Patterns of Attachment: A Psychological Study of the Strange Situation - Mary D. Salter Ainsworth 2015
A Review of Strange-Situation Studies of One-Year-Olds
Measures and Methods of Assessment
Introduction
We believe that our strange-situation procedure will be very useful in research into the attachment of a child to his mother figure and its relationship to other facets of development. Indeed it has already been a point of departure for a substantial number of studies of attachment and/or attachment behavior.
In many of these projects, modifications in the procedure have been introduced that make it difficult to make direct comparisons between their findings and those presented in this volume. Attachment behavior can be strongly affected by the situation in which one attempts to observe it; some situations offer strong instigation to it, whereas others do not. As we have shown in earlier chapters, behavior toward the mother in Episode 8—whether this be attachment behavior or behavior antithetical to it, such as avoidance or resistance—particularly highlights the kind of qualitative differences in attachment that are reflected in our classificatory system. The most common modification of the strange-situation procedure has been to eliminate a second separation and consequently also a second reunion episode, thus omitting the very episode that we have found to be most crucial in highlighting individual differences that prove stable across a variety of situations. However interesting and fruitful it would be to review the findings of these studies in order to interpret each in the light of the whole, this enterprise would take us too far afield.
Here we review only those studies that were intended to be directly comparable to ours, in that the strange-situation procedure was used with little or no departure from our format—at least in the arrangement of the conditions under which observations were undertaken. These studies fall into two main classes: those that examined the strange-situation behavior of 1-year-olds and related it to behavior under other conditions, to other behaviors, or to antecedent factors that might have affected it; and those that dealt with the strange-situation behavior of 2-year-olds or older children. Some of the latter studies were directly concerned with age-attributable changes that may take place in strange-situation behavior itself, and demonstrated that such changes are substantial. Therefore we propose to defer a review of studies dealing with the older preschoolers to the next chapter and to consider here only those that focus on the strange-situation behavior of 1-year-olds.
The studies of 1-year-olds may be divided into five classes: (1) those that view qualitative differences in infant—mother attachment as a “dependent variable” and examine their relationship to antecedent conditions; (2) those that view either mother—infant interaction or infant—mother attachment as an “independent variable” and investigate other facets of infant behavior to which it may be concurrently related; (3) those that similarly view either mother—infant interaction or infant—mother attachment as an “independent variable” and investigate other facets of behavior some months later; (4) those that compare behavior in our standard situation with behavior in another situation of parallel design but with either the setting or the adult figure changed; and (5) those that are concerned with the stability of strange-situation behavior and/or classification over time. Because these last-mentioned studies are clearly relevant to the topic considered in Chapter 11, they are discussed there. This present chapter is concerned with the first four groups of investigations mentioned here. We consider first those that address the issue of antecedent conditions that may be related to qualitative differences in attachment.
Patterns of Attachment at One Year Related to Antecedent Variables
Neonatal Separation. Hock, Coady, and Cordero (1973) compared the strange-situation behavior of 31 infants who had been born prematurely with that of 30 full-term infants when the latter were about 11 months old and the former of comparable age after a correction was made for prematurity. The mean birthweight of the premature group was 1,500 grams or less—that is, less than 3 pounds, 4 ounces. The antecedent condition of interest was the extent of separation from the mother experienced by the premature babies—a mean of 40 days following birth in contrast to only 3 or 4 days for the full-term neonates. The separation was complete in neither case, for the hospitals concerned encouraged the mothers of the prematures to visit them, to touch them, and later to participate in caretaking. Nevertheless the early mother—infant interaction was necessarily more limited with the premature than with the full-term babies. Because of the age correction, both groups had experienced about the same length of time with their mothers at home after discharge from the hospital and before observation in the strange situation—namely, about 46 weeks.
No significant differences were found between the premature and full-term groups. The authors concluded that the 46 weeks that the prematures had spent at home with their mothers after the prolonged neonatal separation had quite overcome any effects the latter might have had on either quality of mothering or on eventual infant—mother attachment.
Twins Versus Singletons. In the premature sample studied by Hock, Coady, and Cordero (1973), there were 14 twins and 17 singletons. Significant differences were found between these two groups. Specifically, the premature twins showed more resistant (p < .01) and more avoidant (p < .05) behavior in the strange situation than the premature singletons. The authors suggested that it is plausible that the mothers of twins would be less able than mothers of singletons to respond promptly to infant signals—a suggestion that is clearly congruent with our findings of behavioral differences among the mothers of the babies in our classificatory groups.
Demographic and Other Variables. Connell (1974) observed the strange-situation behavior of 46 infants from white, middle-class families when they were approximately 51 weeks of age, and later (1976) added 55 more infants to his sample, making a total of 101. For this total sample, as well as for the original 46, he checked a variety of different variables that might have influenced infants’ development or their behavior in the strange situation, in order to “rule them out.” He indeed found that the following had no significant relationship to strange-situation classification in his sample: social class of parents, number of siblings, precise age at the time of observation, time of day of observation, and identity of the stranger in the strange situation.
It must be pointed out that Connell used a “multivariate classifier” based on the classificatory system we developed. It was devised by Connell and Rosenberg (1974) on the basis of the data from our total sample of 105. Their method of reduction of the variables from 72 to a more acceptable number for multivariate analysis differed substantially from the method we used for the same purpose (see Chapter 6). One step of their method involved cluster analyses, in the course of which they concluded that our Groups A, B, and C were highly distinct clusters, thus confirming our classificatory system, but noted “two minor anomalies.” Subgroup-B1 subjects seemed to fit better with Group A than with Group B, and B4 subjects seemed to fit better with Group C. The subjects of these two small subgroups were, therefore, not considered in the development of the classifier.
After the reduction of measures had been completed, a discriminant function analysis was then undertaken to yield a set of discriminant weights that could be used to assign subjects of a new sample to Groups A, B, and C. A “design set” of 65 subjects randomly selected from our total data base (after removing B1 and B4 subjects) yielded a 97%-correct classification rate. The remaining set of 26 subjects, which had been held out of the classifier-design computations, was classified correctly at a 96.2% rate. When the weights were applied to Connell’s own sample, the distribution into Groups A, B, and C was congruent with the distribution we reported in Chapter 6, Table 7.
Although we have chosen other methods of data reduction for our discriminant function analysis, it is evident that the data are sufficiently robust to withstand diverse methods of analysis without undue loss of information. As to the anomalies presented by B1 and B4 infants, we acknowledge that the discriminations between A and B1 infants and between C and B4 infants are the most difficult to make. We have chosen, nonetheless, to retain both subgroups in Group B, for reasons that are discussed in Chapter 12.
Low Birthweight and Low APGAR Scores. Connell (1974, 1976) found a significant tendency for Group-C infants to have lower weight and lower APGAR ratings at birth than the A and B infants in his sample. One child in his sample weighed 4 pounds, 5 ounces at birth; otherwise none weighed less than 5½ pounds. Presumably, all of the low-birthweight infants in his samples were born at term, and none of them weighed as little as the prematures observed by Hock, Coady, and Cordero (1973). The two samples of low-birthweight infants are not comparable, therefore; Connell’s were small for their gestational age, whereas Hock’s were of a size appropriate for their gestational age. The implication is that low-birthweight infants in Connell’s sample may have been retarded in intrauterine growth, and thus perhaps more predisposed toward postnatal developmental anomalies than the “normal” prematures of Hock’s sample. If so, it is indeed interesting that one of these anomalies is related to the development of an anxious attachment relationship with the mother.
Maternal Attitudes and Mother—Infant Interaction. Rosenberg (1975) undertook to evaluate the validity of our inference (Ainsworth, Bell, & Stayton, 1971) that the strange-situation classifications reflect the characteristic harmony/disharmony of mother—infant interaction. Because of this, we have included his study with those concerned with antecedent conditions. It should be emphasized, however, that he views mother—infant interaction as a highly reciprocal matter, to which both partners make an important contribution.
His sample consisted of 46 of the infants with whom Connell worked, and the identification of classificatory groups was done by means of the “multivariate classifier” (Connell & Rosenberg, 1974) to which reference was made earlier. In choosing measures of mother—infant interaction, he focused on Sander’s (1964) concept of reciprocity, and therefore chose to use the Reciprocity Factor Scale of the Maternal Attitude Scale constructed by Cohler, Weiss, and Grunebaum (1970), which had been based on Sander’s work. In addition, he observed mother—infant interaction in two laboratory situations: (1) a free-play situation lasting 6 minutes in which the mother was told to do whatever she wished; and (2) a directed-play situation, also lasting 6 minutes, in which she was instructed to administer three Bayley Scale items to her infant. Both situations could be described as low-stress situations in which the social interaction tended to be mediated through objects. In each situation the degree of sensitivity and reciprocity of the mother’s behavior was rated on a nine-point scale adapted from our scale of maternal sensitivity to infant signals and communications (see Ainsworth, Bell, & Stayton, 1974), which had originally been designed for the rating of maternal behavior in the home environment. In each situation the infant’s social responsiveness was also rated on another nine-point scale adapted from a three-point scale devised by Brody and Axelrad (1970). To tap the actual interaction between mother and infant, as well as to assess the separate contribution of each, he calculated the time spent in the laboratory situations in each of four mutually exclusive interaction states: reciprocal interaction (M+I+), reciprocal ignoring (M−I−), and two mixed interaction states (M+I−) and (M−I+).
Rosenberg focused on comparisons between Groups A and B. He considered the 6 Group-C infants in his sample to constitute too small and too heterogeneous a group for statistical comparisons. Group-B mothers scored significantly higher on the Reciprocity Factor of the Maternal Attitude Scale, and thus more strongly endorsed encouragement than discouragement of reciprocity with the infant. They were also rated significantly higher on reciprocity—sensitivity in the free-play situation, although they did not differ from Group-A mothers in the directed-play situation. Group-B and Group-A infants did not differ significantly, however, in their rated social responsiveness in either situation. Group-A infants increased significantly in social responsiveness from the free-play to the directed-play situation, however. In the free-play situation, Group-B infants showed more nondistress vocalizations than Group-A infants, whereas Group-A infants explored more than Group-B infants. Group-B mother—infant pairs had reciprocal-interaction states (i.e., M+I+) in significantly more 15-second intervals than did Group-A dyads, whereas the latter had significantly more reciprocal-ignoring states (i.e., M−I−).
The finding that Group-B infants were not found to be more socially responsive than Group-A infants, Rosenberg was inclined to attribute to an overly global measure of social responsiveness; he pointed out that Group-B babies vocalized more than Group-A babies—a behavior that is especially likely to elicit maternal social behavior and thus is conducive to reciprocal interaction.
The rest of Rosenberg’s findings offer substantial support both to his hypotheses and to our findings. Group-B mothers emerged as more sensitively responsive to infant signals and more consciously geared toward establishing reciprocal interaction than Group-A mothers. Such sensitive responsiveness undoubtedly includes encouraging rather than ignoring or interfering with an infant’s interest in exploring a novel environment, as well as responding to infant’s bids for interpersonal interaction. Rosenberg’s findings clearly show that mother—infant interaction is more harmonious in Group-B than in Group-A dyads.
Working Versus Nonworking Mothers. Brookhart and Hock (1976) compared 18 home-reared infants with 15 who attended a day-care center for at least 2 consecutive months before the strange situation. The day-care babies were separated from their mothers for a mean of 33.6 hours per week, whereas the home-reared infants were separated for only 7.3 hours a week on the average. Both samples were middle class and aged 11 months when observed in the strange situation. (Brookhart and Hock, in the same study, investigated the differences for this sample between behavior in the strange situation conducted in a standard fashion in the laboratory and behavior in a parallel situation conducted in the home environment. This aspect of their study is discussed later in this chapter.)
The measures used were our measures of proximity and contact seeking, contact maintaining, resistance, and avoidance (all scored for behavior to mother and stranger), search, and crying. A multivariate analysis of variance was undertaken, with the main effects examined for groups (day-care versus home-reared), episodes, and sex. (The main effects attributable to location—i.e., laboratory or home—were also examined, but are considered in a later section.) There were no main effects attributable to rearing conditions, although there were significant group by sex interactions in regard to proximity-seeking and contact-maintaining behavior directed toward the stranger. These interactions were complex, and the groups were too small for firm conclusions about sex differences to be drawn. Therefore, the chief finding was a lack of difference between infants reared at home and infants with at least two months’ experience in group day care.1
In another report, Hock (1976) was concerned with a comparison of the effects of a variety of different types of infant care. The number of variables that might affect the results were several—including part-time vs. full-time work, age of infant when mother began work, how consistently she worked, whether the baby was placed in individual or group care, whether individual care was in the baby’s own home or elsewhere, and how many different caretakers were involved. Hock attempted to control for (or test the effects of) these variables in her several analyses, but even a total sample of 83 working mothers was not large enough to deal with all variables.
Her first analysis of variance concerned a comparison of 74 nonworking mothers with the total sample of 83 working mothers. There were no significant main effects in regard to strange-situation variables, although both resistant and avoidant behavior directed toward the stranger approached significance. The infants of nonworking mothers were somewhat more disturbed in their behavior toward the stranger. The 27 babies whose mother worked part-time were not found to behave significantly differently from those whose mothers worked full-time, and there were no effects attributable to the baby’s age at the time the mother began working.
Two samples of working mothers that met more homogeneous criteria were selected, both of which consisted of mothers who began working and placed the infant in some kind of substitute care before he was 7 months of age, but who stuck to one type of care within which the infant had no more than two caretakers. In other words, babies of these two samples experienced continuity and a reasonable degree of consistency in substitute care. One of these samples consisted of 31 infants who were cared for individually, whether at home or elsewhere. These infants showed significantly less resistance to the stranger in the strange situation than the infants of the 74 nonworking mothers, but otherwise there were no main effects attributable to mothers’ working. No main effects were found within the individual-care sample attributable to whether infants were cared for in their own homes or elsewhere.
The second sample consisted of 28 infants who were cared for outside of their own homes, 17 individually and 11 in group care, whether in a day-care center or in a babysitter’s home where at least three children constituted the group. Those in group care showed significantly more resistant behavior to the mother in the strange situation than those in individual care; they also showed somewhat more crying and less search behavior in the separation episodes. Hock suggested that they seemed to exhibit a disturbance best described as an angry mood. In this report, Hock did not compare the group-care sample with a home-reared group because group care was confounded with care away from home.
Discussion. This group of five studies, primarily concerned with antecedent conditions that might influence the later quality of the infant—mother attachment relationship, tend to support our findings. In Chapter 8, which dealt with the relationship between maternal behavior and infant attachment, we emphasized that the pervasive variable of maternal responsiveness to infant signals emerged as the most important antecedent. We concur in the suggestion made by Hock, Coady, and Cordero (1973) that mothers of twins (and probably not merely mothers of premature twins) are likely to find it more difficult than mothers of singletons to respond promptly to infant signals. Rosenberg’s (1975) findings about the importance of maternal reciprocity and sensitivity are also congruent with ours. Although Connell (1974, 1976) found no relationship with quality of attachment within his sample of either social class or number of siblings, this would not necessarily be the case for other samples in which these variables were related to conditions likely to reduce maternal responsiveness to signals or to make the mother relatively inaccessible to the infant. Thus, for example, Bell (1978) reported a higher incidence of secure attachment among white, middle-class babies than among black, disadvantaged babies, and attributed it largely to the chaotic substitute-care arrangements experienced by a number of infants in the disadvantaged sample. (Other findings of Bell’s study are reported later in this chapter.) Similarly, one could expect that infants in large families might find their mothers less accessible than infants in small families if multiple births or other conditions of high “density” were associated with large family size.
We thought it a reasonable hypothesis that extended, early neonatal separation might affect the quality of mothering and hence the quality of the eventual attachment relationship, especially because evidence has been accumulating that close bodily contact immediately postpartum facilitates the emergence of the attachment of mother to infant and is associated with more harmonious mother—infant interaction in later months (Klaus & Kennell, 1976). It may be, however, that the encouragement given by the hospitals to the mothers of the prematures studied by Hock, Coady, and Cordero (1973) to visit and to interact with their babies, together with the long subsequent opportunity to interact with them fully at home, may have overcome any initial adverse effect of the extended but often partial neonatal separation—especially if the mothers of the full-term control babies were given no particular encouragement toward close interaction in the neonatal period.
The trends found by Connell for his small-for-date babies (i.e., those with low birthweight and low APGAR ratings, but who were not born prematurely) to be classified as Group C and hence as anxiously attached invites more intensive longitudinal research into the development of mother—infant interaction in the case of babies with this kind of “at risk” factor.2
Rosenberg (1975) has implied that infant responsiveness may itself be an antecedent variable that could influence strange-situation behavior and hence presumably quality of infant—mother attachment. It is indeed difficult to disentangle this variable from variables of maternal attitudes and behavior that influence the course of mother—infant interaction. It is, however, too complex a research issue to review and discuss here the question of the ways and extent to which an infant influences the behavior and attitudes of his mother in contradistinction to the ways and extent to which a mother influences the behavior and development of her baby.
The issues raised in Hock’s (1976) study of working mothers are also very complex. Of these, most controversy has focused on the issue of group day care and its effect on the child—mother relationship. Research into this issue has been sparse, and it inevitably raises questions about the nature of the day-care experience itself. The larger issue of the effects of a variety of different methods of substitute care had scarcely been addressed before Hock’s pioneer study. This study had a longitudinal design with mother—infant dyads selected at birth in terms of the mothers’ declared intentions to work or not to work. Even though a large sample was initially assembled, it is clear that crucial comparisons between conditions of infant rearing are difficult to accomplish with a preselected sample. The chief difference between groups that emerged from Hock’s comparisons was that infants in group care tend to be more resistant to their mothers than infants in individual care. Hock suggested that caretakers responsible for more than one infant are inevitably less prompt and consistent in their response to infant signals and communications than are caretakers of only one infant. This interpretation is obviously congenial with our position. The implication is that an anxious (i.e., Group-C) pattern of attachment is more frequent among the infants of working mothers who are cared for in groups than among those cared for individually, because resistant behavior is a cardinal characteristic of the strange-situation behavior of Group-C infants.
On the other hand, Brookhart and Hock (1976) concluded that there is no evidence that experience in day-care centers adversely affects the infant—mother relationship—a conclusion also reached by Caldwell, Wright, Honig, and Tannenbaum (1970) even though their day-care centers undoubtedly differed in their methods, and despite different bases of assessment of the infant—mother attachment relationship. It must be pointed out that Hock’s (1976) group-care sample included infants cared for in informal groups away from home as well as those cared for in formally constituted day-care centers. Her contrast group was with infants cared for individually but also away from home, whereas Brookhart and Hock’s contrast group consisted of infants reared by their own mothers at home. It is not clear which of these variables might have accounted for the discrepancy in findings—or indeed whether the network of potentially significant variables is so complex that critical comparisons of any two sets of conditions could only be made with samples much larger than those so far assembled. Brookhart and Hock’s conclusions also differ from those of a study by one of us (MCB) of children who started day care at a later age; but a discussion of this discrepancy is deferred until Chapter 10.
It is inappropriate in the present context to discuss in further detail research into the effects of home rearing versus day care, or indeed the other alternatives that a working mother may choose. Nevertheless, we wish to point out that Hock (1976), in two of her analyses, selected samples that emphasized consistency of substitute care arrangements and continuity of caretakers. Moore (1964, 1969), in his longitudinal study of 223 London children, found many variations in child-care arrangements among the working mothers who constituted half of his total sample. He found that the degree of stability of the arrangements for substitute care was the chief variable to be associated with outcome. We have no doubt that the most adverse outcomes of substitute care stem from a sequence of arrangements, any of which in themselves might have been adequate, but that are disturbing because of discontinuities.
Patterns of Attachment at One Year Related to Other Concurrent Behavior
There are three studies that examine the relationships between patterns of attachment manifested in strange-situation behavior and other aspects of infant behavior and/or development assessed at or about 1 year of age—two studies by Silvia Bell (1970, 1978) and one by Connell (1974). All of these addressed themselves to the relationship between patterns of attachment and aspects of cognitive function. One of Bell’s studies also investigated social behavior. It is convenient to discuss separately the relationships between attachment and cognitive function and social behavior. Because the three studies had their major focus on cognitive function, we consider that topic first.
Attachment and Cognitive Function at One Year of Age
Let us first consider Bell’s (1970) study, in which she explored a suggestion of Piaget’s (1936) that infants develop a concept of persons’ existing independent of self, even when not present to perception, more quickly than they develop a homologous concept of the permanence of inanimate objects. Piaget suggested that the acquisition of person permanence would be accelerated because the baby finds persons, and especially the mother, the most interesting and important among objects. As Saint-Pierre (1962) showed, however, not all infants are more accelerated in the acquisition of person permanence. Bell hypothesized that the degree of harmony/disharmony in mother—infant interaction would affect the degree and direction of discrepancy between the developments of persons and of inanimate objects as permanent entities. Because such differences in the nature of interaction between mother and infant were also related to qualitative differences in the infant—mother attachment relationship, she further hypothesized that there should be a relationship between patterns of attachment and the development of the object concept. (Bell had access to our then-unpublished findings on the relationship between maternal behavior and subsequent quality of attachment in the first year of life, which are reported in Chapter 8.)
Bell devised parallel scales to assess the development of object and person permanence, based on the details of Piaget’s own (1937) account of this development. These scales were administered to 33 infants from white, middle-class families (Sample 2 in the present report) three times between 8½ and 11 months of age; a fourth testing took place for a subsample at 13½ months. A week after the 11-months’ testing, the infants were observed in the strange situation.
Infants who, with some consistency in the course of the three testings between 8½ and 11 months, were more advanced in the concept of persons than of inanimate objects as permanent were identified as having a “positive décalage.” Those who, with some consistency, were more advanced in “object permanence” than in “person permanence” were identified as having a “negative décalage.” Those who showed no consistency in this regard were identified as having “no décalage.”
Twenty-three infants fell into the positive-décalage group; all of these were classified in Group B (securely attached) on the basis of their behavior in the strange situation. Seven babies fell into the negtive-décalage group; all were classified in either Group A or Group C, indicating anxious attachment. The three babies who had no décalage were distributed among all three strange-situation classificatory groups. Furthermore, infants in the positive-décalage group were significantly more advanced in the development of person permanence than were the infants in the negative- and no-décalage groups (combined) in the development of object permanence; and by 13½ months they were also more advanced in object permanance. It thus appears that the conditions of mother—infant interaction that foster the development of secure attachment also facilitate the development of the concept of the object—a facet of cognitive development that is generally acknowledged to be a highly significant one and an acquisition necessary for several other aspects of cognitive development.
In her second study, Bell (1978) began with a replication of her first study, using 33 black infants from socioeconomically deprived families; but she also observed infant exploratory behavior, maternal behavior relevant to it, and she assessed developmental quotients. She aimed to determine the relative importance of three variables in affecting cognitive development: the infant—mother attachment relationship, socioeconomic level of parents, and degree of environmental stimulation. The design was a short-term longitudinal study that began when the babies were 8½ months old and continued until they were 36 months old. Infants and mothers were seen together on 11 occasions, all sessions taking place in observation rooms in a local hospital. In this section we are concerned only with the sessions between 8½ and 15 months.
The schedule of sessions was as follows: (1) At 8½ months, infants were administered the object- and person-permanence tests and the Griffiths scale of infant development. After the testing the baby was observed for approximately 1 hour in a free-play session, with his mother present, in order to observe the baby’s exploratory play and his mother’s behavior with respect both to interaction with the baby and to the kind and degree of stimulation she gave to his play with toys. During part of the play session, a black research assistant engaged the mother in interview in order to obtain information about demographic variables and living conditions, history, attitudes toward infant-care practices, amount of mother—infant interaction, amount of home stimulation, daily separations, and the number and stability of substitute caregivers. (2) Three days later all of these procedures were repeated. (3) At 11 months the same procedures were again repeated. (4) One week later the strange situation was conducted in another experimental room using toys that were new to the child. (5) At 15 months the same procedures were repeated that had been used in Sessions 1, 2, and 3. (Later sessions are discussed in a later section.)
In regard to object and person permanence and its relation to infant—mother attachment as assessed in the strange situation, the findings of Bell’s (1970) earlier study were confirmed in all essentials, despite the differences between a white, middle-class and a black, disadvantaged sample. Thirteen infants displayed a positive décalage in regard to person and object permanence; all of these were classified in Group B. Twelve showed a negative décalage; eight of these were classified in Group A and four in Group C. Eight showed no décalage, and these were distributed among all three strange-situation groups. The 13 who showed a positive décalage had, at all ages of testing, reached a level that was significantly higher than that reached by the negative- and no-décalage groups for the concept of the permanence of either inanimate objects or persons. Thus the relationships between quality of attachment and the development of the object concept that had been found for the middle-class sample held also for the black, disadvantaged sample.
On the other hand, there were more infants in the white, middle-class sample who had positive décalage. Furthermore, at 8½ and 11 months of age the middle-class whites were significantly superior to the disadvantaged blacks in person permanence, and at 11 and 15 months in inanimate-object permanence as well. These differences are attributable entirely to the negative- and no-décalage groups. There were no significant differences between the black and white samples when only the groups showing positive décalage were considered. It will be recalled that the latter consisted entirely of securely attached (Group-B) babies in both samples. Thus the differences between the samples may be largely attributed to the fact that there were more securely attached babies in the white, middle-class sample and more anxiously attached babies in the black, disadvantaged sample. Among the “living-condition” factors that may be involved in making for less harmonious mother—infant interaction in the dyads of the disadvantaged sample, and hence more frequent anxious attachments, are father absence, mother absence from the home for long daily periods, and multiplicity and discontinuity in regard to substitute caregivers.
The biserial correlations (Group B versus non-B) between the strange-situation classification and both person permanence and DQ were significant (p < .01) at 8½, 11 and 15 months, the former ranging from .65 to .61, and the latter from .55 to .45. The correlations of strange-situation classification and (inanimate) object permanence were lower and not significant—except for a coefficient of .39 at 15 months. It appears that the development of the concept of inanimate objects as permanent is not sensitive to mother—infant interaction in the same way as are person permanence and DQ.
Bell also correlated the cognitive measures with scores in the strange-situation behavioral variables. The “positive” variables, consisting of proximity seeking and contact maintaining, were not significantly correlated with any of the cognitive measures at any of the ages in question. On the other hand, the “negative” variables, consisting of resistance and avoidance directed to the mother, had substantial negative correlations with person permanence and DQ scores at every testing, and also with (inanimate) object permanence at 15 months. (This is another piece of evidence that meaningful individual differences in infant—mother attachment are not well reflected by strength of attachment behavior, even at high intensity of activation, but are reflected both by the negative behaviors of avoidance and resistance that supplant or accompany high-intensity attachment behavior, and by the strange-situation classification that depends heavily upon the patterning of both positive and negative behaviors.)
Infant—mother attachment, hypothesized to reflect the degree of harmony of mother—infant interaction, was not the only variable that Bell found to be significantly correlated with cognitive function. The amount of floor freedom the mother permitted the infant was substantially correlated with person permanence at all ages of testing, and significantly correlated but to a lesser degree with (inanimate) object permanence. A variable entitled “exploratory potential of the environment” had significant correlations with person permanence at all testing, but the correlations were weaker than those with either strange-situation classification or floor freedom. The amount the mother played with the baby was significantly correlated with object permanence at 11 and 15 months. Mother’s educational level had negative correlations with cognitive measures, except with person permanence at 15 months.
Connell, in his 1974 study mentioned earlier in this chapter, was directed by the hypothesis that the quality of the infant’s attachment relationship would affect his learning capacity, as indicated by response decrement to a redundant stimulus. He used a variant of the habituation paradigm designed by Schaffer and Parry (1969, 1972) to investigate the development of wariness, a variant that permitted infant manipulation of the stimulus object. A complex and attractive nonsense object emerged from a sliding door and moved within the infant’s reach. This was presented for 30 seconds in seven trials separated by 20-second intervals. On an eighth trial a “novel” object was presented, which was identical with the first nonsense object except for its color. On the ninth trial the original object was shown again. Measurements included length of first visual fixation of the object, total visual fixation, latency to first manipulative approach, and total manipulation. The sample consisted of 46 white, middle-class 1-year-olds.
Let us consider the findings for the three strange-situation classificatory groups in turn, beginning with Group C. All six Group-C infants manifested so much distress in response to the repeated stimulus presentations that observations had to be discontinued, whereas only two of the 40 non-C infants evinced such responses. This finding is in line with Main’s (1973) hypothesis, discussed in a later section, that the anxiously attached infants of Group C are too anxious to take advantage of opportunities to explore and thus to learn from their explorations. It should be recalled also that Connell had found that Group-C infants in his sample tended to have been small-for-date and to have low APGAR ratings at birth. On both scores it is a reasonable hypothesis that learning disabilities may emerge as more frequent among Group-C infants than among infants of either Group B or A.
The 12 Group-A infants tended to play with the stimulus-object throughout all nine trials, thus showing a low rate of habituation to the repeated stimulus. Although they showed a greatly increased initial fixation of the novel stimulus on Trial 8, they also had a very short latency to manipulate it—indicating a lack of wariness, by Schaffer’s definition. At least on a superficial level, therefore, their behavior in the habituation—wariness test resembled their behavior in the strange situation, where they explored actively throughout and evinced little wariness of either the unfamiliar situation or the stranger.
The 28 Group-B infants showed marked habituation to the repeated stimulus, giving it long visual fixation in the early trials, but very little visual attention during the later trials. Indeed during the latter they showed some behavioral disruption—minor fretting, attempting to climb out of the chair, reaching for the mother, and the like. They showed a significant increment in visual response to the novel stimulus in Trial 8, but also substantial wariness in terms of a relatively long latency before manipulating it. The relationship between wariness defined as latency to manipulate a novel object and wariness of the stranger in the strange situation (indicated by withdrawal to the mother in Episode 3) was found to be significant.
Thus, both A and B infants showed dishabituation to the novel-stimulus object, but in different ways, the former by prolonged visual attention and the latter by inhibition of manipulative response. Only Group B infants showed the clear-cut habituation to the repeated stimulus that Lewis, Goldberg, and Campbell (1969) suggested to be indicative of higher learning capacity. However one may interpret the differences in habituation between A and B babies, Connell’s findings demonstrate that infants who differ in the patterning of their responses to the strange situation—and hence, it is hypothesized, in the quality of the infant—mother attachment relationship—differ also in the patterning of their responses to an entirely different type of laboratory situation, which focuses on their responses to inanimate stimulus objects.
The findings of the studies by both Connell and Bell are congruent with our findings (see Chapter 7) that the Griffiths DQs of the infants in sample 1 were significantly correlated with maternal sensitivity to signals; and if other maternal variables, especially arrangements in regard to floor freedom, are also considered, they may be seen to account for about 50% of the DQ variance. Clearly, differences infants show in interaction with their mothers in the first year and in quality of the infant—mother attachment relationship are related to some, if not all, differences that emerge in their cognitive development and learning capacity.
A second important congruence between Bell’s and Connell’s findings is that not all aspects of strange-situation behavior were found to be related to various aspects of cognitive function. Connell used only the strange-situation classification as a variable. Bell used both classification and two sets of composite behavioral variables. Of these she found that the classification and the “negative” variables of avoidance and resistance were related to the cognitive measures, whereas the “positive” variables of proximity seeking and contact maintaining did not discriminate. Our interpretation of these findings is that seeking to gain and to maintain proximity and contact under even mildly stressful conditions is “built into” the behavioral repertoire of human infants, hence, even though there may be individual differences in the strength in which these behaviors are displayed under stress, these differences are not nearly as differentiating as those in the way in which and the extent to which these proximity/contact behaviors are interlocked with antithetical behaviors, such as avoidance and resistance. It is the patterning of the “positive” and “negative” behaviors that is important, and this is precisely what is reflected in the classificatory system.
Attachment and Quality of Mother—Infant Interaction at One Year of Age
Only Bell’s (1978) study is reported here, although some aspects of Rosenberg’s (1975) investigation, which was described in an earlier section, are also pertinent. Bell derived 14 measures of interaction between mother and child in the free-play sessions, at each of five ages: 11, 15, 24, 30, and 36 months of age. (Of these only those at 11 months are considered here.) These measures dealt with the number of “episodes” of interaction, prorated for the length of the session. They included: the proportion initiated by each to which the other responds (or resists or ignores); the proportion of responses that involve verbal/vocal behavior in distinction to bodily contact; the nature of the affect that might be inferred from the behavior of each partner; and the extent to which the mother attempted to prolong the interaction with the child and used this as an occasion to teach or to explain to the child about the things around him. These measures were correlated with three strange-situation variables: positive behavior (i.e., proximity seeking and contact maintaining summed), negative behavior (i.e., avoidance and resistance summed), and, biserially, strange-situation classification (i.e., B versus non-B).
Interaction measures at 11 months were all significantly correlated with negative behaviors in the strange situation that took place 1 week later, and all but one also with strange-situation classification. Only four were significantly correlated with positive behaviors. These findings suggest an impressive degree of congruence between the quality of mother—infant interaction (including both infant variables and maternal variables) and the child’s behavior in the strange situation when the two are assessed at the same age. Even so, it must be noted that it was the avoidant and resistant behaviors (scored only in the reunion episodes of the strange situation) that were more strongly related to the interaction measures—and, of course, the strange-situation classification in which the negative behaviors are crucial.
Bell then undertook three principal components analyses based on correlation matrices of the behaviors of mother and infant in the free-play session at 11 months. In each, only the first factor was used in subsequent analyses. The first principal component analysis was based on intercorrelations among the measures of infant behavior. The first factor referred to the baby’s social interactions with his mother, with the factor loadings ordered in terms of positive versus negative affective tone in these interactions. Factor scores significantly differentiated infants who had been classified as Group B from those who were not, in terms of their positively toned interactive behaviors. A second principal components analysis was based on intercorrelations among maternal behaviors in social interaction with the baby. The first factor was very similar to the first factor that emerged from the analysis of infant behavior, in that factor loadings reflected a range of behaviors from affectively positive and responsive to affectively negative and unresponsive. These factor scores also significantly distinguished the mothers of Group-B babies from mothers of non-B babies. The third principal components analysis was based on intercorrelations among maternal behaviors having to do with “teaching” the baby and/or stimulating his interest in exploratory play. The first factor to emerge from this set of “didactic” maternal behaviors was labeled the “Super Teacher” factor. It defined all the desirable things a mother might do to induce her young child to take an interest in the world about him and in how it works. This factor also clearly differentiated Group-B from non-B mothers, with the B mothers having significantly higher scores on the “Super Teacher” factor.
Discussion. It is of interest to compare three sets of findings that consider differences among infants of different classificatory groups and among their mothers, in regard to behavior of both infants and mothers in other situations at about the same time as the strange situation was conducted or somewhat earlier. Our own findings are the most naturalistic, having been based on observation of mother—infant interaction in the home environment, without any intervention by the visitor. They also have the broadest data base. The fourth-quarter measures were based on approximately 16 hours of observation on four or five different occasions. Rosenberg’s observations took place in the laboratory and included 6 minutes of highly structured interaction (his “directed-play” situation), as well as 6 minutes of unstructured free play immediately preceding. In terms of length of observation time, his study yielded the narrowest data base. Bell’s study was intermediate between the two. It also took place in a laboratory, but there was no attempt to direct the mother’s behavior and no intervention except that, for part of the time, the mother’s attention was at least partially taken up with responding to an interviewer. Furthermore, her free-play session lasted for 60 minutes—substantially longer than Rosenberg’s total observation time—and thus had a somewhat broader data base.
There were also differences in methods of recording observations and of subsequently reducing data. Rosenberg’s sessions were videotaped, and subsequently two rating scales and one coding system were used for data reduction, all three of which had been devised in advance. Our records of home visits consisted of transcriptions of dictated “play-by-play” accounts of behavior originally recorded in the form of jotted notes. Because ours was a pioneer study, we did not commit ourselves in advance to rating or coding systems, but rather allowed the data themselves to suggest variables that seemed important to examine systematically. Again Bell’s study was intermediate. She viewed the play session through a one-way-vision glass, and concurrently dictated a play-by-play account, thus managing to record more detail than our home visitors could with their jotted notes. She also did not commit herself in advance to coding systems or variables to be derived therefrom.
It is not appropriate here to discuss in detail the relative advantages and disadvantages of these different procedures. What is worthy of note in this context is that despite procedural differences, all three studies yielded data that distinguished between strange-situation classificatory groups in regard to mother—infant interaction; and the kinds of significant differences that emerged are highly congruent among the three studies. In each it emerged that interaction in Group-B dyads was more harmonious and that positive behaviors were generally characteristic of Group-B babies and mothers in contrast with those of non-B. The one exception to this generalization is that Rosenberg did not find support for his hypothesis that Group-B babies would be more socially responsive than Group-A babies. In the light of Bell’s findings and ours, which do support Rosenberg’s hypothesis, it seems likely to us that the very brief duration of his laboratory session did not yield a broad enough data base for stable measures of infant behavior to be derived from it. In addition, he himself questioned whether his predevised rating scale was an adequate instrument for assessing infant social responsiveness.
Patterns of Attachment of One-Year-Olds Related to Other Classes of Behavior at Subsequent Ages
Four investigations were concerned with the relationships between the patterns of attachment reflected in strange-situation classifications and assessments of performance or behavior made in the second or even the third year of life. Bell (1978), Connell (1976), and Main (1973) and her later associates assessed infants in the strange situation at 12 months of age. Matas (1977), however, undertook her strange-situation assessments when the infants in her sample were 18 months old.
These investigations imply a degree of continuity and stability, whether this be in the child’s development, in the mother’s behavior, or in the development of the relationship between child and mother. Bell explicitly stated that her hypotheses were based on a concept of continuity of qualitative differences in mother—child interaction. Main and Connell implied continuity of qualitative differences in the attachment relationship itself, with effects on both cognitive and social development. Matas also implied continuity of qualitative differences in the attachment relationship.
Bell, Connell, and Main made assessments of both cognitive development and mother—child interaction. It is convenient to consider these two aspects of development separately, even though all investigators consider them interrelated. Matas assessed behavior in both a free-play session and in a problem-solving situation. Because it is difficult from preliminary reports to distinguish between findings derived from the free-play session and those derived from the problem-solving situation, we propose to consider them all in our section on mother—child interaction. Main and Connell were also interested in the child’s behavior toward an unfamiliar person. The findings of these four investigations are considered in the next three sections—first findings relating specifically to cognitive development, then findings relating to mother—child interaction and social development, and finally findings related to responses to unfamiliar persons.
Attachment in One-Year-Olds and Cognitive Function at Subsequent Ages
Bell (1978) followed up her sample of black, disadvantaged infants until they were 36 months of age. Connell (1976) followed up his original sample at 30 months of age, and in addition followed through a new sample from 12 to 18 months. Main (1973) assessed her sample at 20½ and 21 months of age.
Bell (1978) used the Bayley Mental Scale with her sample of 33 black infants at 24 months, and the Stanford-Binet Scale at 30 and 36 months. (These tests were administered within 2 days of the free-play sessions, which occurred at each of these ages.) On the Bayley scale the mean scores of children who had been classified as B and non-B were not significantly different at 24 months. On the Stanford-Binet they were significantly different at 30 months but not at 36 months. (As reported in an earlier section, the two groups differed significantly in both Griffiths DQ and in object and person permanence at 15 months.)
Connell (1976) followed up 30 of the original sample of 46 infants who had been studied by Rosenberg (1975) and himself (1974) as 1-year-olds. On the Stanford-Binet scale administered at 30 months, the mean scores of children who had been classified into Groups A, B, and C 18 months earlier did not differ significantly, although there was a slight trend for Group-C children to have lower IQs than children of the other two groups. He also investigated a new sample of 55 children in a short-term longitudinal study from 12 to 18 months. The children were classified on the basis of their strange-situation behavior at 12 months. At 14 months the Cattell Infant Development Scale was administered at home, and at 18 months language was tested. Developmental Age, as measured by the Cattell scale, did not differ significantly between Groups A and B. The mean for each group approximated 16 months. (The three Group-C children in the sample were excluded from this analysis and from the language testing.)
Connell did, however, find significant differences between Groups A and B in respect to both maternal and child measures of language at 18 months. Group-B mothers had longer vocabulary lists, and Group-B children also had larger observed vocabularies. (Significant sex differences in vocabulary were also found, the details of which are not given here.) On a task involving repeating words spoken by the mother, B children were found to imitate her more frequently than A children, even when vocabulary size was controlled for. No differences were found, however, on a task involving language comprehension.
Earlier in this volume we emphasized the dynamic relationship between exploration and attachment; we have shown how even mild stress activates attachment behavior at the expense of exploratory behavior; and we have shown how the mother normally serves as a secure base for infant explorations. We have cited several studies in addition to our own that have shown that the mother’s presence or absence has a strong effect upon exploratory behavior; in particular, her absence has a detrimental effect.
Main (1973, 1977b) hypothesized that when a mother characteristically behaves in ways that prevent her baby from having confidence in her accessibility and/or responsiveness (i.e., ways characteristic of the mothers of anxiously attached babies, as we have reported in Chapter 8), the baby is under chronic stress, and hence unable to devote full attention to objects in this environment other than his mother. She set out to test this hypothesis using our strange situation.
Forty infants who had been observed in the strange situation at 12 months (a sample selected from Sample 3 and 4 of our present study, to represent a reasonable distribution of sexes and strange-situation classifications) were tested on the Bayley Mental Scale at 20½ months of age, and then observed in an hour-long play session at 21 months. Twenty-five infants who had been identified as securely attached (Group B) at 12 months were compared in regard to their later behavior with the 15 who had been identified as anxiously attached (Groups A and C).
At 20½ months, the toddlers earlier judged as securely attached were significantly more advanced in Developmental Quotient than the insecurely-attached toddlers. These differences could certainly be mediated in part by the differences found in exploratory behavior (see later) or by differences in the amount of time their mother had spent with them as Super Teachers (to use Bell’s term). But another mediating variable was simply the significantly greater cooperation that the securely-attached toddlers, in contrast with the anxiously-attached toddlers, showed in taking the Bayley examination (p < .01). They were willing to be tested.
Main’s primary interest in this study was, however, in the observation of unstructured exploratory behavior. The hour-long play session was divided into four episodes: a free-play session of 10 minutes in which the mother was present and responsive but noninterventive, play with an adult playmate for 20 mintues, another free-play session for 20 minutes, and finally a 10-minute episode in which the mother was to play with the child in any manner she found natural and preferable. Here we are interested in exploratory play behavior, and therefore we are concerned only with the 30 minutes devoted to free play. These episodes, as well as the others, were videotaped, and coding of preselected variables was made from the videotaped record.
Five measures of exploratory behavior were made. The securely-attached children spent longer in individual bouts of exploratory behavior (40 seconds vs. 29 seconds, p < .02); showed a more intense interest and attention to objects per bout (p < .001); attended more to the details of complex objects (p < .02); and more frequently laughed or smiled in relation to the toys and other objects in the room (p < .01). The total amount of time spent in exploration did not differentiate the groups significantly, although the secure children spent a longer time (p < .13).
Main also made six assessments of “semiotic function.” Two measures dealt with the presence and level of symbolic play; differences between the groups did not reach significance. The secure toddlers showed some tendency to issue more Vygotskian self-directions than other toddlers (p < .10). Although sheer number of utterances in words did not significantly distinguish the groups, they differed in the number of different words used (15 vs. 9,p < .09) and in mean morpheme length (3.1 vs. 2.1, p < .03).
Main found avoidance and resistance, which we have found associated with different patterns of infant—mother attachment, associated to different degrees with exploration and cognitive functioning. Avoidance was not significantly related to any deficits in cognitive functioning, whereas resistance was significantly negatively related to DQ, to length of exploratory bout, and to the intensity of interest in and attention devoted to the object during the exploratory bout. As we see later in this chapter, avoidance, in contrast, was related to disturbances in social and emotional behavior. In a more recent analysis of her 1973 data, Main (personal communication) has found that it was the Group-C babies rather than the Group-A babies who were lagging behind in tested (and also, to some extent, in observed) levels of cognitive functioning.
In addition to avoidance, resistance, and classification, five measures of discrete strange-situation behavior were correlated with 20 variables of the testing and play session (Main, 1973). These five—crying, touching, vocalizing, smiling, and looking—are variables that have frequently been used by other investigators of attachment behavior in lieu of the scores of interactive behavior that we have found more useful. Only seven of the 100 correlations reached significance at the .05 level or better—findings congruent with chance expectations and not otherwise interpretable.
Discussion. At 14 months Connell found no differences between Group-A and Group-B babies in regard to Cattell Developmental Age. At 20½ months Main found significant differences between Group-B and non-B babies in regard to Bayley DQ, but later reported that this difference was attributable to Group-C infants and that there was no significant difference between Groups A and B, thus confirming Connell’s findings. At 24 months Bell found no significant difference between B and non-B toddlers in regard to Bayley DQ. She found a significant difference to be yielded by the Stanford-Binet at 30 months but not at 36 months; but Connell did not find the same at 30 months.
These findings may be compared with those obtained toward the end of the first year or shortly afterward. Ainsworth and Bell (1974) reported a substantial multiple correlation coefficient (r = .70) between the mean fourth-quarter Griffiths DQ and several maternal variables, of which the two most important were maternal sensitivity to infant signals (a variable clearly characteristic of Group-B mothers in contrast to non-B mothers) and floor freedom permitted by the mother (which did not distinguish between B and non-B mothers). Beckwith (1971) reported similar findings for 24 adopted infants tested on the Cattell scale. Bell found significant differentiation between B and non-B infants in regard to Griffiths DQ and person permanence at 15 months.
There is thus substantial evidence that strange-situation classification and/or certain aspects of maternal behaviors are significantly related to measures of infant cognitive development when both sets of measures are obtained more or less concurrently. As for the value of strange-situation classification for “predicting” cognitive development, the findings are equivocal.
There are three major possibilities that might account for the fact that different investigators found different levels of “predictability” for different ages. One possibility, congruent with the test—retest findings reported in the literature, is that the correlation of performances at two different ages attenuates in proportion to the discrepancy in time between the two assessments. Although this explanation, by itself, does not wholly account for the discrepancies in the reported test findings, it seems likely to have had some effect on them. A second possibility is that different scales of infant “intelligence” tap different facets of cognitive function, and may even do so differentially at different age levels, and that some facets are more closely related than others to whatever is reflected by strange-situation classification. This argument resembles the oft-stated claim that the insignificant correlations between infant DQ and IQ later in childhood can be accounted for by the fact that the test instruments used test different functions. This argument does not altogether fit the facts here reported, but it may have some pertinence.
A third possibility is that the samples used by the various investigators differed sufficiently in regard to distribution of subjects among the three strange-situation groups, so that real differences in some may have been obscured by too few subjects representing one or other of the groups. Main’s finding that her B vs. non-B differences were accounted for largely by Group-C subjects lends support to this possibility. The fact that Bell’s disadvantaged sample contained a larger number of non-B subjects than her middle-class sample may have made it more likely that she could observe effects associated with strange-situation classification than did other investigators who used middle-class samples (e.g., Connell). (It will be recalled that Main selected her sample from a larger pool in order to include a larger proportion of non-B subjects than was the case in the pool from which she drew.) It seems likely that all three of these possible explanations have some pertinence in accounting for the discrepancies in the findings we have reported.
Both Connell and Main, however, did find differences between B and non-B toddlers in respect to language development—specifically, in regard to number of words used, and, at least in Main’s study, mean length of morpheme. Only Main specifically reported assessments of exploratory behavior. Her findings that securely-attached infants differed from anxiously-attached ones (and especially from Group-C infants) in a variety of different measures reflecting involvement in and enjoyment of exploration not only supported her hypothesis that this would be so, but thus also yielded findings strongly confirming our interpretation of the dynamics of behavior in the strange situation.
The studies reported or cited in this section, together with those reported in the earlier section dealing with concurrent assessments, form part of a promising new trend of investigation into the interrelations between social behavior and cognitive performance. This particular set of studies has implied a direction of effects from quality of attachment and social experiences associated therewith to facets of cognitive development and function. At least one study reported in Chapter 10 implies an opposite direction of effects—namely, from developing cognitive processes to changes in the nature of the attachment of child to mother. We do not consider these two sets of studies incompatible in their emphases; but we feel that both represent highly significant new ventures in research that, we hope, will stimulate further investigations of the interlocking of social and cognitive development.
Attachment in One-Year-Olds and Quality of Mother—Child Interaction and Social Development at Subsequent Ages
Mother—child interaction was observed by Bell at 15, 18, 24, 30, and 36 months, by Main at 21 months, and by Connell at 30 months for one sample and at 14 and 16 months for another. Matas (1977; Matas, Ahrend, & Sroufe, 1978) assessed interaction at 24 months. All four studies used a laboratory play session in which social behavior was observed, although they differed in length of play session, in whether or not the session was repeated at different ages, and in the proportion of the session in which the mother either was instructed to be responsive but noninterventive or was left free to play and/or interact with the child as she wished. For one of his samples, Connell observed free play in the home environment. Matas used both a problem-solving situation and a free-play session.
Bell (1978) observed mother—child interaction in an hour-long free-play session in which no constraints were put on the mother’s behavior, except for the distraction entailed in talking briefly with an interviewer midway through the session. We have already discussed the findings of the free-play session at 11 months. Here we are concerned with the repetitions thereof at 15, 24, 30, and 36 months, in which the same data-analysis procedures were used.
Each of 14 measures of the behavior of mother and child in interaction was correlated at each age level with three assessments of behavior in the strange situation that had taken place at 11 months: positive behavior toward the mother (proximity seeking and contact maintaining), negative behavior toward the mother (avoidance and resistance), and classification (Group B vs. non-B). As at 11 months, there were higher correlations of the 14 variables with strange-situation classification and negative behavior than with positive behavior. There were fewer significant correlations than there had been at 11 months, however. Moreover the variables fell into three groups.
One group consisted of variables that were not significantly correlated with strange-situation patterns at any of the later age points, although they had been at 11 months—specifically, the number of child’s initiations of interactions to which the mother responds and the number of child’s initiations to which the mother responds physically rather than socially or verbally. A second group of variables were correlated with strange-situation patterns at 15 months and in one case also at 24 months, but not at 30 or 36 months. These included total number of episodes of interaction, number of episodes initiated by the mother, the proportion of the latter to which the child responds, and the proportion that he ignores or resists. A third group of variables were correlated with strange-situation patterns at 24 months and in most cases also at 30 and 36 months, but not at 15 months. These were: child’s positive verbal/social and physical behaviors toward the mother, ratings of mother’s and child’s affect (both interpreted as indicating warmth and affection), and ratings of mother’s level of communication with the child in regard to interaction as a means of teaching the child about objects, playing with him, or in some way both expanding his awareness of things around him and keeping the interaction going. In addition one variable was significantly correlated with strange-situation patterns at all age points—namely, the proportion of maternal initiations of interaction that were positive, whether these were verbal or physical.
Bell’s basis of interpretation of these significant correlations was that the strange-situation classification (together with avoidance and resistance, the relative absence of which is crucial in distinguishing Group-B infants from others) reflects clear-cut and lasting differences in mother—child interaction. On this basis, the correlations reported above may be summarized as follows. Throughout the second year of life, Group-B dyads will continue to be characterized by frequent interaction, usually initiated by the mother, in which the child is positively responsive, rather than resisting or ignoring. Throughout the third year of life (but, for some reason not immediately explicable, not at 15 months) Group-B dyads continue to be conspicuous for “positive” behaviors—both those that reflect mutual warmth and affection and those on the part of the mother that actively stimulate the child’s interest in play and in objects in the world around him. Throughout the entirety of the second and third years, the social behavior of Group-B mothers toward the child continues to be positive, in both verbal and physical modes. In other words, the degree of harmony/disharmony in mother—infant interaction that was predicated by Bell to be reflected in patterns of strange-situation behavior is not a transitory phenomenon but rather tends to be stable over a long period of time.
Bell also found significant differences between Group-B and non-B dyads in the first-factor scores in each of the three principal components analyses done for behavior in the free-play sessions at 15, 24, 30, and 36 months. (See an earlier section for the findings at 11 months.) The factor reflecting positive versus negative affective tone in the child’s behavior to his mother significantly differentiated B from non-B dyads at 15, 30, and 36 months. The difference was in the same direction at 24 months but fell below the .05 level of significance (p < .07). The factor reflecting affectively positive and responsive versus affectively negative and unresponsive maternal behavior significantly differentiated B from non-B dyads at all four age levels. The mother-as-super-teacher factor differentiated significantly between B and non-B dyads at 15 and 24 months (p < .01); at 30 months it yielded no discrimination between groups, but at 36 months again there was a tendency (p < .09) for Group-B mothers to be super teachers. Although Bell had originally discriminated “didactic” from social/affective modes of interaction with the child when she subjected them to separate principal components analyses, she found that they tended to be associated together; the mother whose interactive behavior reflects a positive affective tone tends also to display didactic features in her play with her child, taking care to interest him in learning how to cope with objects in his environment. Nevertheless, Bell proposed that the didactic features of mother—child interaction become increasingly important toward the end of the first year of life—continuing on throughout the second year at least—in stimulating her child’s cognitive development.
Connell (1976) followed up two samples of children who had been observed in the strange situation at 12 months of age. One sample, consisting of 30 of the original sample of infants who had been studied by Rosenberg (1975) and himself (1974), were introduced at 30 months to a laboratory playroom with their mothers. The room was partitioned into a playroom section equipped with toys and a “living room” with a sofa and magazines. For 10 minutes they were left by themselves for the child to become accustomed to the playroom. Then the investigator placed a chair in which the mother was to sit, responding to the child’s initiations of interaction but otherwise being noninterventive. After 5 minutes the mother was cued to move to the living-room sofa, where she was to behave in the same noninterventive but responsive way and where she remained for 5 minutes.
At 12 months, eight of the 30 children had been classified in Group A, 17 in Group B, and five in Group C. There were significant differences among these groups in regard to mother—child interaction at 30 months. The Group-B dyads had more interaction than the A dyads, and the mean length of a bout of interaction was longer. The C dyads had more interaction than the B dyads during the episode when mother was on the sofa. Group-B children spent more time within 6 feet of the mother than the A children, under all conditions. When mother was on the sofa, C children spent more time within 6 feet of her than did B children. It may be noted that the sofa was in the living-room area, out of sight of a child in the play area. Presumably the C children interpreted the mother’s move to this more distant position as indicating a decrease in her accessibility, whereas the B children considered her still accessible.
Connell also examined the correlation of six strange-situation variables with the following four measures of interaction at 30 months: mean length of interaction, type of interaction, total proximity, and total vocalization. None of the strange-situation variables was significantly correlated with the proximity measure; only contact maintaining in the reunion episodes even approached significance. Both proximity/contact seeking and contact maintaining in the reunion episodes were significantly correlated with length of interaction, type of interaction, and vocalization. Exploration was significantly but negatively correlated with quality of interaction and vocalization. Crying in episodes 7 and 8 was significantly correlated only with vocalization. Neither avoidance nor resistance in the reunion episodes was significantly correlated with any of the later measures of interaction, although a negative correlation between resistance and vocalization approached significance.
Connell’s second (1976) sample consisted of 55 infants observed in a short-term longitudinal study from 12 to 18 months of age. The strange situation was administered at 12 months. Two observers watched mother—infant interaction at home during a 1-hour session at 14 months and again at 16 months. At 12 months, 19 infants were classified in Group A, 33 in Group B, and 3 in Group C. Because they were so few, the C infants were not included in the home observations.
A multivariate analysis of variance yielded significant (p < .02) effects for Group-A versus Group-B dyads for the following measures of interaction in the home visits: time spent in interaction, number of interactions initiated by the child, distance between child and mother, and mean length of bouts of interaction—at both 14 and 16 months. In each case the means for the B dyads were higher, except for distance between child and mother, which was less for the B dyads. No differences between A and B dyads were found for the following variables: mother’s ignoring of the child’s signal for interaction, child’s crying, mother’s restriction of the child, and either mother’s or child’s anger—at either 14 or 16 months.
Thus the findings for his two samples—one was observed at 30 months in the laboratory and the other at 14 and 16 months at home—are congruent. Children classified in Group B in the strange situation at 12 months, in comparison with children classified in Group A, maintained closer proximity to their mother and sought and maintained more interaction with them, both in terms of total time spent in interaction and longer bouts of interaction.
Main’s 1973 report did not deal with maternal behavior or with mother—child interaction; these aspects of her study were subjected to later analysis by her students at the University of California at Berkeley. Tomasini (1975), Tolan (1975) and Tolan and Tomasini (1977) examined Main’s play-session videotapes of her sample of 40 children, aged 21 months, and their mothers with respect to maternal behavior. Both Tomasini and Tolan (and later Main & Londerville, 1978; Londerville, 1977) compared three groups of mothers in terms of the strange-situation classification of their babies at 12 months of age: mothers of B3 infants, mothers of B1 and B2 infants, and mothers of A and C infants.
Tolan (1975) focused on mother’s facial expression, hypothesizing that differences in her “facial affect communication” should be related to differences in the security of the infant—mother attachment relationship. To test this hypothesis, slides were made of the faces of the mothers videotaped in Main’s study at two points in the session—when the adult playmate first entered the room to invite the toddler to participate in a game of ball, and during the first minute of the game. The toddler’s face was masked in the slides. The slides were rated (by raters who were “blind” in regard to both the hypothesis and strange-situation classification) for expressiveness—the degree to which the mother’s face seemed to express any emotion whatsoever—and for pleasure in her expression. Analyses were conducted using both the first slide taken and the mean rating over several slides. All eight resultant correlations were significant. Mothers whose infants were more securely attached (as assessed by the three-point measure implicit in the B3 vs. B1/B2 vs. A/C groupings) had more expressive facial behavior and more frequently expressed pleasure than did mothers whose infants were less securely attached. Even ratings given to the first slide of the mother’s face taken at the playmate’s entrance differed significantly, even though at this moment the mothers were reacting to the adult playmate rather than to their (secure or anxious) toddlers.
Tomasini (1975) was interested in replicating our findings (see Chapter 8) that mother’s degree of sensitivity to infant signals and communications and her degree of acceptance vs. rejection of the child were related to the patterns of strange-situation behavior that are reflected in the classifications. She adapted the two Ainsworth scales of sensitivity—insensitivity and acceptance—rejection to fit videotaped mother—child interaction for the two-year-old, and rated the mothers of Main’s sample on the basis of repeated viewings of the play-session tapes—the first 10 minutes of free play and the last 10 minutes, which was made up of mother—child play. Both sensitivity (p <.05) and="" acceptance="">p < 0.1) were positively related to the degree to which the child had shown secure attachment in the strange situation 9 months earlier.
<.05) and="" acceptance="">
Tomasini also made extensive and extremely detailed narrative descriptions of the mother’s behavior with special reference to three variables of particular interest to Main; the mother’s apparent attitude toward physical contact with her child, her general emotional (rather than merely facial) expressivity, and her anger. An assistant, unfamiliar with Tomasini’s other ratings and any other assessments, rated these from Tomasini’s narrative descriptions. The mother’s apparent attitude to physical contact was not related to strange-situation classification, but mothers of more secure babies were significantly more expressive (p <.05) and="" much="" less="" angry="">p < .001) than other mothers.
<.05) and="" much="" less="" angry="">
Main and Londerville (1978), in a new and elaborate analysis of the Main videotapes, were not so much interested in mother—child interaction as such as in toddler socialization—whether the toddler was developing into a cooperative, “easy” child, with internalized controls, or an actively disobedient, difficult, and antisocial child—and its relationship to strange-situation behavior in infancy. The three groups compared in their analysis were: B3, B1/B2, and A/C. Londerville devised the coding system for each of six “socialization” variables: obedience to maternal verbal commands, active disobedience to such commands, “internalized controls” (shown by self-inhibition of forbidden behavior), “baby rescue” (acceding to the adult playmate’s urgent request not to let a toy dog bite a baby doll), physical attack on the mother, and maternal description of the child as being very difficult vs. easy to live with. Londerville undertook the coding of the videotapes (and rating of the maternal interview) without knowledge of strange-situation behavior. A seventh variable was a rating of the child’s cooperation with the Bayley examiner (Main, 1973). A trend analysis was used to determine the presence of a linear ordering in the expected direction for each of the seven variables.
For six of the seven variables, the means for the B1/B2 toddlers fell between the means for B3 and for A/C; the exception consisted of a minimal crossover between B1/B2 and B3 with respect to active disobedience. Trend analysis showed five significant linear orderings. The more securely attached the toddler had been as a 1-year-old, the more likely he was as a toddler to self-inhibit forbidden behavior, to obediently “rescue” a toy “baby” when urgently asked to do so by the adult playmate, to be reported easy to live with by his mother and, to cooperate with the Bayley examiner, and the less likely to physically attack or threaten to attack his mother. There was also a positive association between degree of security of the attachment relationship and the percentage of maternal commands obeyed (p < .10), and a negative association with the percentage of maternal commands actively disobeyed (p < .10). When the seven variables were converted into standard scores and summed into a single “socialization score,” trend analysis showed an extremely strong ordering in terms of the original security groupings: F (1,33) = 21.13, p < .001.
Main and Londerville also considered strange-situation avoidance and resistance in relation to toddler socialization. They entertained the hypothesis that these two variables might reflect some temperamental characteristic of the infant first manifested at 12 months, rather than be the outcome of important differences in mother—infant relationships. If this hypothesis were true then behavior with the stranger at 12 months should relate as strongly to toddler socialization as does behavior with the mother in the strange situation. Correlations between the toddler-socialization variables and strange-situation avoidance and resistance shown both to the mother and to the stranger were therefore computed. To summarize, the relationships between avoiding and resisting the stranger in the strange-situation and the toddler-socialization scores were insignificant. Avoidance of the mother, however, was significantly related to toddler socialization: r(36) = −.54, p < .001. Resistance to the mother was not significantly related to toddler socialization, except that it was related to failing to cooperate with the Bayley examiner at 20½ months, whereas resisting the stranger in the strange situation was not related.
Finally, Main and Londerville computed the product-moment correlations between the five discrete attachment measures in the strange situation—crying, touching, looking, smiling, and vocalizing—and each of the toddler-socialization variables, and found none of them significant.
Matas (1977; Matas, Ahrend, & Sroufe, 1978) observed a sample of 45 infants at 24 months of age in a free-play session and in a problem-solving situation, in both of which their mothers were present. The problems involved the use of tools, and some of them were very difficult for 2-year-olds. The mothers were instructed to let the child work on the problem for a while before giving whatever help they thought necessary. Both sessions were videotaped, and later both child’s and mother’s behaviors were coded or rated by two independent judges.
All children had been earlier observed in the strange situation and classified according to our procedure—33 on the basis of their performance at 18 months, and 12 at both 12 and 18 months. (In Chapter 11 we cite findings—Waters, 1978—of the degree of agreement between classifications at 12 and 18 months—and it is impressive.) Those who had been identified as securely attached (i.e., Group B) 6 and/or 12 months earlier, at 24 months of age were significantly more enthusiastic, affectively positive, and persistent. They showed less frustration behavior and less nontask behavior in the problem-solving situation than did those who had been classified as insecurely attached (Group A or C). In comparison to the latter, the securely-attached children also showed less ignoring of the mother and less noncompliance, negativism, and negative affect. It was the avoidant children (Group A) who were especially noncompliant. In addition the Group-A children tended to seek help from the experimenter rather than from the mother, and showed unprovoked aggression to the mother. Ambivalent babies (Group C) were conspicuous for showing extreme reliance on the mother, giving up quickly in the problem-solving situation, exhibiting such frustration behaviors as whining and stomping, and appearing generally incompetent. In the laboratory situation the mothers of anxiously-attached children (i.e., non-B) were rated as significantly less supportive and as offering a lower quality of assistance.
Discussion. There are two ways in which the findings of these four studies may be interpreted. One is that the behavior of a 1-year-old in the strange situation reflects the degree of harmony/disharmony experienced in the interaction with his mother during previous months and that individual differences in the degree of harmony in mother—child interaction continue throughout at least the next year or two. The other interpretation is that the way in which an infant has organized his attachment to his mother, in response to the nature of his previous interaction with her, tends to persist and to influence his behavior in predictable ways. There is no way to distinguish decisively between these two interpretations, for none of the studies report changes in maternal behavior that might influence the nature of the child’s attachment to her. As Sroufe and Waters (1978) have pointed out, however, assessments of the child’s behavior in the strange situation are independent of maternal behavior, so that it is not merely mother—infant interaction that is being assessed in that situation.
Nevertheless, the implication of the first interpretation is that a securely attached infant continues to be securely attached not so much because the quality of attachment is an enduring characteristic as because his interaction with his mother continues to support a secure-attachment relationship. Both Bell and Connell reported more interaction among Group-B dyads throughout the second year of life, and Connell also reported more in the third year. He noted longer bouts of interaction among B dyads and a tendency for the child to maintain close proximity to his mother.
Three of the four investigators found that Group-B mothers differed from A and C mothers in the nature of their input into interaction with their children at times during the second and/or third year. Bell found that they initiated more interaction, were more positive in the affective tone of their behavior, and were more active in stimulating the child’s interest in exploratory play. Main and her associates found that they expressed feelings and emotions more readily, showing more pleasure. They were also more sensitive in their response to the child’s communications, as well as less rejecting and less angry. (Connell, however, did not find differences between A and B mothers in terms of anger.) Matas found them more supportive to their children in a problem-solving situation, and as offering a higher quality of assistance than did A or C mothers.
All four investigators found that Group-B children differed from A and C children in terms of their input into interaction with their mothers. Bell found that Group-B children were more positive in the affective tone reflected by their behavior throughout the second and third years, and Matas reports the same finding for 24-month-olds. Connell found that they maintained less distance from their mother and initiated more bouts of interaction. Main and her associates found that they were more socialized—more capable of “self-inhibiting” forbidden behavior, easier to live with, more cooperative, and more empathetic. Both Main and Matas reported that Group-B children are less aggressive toward their mothers and more compliant to maternal commands. In addition Matas’ findings that Group-B children showed less frustration and nontask behavior in the problem-solving situation are congruent with Main’s findings that were reported in an earlier section.
Connell found no differences between Group-B and non-B dyads in terms of anger. Main found Group-B mothers less angry and Group-B children less aggressive. Matas found that Group-A children were particularly likely to display unprovoked aggression toward their mother. It will also be recalled that Main found more evidence of anger in Group-A mothers and babies in our Sample 1 than among Group-B dyads. The differences between Connell’s findings and those of Main and Matas may reflect different lengths of observation, different conditions of observation, and differences in criteria of anger and/or aggression. We are sufficiently impressed with Main’s and Matas’ positive findings to urge future investigators to attend to this variable.
It may be noted that the four investigators used different degrees of refinement in their comparisons of variables with strange-situation classification. Bell limited herself to B vs. non-B comparisons; Connell compared A, B, and C groups in one sample, and A and B in the other. Matas compared both B and non-B groups, Group B with both A and C, and A and C with each other. Main compared B and non-B in some of her analyses and dealt indirectly with A and C groups in her analyses of avoidant and resistant behavior. In some of the comparisons of maternal behavior, Main’s associates compared B3, B1/B2 and A/C. It is of interest that the findings of these studies are congruent on the whole, and that Matas’ differentiations between A and C and Main’s differentiations between B3 and B1/B2 were also significant.
Main, Bell, and Connell all considered strange-situation interactive measures in addition to classification. Bell pooled avoidance and resistance and found that, combined, they tended to correlate more significantly with later mother—child interaction variables than did the combined “positive” variables of proximity/contact seeking and contact maintaining. Main dealt only with avoidance and resistance, treating them separately; she found that avoidance was significantly (and negatively) related to toddler socialization, whereas resistance was not. Connell, on the other hand, found that both proximity/contact seeking and contact maintaining were significantly correlated with measures of mother—child interaction at 30 months, whereas avoidance and resistance were not. (The discrepancy between Connell’s findings and those of Main and Bell is not immediately explicable.) Main also checked out five “discrete” measures of strange-situation interactive behavior and found them unrelated to later socialization.
Attachment at One Year and Behavior Toward Unfamiliar Adults in the Second and Third Years
Two studies, one by Main (1973) and another by Connell (1976), examined the extent to which strange-situation behavior or classification at 12 months was related to later behavior toward adult figures with whom the child had only slight familiarity. In the sample followed up at 30 months, Connell at the end of the mother—child free-play session invited the child to accompany him to another room to play some games. If the child would not leave with him within 1 minute, the child was asked if he would come with his mother. In either case the mother accompanied them to the other room, where a Stanford-Binet test was administered.
Scores were assigned to the response to the experimenter/stranger, as follows: a score of 1 for leaving with the stranger, without the mother and without urging; a score of 2 for leaving with the stranger, without the mother after urging; a score of 3 for refusing to leave without the mother; a score of 4 for uncooperative behavior in the test situation. Group-A children received the lowest mean score, and Group C the highest (F[2,27] = 5.2, p < .02). Most of the Group-A children left without their mothers, whether with or without urging. Most of the Group-B children refused to leave the room without the mother, although they subsequently participated willingly in the test; a substantial minority did, however, agree to leave without their mothers. None of the group-C children left without their mothers, and most of them were uncooperative in the test situation.
Main’s (1973) data came from two sources: from the child’s interaction with the Bayley examiner at 20½ months and from the 20-minute episode of her play session at 21 months, in which an adult playmate attempted to engage the child in a sequence of different kinds of play behavior. As we mentioned earlier, Main found Group-B toddlers to be significantly more cooperative with the Bayley examiner than the non-B toddlers. They were also more likely to treat the test as an opportunity for playful interaction with the Bayley examiner—that is, to show more of a “game-like spirit.” This greater friendliness and playfulness was again affirmed 2 weeks later with another person, the female playmate, who also had entertained the child for a few minutes before the laboratory play session began. When the adult playmate invited Group-B toddlers to engage in a game of ball, they tended to approach her and to return the ball to her in a game-like manner, whereas non-B toddlers tended to avoid her. Main further found that avoidance of the mother in the strange situation was positively related to avoidance of the playmate 9 months later and negatively related to a game-like spirit in the episode with the playmate.
Main (personal communication) does not consider avoidance of the playmate in the play session to be the kind of fearful or wary avoidance young children may display toward a complete stranger. Although, like fearful avoidance, it was manifested by gaze aversion and/or turning away from the adult, it seemed to express merely unwillingness to interact with the playmate rather than fear/wariness. She further suggested that such avoidance seems to reflect an entirely different phenomenon from unwillingness to leave the mother to accompany an unfamiliar person. She based this suggestion on informal observation of the behavior of her sample in the 15-minute familiarization session, which preceded the formal laboratory play session and during which the adult playmate attempted to interact with the child in the investigator’s office and in the adjoining hallway. As in Connell’s sample, her Group-B toddlers were reluctant to leave their mothers to go into the hall with the playmate, although later in the play session they were friendly and playful with her; and as in Connell’s sample, her Group-A toddlers were willing to accompany the playmate away from the mother, even though they were unwilling to enter into reciprocal play with her in the play session. One A toddler, alarmed by thunder during the familiarization session, dashed to the playmate rather than to his mother. Two A toddlers protested the playmate’s departure at the end of the playmate episode of the play session, even though they had been unwilling to interact playfully with her; and one of these also protested when first left in the playroom with his mother for the beginning of the play session.
In conjunction with Main’s findings pertaining to behavior toward the mother in the play session, reported earlier, it would appear that the willingness of the Group-A toddler to leave his mother with an unfamiliar person in an unfamiliar environment reflects disturbance in his relationship with her rather than greater friendliness toward other persons than Group-B toddlers tend to show. Indeed, as long as the mother is present, Group-B toddlers interact more positively with unfamiliar persons than do Group-A toddlers, just as they interact more, and more positively, with their mothers—as Bell and Connell, as well as Main, have shown.
Comparison of Behavior in the Strange Situation with Behavior in Other Situations of Parallel Design
Whereas the other studies that we have considered in this chapter have focused on individual differences in strange-situation behavior and their relationship to individual differences in other variables, the four studies that we consider in this section are normative in their thrust, concerned with the effect of variations in environmental conditions upon strange-situation behavior. One of these, by Brookhart and Hock (1976), is concerned with the setting in which the “strange” situation is conducted, whether in the unfamiliar setting of the laboratory or in the familiar setting of the home. The other three deal with the issue of whether the behavior directed by the child to the adult who accompanies him differs according to the identity of that figure.
Behavior in the Laboratory Versus Behavior at Home. Brookhart and Hock (1976) designed a set of episodes, similar to those of the strange situation, that could be staged in the infant’s familiar home environment. Some episodes had to be altered substantially in order to fit the conditions at home, although the investigators aimed to make them as closely matched as possible to those of the standard laboratory situation. (Obviously, however, the situation was not “strange” in the sense in which we originally intended it—as taking place in an unfamiliar environment.) Thirty-three infants were introduced to the home and laboratory situations, in counterbalanced order, with the first session occurring at a mean age of 11.3 months. The measures used were our measures of proximity and contact seeking, contact maintaining, resistance, and avoidance (all scored for behavior to mother and stranger), together with search and crying.
A multivariate analysis of variance was undertaken with main effects examined for location (i.e., home vs. laboratory), episodes, and sex. (Because the subjects included both day-care and home-reared groups, the main effects attributable to group were also examined as reported earlier.) Here we consider only the main effects attributable to location. Four of these emerged as significant: The infants showed more proximity seeking and contact maintaining toward the stranger in the laboratory than at home, and more avoidance of both mother and stranger at home. There were significant location by episode interactions for maintaining contact with both mother and stranger, indicating a greater increase across episodes in the laboratory than at home.
The findings relating to proximity and contact are readily explicable by the assumption that the separation episodes in the unfamiliar environment of the laboratory are more upsetting than those occurring at home; hence stronger instigation to attachment behavior occurs both in these and in the reunion episodes that follow them. The lesser avoidance of the stranger in the laboratory situation may be viewed in similar terms. The greater avoidance of the mother in the home environment has no such obvious explanation.
Brookhart and Hock acknowledge that avoidance of the mother at home could not be interpreted as a defensive “detachment” reaction, as Ainsworth and Bell (1970) had interpreted such avoidance in the reunion episodes in the laboratory. Rather, they interpreted it as an “independent gesture” by infants who had not become anxious about their mother’s comings and goings at home. In Chapter 7 we reported that the securely-attached (Group-B) infants showed least separation anxiety at home; they were also notable for lack of avoidant behavior in the laboratory strange situation. On the other hand, Group-A babies, conspicuous for avoidance in the mother in the laboratory reunion episodes, show relatively strong separation anxiety in the home environment (in fact, A1 babies did so, although A2 did not). If indeed the securely attached babies were scored as most avoidant at home, we certainly agree with Brookhart and Hock that mother-avoidance in the two different situations must be interpreted differently. Indeed, we are not convinced that what Brookhart and Hock scored as avoidance of the mother in the home situation should be considered as such. It seems likely to us that the behavior of the securely attached infants at home would be comparable to behavior in Episode 2 in the laboratory, when the child is using his mother as a secure base from which to explore the world and is thus absorbed in his own activities. At home, he may be confident enough of his mother’s accessibility so that even when she leaves the room he can continue to use her as a secure base and so that when she returns, attachment behavior may not be activated. Under these circumstances, to continue exploratory play and to make only brief acknowledgment (if any) of the mother’s return seems unlikely to imply avoidance, for we have observed such behavior to occur commonly among securely attached infants in the course of our longitudinal study conducted in the home.
Comparisons of Two or More Figures as Accompanying Adults. Feldman and Ingham (1975) held that the “measures of attachment derived from the Ainsworth strange situation suffer from inadequate validation.” If they are indices of attachment, it is necessary to demonstrate that they are exhibited more frequently or more intensely toward attachment figures than toward others. The fact that attachment behaviors are indeed exhibited differentially to the mother in comparison with the stranger in the strange situation was not considered to be adequate evidence of their specificity; the child might have behaved similarly to any accompanying adult. (We agree that the mother vs. stranger comparisons in the strange situation do not provide crucial evidence of specificity of attachment behavior, but we point out that the strange situation was not designed to explore the issue of behavior that is differentially directed to the mother in comparison with other figures.)
To examine this issue of differentiality, Feldman and Ingham undertook two comparable studies, one with 1-year-olds and another with 2½-year-olds, in each of which the subjects were randomly assigned to one of three conditions, in terms of whether the accompanying adult was the mother, the father, or a relatively unfamiliar acquaintance. The third was a woman whose total familiarity to the child was 1 hour’s free play with him at home immediately before accompanying him to the strange situation. Only the 1-year-old findings are considered here; the findings for the older group are presented in Chapter 10.
The 1-year-old sample numbered 56. The procedure was intended to be identical with our strange situation (except for extending Episode 2 to 4 minutes, during the first two of which the adult was attentive to the child, and during the last two of which he filled out a questionnaire).3 Feldman and Ingham used the same measures used by Maccoby and Feldman (1972), some of which differed substantially from our measures. Their measure of proximity was a composite score roughly comparable to our score of proximity and contact seeking. All of their other measures were frequency measures: playing (similar to our exploratory manipulation), activity (which included all locomotion, whether exploratory or otherwise), crying, looks, and distance bids (a composite measure combining speaking, smiling, and showing a toy—roughly comparable to our measure of distance interaction). None of these measures was scored for the reunion episodes. Kruskal—Wallis analyses of variance were carried out for each of these measures for each relevant episode (excluding the reunion episodes). In the case of significant findings, Mann—Whitney U tests were performed to examine the differences between the mother, father, and acquaintance conditions.
For the reunion episodes a tally was made of the presence or absence of certain behaviors within 10 seconds of the reentrance of the accompanying adult, including the following: looking, smiling, talking, gaze aversion, increase or decrease of crying, movement toward the entering adult, and bids for comfort. Although the presence or absence of certain other behaviors was noted for the remainder of the reunion episode—for example, whether the child initiated physical contact, and whether he rejected a toy offered by the adult—nothing comparable to our measures of proximity and contact seeking, contact maintaining, avoidance, and resistance was undertaken for the reunion episode as a whole.
Let us turn to the findings. In regard to several measures, it seemed that mother and father were interchangeable as attachment figures. In comparison to those accompanied by an acquaintance, babies accompanied by a parent cried less in the preseparation episodes, were more active in all but one episode, more frequently acknowledged the stranger’s first appearance in Episode 3, and sought more proximity in the preseparation episodes. There were some differences between mother and father as accompanying adults, however. Infants accompanied by the mother played more in Episode 2 than those accompanied by father or by the acquaintance, whereas there was no difference between the latter two groups. In Episode 3 those accompanied by the mother made more distance bids to her than did those accompanied by the father to him, but both groups in that episode made more distance bids to the stranger than to a parent.
Differences between conditions in regard to reunion behavior were examined by means of chi-square tests. Although approach and wanting to be picked up were both more frequent in the case of parents than in the case of the acquaintance, the differences were evidently not large enough to be significant.
Feldman and Ingham concluded that reunion behaviors in the strange situation fail to yield significant differences between accompanying figures and hence challenged our claim as to their significance as attachment behaviors. The ways in which they measured reunion behavior were so different from the ways in which we did, however, that we do not consider that they have properly tested our propositions.
Let us consider the ways in which our measures differ from theirs. First, our measures of interactive behavior recognize that even the 1-year-old is capable of goal-corrected behavior, so that different children (and the same children at different times) may adopt different and yet perhaps equally effective modes promoting contact and/or proximity—whether by approaching and clambering up, by reaching, or by other modes of signaling the adult to approach, or by some combination of these. To consider each of these modes as discrete behaviors, rather than as more or less equivalent alternatives, would tend to obscure differences in proximity and contact seeking that might well exist. Second, Feldman and Ingham (and Maccoby and Feldman in their earlier investigation, which is discussed in Chapter 10) did not attempt to take account of the contingencies of the adults’ behavior, as did our systems for scoring interactive behavior. In the reunion episodes the adult’s behavior is variable and largely uncontrolled by instructions; they therefore felt it necessary to limit the use of most of their presence—absence measures to the first 10 seconds of the episode, which constitutes a very small segment of reunion behavior. Our measures of interactive behavior were, in contrast, designed to allow for differences in the behavior of the adult, and therefore could comprehend the entire episode instead of a fraction of it.
The fact that significant differences in behavior toward attachment and nonattachment figures were found in the preseparation episodes makes it seem likely to us that they would also be found in the reunion episodes if our scoring system had been used. It may be noted that it was proximity seeking rather than “distance bids” or looking that differentiated between figures in Episodes 2 and 3, and this would also be our expectation for the reunion episodes.
Furthermore, we have stressed reunion behaviors for their role in distinguishing qualitative differences in the attachment relationship with the mother rather than indices of “strength of attachment” in terms of which attachment toward one figure might be compared with another. In advance of undertaking a study such as Feldman and Ingham’s our expectation would be that both seeking to gain and to maintain contact and/or proximity and the negative behaviors (avoidance and resistance) would tend to be greater when the accompanying adult was the mother rather than an acquaintance, and that although proximity and contact might be sought with the father as strongly as with the mother, babies would tend to show less avoidance of or resistance to the father than to the mother.
Lamb (1978) was concerned with Freud’s assertion that the infant—mother relationship was “the prototype of all later love relations,” because his previous work (e.g., 1976a, 1976b, 1976c, 1977a) supported the view that attachment to mother and attachment to father are qualitatively different. In one context and through one set of behaviors a child might show preference for his father, whereas in another context and through another constellation of behaviors he might show preference for his mother. Regardless of such differences, however, both attachment relationships in their own ways are significant for social and personality development. This being so, then one relationship could scarcely be the prototype for the other.
Lamb chose the strange situation as a procedure for further investigating this issue—for rounding out the picture presented by his earlier research into this complementary nature of an infant’s attachments to father and mother. He used 32 1-year-olds who were brought to the strange situation twice, with an interval of 1 week, once accompanied by the mother and once by the father, half of the subjects coming first with mother, half first with father. Two strangers were also used in counterbalanced order, resulting in four procedural groups, each of which consisted of four girls and four boys.
He used videotapes rather than dictated narratives as records. He recorded the duration per episode for each of the following: crying, exploration, and oral behavior. His measure of distance interaction (with parent and with stranger) resembled ours. Perhaps because the videotape records yielded an overwhelming amount of detail (see our discussion of this issue in Chapter 13), he did not use our measures of proximity/contact seeking, contact maintaining, avoidance, resistance, or search behavior. Instead he tallied the presence or absence in each episode of a specific list of relevant behaviors, including protest, search, and soothability (by the stranger) in the separation episodes, and approach (full or partial, delayed or otherwise), touching, pick-up appeal, positive or negative greeting, resistance to the put-down, avoidance, and resistance to contact and to interaction in the reunion episodes. Finally each baby was classified in one of Groups A, B, or C, according to our classificatory criteria.
Sixteen infants were classified as secure (Group B) and seven as insecure (Group A or C) in the relationship to both parents, whereas nine emerged as secure with one parent and insecure with the other. It is not clear whether the counterbalanced design of the study confounded the order effects with differences in classification of attachment to the two parent figures. Lamb reported that he found no order effects for measures of behavior in the strange situation. Nevertheless, there was a marginally significant similarity of attachment relationships (p < .055).
We cannot go into the details of the findings in regard to differences in the percentage of babies showing and not showing each item of behavior to father vs. mother. There were a number of similarities in responses to parents in the first reunion episode, but very few in the second. In the separation episodes the only consistent similarity was distress so acute that the episode had to be curtailed. Lamb concluded that this study gave equivocal support to the hypothesis that attachment relationships to mother and to father are similar in nature. Indeed, we suggest that Lamb’s earlier procedures, which compared the behavior of infant to each parent figure when both were present, did more than the strange—situation procedure to clarify differences (and similarities) in regard to attachment behavior to different parent figures in a variety of situations.
Willemson, Flaherty, Heaton, and Ritchey (1974) predicted, from their understanding of attachment theory, that various indicators of attachment should be intercorrelated not only with each other but also with indicators of exploratory behavior and furthermore that these correlations should be greater for the infant when with his mother than with his father, who may presumed not to be an object of strong attachment. Their subjects numbered 24 (12 boys and 12 girls), divided into four groups of six each. Two groups experienced the strange situation first with the mother, two first with the father; and each was exposed to two arrays of toys in counterbalanced order, one “more interesting” and one “less interesting.”
Neither dictated narratives nor videotape records were kept. Instead, scoring was done “instantaneously” by two observers. For crying and exploratory behavior, presence or absence was scored for each 15-second interval. For our measures of proximity/contact seeking, contact maintaining (to mother and stranger) and search, there was a quick rating at the end of each relevant episode. Neither avoidant nor resistant behavior was scored, and exploratory manipulation (which we found to be the most useful of the three exploratory measures) was dropped from analysis because of low intercoder agreement.
The findings may be summarized as follows. There was a strong order effect; attachment behavior, especially crying, increased from Session 1 to Session 2. There were few significant differences in behavior toward mother vs. father. Slightly more exploratory behavior to toys and less attachment behavior to parent was evinced in the presence of the “more interesting” in comparison with the “less interesting” set of toys. Proximity-seeking and contact-maintaining behaviors were positively correlated with each other, but the two exploratory behaviors (locomotor and visual) were not significantly correlated with each other or with the attachment behaviors. There were no “main effect” sex differences, although there was an inverse relationship between attachment and exploratory behaviors for girls, but not for boys.
The findings that behavior toward mother and father is essentially the same the authors take to be inconsistent with ethological attachment theory, as is the finding of an inverse correlation between exploratory and attachment behavior (albeit only in girls), although the authors acknowledge that the latter was their own prediction. They perceive some basis for speaking of attachment behavior as an entity. On the other hand, their findings in regard to the two sets of toys lead them to conclude that attachment behavior is probably more a function of situational variables than evidence of a focused, specific bond with the mother.
Despite disparities in approach and hypotheses there are some similarities in the findings of these three studies. The first and most conspicuous is that 1-year-olds in the strange situation behave toward mother and father in much the same way. This finding is not surprising to us, we would have expected that any attachment figure could support exploration in the nonstressful episodes of an unfamiliar situation and that intensified attachment behavior would be directed toward such a figure under stress. Although Bowlby’s concept of “monotropy” suggests that the principal attachment figure may be preferred under stress conditions to other attachment figures (and that the figure who plays the mother role is likely to be the principal attachment figure), such preference can scarcely be assessed unless the comparison figures are simultaneously present.
Furthermore, in our opinion, none of these studies provides an adequate basis for assessing the attachment relationship to mother and to father (separately) and then comparing them. Both Feldman and Ingham and Willemsen and associates implied in their hypotheses a trait theory of attachment, so that differences in behavior to mother and father might be indicators of different strengths of attachment, although their findings gave little support to these hypotheses. Lamb, on the other hand, expected to find qualitative differences in response to the two figures, but used measures too insensitive (in our opinion) to identify them if indeed they do exist. We think it is likely that the “main effect” of the strange situation is to progressively heighten attachment behavior toward an accompanying attachment figure, and this effect is so strong that it tends to overshadow quantitative and qualitative differences that might otherwise differentiate one figure from another in regard to specific attachment behaviors or their organization together.
It is of interest to consider Feldman and Ingham’s findings regarding the behavior of children accompanied to the strange situation by a mere acquaintance. First, they attest to the efficacy of a “familiarization” period to turn a possibly alarming stranger into a comfortable companion. Second, they show that a barely familiar but benign figure can serve remarkably well as a substitute for an attachment figure in a stressful situation. (This matter is discussed further in Chapter 13.) Even so, high-intensity behavior (e.g., proximity seeking) was found more likely to be directed toward parents as accompanying figures.
Willemsen and associates are correct in our opinion to emphasize the situational variables in the strange situation. Our interpretation, however, is that the succession of episodes tends to progressively heighten attachment behavior, and that this heightening will tend to be expressed toward the accompanying adult, whether principal or secondary attachment figure, and whether attachment figure or merely benign conspecific. On the other hand, for reasons that we elucidate in Chapter 13, we acknowledge that the extent to which an unfamiliar situation is “interesting,” and thus activates exploratory behavior, may modify the expression of attachment behavior toward the accompanying adult.
Finally, although Lamb did concern himself with the patterning and organization of behaviors insofar as he employed our A-B-C classificatory system, the other two mother vs. father studies ignored this consideration. We agree with Lamb that the crux of such comparison should rest with qualitative differences reflecting differences in the way attachment relationships are organized.
Summary and Discussion
In summary, the findings of 13 different studies using our strange-situation episodes with little or no modification interlock to an impressive extent both with each other and with our findings as presented in Chapters 7 and 8. The evidence strongly suggests that our strange situation, at least when used with 1-year-olds, is a useful instrument for studying individual differences in infant—mother attachment as they relate to: (1) antecedent and possibly causative variables; (2) concurrent behaviors in a variety of different situations; and (3) subsequent development.
A number of studies found that strange-situation classification and/or avoidant and resistant behaviors in the reunion episodes were significantly correlated with other measures of infant behavior and with maternal behavior (as hypothesized); whereas attachment behaviors, such as proximity seeking and contact maintaining, were not. Frequency measures of discrete behaviors—for example, such as looking, touching, smiling, vocalizing, and crying—also tended not to be significantly correlated with other measures. Findings such as these have led us to conclude that strength and/or frequency of attachment behavior are not the crucial variables when relating strange-situation behavior to other variables external to the strange situation. The patterning of attachment behaviors with avoidant and resistant behaviors is of primary importance, however, and this is what is reflected in strange-situation classification. It is our hypothesis that qualitative differences in the attachment of infants to their mothers are significantly related to differences in antecedent experience and to differences in subsequent development. The strange-situation classification, we propose, highlights some of these qualitative differences in attachment, and is therefore a useful variable in research into their antecedents and consequences, together with the measures of avoidance and resistance, which are crucial to the classification.
Those studies that use strange-situation classification as a variable yield findings highly congruent with each other and with ours. Those that use measures of resistant and avoidant behavior in the reunion episodes yield findings that tend to be congruent with those that use classification. The patterning of behaviors that our classificatory system identifies is robust enough to be approximated by different multivariate approaches with different samples, as demonstrated by a comparison of our findings (Chapter 6) with the “multivariate classifier” devised by Connell and Rosenberg.
The findings of the studies reported in this chapter plainly show that 1-year-olds who are identified as securely attached on the basis of their strange-situation behavior have experienced and concurrently experience more harmonious interaction with their mothers than those who are identified as anxiously attached, whether avoidant or resistant. Their mothers are more sensitively responsive to their signals and communications and are more keyed to reciprocity. These findings emerge from studies representing a wide variety of conditions of observing mother—infant interaction. Furthermore Bell’s findings (as well as our own) suggest that there are also comparable differences in infant responsiveness when interacting with the mother, although Rosenberg did not find such differences, perhaps because his observation time was too short.
The findings of these studies also suggest that the behavior of mother and child and the interaction between them differ in the second and third year in the case of children who had been identified as securely or anxiously attached at the end of the first year. These findings suggest that patterns of mother—child interaction established in the first year of life tend to persist.
Relatively little has yet been done to relate strange-situation behavior to the child’s response to figures other than his mother, whether these be other attachment figures, more or less familiar nonattachment figures, or strangers. Main’s and Connell’s findings suggest, however, that securely attached and anxiously attached children behave differently to relatively unfamiliar figures who propose or initiate play.
Three studies reported here compared strange-situation behavior toward mother versus father, and these figures emerged as fairly interchangeable as attachment figures in the strange situation. Further research comparing behavior toward different attachment figures, as well as to different classes of nonattachment figures, would be highly desirable, although we do not consider the strange situation to be an ideal procedure for such comparisons.
It is clear that the environmental context makes a great difference in resulting behavior. One such context that has been considered here is that of the accompanying adult. However, as Brookhart and Hock have shown, the degree of familiarity of the environment (home vs. laboratory) affects behavior in episodes that are also “environmentally” defined. Furthermore, as Willemsen and associates have shown, the degree of “interestingness” of toys provided in the strange situation is a variable that should not be ignored.
Finally—and to return to individual differences—an impressive amount of evidence has accumulated to demonstrate that strange-situation classification and/or associated maternal behavior are related to various aspects and measures of infant cognitive development. Furthermore, both Main and Bell have reported findings that suggest that qualitative differences in attachment are associated with differences in subsequent cognitive development in the second and third years. These positive findings point toward the desirability of further intensive research intended to elucidate the effect of qualitative differences in infant—mother attachment in the development of cognitive processes, or vice versa.
Notes
1 Because significant effects were found for location, and because the effects attributable to rearing were based on observations in both home and laboratory contexts, it occurred to us that the effects attributable to rearing might have been obscured by the effects of location. Brookhart and Hock (personal communication) reported, however, that a separate analysis of variance based solely on the laboratory data also failed to yield significant main effects attributable to rearing conditions, and replicated the findings in regard to interactions.
2 As we go to press, E. Waters, B. Vaughn, and B. Egland (personal communication) offer preliminary findings pertinent to Connell’s in regard to the neonatal status of Group C infants. At the University of Minnesota, 72 of a projected sample of 100 infants from families of very low socioeconomic status have been studied. They were given the Brazelton Infant Scales on the seventh day of life and again three days later and were observed in the strange situation at about 12 months. Of the 72 infants, 12 were classified in Group A, 42 in Group B, and 18—an unusually high proportion—in Group C. Whereas the Day 7 mean Brazelton scores were normal for A and B babies, C babies departed from normal expectations in a number of ways. They were unresponsive to both auditory and moving visual stimuli. They had low muscle tone. When crying, they were difficult to soothe. When excited, their level of excitement was higher than average, and they built up to a peak of excitement faster. They were more irritable and showed more startle. There were, however, no “hard” signs of neurological damage; furthermore, three days later their performance on the Brazelton Scales was no longer significantly different from those of the other two groups. Nevertheless, these preliminary findings suggest some constitutional basis for difficulty in coping with moderate stresses that other infants do not find unduly difficult. Should such an infant experience insensitive mothering of the type given by the Group C mothers of our Sample 1, it is especially easy to understand that he would develop anxious attachment. Indeed, unless the mother were strongly predisposed to respond promptly and appropriately to infant signals, it seems likely that she would find such a baby “difficult” and react to him with less sensitivity than she might have managed with a less vulnerable child.
3 The summary of episodes given by Feldman and Ingham (1975) suggested that their procedures may have differed substantially from ours, but upon examination of their detailed instructions, which they kindly sent us, we are convinced that there were in fact no substantial differences.