Individual Differences - Discussion

Patterns of Attachment: A Psychological Study of the Strange Situation - Mary D. Salter Ainsworth 2015

Individual Differences

In Light of Contrasting Paradigms


Most of the criticisms of attachment theory that have arisen since Bowlby (1958) offered his preliminary formulation of it have focused on the issue of individual differences. If we may liken the programatic theories that have so far guided psychological research to what Kuhn (1962) calls paradigms, the controversies about individual differences in attachment and attachment behavior constitute a good example of what he describes as the difficulties that face the adherents of an earlier paradigm when trying to come to terms with a new paradigm. The new paradigm in this case is Bowlby’s evolutionary—ethological attachment theory, and the earlier paradigms are variants of sociallearning theory.

In Chapter 1 we sought to summarize the basic components of the new paradigm as it applies to the attachment of an infant to his mother figure. Elsewhere (Ainsworth, 1969) we reviewed two major variants of the social-learning-theory view of the origins and development of a child’s relationship to his mother, a relationship characterized, before Bowlby’s 1958 paper, as “dependency.” Since then, some social-learning theorists have adopted Bowlby’s term “attachment” for the child’s tie to his mother, but have attempted to rework attachment theory so that it is in harmony with the earlier dependency paradigm. Among the first of these were Maccoby and Masters (1970), whose review includes both an authoritative account of dependency research and theory and a discussion of attachment theory from a social-learning point of view—a discussion that has had substantial influence in shaping a type of social-learning attachment theory differing in important respects from Bowlby’s ethological—evolutionary theory.

Before considering some of the influential implications of the view suggested by Maccoby and Masters, let us very briefly summarize some essentials of an earlier social-learning dependency paradigm. Following Hullian theory, dependency was initially viewed as a secondary or learned drive, derived from such primary drives as hunger, cold, and pain. Because his mother is associated with the reduction of such drives, the infant learns to attach strong reinforcement value to her proximity, and thus to be dependent on her. This learned dependency drive was held to generalize readily from the mother to other people. (Indeed most research into dependency focused on the child’s relations with nursery-school teachers and age peers.) Behavioral indices of such dependency in young children were generally held to be: seeking physical contact, seeking to be near, seeking attention, seeking approval, and seeking help. Research focused on individual differences in the strength of the dependency drive.

The drive model eventually lost ground, in part because of the criticisms implicit in other emerging variants of learning theory, perhaps particularly Skinnerian theory. Bijou and Baer (1965) and Gewirtz (e.g., 1969) described the origins of the child’s relationship with his mother in terms of her acquisition of positive reinforcing function as a result of her association with primary reinforcers, such as food, relief from pain, stimulus change, and the like. They considered dependency as a convenient label for certain kinds of learned behavior, and as neither a drive nor a trait. We consider the implications of this view for attachment later in this chapter. Here we are concerned with those who shifted from a view of dependency as a generalized acquired drive to viewing it as a generalized personality trait. Focal to the trait view is the matter of measuring the strength of the trait in different individuals. The shift from drive to trait made little apparent difference in either measures of dependency or in the criteria considered necessary to validate the concept.

When attachment theory emerged, it was immediately perceived to be somewhat akin to dependency theory. Both theories were concerned with the origins of a child’s tie to his mother, and contact and proximity seeking were focal to both formulations. Maccoby and Masters (1970) suggested that attachment might be viewed as a trait or central motive state, thus obviously attempting to assimilate attachment to the dependency paradigm. This implied that a major dimension of attachment was its strength, and this was to be inferred from the strength of attachment behavior. They suggested 10 possible measures of behavior strength, including the number of persons toward which the behavior is shown—a criterion obviously at variance with evolutionary—ethological attachment theory—as well as frequency with which the behavior is shown. The traditional criteria for testing the validity of the concept of a generalized trait were conceived as applicable to attachment. Thus, to support a trait hypothesis of attachment, Maccoby and Masters suggested that the following criteria should be met: (1) that all the behavioral indices of attachment should be positively correlated; (2) that there should be stability of measures of attachment across situations; and (3) that there should be stability of such measures across time (i.e., stability in the course of development).

Because the theoretical base from which we are working does not conceive of attachment as a personality trait (or central motive state), these criteria are not applicable to it. Indeed they are largely irrelevant. We can see no theoretical basis for expecting all attachment behaviors to be positively correlated—in other words, that a baby who cries much, for example, should also smile, vocalize, approach, and cuddle in when picked up significantly more often than others. This is not to say that we view the ways in which a baby mediates his attachment to his mother as unrelated to one another; but we view their intercorrelations as complexly patterned rather than in any simple, unidimensional relationship implying strength of attachment. We do not believe that attachment behaviors, considered as individual measures, should necessarily be positively correlated across situations. Thus, we can see no reason to expect that a baby who seeks contact with his mother when his attachment system is at a relatively high level of activation will necessarily do so proportionally often when his attachment system is at a low level of activation. On the other hand, we can expect that two infants who differ in the patterning of their behavior to the mother in one situation may well also differ in the patterning of their behavior in another situation, and that through research we can discover how different patterns of attachment manifest themselves in behavior across a variety of situations. Thus, it is obvious that our position implies stable individual differences; but these differences concern the ways in which several forms of behavior—including behaviors other than attachment behavior—are organized. Thus, in the strange-situation research described in this volume, we have reported the ways in which proximity-seeking and contact-maintaining behaviors are organized vis-à-vis angry, resistant behaviors and avoidant behaviors that are also directed to the mother; vis-à-vis wary/fearful behaviors and friendly affiliative behaviors directed to a stranger; and vis-à-vis exploratory behavior.

Let us consider the evidence for such stability of organization, as well as the evidence for the competing theory of attachment as a trait.

Stability of Individual Differences Reflected in Strange-Situation Behavior

As reported in Chapter 10, Maccoby and Feldman (1972) noted that they found no evidence that “attachment” to the mother could be considered stable across time—that is, from 2 to 2½ and from 2½ to 3 years of age—and that more stability was found in regard to reactions to the stranger. Their measures of attachment consisted of “distal” behaviors, except for one “proximal” behavior—proximity—in the preseparation episodes only.

Coates, Anderson, and Hartup (1972a, 1972b) observed attachment in a laboratory situation using two samples and two periods of observation. The first sample was observed first at 10½ months of age and again at 14½, the second sample at 14½ and later at 18½. At each age there were two sessions and two conditions. One condition was a “nonseparation” condition consisting of 10 minutes when mother and infant were together under circumstances similar to our Episode 2. The other condition, the “separation” condition, consisted of three episodes: a 3-minute preseparation episode similar to the nonseparation condition, a 2-minute separation, and a 2-minute postseparation episode. The infants were randomly assigned to one of the two conditions in the first session at the first age point, and then the next day experienced the other condition; at the second age point the conditions were experienced again in the same order. The measures of attachment were time-sampled frequency measures—visual regard, vocalizing, smiling, touching, crying, and proximity—the last indicated by the child’s presence in or absence from the same cell of a floor-grid as that occupied by the mother.

Stability of behaviors was assessed in three ways (Coates et al., 1972b): (1) within a session—which, in the nonseparation condition, compared each third of the session with the other thirds, and which, in the separation condition, compared pre- and postseparation behavior; (2) between sessions, which compared behavior in the nonseparation condition in one session with behavior in the preseparation episodes of the other, and (3) long-term stability, which compared behavior at one age with that at a later age—that is, 10½ vs. 14½ months for the first sample and 14½ vs. 18½ months for the second—both for the nonseparation and for the separation conditions. Neither visual regard nor vocalizing was consistently stable in the three comparisons for the two samples. Smiling was apparently too unstable for the authors even to report. Touching the mother and proximity to her were “moderately stable,” with coefficients of correlation that were mostly significant. The measures relevant to separation behavior (crying and orienting to the door) were essentially unstable.

Masters and Wellman (1974), reviewing the two studies just cited, as well as a study of home behavior over 9 months of the first year (Stayton & Ainsworth, 1973), counted the proportion of significant to nonsignificant correlations in each study and concluded that “there is little stability and functional equivalence among many attachment behaviors . . . it is their [the authors’] conclusion that the correlational analysis of human infant attachment behaviors does not provide substantial support for the concept of attachment as a psychological trait or central motive state [p. 228].”

In contrast, let us consider the evidence presented in this volume for the stability of both attachment behaviors (and certain behaviors antithetical to attachment behavior) and the patterns of behavior that we have proposed as indicative of qualitative differences in the infant—mother attachment relationship. This evidence is of three main kinds: evidence from repeated administrations of the strange-situation procedure, evidence of the relationship between strange-situation behavior and prior or contemporary behavior in other settings, and evidence of the relationship between strange-situation behavior and subsequent infant behavior in other settings.

Stability of Strange-Situation Behavior and Classification. In four studies the strange situation was repeated after varying lapses of time with the mother as the accompanying adult. One of these was Maccoby and Feldman’s (1972) study, which has already been discussed. The other three were reported in Chapter 11. All of these three first administered the situation at 12 months. In our study it was repeated 2 weeks later; in Connell’s (1976) and Waters’ (1978) it was repeated 6 months later. From these studies we may conclude that our four chief measures of interactive behavior were stable during periods of the second year of life—contact maintaining, proximity/contact seeking, avoidance, and resistance, as directed toward the mother, especially in the reunion episodes. In addition, there was some evidence, especially from Waters, of stability of contact maintaining and proximity/contact seeking in the preseparation episodes. Waters and we also found some stability for crying; Connell did not report findings for crying. Our analyses and Waters’ further showed that the “discrete” behaviors so commonly used by others in attachment research—looking, smiling, vocalizing, gesturing, approaching, and touching—tended not to be stable across sessions (with the exception of crying, mentioned earlier). We refer to these as “discrete” behaviors because the measures thereof are simple frequency measures that do not take into account the fact that different behaviors may serve the same purpose (i.e., gaining or maintaining proximity/contact) or allow for the contingencies of the situation, including the reciprocal behavior of the adult(s) in question. We consider our “categorical” measures of interactive behavior superior to frequency measures because they take into account the way behavior is organized with reference to the situation.

Furthermore, some of these discrete behaviors are also “distal” behaviors—looking, smiling, vocalizing, and gesturing. Not only were these unstable across time, but so was our categorical measure of distance interaction (Waters, 1978). Looking, as we have often pointed out (e.g., Ainsworth, 1973) probably should not be classed as an attachment behavior, for it serves so many behavioral systems from birth onward. As for smiling and vocalizing, we suggested in Chapter 13 that these may be considered to belong to a subclass of attachment behaviors that appear chiefly when the attachment-behavioral system is activated at low intensity. Under such conditions of activation, attachment behaviors tend to appear intermittently and irregularly. Without much longer periods of observation than were undertaken in the laboratory studies under consideration, there is no reason that one could expect them to appear stable; and one should especially not expect them to be stable when comparing a preseparation or nonseparation condition in which attachment behavior is only weakly activated with a separation or reunion condition in which it is intensely activated.

In contrast with Coates and associates (1972b) and Maccoby and Feldman (1972), Connell (1972), Waters (1978), and we also examined the stability of A-B-C classifications, these reflect the way in which an infant’s behavior is organized toward his mother and hence, we have suggested, qualitative differences in attachment. Although there was too much carryover of anxiety (with a time interval between sessions of only 2 weeks) for us to find stability in classifications, both Waters and Connell reported a very impressive degree of stability over the 6-month period from 12 to 18 months. Individual differences in quality of attachment of infant to mother thus appear to be strikingly stable over a relatively long period of time and despite the possibility of occurrence of life events that might have intervened to change the attachment relationship in some cases.

We have already suggested three reasons for the apparent discrepancy between the findings of Coates and associates and Maccoby and Feldman, on the one hand, and those of Waters, Connell, and ours, on the other hand:

  1. Use of measures of discrete behaviors vs. measures that reflect some degree of organization and interchangeability among behaviors.
  2. Emphasis on attachment behaviors that are characteristic when the system is activated at low levels of intensity (which tend to be “distal” behaviors) vs. those characteristic when the system is activated at high levels of intensity (which tend to be “proximal”).
  3. Search for stability of separate measures vs. attention to the stability of patterns of behavior such as those reflected in our classificatory system.

There are several other possible sources of discrepancy:

  1. Neither Coates and associates nor Maccoby and Feldman capitalized on the fact that repeated separations raise the level of activation of attachment behavior to especially high intensity. Coates and associates used only one 2-minute separation. Maccoby and Feldman, although using the standard strange-situation procedure with two separations and two reunions, used only discrete and/or distal measures for behavior in the reunion episodes.
  2. Neither Coates and associates nor Maccoby and Feldman used our measures of resistance and avoidance when reporting a child’s interaction with his mother, whereas Connell and Waters found these measures especially stable over time.
  3. Maccoby and Feldman dealt with children older than those of other studies.

Let us comment on the last two points of difference. Studies that examine the stability of the A-B-C strange-situation classifications across time intervals inevitably examine avoidant and resistant behavior, which, although antithetical to attachment behavior, feature conspicuously in classification. It is not merely the way in which attachment behaviors are organized together that is of moment for individual differences, but, more saliently, the way in which behavior as a whole is organized vis-à-vis the attachment figure; and avoidant and resistant behaviors have emerged to be especially important among the nonattachment behaviors.

In Chapter 10 we considered developmental changes in strange-situation behavior between the ages of 12 months and 4 years. These developmental changes undoubtedly affect the ways in which a young child interacts with an attachment figure, and hence may well affect the behavior that mediates the relationship, including attachment behavior. Of perhaps crucial importance is that the features of the strange situation that activate the behavior of 1-year-olds at high intensity tend not to do so as the child grows older. This is not to imply that the degree of activation can be judged solely in terms of overt behavior, for heart-rate studies (e.g., Sroufe & Waters, 1977b) suggest that there may be quite intense activation internally while overt behavior appears to remain in a low key. Nevertheless, the “strange” or unfamiliar may have become less strange to the older preschooler, and, especially, cognitive development is likely to have enabled him to endure his mother’s absence over longer periods of time. From Waters’ and Connell’s work it appears that 18-month-old infants respond to the strange situation in much the same way as 12-month-olds. Marvin (1972) found that 2-year-olds also organized their behavior in much the same way as did 12-month-olds, but that 3- and 4-year-olds did not. This being so, we must know much more about how preschoolers from 2½ onwards organize their behavior toward their mothers, both in situations in which attachment behavior is activated at high intensity and in situations in which it is activated at low intensity, before we can assess the stability of individual differences in their attachments to their mothers.

The Relationship Between Strange-Situation Behavior and Prior or Contemporary Infant Behavior in Other Settings. The evidence presented in Chapter 7 that infant behavior at home, both in the first and in the fourth quarters of the first year, is significantly related in many ways to strange-situation assessments is of even more importance for our proposition of stability of organization or patterning of behavior than is the test—retest evidence cited above. Behaviors at home that were found by Stayton and Ainsworth (1973) to reflect a secure attachment to the mother bear a significant relationship to the patterning of strange-situation behavior that we have identified as characteristic of Group B. Mother-avoidant behavior in the strange situation, characteristic of Group A, is significantly related not only to the cluster of home behaviors that signify anxious attachment, but also specifically to another cluster of home behaviors not shown by nonavoidant but anxiously attached infants—namely, Group C.

Bell’s (1978) finding that Group-A and Group-C babies, more frequently than Group-B babies, in a free-play situation showed a cluster of behaviors judged to display negative affect in interaction with their mothers is congruent with our findings; it seems likely that her positive vs. negative affect factor is equivalent to our secure vs. anxious attachment factor. Rosenberg’s (1975) findings that Group-B infant—mother dyads had more reciprocal interaction than Group-A dyads in a free-play situation, and fewer reciprocal-ignoring states, are also congruent.

Despite these close relationships between strange-situation patterning of behavior and behavior in low-stress situations, separate behavioral variables are not necessarily positively correlated across settings. Thus, for example, crying in the strange situation is not significantly correlated with crying at home. In particular, crying in the separation episodes of the strange situation is not significantly correlated with crying in the brief, everyday separations that occur in the home environment. Securely attached infants show little separation distress in the familiar environment of the home but tend to protest separation in the strange situation—especially the second separation. Group-C infants protest separation in both settings, especially intensely in the strange situation. Group-A infants, however, who behave anxiously when mother leaves the room, tend not to cry when separated in the strange situation. On the other hand, behaviors related to close bodily contact with the mother are positively correlated with comparable behaviors in the strange situation.

Significant relationships between cognitive measures and individual differences in the organization of strange-situation behavior were found by Bell (1970, 1978) and Connell (1974). Connell found striking differences among the A-B-C groups in terms of behavior in his habituation experiment. Bell found that Group-B infants in both her white, middle-class and black, disadvantaged samples were significantly advanced in comparison with non-B infants in the development of the concept of the object, especially the concept of a person as having permanence. She also found that Group-B infants had a significantly higher mean DQ than non-B infants, although in our smaller Sample 1 this difference fell short of statistical significance (Ainsworth & Bell, 1974).

The only separate strange-situation measures that consistently yield significant correlations with behavior in other settings are our scores for resistant and avoidant behavior in the reunion episodes, which are themselves “categorical” rather than “discrete” measures. In Chapter 7 we reported positive correlations between resistance in the strange situation and amount of crying, including separation distress, at home. We also reported significant correlations between avoidance in the strange situation and both anger and various measures of behavior relevant to close bodily contact at home. Bell (1978) combined avoidance and resistance into a measure of “negative” behavior, and found this to show essentially the same relationships with behavior in other settings as did the B vs. non-B comparisons. It may seem paradoxical that these two classes of behavior, which are clearly not attachment behavior because they do not promote proximity/contact to attachment figures, are more closely related to attachment behavior in other settings than are separate components of strange-situation attachment behavior. The paradox is resolved, however, when one considers that these two measures give important clues to the way in which behavior is organized to mediate the infant’s attachment to his mother, and are indeed key behaviors in our classificatory system.

The Relationship Between Strange-Situation Behavior and Subsequent Behavior in Other Settings. In Chapter 9 we reviewed evidence by Bell (1978) Connell (1976), Main (1973, 1977; Main & Londerville, 1978), and Matas (1977) that the patterning of strange-situation behavior is significantly related to social, emotional, and cognitive development in the second year of life, and in the case of studies by Bell and Connell the third year of life as well. In regard to social and emotional development, the evidence is plentiful. In summary, Group-B children, in comparison with non-B children, emerged as more responsive to and initiating more interaction with their mothers, directing more positive behavior and less avoiding, ignoring, aggressive, and/or resistant behavior to their mothers, displaying more positive affect, maintaining more proximity to mothers, and being more cooperative and willing to fit in with their mothers’ wishes, and generally easier to live with. In addition, B children, in comparison with non-B children, are friendlier, more cooperative, and more participant in interaction with relatively unfamiliar persons.

Although there were nonsignificant relationships between strange-situation classification and developmental test scores at some age points, there was nevertheless suggestive evidence of a relationship between quality of infant—mother attachment and subsequent DQ or IQ. In addition, Bell found that Group-B babies at 14 months continued to be more advanced in the development of person- and object-permanence, while Connell and Main both found Group-B toddlers to be more advanced in language acquisition. Main reported that B toddlers engaged in superior exploratory activity. Matas found them to be more enthusiastic and persistent in a problem-solving situation, and to show less frustration behavior and less nontask behavior than non-B children.

Furthermore Main, Connell, and Matas all found some significant differences between Group-A and Group-C children. Connell found that whereas A children clearly maintained more distance between themselves and their mothers than B children, C children did not, and indeed in an episode in which the mother moved to a less accessible place, C children showed more proximity seeking than B children. Matas found that C children showed extreme reliance on the mother in a problem-solving situation and were also likely to give up quickly and show frustration behavior, whereas A children sought more help from the experimenter than from the mother and were aggressive to the mother. Main reported that resistant strange-situation behavior, which is characteristic of Group C, was negatively correlated with DQ, intensity of exploratory play, and length of play bouts, whereas avoidant behavior (characteristic of Group A) was not significantly related to these variables. Matas identified Group-C children as relatively incompetent in problem-solving situations, whereas Group-A children were especially noncompliant. The dynamics that we believe to underlie these differences are discussed in Chapter 15.

Conclusions Regarding Stability of Individual Differences. Whereas Masters and Wellman (1974) found so little evidence of individual differences in attachment behavior that they suggested that these ought to be disregarded in studies of attachment, the evidence that we have reported here leads us to conclude that there is substantial stability of individual differences in attachment across time and across situations. It is clear that the A-B-C classifications of strange-situation behavior yield the most striking evidence of stability. This implies that it is the way in which an infant organizes his behavior in directing it toward his mother figure that is stable. Individual differences in such organization reflect what we have termed differences in the quality of the infant—mother attachment relationship. The focus is on the organization—the attachment—rather than on the separate components of behavior that enter into the organization.

There is substantial evidence also for the stability of our “categorical” measures of strange-situation behavior, such as proximity/contact seeking, contact maintaining, resistance, and avoidance, across time in similar situations. This stability we attribute to the fact that in dealing with classes of behavior that have the same “predictable outcome” (Bowlby, 1969), we take into account the goal-corrected nature of behavior. Thus in the category of contact-maintaining behavior, we acknowledge a variety of ways in which a baby can maintain close bodily contact with his mother when there is a threat of interruption, rather than consider clinging, clambering up, holding on, and crying in protest to the interruption as separate behavioral items.

There is little evidence, however, for the stability of what we have termed “discrete” behaviors. As our previous discussion has implied, we attribute this to the fact that measures of such behaviors do not take into account either the way in which behaviors are organized or their goal-corrected nature. When, in addition, context is ignored—as it often is when testing for stability across situations—it is small wonder that such isolated behavioral items tend to emerge as unstable. This tends especially to be the case with “distal” behaviors that tend also to be “low-intensity” behaviors characteristically but intermittently occurring when the attachment system is at a low level of activation and especially subject to competition from other behavioral systems, such as exploratory behavior or affiliative behavior directed toward persons other than attachment figures. As suggested earlier, the intermittent nature of such behaviors would require much longer samples of time than afforded by the usual laboratory study for their stability to be tested.

We conclude therefore that individual differences in the quality of attachment tend to be stable across time and across situations because they reflect underlying differences in the organization of behavior. Nevertheless we cannot expect to find similar consistency across situations in the behaviors that mediate attachment unless we take into account the way the exigencies specific to the situation interact with the underlying organization of attachment. Sroufe and Waters (1977a) have also emphasized that our concept of attachment is an organizational construct. According to this view the specific behavior toward an attachment figure in any given situation will be determined both by the underlying organization and by the situational context.

Covariation of Attachment Behaviors

In addition to stability of the behavioral indices of attachment, covariation of these indices must be demonstrated, according to Maccoby and Masters (1970) and Masters and Wellman (1974), to validate the concept of attachment as a trait or central motive state. Although we consider this criterion—at least as narrowly conceived—to be irrelevant to the ethological—evolutionary theory of attachment, it is useful to consider some of its implications. There may be at least three ways in which two or more behaviors might be conceived as covarying: (1) they tend to occur together; (2) they are positively and significantly correlated; and (3) they are organized together in stable ways, which may result in a complex matrix of positive and negative correlations. We consider that only (3) is relevant to our concept of attachment.

Concurrence of two behaviors would certainly satisfy the requirements of a trait model (although one of the behavioral indices would be redundant if concurrence was invariable or nearly so). Thus, proximity to the mother and touching her concur; it is impossible for a baby to touch his mother if he is not already close to her, although of course he may be close to her without actually making physical contact. Another example comes from face-to-face behavior in the early months (Blehar, Lieberman, & Ainsworth, 1977). Smiling, vocalizing, and bouncing often occur together when an infant is face-to-face with his mother. However, if one accepts the definition of attachment behavior as behavior that promotes proximity to an attachment figure (or caregiver), it is evident that concurrence cannot be required as a criterion. Thus it can be demonstrated that both crying and smiling promote proximity, but they are obviously unlikely to occur at the same time.

Maccoby and Masters (1970) specified positive correlation among behavioral indices as their criterion of covariation. Obviously concurring behaviors, even when the concurrence is only partial, will be positively correlated more or less strongly. Even though behaviors never concur, they might nevertheless be positively correlated if the same individuals who show one behavior in one kind of situation also tend to show the other in another kind of situation. Thus, for example, babies who respond positively when picked up and held by the mother also tend to greet her positively when she returns after a brief absence from a familiar environment (Stayton & Ainsworth, 1973). In our strange-situation findings, other examples may be found. For example, infants who seek proximity/contact with their mothers in the reunion episodes also tend to resist any attempts by their mothers to put them down after being held for “too short a time.”

On the other hand, there are many instances in which behaviors that may be classed as attachment behaviors (in that they are proximity-promoting) are negatively rather than positively correlated, and hence speak against the concept of attachment as a generalized trait. For example in the strange situation, measures of distance interaction are negatively correlated with both proximity seeking and crying, even though they include such obvious proximity-promoting behaviors as smiling and vocalization. At home both positive and negative greeting of the mother after an absence may be considered proximity promoting, for behaviors such as smiling, vocalization, reaching, and approach are classed together as positive, while crying is the essential feature of the negative greeting. Nevertheless, positive and negative greetings are negatively correlated; not only are they rarely concurrent, but the children who tend to show negative greetings, relative to other children, infrequently tend to show positive greetings (Stayton & Ainsworth, 1973). Furthermore, some attachment behaviors are positively correlated with behaviors antithetical to proximity promotion; thus at home positive response to being held is positively correlated with positive response to being put down (Ainsworth, Bell, & Stayton, 1971), and in the reunion episodes of the strange situation, proximity/contact seeking is positively correlated with resistant behavior (because they concur in Group-C babies). These few examples suffice to reflect the complex pattern of intercorrelations that may be found among the behaviors infants and young children direct toward their attachment figures.

It is nevertheless possible to conceive of covariation in terms of behavioral organization rather than as a matter simply of concurrence or positive correlation among all behavioral indices. As Blurton Jones and Leach (1972) commented: “Ethologists, asking themselves what they mean by words like ’attachment’, find that the only use for such a term is as shorthand for a number of behaviour items which vary together, or are found to be related together in a more complex way in a causal system [italics ours] [p. 218].” We believe that the data presented in this volume indeed suggest that the behavioral items relevant to attachment are related in a complex but systematic way that suggests an underlying “causal system.” They do not, however, suggest a unitary, generalized trait.

Let us not, however, belabor the issue of attachment as trait. It is now generally agreed that attachment cannot be conceived as a trait or central motive state (e.g., Coates et al., 1972b; Masters & Wellman, 1974), although it is erroneous to conclude that the concept of attachment is therefore an invalid concept, as others have done (e.g., Rosenthal, 1973). We should like, however, to pursue the issue of covariation in order to discuss further the implications of intercorrelational analysis of behavioral measures of attachment. We have three main points to make.

First, the dimension of strength of attachment (or strength of attachment behavior), which is so focal to the trait concept, is of relatively little importance for ethological—evolutionary attachment theory. We can think of only two contexts in which strength of attachment is relevant. If one were attempting to distinguish a principal attachment figure from other supplementary and secondary attachment figures, we can at present conceive of no criterion other than the strength of preference for one over the others, and no way to assess the strength of preference except the strength of attachment behaviors displayed to one in comparison with another in a free-choice situation. (Even then, one could argue that one set of attachment behaviors is more relevant than others for an assessment of such preference.) Or in a practical situation in which the issue is whether or not to remove a child from his natural parents and place him in a foster or adoptive home, it might be of moment to ascertain whether he has become strongly enough attached to his parent(s) that it would be more traumatic to him to be separated from them or to remain with them. Certainly, in the context of a paradigm that views attachment as specific to the figure to whom one is attached, it is not pertinent (as Maccoby and Masters suggested) to consider the number of figures to which attachment behaviors are directed as an index of the strength of attachment. Clinical findings suggest that the relationship a child has with his principal attachment figure may to some extent color his subsequent relations with other attachment figures, but this is a proposition that requires further research. In the meantime the conservative position would be to assume that the relationship with each attachment figure depends upon the history of the interaction between the child and that figure.

Second, the most conspicuous dimension that has emerged so far in our attachment research is not strength of attachment but security vs. anxiety in the attachment relationship. This does not imply substitution of degree of security for degree of strength in a unidimensional concept of attachment. On the contrary, insofar as individual differences are concerned, we conceive of security-anxiety as being only one dimension in terms of which an attachment relationship might be assessed. Our concept of patterning has implicit in it that there may well be several other dimensions relevant to such assessment. Obviously, from the data reported in this volume, avoidance and conflict relevant to close bodily contact constitutes a second dimension. Only our limited number of subjects has held us back from identifying other dimensions. In short, our concept of attachment is multivariate insofar as individual differences are concerned. A multivariate model implies complexity in the correlations among behavioral indices of attachment and not a simple “positive correlation” model.

Third, any frequency measure of behavior implies an assumption that all instances of the behavior in question are equivalent. This assumption is highly questionable in many cases. Let us consider approach behavior, for example. As Tracy, Lamb, and Ainsworth (1976) have pointed out, it cannot be assumed that a given instance of approach behavior serves the attachment-behavioral system rather than some other system, such as food seeking, exploration, or mere friendly affiliation, unless the context—both behavioral and environmental—is taken into account. We cannot even assume that all instances of approach to an attachment figure serve the attachment-behavioral system; a child may approach his mother for a variety of reasons on different occasions—for example, because she offers him food, because she shows him an interesting toy, or because he wants to be near to her. Even among instances of approach that may be assumed to serve the attachment system, there is a miscellany of contexts in which approach might appear—for example, following a departing figure, greeting a returning figure, approaching when frightened, and approaching because the current proximity set-goal has been exceeded. A total score consisting of the frequency of all approaches within a given period of observation tends to contain a great deal of “noise.” Because frequency measures have commonly been used in studies of both attachment and dependency, it is not surprising to find that intercorrelations among behavioral indices tend to be low.

Such difficulties may be reduced in several ways: (1) by controlling the context in which behavior occurs, as in the design of the episodes of the strange situation; (2) by taking context into account when devising measures, rather than relying on frequency alone, as in our categorical measures of strange-situation behavior; (3) by including within one measure several behaviors that may be considered equivalent, as we have done both in our categorical measures of strange-situation behavior and in some of our measures of behavior at home—for example, responses to being picked up and being put down; and (4) by using ratings rather than frequency measures of behavior, presumably because an intelligent rater using a well-designed rating scale tends not to give equal weight to all instances of a behavior but to take context and behavioral equivalents into account; and finally, and infinitely more difficult, (5) by abandoning correlational methods in favor of a detailed analysis of the environmental and behavioral contingencies of each item in a prolonged sequence of interactions—a type of analysis that Gewirtz (1961) has proposed. So far such detailed analyses have been undertaken only for very brief sequences of interaction between mother and infant in a face-to-face situation (Brazelton, Koslowski, & Main, 1974; Stern, 1971). Considering the bewildering complexity of data yielded by such detailed analyses, it seems reasonable that they be confined to behavior in specific contexts and that, for a study of attachment behavior, they be confined to contexts known in advance to elicit such behavior with fair consistency.

Giving no credence to the concept of attachment as a trait, some of those approaching attachment from a social-learning position (e.g., Cairns, 1972; Gewirtz, 1972a, 1972b; Rosenthal, 1973), rather than turning toward attachment as an organizational construct, have argued that the study of the phenomena of attachment requires no construct at all but can best be understood as stimulus—response contingencies in the interaction of an infant with his mother or other caregiver. Before discussing this proposal let us review the findings earlier reported in Chapters 8 and 9 on the relationship between strange-situation behavior and maternal behavior.

Relationship Between Strange-Situation Behavior and Maternal Behavior

Because our hypothesis is that different experiences in interaction with the mother are largely responsible for qualitative differences in infant—mother attachment, the relationship between maternal behavior and patterns of strange-situation behavior is of particular interest. Of most relevance to our hypothesis are studies of maternal behavior prior to or at least contemporaneous with the strange-situation assessments of patterns of infant behavior. Here we are concerned solely with maternal behavior at home or in other “uncontrolled” situations, for the behavior of the mother in the strange situation was at least partially controlled by instructions and by the structure of the situation, so that there was relatively little scope for individual differences to be manifested.

The findings reported in Chapter 8 for our Sample 1 are of particular importance because they are based on extensive observations of mother—infant interaction at home throughout the first year of life. In comparison with the mothers of A and C babies, the mothers of Group-B infants were found to be more sensitively responsive to infant signals and communications, including crying signals. In the first quarter of the baby’s first year, their sensitivity to signals was specifically shown in their behavior relevant to feeding, in their contingent responsiveness in face-to-face situations, and in their “tender, careful holding” when in close bodily contact with the baby. They were relatively mobile in emotional expression and tended to lack rigidity and compulsiveness in dealing with the baby throughout the first year. When rated in regard to fourth-quarter behavior, they were also found to be psychologically accessible to their infants, accepting rather than rejecting, and cooperative rather than interfering. They continued to be responsive to infant crying signals, and showed more affectionate behavior when in contact with their babies than did the mothers of non-B babies.

Group-A mothers were clearly more rejecting than non-A mothers; they more frequently had their positive feelings toward the infant overwhelmed by anger and irritation. They also expressed their rejection in terms of aversion to close bodily contact with their infants. They gave them more unpleasant experiences in the context of bodily contact. They showed a relative lack of emotional expression, which was interpreted as reflecting a way of controlling the expression of anger. They were rigid and compulsive in dealing with their babies. Their insensitivity to infant signals, as well as their rigidity, seems to have fed their frequent tendencies to interfere with the baby’s activity in progress.

Group-C mothers, like Group-A mothers, were relatively insensitive to infant signals, but they were clearly less rejecting. They showed no aversion to close bodily contact; yet they were inept in holding their babies and manifested little affectionate behavior when in contact with them, but rather used holding time largely for routines, even in the fourth quarter.

Studies that followed mother—infant dyads beyond the strange situation into the babies’ second or third years of life report subsequent maternal behavior that is highly consistent with our reports of maternal behavior during the first year. Thus Bell (1978) reported that Group-B mothers, in comparison with non-B mothers, tended to be more positive and appropriate in their interactions with the child, to manifest more positive affect, and to have superior communication. Tomasini (1975) found that maternal anger, rejection, insensitivity, and lack of emotional expressivity when the child was 21 months old were significantly correlated with strange-situation classification at 12 months; A/C mothers showed these behaviors most conspicuously and B3 mothers the least. Tolan (1975) confirmed that the facial expressions of B3 mothers while watching their toddlers were more expressive of both pleasure and a wide range of emotions than were the expressions of mothers of A/C babies. Main (1977a) found mother avoidance in the strange situation to be positively correlated with the following maternal behaviors at 21 months: anger, avoidance of proximity to and contact with the child, and lack of emotional expressivity.

Connell (1976) found that Group-B dyads had more interaction and longer bouts of interaction than did non-B dyads. In this follow-up of one sample at 14 and 16 months, he found no differences between Group-A and Group-B mothers in restrictiveness, ignoring the child’s signals for interaction, or anger. The latter two findings are disparate with the findings of the other studies. In regard to ignoring signals, we would suggest that Connell’s mothers, who were visited for 1 hour on two occasions, may have been making an effort to “do a good job” and hence may have been more than usually responsive to signals, whereas our mothers, who were observed more frequently and for longer periods of time (as well as Bell’s mothers, observed in the course of eight long visits), could not keep up such unusual effort; hence significant differences in maternal responsiveness to signals and to initiations of interaction were more likely to emerge. As for anger, Main (personal communication) acknowledges that maternal anger is very difficult to assess. Overt angry display tends to be inhibited in the presence of observers, and this is especially the case in the relatively brief span of a laboratory session or in a first or even second home visit. When anger is under tight control, it can only be inferred through subtle or indirect cues, as was the case in Tomasini’s study.

In summary, we may conclude that different patterns of infant strange-situation behavior are associated with different constellations of maternal behavior both before the strange situation and subsequent to it. During the strange situation, however, maternal behavior was controlled both through instructions and through the structure of the episodes themselves. Therefore, in that situation infant behavior was largely freed from its usual contingencies with maternal behavior. Nevertheless individual differences in infant behavior emerged under these circumstances that, although consistent with individual behavioral differences shown in previous interaction with their mothers, could not be attributed to individual differences in the contingencies provided by maternal behavior in the strange situation. This kind of continuity in patterns of infant behavior, despite control of maternal behavior, suggests that the determinants of infant behavior toward an attachment figure include an inner organizational component, as well as situational determinants. We have no doubt that the long experience an infant has in interaction with his mother in the course of his first year of life is chiefly responsible for the way in which he organizes his behavior toward her; but the resulting organization becomes to some extent independent of the particulars of his interaction with her in any given situation.

Attachment as Distinguished from Attachment Behavior

In Chapter 1 we distinguished between infant—mother attachment and an infant’s attachment behavior. By attachment we mean the affectional bond or tie that an infant forms between himself and his mother figure—a bond that tends to be enduring and independent of specific situations. By attachment behavior we mean the class of behaviors that share the usual or predictable outcome of maintaining a desired degree of proximity to the mother figure—behaviors through which the attachment bond is first formed and then later mediated, maintained, and further developed. Further, we refer to the attachment-behavioral system, which implies that the behaviors that may be classed together as attachment behavior come to operate systematically together. Specifically, the behavioral system in question is highly responsive to situational factors. Thus some situations activate the behavioral system at higher levels of intensity than other situations. The intensity of activation of the system may affect not only the intensity with which a specific behavior is shown, but also which specific attachment behaviors are activated—whether smiling or approaching or tightly clinging, for example. Furthermore, the attachment-behavioral system is only one of a number of behavioral systems, and the extent to which it is manifest in behavior is conceived to depend on its intensity of activation relative to the intensity of activation of other behavioral systems that may be either in competition or compatible with it. Thus whereas attachment behaviors may be manifested only intermittently and are closely tied to situational determinants, attachment as a bond is conceived as more or less constant and little affected by situational factors (except perhaps over a very long period of time). The fact that a baby is busily exploring his environment at time A and showing no overt attachment behavior does not mean that he is not attached to his mother at that time, or that he is less attached than at time B when he is alarmed, for example, and wants to be in close physical contact with her.

These definitions and distinctions are not shared by most of those who have approached a study of attachment from a social-learning point of view. They neither distinguish between attachment and attachment behavior nor espouse the construct of a behavioral system. Those who explicitly or implicitly view attachment as a trait (or general motive)—and this includes those for whom the dimension of strength or intensity of attachment has primary salience—consider attachment behaviors as indices of attachment, but then define attachment solely in terms of its indices. Those who hold that infant—mother attachment is neither more nor less than the stimulus—response contingencies implicit in mother—infant interaction explicitly disclaim the need for a construct of attachment in distinction from attachment behavior.

It seems to us self-evident that infants become bonded to attachment figures not only in the human species but also in many other species. It seems to us unnecessary to “validate” that such a phenomenon exists. The problem is to understand how the bonding takes place, how different experiences affect the nature of the bond, and what effect differences in the nature of the bond have upon subsequent development. To be sure, different constructs of attachment may imply different hypotheses about the formation, nature, and effects of the attachment bond, and these are subject to the usual kind of hypothesis-testing procedures. But to say that the notion of attachment—in general, and presumably as a phenomenon rather than as a particular set of hypotheses about its nature—is proven invalid or unnecessary, as Maccoby and Feldman (1972), Rosenthal (1973), and Weinraub, Brooks, and Lewis (1977) have concluded, appears to us as absurd as it would be to deny the existence of some kind of thermal regulatory system in mammals.

Let us restate our view of attachment as distinct from attachment "font-size:9.0pt">We infer the existence of an attachment from a stable propensity over time to seek proximity and contact with a specific figure, even though attachment behavior may appear only intermittently, or—in the case of major separations—may be absent for long periods. The term “attachment” refers to the propensity, whereas the term “attachment behavior” refers to the class of diverse behaviors which promote proximity and contact, at first without discrimination of figure, but later with increasing specificity in regard to the figure(s) to whom the child is or is becoming attached.

It is further suggested that it is useful to view attachment—as a construct—as an inner organization of behavioral systems which not only controls the “stable propensity” to seek proximity to an attachment figure, but also is responsible for the distinctive quality of the organization of the specific attachment behaviors through which a given individual promotes proximity with a specific attachment figure. Such an hypothesis implies some kind of stable intraorganismic basis for individual differences in the organization of attachment behaviors. Such a relatively stable inner organization must be conceived as interacting with environmental conditions and other “situational” intraorganismic conditions—neurophysiological, hormonal, and receptor processes—to activate, terminate, and direct attachment behavior in any specific situation. It is conceived as a hierarchical organization that permits more or less interchangeable behaviors to be directed by any one of several general plans or strategies that may be specifically tailored to fit the requirements of different situations. A hierarchical organization of this kind suggests internal structure [Ainsworth, 1972, p. 123].

Obviously this kind of hypothesized intraorganismic structure is alien to social-learning-theory formulations—at least to those that have attempted to grapple with the notion of attachment, whether from their own data, other people’s data, or without any data at all.

Implicit in the aforementioned view is the notion that there are individual differences in the intraorganismic structure that constitutes attachment—differences attributable to differences in long-term interaction with the attachment figure. It is to such differences that we refer when speaking of differences in the quality of attachment.

To assert the theoretical distinction between attachment behavior and attachment, we have often used the term “attachment relationship” when referring to the bond. Hinde (1976a, 1976b) views a “relationship” between two individuals as an abstraction from a multiplicity of interactions between them. It is anchored in neither of the individuals concerned but is a convenient construct for characterizing the nature of the interactions between them. Ordinarily, an attachment relationship would be a relationship between two individuals who are attached to each other.

It is conceivable, however, that an infant might be attached to his mother but that the mother might not be bonded in a complementary way to her infant—as perhaps in the case of Harlow’s (1963) “motherless mothers.” It is also conceivable that a mother might be attached to her infant but the infant not bonded to her; indeed this is likely before he has become attached to his mother. Further, an infant may behave with reference to his mother figure in certain ways consonant with the nature of his attachment to her during certain periods when he is not in interaction with her—as, for example, when he is exploring away from her, using her as a secure base, or when he is separated from her but attempting to regain proximity to her. Likewise a mother may behave with reference to her baby in certain ways consonant with the nature of her attachment to him when they are separated.

As a consequence of her relationship to her baby, a mother has an inner representation of him that is not contingent upon his actual presence; and in the course of his development, an infant comes to have an inner representation of his mother. The inner representation that each member of the dyad has of the other is a consequence of the relationship which each has with the other, and these are plainly not identical. Similarly, and underlying the inner representation of the partner, each has built up some kind of intraorganismic structure that we have hypothesized as attachment. Although such a structure can be conceived to be influenced also by the interactions that constitute the relationship, it is obviously different for each partner. On these grounds we hold that the attachment of child to mother is by no means identical to the attachment of mother to child, even though they both share an attachment relationship.

Attachment in Older Preschoolers

As we pointed out in Chapter 1, it is a misconception to believe that Bowlby was not concerned with the development of child—mother attachment beyond toddlerhood. To be sure, there had been very little research relevant to the later stages of development of child—mother attachment; thus, his formulations (1969, 1972) of such development were necessarily sketchy and programatic. He acknowledged that proximity-seeking behavior becomes less conspicuous in the child’s interaction with his mother as development proceeds. He did not equate this, however, with an attenuation of attachment itself. He emphasized the significance of the develoment of “working models”—inner representations—that the child builds up both of himself and of his attachment figure, and the development of the capacity for making plans, both of which developments begin no later than the second year of life. In the final phase of development, in which a “goal-corrected partnership” is formed and sustained, the partners develop “a much more complex relationship with each other” than is characteristic of a 1-year-old (Bowlby, 1969). In this phase the development of the capacity to take the perspective of another is crucial. As this capacity develops, a child gains insight into his mother’s plans, set-goals, and motivations, so that he can form increasingly complex plans that include influencing his mother to fit in with his plan. Indeed Bowlby’s notion of “partnership” implies that both partners can negotiate mutual plans that comprehend the set-goals of each.

Obviously a child’s cognitive development profoundly changes the specifics of the behaviors that mediate attachment in the older preschooler, as well as in still older children and in adults. Nevertheless, Bowlby (1973) conceived of the attachment of a child to his mother as enduring through a substantial part of life, even though it undoubtedly becomes attenuated, especially in adolescence, and supplemented with other relationships, including a number (a limited number) of other attachments. Furthermore, the fundamentally proximity-promoting nature of attachment behavior does not altogether disappear with increasing sophistication. Bowlby (1973) makes clear that even in infancy, proximity to the mother figure may come to be conceived in terms of her apparent availability—the degree to which she is believed by the child to be accessible to him and responsive to his signals and communications. Increasingly, therefore, proximity becomes less a matter of literal distance and more a matter of symbolic availability. Nevertheless, even in adult life, when the attachment system is activated at a high level of intensity—for example, by severe illness or disaster—the person seeks literal closeness to an attachment figure as an entirely appropriate reaction to severe stress.

To our knowledge, the only body of research that has picked up the threads of Bowlby’s discussion of the development of attachment beyond the first year or two of life is that conducted by Marvin and his associates (Marvin, 1972, 1977; Marvin, Greenberg, & Mossler, 1976; Mossler, Marvin, & Greenberg, 1976), discussed in Chapter 10. They have shown that shifts in strange-situation behavior from one age level to another are associated with certain cognitive acquisitions. In particular, they have shown that the ability to take the perspective of another—at least in simple conceptual tasks—generally emerges between the third and fourth birthday. In recent, as-yet-unpublished research, Marvin (personal communication) has been investigating the way in which a child and his mother may negotiate a mutual plan—specifically, one in which the mother’s plan (suggested by instructions) is to leave the child alone in a laboratory playroom for a few minutes. He demonstrated that when a mutual plan is negotiated, a 4-year-old shows no separation distress, although if (again according to instructions) the mother does not negotiate in response to the child’s attempts to do so, the child is upset. The distress seems more likely to be angry distress, as a result of the mother’s arbitrary unresponsiveness to his attempts to communicate his plan to her and to influence her plan, than attributable to mere separation. Furthermore, in the case of dyads who do successfully negotiate a mutual plan, a common compromise is the mother’s acceding to the child’s request to leave the door open, if only by “just a crack.” The implication is that the child does not require his mother’s actual presence as long as he feels that she would be accessible to him if he wanted to go to her. All of this is clearly in line with Bowlby’s hypothesis about developments in the later preschool years.

As we have already pointed out, the strange situation does not activate attachment behavior at the same high level of intensity in 3- and 4-year-olds as in 1-year-olds. Consequently the patterning of behavior reflected in our classificatory system, dependent as it is on high-intensity activation of the attachment behavioral system (and also upon associated avoidant and resistant behavior), does not occur in older preschoolers in the same way that it does in 1-year-olds. In Chapter 10 we suggested several solutions to this problem. One solution is to use our categorical measures of interactive behavior, as Blehar (1974) did instead of employing our classificatory system—although this implies some loss of the patterning highlighted in classification. Another solution is to modify the classificatory system to make it more applicable to the behavior of the older preschoolers, as Marvin (1972) did. The other possible solutions considered in Chapter 10 involved devising new ways of assessing the attachment of older preschoolers to their attachment figure(s). Clues that might be useful might be found in the results of investigators such as Main, Bell, Connell, and Matas (reported in Chapter 9), who examined individual differences in later behavior of children who had been assessed in the strange situation at the end of the first year. Similarly, Lieberman’s study (see Chapter 10) might give leads to variables relating to mother—child interaction at home that might substitute for strange-situation variables in the older preschooler. Marvin’s current unpublished work seems likely to yield suggestions for ways in which laboratory assessments might be made more appropriate for the older child.

All of the foregoing implies that the situation-specific behaviors that reflect important qualitative differences in attachment in 1-year-olds may be replaced by a number of equally situation-specific behaviors in older preschoolers. Such a suggestion is akin to the concept of “transformation,” proposed by both Maccoby and Feldman (1972) and Lewis and his associates (Lewis & Ban, 1971; Weinraub, Brooks, & Lewis, 1977); but it demands something less simplistic than their assumption that “proximal” behaviors become transformed into “distal” behaviors in the course of development. Both proximal and distal behaviors are involved in mother—infant interaction throughout the first year of life, and both may be viewed as contributing to the formation and later mediation of the attachment bond. Even though the relative balance between proximal and distal behaviors shifts with increasing age, the distal behaviors remain those that emerge only intermittently and for the most part under conditions of low-level activation of the attachment system, and hence less useful as indices of qualitative differences in attachment, even in the older preschool child. A more important consideration is that the most crucial differences in patterning, even in the 1-year-old child, pertain neither to proximal nor to distal attachment behaviors but to the way in which such behaviors are organized together with key nonattachment behaviors—specifically those that reflect avoidance of or resistance to the attachment figure. Our prediction is that those patterns of behavior in the older preschooler that will be found to link up with earlier strange-situation-based differences in attachment quality are patterns that include negative nonattachment behavior related to avoidance and resistance—and thus to anxiety and anger.

Attachments to Figures Other Than the Mother

One of the reasons that the concept of attachment has captured so much of the interest of developmental researchers and clinicians regardless of their initial theoretical starting-points is the implicit hypothesis that the nature of a child’s attachment relationship to his mother figure has a profound effect on his subsequent development. (We, as well as Bowlby, emphasize the term “mother figure” to assert our belief that the child’s principal caregiver in infancy and early childhood is most likely to become the principal attachment figure—and thus the most important initial influence on subsequent development—whether such a figure be his natural mother, a foster or adoptive mother, a grandmother, a “nanny,” or father.) In the beginning stages of research into attachment, it made good sense to focus on attachment to the mother figure, without thereby implying that attachments to other figures were of no consequence, or that other later relationships, whether or not they could be classified as attachments, had little significance in influencing a child’s development. It ought to be possible to assert the importance of research into other attachments and other relationships without thereby impugning the value or validitiy of the attachment theory. Thus it seems naive of Willemsen and associates (1974) to have concluded that their finding that the father serves as an attachment figure in the strange situation essentially as the mother does demonstrates the invalidity of attachment theory. It is undeniable that the young child, and indeed also the young infant, develops within the framework of a “social network,” as Weinraub, Brooks, and Lewis (1977) have eloquently described. Undoubtedly it is important to trace through the characteristics and effects of relationships other than the child’s attachment to his mother figure. It is clearly important to investigate children’s relationships with siblings, playmates, teachers, and so on. But this does not mean that attachment theory is of no value.

It seems to us to be of more urgent importance, however, to investigate relationships an infant has with those figures who share the caregiving role with the principal caregiver (usually the mother)—whether these figures include the father, other adults resident in the household, or supplementary or substitute figures such as day-care personnel, long-term “babysitters,” and the like. We need to take advantage of cross-cultural studies and “experiments of opportunity” within our own culture in order to investigate how different patterns of infant care affect the attachments of the infant to those involved in a caregiving role, and how variations on the theme of principal caregiver with supplementary and secondary figures show support and reinforcement for each other, compensatory function, or conflict; and we need to show how at least the more common of the many possible variations affect the development of the child.

Let us pose a few of the questions that readily emerge when one contemplates investigating a child’s social network, while still concerning oneself only with his major caregivers. Can a “good” relationship with the father compensate for a conflicted and anxiety-provoking relationship with the mother? Can a few hours of high-quality interaction with the mother compensate for the fact that she leaves the major responsibility for daily infant care to substitute or supplementary figures? If both parents share equally in the care of the infant or young child, does he become equally attached to both, and what influence does this pattern have on his subsequent social development? Does the nature of the attachment a child has to his principal caregiver (mother figure) affect his relationship with other attachment figures, and in what ways? Or is the nature of his relationship with different attachment figures affected only by the nature of his interaction with each figure in isolation from and unaffected by his relationship with other figures? Does a child form significant attachment relationships with day-care personnel, and how do such relationships affect his relationship with his principal attachment figure and indeed his subsequent development? Each of these questions would require very time-consuming and difficult research projects before we begin to know as much about them as we already know about infant—mother attachment, which indeed is all too little.

In short, the fact that ramifications of research into a wide variety of attachments and other relationships have been indeed sparse denies neither the importance of undertaking such research nor the commonsense of beginning with the infant’s attachment to his principal caregiver, which, across many cultures and throughout history, implies attachment to his mother.