHD Early Child Care Research Network, 1994), a research project in which exactly the same research protocol is being implemented at ten different research sites across the United States in order to illuminate the conditions under which early child care enhances or compromises children's development. More than 1,300 children and their families were recruited into this work when infants were 1 month of age, after identifying children and families at their local hospitals shortly after their births. The sample is quite varied demographically and ethnically, but does not include any families in which mother does not speak English fluently or in which the mother is under 18 years of age. A rather extensive research protocol has been implemented to study infant and family development, as well as child care, in this extensive project which is following the children from age 1 month through middle childhood (with pending prospects for further extension). Most importantly, the quality of child care received is measured in detail using observational methods that evaluate both the general child care setting, whether it be a private home or center, and the moment-to-moment experiences that the child has with caregiver(s) and other children when the child is 6, 15, 24, 36, and 54 months of age. At these same ages, multiple features of the family are measured; most important for purposes of this chapter are assessments of the sensitivity of mothering. Such assessments take the form of videotapes of mother-child interaction under free play conditions which are coded using Ainsworth-like rating scales and more naturalistic observations of maternal attentive responsiveness to the child while being interviewed. At all but 6 months of age, children are evaluated in the university laboratory using a variety of methodologies; Strange Situations were administered when children were 15 months of age (and again when children were 36 months old).

With respect to infant-mother attachment security and the effects of early child care, the NICHD Study of Early Child Care had two specific goals. The first was to assess the validity of Strange Situation measurements in the case of children with repeated experiences of being separated from mother due to concerns raised about this methodology with these particular children. The second goal concerned relations between early nonmaternal care and attachment security. Specifically, we sought to determine whether, or under what conditions, experience in child care increased or decreased the probability that a child would establish a secure attachment to his mother.


Ever since linkages have been chronicled between nonmaternal care and attachment insecurity, questions have been raised about whether the infant's independent and exploratory behavior in the Strange Situation is mistakenly judged to be evidence of avoidance, especially in the case of infants who experience routine nonmaternal care. Clarke-Stewart (1989), Thompson (1988), and others have argued that these children's daily separation experiences may lead them to behave more independently in the Strange Situation because they are less stressed by the separations purposefully designed into the procedure. Such a methodological artifact could then cause them to be (erroneously) classified as insecure-avoidant more frequently than children without early and extensive infant day care experience. Not only is it the case that a comprehensive review of relevant studies by Clarke-Stewart and Fein (1983) showed that infant day care experience was not related to less (or more) stress in the Strange Situation (as indexed by distress), but when I carried out a small study using subjects from my second and third longitudinal investigations to directly address this proposition, I found no support whatsoever for it (Belsky & Braungart, 1991). More specifically, infants with extensive nonmaternal care experience who were classified insecure-avoidant did not evince less distress and more play than those similarly classified with limited or no nonmaternal care experience; in fact, the former actually played significantly less during reunions—exactly the opposite of what critics of the Strange Situation methodology had propositioned. Just as noteworthy as our findings are those of Berger, Levy, and Compaan (1995); these investigators found that classifications of children's attachment security based on their behavior during a standard pediatric exam (in which no separation of infant from parent occurred) are highly concordant with Strange Situation classifications for both infants with extensive child care experience in the first year (81.5% concordance) and those without such experience (76%). In others words, my own work as well as that of Berger et al. (1995) failed to find any evidence to suggest that the Strange Situation was an invalid methodology for studying one group of children in particular, namely those with repeated separation experiences due to routine nonmaternal care.

Because these evaluations of the validity of Strange Situation classifications in the case of infants with lots of experience being separated from their mothers involved small samples, the NICHD Study of Early Child Care provided an ideal opportunity to address this issue again. Comparisons were made between children who averaged more than 30 hours per week of care from 3-15 months of age with others who experienced less than 10 hours per week of nonmaternal care during the same developmental period on measures of distress when separated from mother. Not only was it the case that the two groups were not different in terms of how upset infants became when separated, but it was also true that coders' ratings of their confidence in classifying children in terms of the three primary attachment categories (i.e., avoidant, resistant, secure) did not vary as a function of nonmaternal care experience. Thus, it was concluded that there were no empirical grounds to question the internal validity of the Strange Situation in the large data set; therefore, classifications of children in this separation-based procedure could be regarded as valid even in the case of children with extensive experience with separation (NICHD Early Child Care Research Network, 1997).

Effects of Early Child Care

Having established empirically that one could have confidence in the Strange Situation classifications of children with extensive separation experience, the NICHD Early Child Care Research Network (1997) was now in a position to powerfully address a question that had stimulated much debate and controversy. Interestingly, results of this comprehensive investigation revealed that neither the quantity of nonmaternal care, the quality of such care, the stability of care, nor the child's age of entry (within the first 15 months of life), as isolated factors, accounted for variation in infant-mother attachment security. Even though maternal sensitivity when considered by itself did predict attachment security, yet features of child care, when considered individually, did not, it was not the case that features of early child care proved to be totally unrelated to attachment security. Consistent with Bronfenbrenner's (1979, p. 38) dictum that "in the ecology of human development the principal main effects are likely to be interactions," we found that even though most children with early care experience were not more (or less) likely to develop insecure attachments to their mothers, this was not the case when certain ecological conditions co-occurred. More specifically, rates of insecurity were higher than would otherwise have been expected (on the basis of maternal sensitivity alone) when infants received poorer quality (i.e., insensitive) care from their mothers and (a) low quality nonmaternal care, or (b) more than 10 hours per week of nonmaternal care, or (c) more than one nonmaternal care arrangement in their first 15 months of life. In other words, it was under conditions of "dual-risk" that early care was associated,

Positive Thinking Power Play

Positive Thinking Power Play

Learning About A Positive Thinking Power Play Can Have Amazing Benefits For Your Life And Success. Learn About Positive Thinking Power Play -And Have A Look At 10 Steps to Success To Create Amazing Results.

Get My Free Ebook

Post a comment