Review of Strange-Situation Studies of Two- to Four-Year Olds - Measures and Methods of Assessment

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

Review of Strange-Situation Studies of Two- to Four-Year Olds
Measures and Methods of Assessment


The strange situation was designed especially for 1-year-olds. Because it depends on the baby’s being old enough both to have become attached to a mother figure and to have become adept at some form of locomotion, it is not applicable to infants much younger than 11 months of age. Indeed, it has not been used—at least not without substantial modification—with younger infants. It has been used, however, for children older than 12 months—in particular, for children between the ages of 2 and 4 years, inclusive.

This volume has assembled all the available information known to us about the behavior of 1-year-olds in the strange situation, together with its correlations with behavior elsewhere. On the basis of this, we have a reasonably good normative picture of how white, middle-class American infants behave in it. Furthermore, the information contained in Chapters 7, 8, and 9 gives us a reasonably satisfactory basis for interpreting the significance of individual differences in strange-situation behavior. It could be assumed that the norms would differ with increasing age, however, if only because we could expect that developmental processes would bring about increasing tolerance for brief separations from an attachment figure, and that attachment behavior activated by the situation would be both less intense and somewhat different in form. This being so, it is by no means certain that individual deviations from the norm may be interpreted in the same way as they would be with 1-year-olds. We must rely on empirical evidence on both scores—both to establish norms for children aged more than approximately 12 months and to examine the correlates of individual differences among children of different ages.

Eight studies to date have yielded evidence relevant to these issues. They may be classified as follows, with some studies classifiable under more than one heading: (1) those that provide normative comparisons for different age groups; (2) those that explore the developmental processes associated with age differences; (3) those that are concerned with antecedent conditions that might influence attachment and hence strange-situation behavior; and, finally (4) those that use the strange situation as a basis for identifying qualitative differences in child—mother attachment, and consider these as antecedents possibly related to other aspects of subsequent development.

Normative Studies of the Development of Strange-Situation Behavior

Three studies have specifically addressed the issue of developmental changes in strange-situation "font-family:"LinLibertine",serif">Marvin’s (1972) study used a cross-sectional design to explore certain aspects of the development of child—mother attachment beyond 12 months of age. In particular, he was interested in the transition from Bowlby’s (1969) Phase 3, in which the child is capable of active, goal-corrected behavior in maintaining his desired degree (set-goal) of proximity to his mother figure, to the Phase-4 level of “goal-corrected partnership.” His sample included 48 young children from white, middle-class families, 16 of each of three age levels—2, 3, and 4 years. He used our standard strange-situation procedure, except that at the beginning of Episode 61 the mother said, instead of merely “Bye, bye!”: “I have to make a phone call; I’ll be back!” He employed the same measures as we, and attempted to use our classificatory procedure, although he found it necessary to modify this for 3- and 4-year-olds.

In general, Marvin (1972, 1977) found that his 2-year-olds behaved in much the same way as our 1-year-olds. They tended to cry in the separation episodes, especially in Episode 6, although somewhat less than the 1-year-olds, and tended to continue to cry in Episode 7, after the stranger returned. Even more strongly than the 1-year-olds, they sought proximity to the mother in the reunion episodes, although they were more content with mere proximity and did not seek to maintain contact as strongly. More than any other group, including the 1-year-olds, they sought proximity to the mother in Episode 3, after the stranger had entered. Avoidant and resistant behavior in the reunion episodes had about the same incidence as in our sample of 1-year-olds. (Indeed, this was also the case with 3- and 4-year olds, suggesting that these may be characteristics of individual children that change little during the preschool period of development.)

The 3-year-olds seemed little disturbed by the first separation episode, and maintained their exploratory behavior at a high level until Episode 6, when they were left alone. They tended to cry while alone, but, unlike the younger children, they were reassured when the stranger entered in Episode 7. Nevertheless, they sought proximity to the mother in the final reunion episode almost as strongly as the 2-year-olds.

The 4-year-olds were the least disturbed by separation, on the whole, and maintained their exploratory play well over all episodes. They seemed content to be with the stranger when the mother was absent, but some of them were indeed distressed when the mother left in Episode 6. These distressed children begged the mother to take them with her, arguing that they did not want to be left alone; but the mother, acting on instructions, had no option but to leave the child behind—which must have seemed very arbitrary to the child. Marvin suggests that it was this, rather than acute separation distress, that caused 4-year-olds to cry in Episode 6. Those who had not begged to go with the mother were not distressed in Episode 6, greeted the mother cheerfully in Episode 8, and then went on with their play, maintaining communication with her across a distance. The distressed group stopped crying when the stranger entered in Episode 7, but when the mother returned in Episode 8 they strongly sought proximity to her, whining, angry, and demanding.

Maccoby and Feldman (1972) undertook a longitudinal study of white, middle-class children, observed in the strange situation first at 2 years of age, then later at 2½ and 3 years of age. Forty-eight children were observed in at least two of these sessions, and 23 of them were followed through to nursery school. As mentioned in Chapter 9, in conjunction with the study by Feldman and Ingham (1975), they used different measures from either ours or those of other investigators reported in this volume. Therefore their findings are not entirely comparable with Marvin’s, especially in the reunion episodes.

Those measures that seem comparable are manipulative play (our exploratory manipulation), crying, looking, and speak—smile—show (our distance interaction.) Finding avoidant and resistant behavior infrequent in their samples (or with their age groups), they did not score them as we did. Their measure of proximity was a composite measure, which proved to be highly correlated with ours in our sample when we checked it, but which they applied only in the preseparation episodes. In addition to their formal measures, they reported percentages of children showing specific, discrete behaviors in the first few moments of the reunion episodes, before the mother intervened.

They found that some behaviors increased with age: amount of manipulative play in Episodes 2, 4, 6, and 7, speak-smile-show to the mother in Episode 2 and to the stranger in Episodes 4 and 7, and attention span in play. Other behaviors decreased with age: proximity to the mother (i.e., in the preseparation episodes), crying, calling to the mother, going to the door, and proximity to the stranger in the separation episodes. In addition, in the reunion episodes “movement toward the mother” was most frequently seen at age 2, whereas 2½ and 3 the more “distal” forms of greeting were more common. These findings are generally congruent with Marvin’s.

In an earlier review of attachment and dependency, Maccoby and Masters (1970) entertained the hypothesis that attachment might be a trait or central motive state, as dependency had originally been viewed. To establish the validity of such a hypothesis, it would be necessary to demonstrate intraindividual consistencies, across measures of attachment behavior, across episodes, and across time intervals. Measures of dependency have yielded equivocal evidence of such consistencies, but measures of attachment behavior might yield more. For such an analysis, Maccoby and Feldman confined themselves to five measures assumed to reflect attachment to the mother: proximity; frequency of speak-smile-show (both only in the preseparation episodes); frequency of looking at her; crying in Episodes 4 and 7, when the stranger was present; and crying in Episode 6, when the child was left alone.

They found that proximity to the mother in Episode 3 correlated significantly with crying when mother leaves in the next episode. Looking and speak-smile-show were not correlated with either proximity or separation protest, although these two distal modes were correlated with each other between ages 2 and 2½. Similar findings for behavior with the stranger led the authors to conclude that “proximal” and “distal” attachment behaviors are independent of each other.

Neither “distal” attachment behavior nor proximity in Episode 2 was stable across time (i.e., from 2 to 2½ to 3). Significant crossage stability was found for: crying in the separation episodes, proximity to mother in Episode 3, manipulative play in Episodes 3, 4, and 7 (between 2½ and 3), speak-smile-show to the stranger (between 2½ and 3), and looks at stranger (between 2 and 3). In general the authors concluded that the child’s reaction to the stranger is more stable than “attachment to the mother.”

They next examined their data for evidence of “transformations” from one age to another in regard to the form that attachment behavior might take. The implication was that forms of behavior characteristic of 2-year-olds were “immature,” whereas those forms characteristic of older children were “mature.” Consistency would be shown if those displaying immature forms of attachment behavior most strongly at age 2 showed mature forms most strongly at age 2½ or 3, thus demonstrating that the same children were among those most strongly attached to their mothers at all age levels. The findings were equivocal in regard to this issue.

Feldman and Ingham (1975), in a comparison of behavior toward mother, father and an adult acquaintance as adults accompanying the child to the strange situation, used a sample of 79 2½-year-olds, as well as their sample of 1-year-olds (discussed in Chapter 9). They expected the 2½-year-olds to show less attachment behavior toward parents than did the 1-year-olds because presumably a “process of progressive detachment” had begun.2 Indeed there were fewer significant differences between the three conditions (mother, father, and acquaintance) in the case of the older sample. The exceptions are as follows: Children stood closer to their parents than to an acquaintance in Episode 2, and in Episode 3 they moved closer to the mother but not to the father; they also moved closer to the acquaintance in Episode 3, but did not achieve the same proximity to her as to the father or mother; children left by their mother with the stranger in Episodes 4 and 7 played less than those left by an acquaintance; and children accompanied by parents looked more at the stranger in Episode 3 than those accompanied by an acquaintance. There were no significant differences among the three conditions in regard to reunion behavior, nor were there such differences with the 1-year-old sample.


In regard to the two groups in which there was overlap—namely, 2- and 3-year-olds—there seems considerable congruity between the normative findings of Marvin and those of Maccoby and Feldman, even though the findings for reunion episodes are not strictly comparable. It is clear that there are developmental changes in strange-situation behavior between the ages of 1 and 4. The general trend of these changes tends to be confirmed by the study by Feldman and Ingham, and also by the work of Blehar (1974), which is reported later in another context. In general older children manifest less distress in the separation episodes than the younger children. Although 2-year-olds seem to seek proximity as strongly as, or sometimes more strongly than, 1-year-olds, proximity seeking decreases in older children. Furthermore, the maintenance of physical contact seems less important with older children at all ages studied than it was with 1-year-olds.

Although the general trend of changes in normative behavior across ages seems common to all studies, there is no similar consensus for stability of individual differences across ages. Marvin’s design did not permit an examination of crossage stability of individual differences in response to either mother or stranger. Maccoby and Feldman noted some stability in response to the stranger, but not in response to the mother. It seems likely to us that the reason for their failure to find stability of individual differences in response to the mother was that they examined discrete measures rather than patterns of behavior. We found that their measures—especially looking, crying, and distance interaction—failed to discriminate in any important way among our strange-situation classificatory groups. Main (1973)—see Chapter 9—found that discrete measures, which included both “proximal” and “distal” measures, failed to predict behavior from 12 to 21 months. In Chapter 11 we note that we found no stability for discrete measures from 50 to 52 weeks of age and that Waters (1978), in an independent study, found no stability for them from 12 to 18 months.

In Chapter 11 we see that both Waters (1978) and Connell (1976) found stability in strange-situation classification from 12 to 18 months. No one, however, has searched for such stability among 2-year-olds and older. Marvin (1972) attempted to use our classificatory system with his sample, but found that modifications were necessary because of developmental changes in behavior patterns. The problem of “transformation,” to which Maccoby and Feldman addressed themselves, would best be examined after devising new classificatory systems for older children and then investigating the relationship between our classifications at age 1 and later classifications.

Maccoby and Feldman’s hypothesis that attachment might be a trait or central motive state is a complex issue, the discussion of which is deferred until Chapter 14. Likewise, the issue of strength of attachment behavior as an index of the strength of the attachment relationship—an issue raised in their discussion of possible transformations of behavior—is considered in Chapter 14.

Developmental Processes Associated with Age Changes in Behavior

Marvin (1972, 1977) hypothesized important cognitive developmental processes as a condition both for changes in attachment behavior with age and for changes in child—mother communication. As the child moves from Phase 3 to Phase 4 in the development of attachment, he and his mother enter into a “goal-corrected partnership” (Bowlby, 1969). Although his attachment behavior becomes attenuated, the change in the nature of the child—mother relationship does not imply a weaker attachment.

Following Bowlby’s original formulation, Marvin argued as follows. Although a Phase-3 child may become increasingly capable of modifying his “plans”—that is, intentionally adjusting his behavior—in accordance with the behavior of his mother figure, he is still too egocentric (in Piaget’s sense of the term) to be able to take the perspective of another person. Consequently, in the course of his attempts to maintain his set-goal of proximity to his mother, he does not realize that she has plans and goals of her own that influence her behavior, sometimes in a direction antithetical to his own. Being unable to infer her plans, the Phase-3 child is unable to undertake any deliberate course of action designed to change her plans so that they are in greater harmony with his own.

The Phase-4 child, on the other hand, becomes increasingly capable of inferring his mother’s plans and goals, and of coordinating them with his own, both conceptually and behaviorally. Whereas the Phase-3 1- or 2-year-old has his goal-corrected attachment behavior specified in terms of literal spatial and temporal proximity to his mother, the Phase-4 child is capable of transcending literal spatial—temporal proximity in his plans, in favor of maintaining a relationship with her in more “abstract” terms. This is not to say that a Phase-4 child no longer wants literal contact with and proximity to his mother. He does sometimes, but at other times he is content with maintaining communication with her, sporadically and across a distance, secure in the knowledge that the relationship continues to exist despite periods of absence and despite lack of actual physical closeness.

Foremost in the development of a mature partnership, according to Marvin, is the development of communicative skills, including but not limited to verbal communication. Thus, although one would expect proximity-promoting behaviors to decrease with age, this would not necessarily imply a disappearance or even an attenuation of the attachment relationship, provided that it could also be demonstrated that this decrease in overt attachment behavior is associated with increasing ability in communication and in perspective taking.

Marvin therefore examined his sample of 2-, 3-, and 4-year-olds in a variety of simple cognitive tasks that implied ability either to defer gratification in the face of frustration or to take another person’s perspective (i.e., role-taking ability.) The frustration task was entitled the “Cookie Test.” The mother showed the child a cookie and told him that he could have it as soon as she finished writing a letter. She then placed the cookie out of reach but still in sight. After 3 minutes she told the child he could have the cookie. Marvin was interested in whether the child could accept the mother’s inserting one of her plans into his plan, and whether he could inhibit his goal-directed behavior (i.e., to get the cookie) in accordance with his mother’s plan. Nearly all the 2-year-olds (81%) failed to inhibit their cookie-seeking behavior, and displayed some combination of crying, reaching for the cookie or attempting to disrupt the mother’s letter writing. In contrast, 75% of the 3-year-olds and all of the 4-year-olds accepted the situation immediately and waited for mother to finish her letter. Marvin suggested that in these cases a mutual plan, incorporating the mother’s, had been implicitly agreed upon.

Four perspective-taking tasks were used. In the simplest of these the children were asked four questions of the following type: “Which do you think your Mommy would like for her birthday—a toy doll or a new dress?” The child’s response was scored as “egocentric,” indicating failure to take the mother’s perspective, if he chose the child-appropriate article instead of the adult-appropriate article in more than one case in four. None of the 2-year-olds answered any of the questions in a nonegocentric fashion, whereas 20% of the 3-year-olds and 75% of the 4-year-olds were judged nonegocentric. From this and other tasks, Marvin concluded that when a child is about 4 years old, he begins to be able to understand his mother’s perspective, and consequently to realize that she has plans of her own, to infer something of what they are, and therefore to be able to communicate with her more effectively in his attempts to get her to accept a mutual plan compatible with his own.

Now, let us consider Marvin’s interpretation of developmental changes in strange-situation behavior. He suggests that the child—mother attachment relationship is organized in much the same way in 2- and 3-year-olds as it is in 1-year-olds. Despite some obvious developmental changes, maintenance of a reasonable degree of proximity remains the major goal in relation to the mother. Separation from her, not initiated by the child himself, can disturb the goal of proximity maintenance. That this goal was disturbed by separation in the strange situation was shown not only by crying and by efforts to regain the mother in the separation episodes (especially in the case of children aged 1 and 2), but especially by the fact that they sought proximity to her in the reunion episodes. The 3-year-olds, however, may be perceived as capable of inhibiting a goal-directed behavior in order to fit in with the mother’s plan—as shown in the Cookie Test. Although being left alone in the strange situation was disturbing to most of them, they tended to be able to inhibit their proximity seeking to the mother while the stranger was present, and to wait until the mother had returned before releasing it.

Four-year-olds—having begun to be less egocentric, more capable of perspective taking, and more able to sustain a relationship on the basis of communicative skills, sharing of mutual plans, and internalized models of self and mother and their relationship—should behave differently in the strange situation. They were expected to maintain exploratory play, show little separation distress, and display little proximity-seeking behavior in the reunion episodes, but rather to be content with communication with the mother across a distance. Indeed about half of the 4-year-olds behaved in accordance with these expectations, as though they no longer had physical proximity to the mother as the overriding goal in their attachment to her.

The other group of 4-year-olds—those, described earlier, who begged to go with their mothers, who subsequently cried in Episode 6, and who were angrily demanding in Episode 8—can also be accounted for by Marvin’s model. The mother, through refusing to negotiate a mutual plan acceptable to both (when she arbitrarily left at the end of Episode 5), abandoned the very process through which the equilibrium of a goal-corrected partnership is maintained. The child cried, then, more in anger at her refusal to negotiate than in distress at being separated from her. When she returned he tried to reestablish the equilibrium that she had disturbed through behavior directed toward controlling her—in a sense reasserting her right to alter her plans by making demands on her that, under the circumstances, were impossible for her to agree to. It would be interesting to ascertain in further research whether this kind of behavior in the 4-year-old reflects an attachment relationship that is anxious despite the child’s cognitive gains, or whether it is entirely situational as Marvin’s account implies.

Antecedent Conditions Possibly Affecting the Attachment Relationship

Three studies of preschool-age children were concerned with the possible effects of antecedent conditions. One of these, by Serafica and Cicchetti (1976), was concerned with the effect of retarded cognitive development attributable to a genetic anomaly. The other two examined the effect of different conditions of rearing. Maccoby and Feldman (1972) compared Israeli kibbutz-reared children with the home-reared children of their American sample. Blehar (1974) compared day-care and home-reared children.

The sample studied by Serafica and Cicchetti (1976) consisted of 12 children with Down’s syndrome and 12 normal controls—all white, middle-class, and family-reared. The dyadic relationship between child and mother was explored through behavior in the strange situation when the children were about 33 months old. Few significant differences were found between the groups. The children with Down’s syndrome cried less in the separation episodes than did their controls, and in Episode 8 they sought contact with their mothers less often. The children of the control group, on the other hand, vocalized more in Episodes 5, 6, and 7, and these vocalizations could be construed as relating chiefly to the separation.

These findings are surprising, for one would have expected the children with Down’s syndrome, who were retarded in development, to behave more like 1-or 2-year olds and thus, in comparison with their 3-year-old controls, to cry more in the separation episodes and to seek more proximity to the mother in the reunion episodes. Serafica and Cicchetti discussed three possible interpretations of these findings: “(1) a lag in the development of attachment among Down’s syndrome children; (2) a difference in the strength of attachment to the mother between the Down’s syndrome and normal groups; and (3) differential interpretations attached by the two groups to being alone in the strange situation [p. 147].” We cannot here report their considerations, except to say that they believed the third explanation to be the most tenable.

Later, Cicchetti and Sroufe (1976) suggested that children with Down’s syndrome are generally low in reactivity to stressful situations and that this appears to be related to their hypotonicity, so that more strength of external stimulation is necessary to produce a given amount of physiological excitation. They would be the last to argue that affective expression is unaffected by cognitive factors, but nevertheless it seems to us that in this study the role of the cognitive factors is obscured by the hypotonicity to which they have drawn our attention.

In the second part of their 1972 monograph, Maccoby and Feldman reported a study of the behavior of 20 Israeli kibbutz-reared 2½-year-olds and the 35 American home-reared children whom the authors had observed at 2½. The kibbutz children were accompanied by their mothers in the strange situation, and the procedure was essentially the same, except that the arrangement of toys was somewhat different and the room was smaller, so that the groups could not be compared in regard to the proximity measures.

Because of the substantial differences in rearing conditions, it might have been expected that the Israeli and American groups would differ significantly in their strange-situation behavior, but the similarities between the groups were much more impressive on the whole than were the differences. The differences were as follows: The kibbutz children tended to display less touch—cling behavior to their mothers in Episode 3, less vocalization in greeting their mothers in the reunion episodes, more activity in Episodes 3 and 4, less looking at the mother in Episode 2 but more in Episode 3, and less speak-smile-show behavior to the stranger in Episodes 3 and 7. Although it might have been expected that the kibbutz children would show less upset over separation than the Americans, this was not so. It might also have been expected that they would be more readily accepting of strangers, but the findings suggest that they accepted strangers less readily. The authors expected that kibbutz children would be more homogeneous in their strange-situation behavior than the Americans, but this did not prove to be the case. Maccoby and Feldman (1972) comment: “Either kibbutz environments are not as uniform as one might suppose, or else . . . there are strong individual differences among kibbutz children that emerge despite the environmental uniformities that do exist [p. 80].”

In any event, Maccoby and Feldman confirmed the finding that has emerged consistently from research into kibbutz rearing—namely, that such rearing does not prevent a child from becoming attached to his parents. On the other hand, qualitative differences in the child—mother attachment relationships within either the American or the Israeli samples were not explored, nor were the groups compared in respect to such differences.

One of the reasons that there has been much interest in the effect of kibbutz rearing is that it is perceived as having some parallels with group day care, which has become a very significant political issue, as well as being a matter of much concern to mothers who wish or need to work in full-time jobs. One of the issues is the effect of full-time group day care on social development in general, and on the child’s attachment to his mother in particular. Because there is a dearth of adequate measures of social development in infancy and early childhood, the strange situation suggests itself as a possibly useful instrument for the whole relevant age range, despite the fact that it was designed to investigate the attachment relationship in 1-year-olds.

One of us (Blehar, 1974) used the strange situation to assess the effect of day care on young children aged approximately 30 and 40 months at the time of observation, comparing their behavior with a matched control group of home-reared children. The total sample consisted of 40 middle-class children, 20 of whom were in full time group day care of a traditional nursery-school type, and 20 of whom were home-reared children matched in sex and age. Several checks were made as to whether the home background of the day-care children and their controls differed; all such checks were negative. The 20 day-care children were divided into two groups that differed in the age at which they first began day care, 10 children having begun at age 2, and 10 at age 3. Five months after day care had begun, and at an equivalent age for the home-reared controls, all groups were observed in the strange situation. Because the groups were initially equivalent in home-background variables, the implication was that any significant differences that emerged 5 months later could be attributable to the effects of day care.

The total day-care group was found to differ significantly from the home-reared group in regard to a number of strange-situation variables. The day-care children interacted less with their mothers across a distance in Episode 2; they cried more in the separation episodes; they displayed more oral behavior, especially in Episode 7; they avoided and resisted their mothers in the reunion episodes. Furthermore, they sought less proximity to the stranger, avoiding her increasingly from the earlier to the later episodes, whereas the home-reared children, having been more wary of the stranger in Episode 3, became increasingly accepting of her as the situation progressed.

Even more interesting than these main group differences was the fact that the 30- and 40-month-old day-care groups showed two distinct patterns of strange-situation behavior. The children in the 40-month group, who had started day care at age 3, were the most disturbed by separation. They cried more than children in any of the other groups, searched more for the mother in the separation episodes, and explored less. They sought proximity to the mother in the reunion episodes more than either their 40-month-old home-reared controls or the 30-months-old day-care children, and in this they resembled the 30-month-old home-reared children. Resistant behavior directed toward the mother in the reunion episodes was stronger in the older day-care group than in any of the others, and occurred in 60% of the children. Resistance to the stranger was also higher among the children of this group. In short, the older day-care children resembled Group-C 1-year-olds in the pattern of their strange-situation behavior; they were anxious, resistant, and conflicted in their relations with their mother, although the intensity of their reactions was less.

The 30-month-old day-care children were particularly conspicuous for avoidance of the mother in the reunion episodes. Furthermore, they tended to approach and touch their mothers less frequently than the home-reared 30-month-olds, whereas the 40-month-old day-care group tended to approach and touch their mothers more frequently than the 40-month-old home-reared children. In short, the 30-month-old day-care children tend to resemble the 1-year-old children of Group A in the pattern of their behavior in the strange situation.

Blehar pointed out that the strange-situation behavior of the two day-care groups paralleled the reunion responses of young children after major separations, in regard to which it is the younger children who are more likely to be detached on reunion, whereas the older children are more likely to respond in an anxious, ambivalent fashion with intensified attachment behavior. She suggested that the results of her study may imply that the many repetitions of minor separation that occur in full-time day care may have effects similar in form, although perhaps not in severity, to those of major separations.


Each of the studies reported in this section raises more questions than it answers. These have already been discussed with reference to the study by Serafica and Cicchetti of children with Down’s syndrome. In both rearing-method studies, we recognize two major lacks: (1) a classification of patterns of behavior, comparable to our classification of 1-year-olds, in terms of which the kibbutz and day-care groups could have been compared with their controls; and (2) a thorough study of the interaction of the child with both mother and substitute figures in the natural environments of the home, day-care center, or kibbutz children’s house, as the case might be. In regard to Maccoby and Feldman’s kibbutz group, one would like to know the characteristic interaction of children with the metapelet in the children’s house (and in the strange situation), and with parents at home, as well as patterns of individual differences in all of these settings. In regard to Blehar’s day-care group, one would like to know the characteristic interaction with each of the caregivers in the day-care center and with the parents at home as well as the patterns of individual differences in each of these settings. Basic to all of these considerations is the need to know the relationship between characteristic behavior at home and behavior in the strange situation for children of differing age levels.

Although Blehar found significant differences between home-reared and day-care 30- and 40-month-olds, it will be recalled that Brookhart and Hock (1976) found no significant differences between home-reared and day-care 1-year olds. A crucial difference between their samples was the age at which day care began. Brookhart and Hock’s sample consisted of those who had begun day care in infancy—at 10 months of age at the latest and in most cases much earlier. The children of Blehar’s two groups began day care at 2 and 3 years of age, respectively. She suggested that 2- and 3-year-olds may interpret the long, daily separations from the mother, implicit in full-time day care, as rejection or abandonment. It is possible that an infant who begins day care in the first year of life, before he has become attached to his mother—or at least before attachment has become well consolidated—may accustom himself more readily than older preschoolers to long, daily separations and be less apt to experience them as implying rejection or abandonment by the mother. Should this prove to be the case, it would help to resolve the discrepancies between Blehar’s clear-cut findings of disturbance in child-mother relationships attributable to day care and findings of lack of disturbance reported not only by Brookhart and Hock, but also by two other sets of investigators of the effects of infant day care (Caldwell et al., 1970; Ricciuti, 1974).

Significant though they may be, these research issues are tangential to the main issue that should concern us here—the suitability of the strange situation as a basis for assessing individual differences in child-mother attachment among 2- to 4-year-olds. Further research is clearly necessary. First, it is necessary to undertake extensive studies of the relationship between behavior in the strange situation and behavior toward the mother at home and/or in various seminaturalistic situations of children in various preschool-age groups. Second, especially in the light of the findings of these studies, it would seem wholly desirable to revise the basis of strange-situation assessment of older preschool children, preferably devising a new classificatory system better suited than the present one to behavior of these older children, or at least to redefine the measures of interactive behavior so that they are more sensitive to differences among older children. Finally, it would be desirable in a longitudinal study to examine the degree of continuity from the assessments we have used with 1-year-olds to these new bases of assessment. With an extension of validation studies of this sort, the strange situation might well prove to have a wider application as a method of examining the effects of rearing methods.

Attachment as Related to Later Behavior

Two studies have used child-mother attachment as it is reflected in strange-situation behavior as an independent variable hypothesized to affect other aspects of development. Pentz (1975) was concerned with its relationship to language development, and Lieberman (1977) with its relationship to competence in peer play.

Pentz (1975) considered qualitative differences in the child—mother relationship to be an indirect indicator of the long-term nature of mother—child interaction, and examined its relationship to language acquisition. He observed 31 mother—child pairs in the strange situation when the children were 28 months old. In a free-play session at about the same age, tape recordings were made of the language of both child and mother in interaction with each other. At 36 months of age a second assessment of language was made in another free-play session, and, in addition, the child was given a test of language comprehension. Pentz’s chief hypothesis was that mothers can facilitate the development of language in their children by providing a simplified model of language to the child, the complexity of which is tailored to the child’s level of linguistic skill. To do so requires some sensitivity to the child’s language level, and consequently Pentz further hypothesized that this sensitivity may be a continuation of sensitivity to infant signals and communications, which we have found to be related to many aspects of an infant’s social development. Therefore, on the basis of strange-situation behavior, the children in his study were classified as either securely or anxiously attached to their mothers; their avoidance behavior in the reunion episodes was also scored.

His hypothesis that a simplified model of language would facilitate language acquisition in the child was confirmed. Another hypothesis was also confirmed—namely, that various “teaching devices” (such as expansions and recast sentences) were significantly related to language acquisition, although different teaching strategies seemed effective at different points in the child’s development. The child was found to play a very active role in his own language learning by adopting learning strategies of his own, frequently reciprocal to the teaching strategies of their mothers. The hypothesis relating quality of attachment to language development was not confirmed, however. The securely and anxiously attached groups did not differ significantly in level of language acquisition at either age. Pentz offered two explanations for the failure of this hypothesis. Language development may be buffered against the emotional context within which it proceeds, as Chomsky (1965) and Lenneberg (1967) might argue from their postulation of an inbuilt language acquisition device. Furthermore it would appear that mothers who were sensitive to their infants’ nonverbal signals and communications are not necessarily sensitive to the verbal abilities and signals of their children a year or two later.

Lieberman (1977) hypothesized that competence in play with age peers was related both to the quality of the attachment relationship with the mother and to the amount of previous experience in play with other children. The quality of attachment was assessed in two ways: (1) through measures of strange-situation behavior (the “interactive” behavioral variables, plus several frequency measures of discrete behaviors); and (2) through an Inventory of Home Behaviors. The latter was scored on the basis of a home visit, which consisted of both an interview with the mother and observation of mother and child in a semistructured, task-oriented play situation. The Maternal Attitude Scale (Cohler, Weiss, & Grunebaum, 1974) was given; two of the component scales were used—encouragement of widening reciprocal exchanges and control of aggression. Social competence with an age peer was assessed in a free-play session with an unfamiliar playmate of the same sex and age. The measures of interactive behavior derived from videotape records of this session were too complex and numerous to be described here.

The subjects were 40 white, middle-class children about 3 years of age, recruited from the waiting lists of nursery schools and day-care centers,3 with 20 destined for each preschool group experience. All of the assessments, except for social competence with peers, were undertaken before the beginning of preschool. Four months after the children had begun their preschool experience, their social competence was tested in a laboratory playroom with which the children had been individually familiarized in advance.

In order to reduce the number of variables, two principal components analyses were undertaken, one of the social-competence measures of behavior in the laboratory play session, and one for the measures derived from both the home visit and the strange situation. Composite measures were constructed on the basis of these analyses: Scores were derived by adding the z scores of variables with high positive loadings and subtracting the z scores of variables with high negative loadings on each of the components yielded by the analysis. The three composite scores (and the components upon which they were based) for social competence with a peer were labeled reciprocal interaction, negative behavior, and conflict behavior. In addition two composite measures, not included in the principal components analysis, were responsiveness and number of chains of exchange. The three composite measures of quality of attachment were labeled low home anxiety, excessive mother-centeredness in the strange situation, and sociability in the strange situation. The low home-anxiety measure was further divided into two: experience with peers and secure attachment.

All of the measures of social competence, except conflict behavior, were correlated significantly with low home anxiety. Only one measure, responsiveness to the playmate, was significantly related to excessive mother-centeredness, and the relationship was negative. None of the social-competence measures was significantly related to sociability. Maladaptive maternal attitudes about the child’s expression of aggression (as measured by the MAS) were correlated negatively with responsiveness, reciprocal interaction, and number of exchange chains in peer play. A restrictive maternal attitude about the child’s widening reciprocal exchanges was positively correlated with negative behavior and negatively correlated with the child’s responsiveness to a playmate. Because the two components of low home anxiety—secure attachment and experience with peers—were substantially correlated with each other, partial correlations were done when examining the relationship of each to the social-competence measures. Secure attachment was positively related to reciprocal interaction and negatively to negative behavior, both measures dealing with nonverbal behavior. Experience with peers was positively related to number of chains of interaction and responsiveness to the playmate, both measures dealing with verbal behavior.

Lieberman concluded that both the quality of the attachment relationship and experience with peers are related to social competence with peers; but because both “independent” variables are substantially correlated, she suggested that mothers who promote secure relationships are also responsive to their children’s growing interest in peers.

Neither of the two composite scores that consisted largely of strange-situation variables was significantly related to the peer-competence measures, except that excessive mother-centeredness was negatively related to responsiveness to the playmate. It is our opinion that somehow the key strange-situation behaviors may have been obscured through the principal components analysis, so that their relationship to social competence could not be assessed. Lieberman (personal communication) has supplied us with correlations of the following strange-situation measures—contact maintaining, proximity seeking, resistance, avoidance (each pertaining to behavior to the mother), search, and crying—with each of the five composite measures of peer competence. The most conspicuous finding is a positive correlation between resistant behavior in the strange situation and negative behavior in interaction with the playmate (r = .57; p < .001). Four other statistically significant correlations were found, but these were low and may reflect nothing more than chance relationships. Particularly interesting to us was that avoidant behavior in the strange situation had no significant correlations with the measures of social competence with peers. Resistant behavior in the strange situation was also negatively and substantially related to low home anxiety, as were search and crying.


Lieberman concluded that quality of child—mother attachment was one of two major variables significantly related to later social competence in play with age peers; but it is clear that the significant attachment measure stemmed from home-visit data and not from the strange situation, except for the fact that resistant behavior in the strange situation was correlated with negative behavior in peer play. Pentz concluded that maternal language behavior had a significant influence on the child’s language acquisition, but that attachment as assessed by strange-situation behavior did not. In both studies it is notable that avoidant behavior, which was related to many variables when assessed in 1-year-olds, was related neither to language acquisition nor to social competence with peers when assessed in the third year of life. Although Lieberman did not use the patternings of behavior that are reflected in our classificatory system, Pentz did, distinguishing securely from anxiously attached children; this distinction also failed to yield differences.

We can scarcely conclude that the mother—child relationship has no bearing on the development of language, especially because both Connell (1976) and Main (1973) found that Group-B toddlers were superior to non-Bs in language function. Connell’s assessment of language took place at 18 months, however, and Main’s at 21 months; Connell’s strange-situation assessment was at 18 months, and Main’s at 12 months. It could be either that the special influence of mother—child interaction on language acquisition is effective during the second year of life, but later wanes, or that the strange-situation procedure as an assessment of child—mother relationship is more valid and/or more sensitive at 12 to 18 months than it is after the child’s second birthday.


There is no doubt from the developmental studies of Maccoby and Feldman (1972) and Marvin (1972) that there are substantial changes in strange-situation behavior from age 2 to ages 3 and 4. Older preschoolers find the strange situation much less disturbing than 1-year-olds. In particular they are better able to sustain their equilibrium during brief separations from their mothers, and consequently attachment behavior is less intensely activated in both the separation and reunion episodes. In 1-year-olds resistant and avoidant behavior toward the mother in the reunion episodes may sometimes be assumed to occur because the attachment-behavioral system has been strongly activated. Children who respond to strong instigation of attachment behavior either with avoidance (which is antithetical to attachment behavior) or with resistance (which, when it occurs, is likely to accompany intense attachment behavior) tend to be those whose experience in interaction with their mothers had been disharmonious, at least in the case of 1-year-olds. There is no reason to believe that avoidance and resistance occurring in the case of older preschoolers has dynamics different from those that occur in the case of 1-year-olds. Marvin found proportions of avoidant and resistant children in each of his three age groups similar to those groups that we found in each of our four samples of 1-year-olds, and that Connell and Rosenberg (see Chapter 9) found in their samples. It is nevertheless reasonable to assume that both avoidant and resistant behaviors may be shown less conspicuously by older preschoolers than by 1-year-olds or even 2-year-olds, for the older the child the less strongly attachment behavior tends to be instigated by the brief separation episodes of the strange situation.

Connell (1976) and Waters (1978) found that 18-month-olds could be classified without difficulty in accordance with our classificatory system and, furthermore that there was a high degree of congruence with the classifications they had received at 12 months of age. (See Chapter 11.) Marvin found that his 2-year-olds could be classified in terms of our scheme without undue difficulty, but that the system had to be modified for 3- and 4-year-olds. He resorted to a two-step classification; the first applied our criteria as strictly as possible, and the second attempted to transform our system into one applicable to child—mother interaction in which proximity/contact seeking was no longer the focus, in any literal sense of the term. Thus he sought to distinguish between wholly positive interaction and interaction showing avoidant and/or resistant complexities.

Pentz (1975) classified his 28-month-olds into securely versus anxiously attached. Maccoby and Feldman (1972) did not attempt to use our classificatory system, nor did they use our scoring system for any interactive behaviors including avoidance and resistance. They watched for avoidant and resistant behaviors, however, but reported them as occurring in very few children. It could be that they minimized the importance of momentary avoidant or resistant gestures that receive scores of 2, 3, or even 4 in our system, and that Marvin took into account when distinguishing avoidant and resistant children from those classifiable as securely attached. Feldman and Ingham (1975) have suggested that the strange situation does not yield a valid assessment of attachment among older preschoolers. This may be so, but before accepting this conclusion we must examine two other possibilities.

One possibility is that the strange-situation measures do yield valid distinctions among older preschoolers but that the behavioral differences among them are smaller and less obvious than they are among 1-year-olds, requiring measures that take small and subtle distinctions into account. We have found that our interactional-behavioral scores yield valid distinctions among 1-year-olds, whereas frequency measures of discrete behaviors, such as looking, vocalization, or even touching, do not. Indeed Blehar (1974) found significant differences between 30- and 40-month-old day-care children and their home-reared controls, using the interactional behavior scores—differences congruent with her hypothesis that the long daily separations experienced after attachment to the mother had become well established made for anxiety in the attachment relationship.

A second possibility is that instead of using the patternings of behavior that differentiate significantly among 1-year-olds there should be a search for new patternings among older preschoolers, these would then be “calibrated” against mother—child interaction in other settings and against other aspects of development. So far no one has attempted this—not even Marvin, who modified the classificatory system to accommodate the behavior of the older children in his sample.

In our opinion both of these alternatives are worth consideration, but both require further research. A third alternative is to search for entirely new methods of assessing the child—mother attachment relationship in children beyond the first 2 years of life. Because home observations are very time consuming (and also present special difficulties in observing and recording the behavior of the child beyond infancy), one would hope to find a laboratory situation that would bear the same kinds of relationships to mother—child interaction in the natural environment that the strange situation does for 1-year-olds. Two strategies suggest themselves as possibilities. One is greatly to increase the stress in the laboratory situation so that strong instigation to attachment behavior is provided for 3- and 4-year-olds—a strategy with obvious disadvantages. Another strategy is to capitalize—as Marvin and his associates are currently doing—on cognitive changes in the preschool years and to relate behavior in laboratory situations tapping cognitive abilities to mother—child interaction in natural and seminaturalistic situations (Marvin, 1977; Marvin, Greenberg, & Mossler, 1976; Mossler, Marvin, & Greenberg, 1976).


1 Marvin excluded the introductory episode when numbering his episodes, so that his Episode 1 is our Episode 2, and so on. We have adhered to our numbering in reporting his work here.

2 Rheingold and Eckerman (1970) use the term “detachment” to refer to an infant’s willingness to separate himself from his mother in order to enter and explore an unfamiliar room. Feldman and Ingham have adopted this term to refer to the decrease of “proximal” attachment behavior with increasing age. Both of these uses obviously differ from the term “detachment” first used by Bowlby (1953) to refer to the last of three phases of response to a major separation. Here a child, having previously shown intense separation distress, now seems indifferent to his mother’s absence and, in her presence (or upon reunion), shows little or no attachment behavior, but instead avoids, rejects, or is indifferent to her. He interpreted this response as a defensive behavior, with perhaps an underlying repressive process. It is in this sense that in this volume and elsewhere (e.g., Ainsworth & Bell, 1970) we liken avoidance of the mother in the reunion episodes of the strange situation to the detachment response of young children in long and disturbing separations.

3 Lieberman has not yet reported her comparisons between nursery-school children and those in full-time day care.