Theoretical Background - Introduction

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

Theoretical Background


Attachment theory was given its first preliminary statement in John Bowlby’s 1958 paper entitled “The Nature of a Child’s Tie to His Mother.” It was fully launched by the first volume of his trilogy on Attachment and Loss in 1969, which was followed by a second volume in 1973. The first two reports of research inspired by Bowlby’s early formulation were by Ainsworth (1963, 1964) and Schaffer and Emerson (1964). Since then there has been an increasing volume of research relevant to infant—mother attachment, including research into mother—infant interaction and into early social development. There is no doubt that the further formulation of attachment theory, as represented in Bowlby’s major works (1969, 1973) was influenced by this research. In the meantime other statements of attachment theory have emerged, some of which (e.g., Ainsworth, 1969, 1972; Sroufe & Waters, 1977) dovetail closely with Bowlby’s evolutionary—ethological approach. In contrast, others (e.g., Cairns, 1972; Gewirtz, 1972a, 1972b; Maccoby & Masters, 1970) have attempted to assimilate attachment theory to other earlier paradigms.

Attachment Theory as a New Paradigm

Bowlby’s attachment theory stemmed from a convergence of several important trends in the biological and social sciences. An initial psychoanalytic orientation was integrated with the biological discipline of ethology and its insistence on viewing behavior in an evolutionary context; with psychobiology and its focus on neurophysiological, endocrine, and receptor processes that interact with environmental stimuli to activate and terminate the activity of behavioral systems; with control-systems theory, which directs attention to “inner programming” and links behavioral theory to an information-processing model of cognition; and with Piaget’s structural approach to the development of cognition. Although this integration was undertaken primarily to understand the origin, function, and development of an infant’s early social relations, that part of Bowlby’s theory that deals specifically with attachment is embedded in a general theory of behavior that owes much to its several origins.

Attachment theory might be described as “programatic” and openended. It does not purport to be a tight network of propositions on the basis of which hypotheses may be formulated, any one of which, in the event of an adequate but unsuccessful test, could invalidate the theory as a whole. Instead, this is an explanatory theory—a guide to understanding data already at our disposal and a guide to further research. “Validation” is a matter of collecting evidence relevant to “construct validity” (Cronbach & Meehl, 1955), with the implication that the “construct” itself can be elaborated and refined through further research, rather than standing or falling on the basis of one crucial experiment.

Despite its lack of resemblance to a mathematicophysical theory, both the general theory of behavior and attachment theory amount to what Kuhn (1962) termed a paradigm change for developmental psychology—a complete shift of perspective. According to Kuhn, such paradigm changes are at the root of scientific revolutions and account for the major advances in science, even though much constructive endeavor must follow the advancement of a new paradigm before it is fleshed out fully.

Kuhn emphasized the difficulty encountered by adherents of earlier paradigms in assimilating the implications of the new paradigm. Such difficulty is unavoidable, for a new paradigm comes into being in an attempt to account for findings that older paradigms could not deal with adequately. For Bowlby the inexplicable findings pertained to a young child’s responses to separation from his mother figure. Although a new paradigm may build on older ones and must also account for the empirical findings that they dealt with adequately, the new paradigm cannot be assimilated to an old paradigm—not without such substantial accommodation that the old paradigm is changed beyond recognition and itself becomes a new paradigm more or less akin to the other new one that could not readily be assimilated. We hold that Bowlby’s attachment theory constitutes a new paradigm for research into social development. It is in terms of this paradigm that we interpret our findings—and indeed we view our findings as helping to flesh out the framework of the new paradigm.

Although in Chapters 9, 10, and 14 we also discuss some researches that stemmed from divergent paradigms, we are cognizant of Kuhn’s warning that it is difficult to move from one paradigm to another. Ainsworth (1969) attempted an elucidation of the differences between three major paradigms relevant to an infant’s relationship with his mother; we shall not repeat this endeavor here. The attachment theory that we shall summarize in this chapter is based on Bowlby’s paradigm, with particular emphasis on those aspects that are most relevant to the research with which this volume is concerned.

The Behavioral System

One of the major features of Bowlby’s general theory of behavior is the concept of a behavioral system. To ethologists this “construct” is so fundamental that it scarcely requires explanation. (Nevertheless, see Baerends, 1975, for a detailed discussion of behavioral systems.) Bowlby holds that the human species is equipped with a number of behavioral systems that are species characteristic and that have evolved because their usual consequences have contributed substantially to species survival. Some of these systems are toward the labile end of an environmentally labile vs. environmentally stable continuum. An “environmentally stable” system manifests itself in much the same ways throughout almost all members of the species (or almost all members of one sex) despite wide variations in the environments in which the various populations that compose the species have been reared and in which they now live. The manifestations of a relatively “labile” system vary considerably across the various populations in the species in accordance with environmental variations.

For those who are not conversant with evolutionary theory, it is perhaps useful to explain that “survival,” in terms of natural selection means species survival or at least population survival. It implies survival of the individual only to the extent that he or she survives to produce viable offspring and to rear them successfully. Natural selection implies that the genes of the most reproductively successful individuals come to be represented in larger proportion in the “gene pool” than the genes of individuals who do not survive long enough to reproduce, who survive but do not produce as many offspring, whose offspring do not survive to sexual maturity, or whose offspring do not reproduce, and so on. Given the natural-selection process, it is scarcely surprising that among the most environmentally stable behavioral systems characteristic of many species (including the human species) are those concerned with reproduction and with care and protection of the young.

It is generally acknowledged that the relatively long period of infantile helplessness characteristic of humans, together with a relative lack of fixed-action patterns, provides the necessary conditions for flexibility and learning—for adaptation to a very wide range of environmental variation. Nevertheless a long period of immaturity implies a long period of vulnerability during which the child must somehow be protected. Bowlby argues, therefore, that human young must be equipped with a relatively stable behavioral system that operates to promote sufficient proximity to the principal caregiver—the mother figure—that parental protection is facilitated. This system—attachment behavior—supplements a complementary behavioral system in the adult—maternal behavior—that has the same function.

Attachment behavior conceived as a behavioral system is not to be equated with any specific bit of behavior. First, the external, observable behavioral components are not the only components of the system; there are intraorganismic, organizational components as well. These are discussed later. Second, there may be a variety of behaviors that serve the system as action components, and indeed a specific behavioral component may, in the course of development, come to serve more than one behavioral system. Nevertheless several behaviors may be classed together as serving a given behavioral system because they usually have a common outcome. The behaviors thus classed together may be diverse in form. They may be classed together because each is an essential component of a series of behaviors that lead to the outcome, such as nest building among birds, or they may constitute alternative modes of arriving at the outcome, as in the case of attachment behavior. Bowlby refers to the outcome as “predictable,” to imply that once the system is activated the outcome in question often occurs, although not invariably. If the outcome did not occur consistently enough and in enough individuals, however, the survival of the species would be at risk.

Predictable Outcome

The predictable outcome of a child’s attachment behavior is to bring him into closer proximity with other people, and particularly with that specific individual who is primarily responsible for his care. Bowlby refers to this individual as the “mother figure,” and indeed in the human species, as well as in other species, this individual is usually the biological mother. The mother figure is, however, the principal caregiver, whether the natural mother or someone else who plays that role. Some behavioral components of the attachment system are signaling behaviors—such as crying, calling, or smiling—that serve to attract a caregiver to approach the child or to remain in proximity once closeness has been achieved. Other components are more active; thus, once locomotion has been acquired, the child is able to seek proximity to his attachment figure(s) on his own account.

Causation of Activation and Termination of Behavior

Several sets of conditions play a part in the activation of a given behavioral system, both specific and general, and within both the organism and the environment. Bowlby notes that the most specific causal factors are the way in which the behavioral systems are organized within the central nervous system and the presence or absence of certain objects within the environment. From the study of other species, we also know that hormones may have a fairly specific influence on behavior, although our knowledge of hormonal influences on human attachment behavior or reciprocal maternal behavior is sparse indeed. Among the more general factors that play a part in the causation of behavior are the current state of activity of the central nervous system—its state of “arousal”—and the total stimulation impinging on the organism at the time. These five classes of causation act together; no one of them may be sufficient to set a behavioral system into action unless one or more of the other factors are also favorable.

Among the various environmental conditions that may activate attachment behavior in a young child who has already become attached to a specific figure are absence of or distance from that figure, the figure’s departing or returning after an absence, rebuff by or lack of responsiveness of that figure or of others, and alarming events of all kinds, including unfamiliar situations and strangers. Among the various internal conditions are illness, hunger, pain, cold, and the like. In addition, whether in early infancy or in later years, it seems apparent that attachment behavior may be activated, sustained, or intensified by other less intense conditions that are as yet not well understood. Thus, for example, an infant when picked up may mold his body to the person who holds him, thus manifesting proximity/contact-maintaining behavior, even though his attachment behavior may not have been activated at any substantial level of intensity before being picked up. Or a somewhat older infant or young child may respond with attachment behavior to a figure—particularly a familiar one—who solicits his response and interaction. Indeed he may seek to initiate such interaction himself, and if the figure is a familiar caregiver or (later) an attachment figure, one could argue that the behaviors involved in the initiation and in the subsequent interaction operate in the service of the attachment system. As for the most specific intraorganismic factor—the organization of behavioral systems within the central nervous system—we shall only say at this juncture that whatever constitutional organization is present at birth becomes substantially modified and elaborated through experience, and that individual differences in experience may be presumed to result in different patterns of organization. Thus, although one may generalize to some extent about the conditions likely to activate attachment behavior, the factor of internal organization is highly specific to the individual and, in addition, specific to his particular stage of development.

The conditions for termination of a behavioral system are conceived by Bowlby as being as complex as the conditions of activation, and as related both to the intensity with which the system had been activated and to the particular behavioral component of the system that was involved. Thus the most effective terminating condition for infant crying is close bodily contact contingent upon being picked up by the mother figure (Bell & Ainsworth, 1972), whereas simple approach behavior in a 1-year-old may be terminated by achieving a degree of proximity without requiring close bodily contact. On the other hand, if the attachment system has been activated at a high level of intensity, close contact may be required for the termination of attachment behavior.

A note on terminology may be helpful to the reader. Bowlby (1969) uses the term “attachment behavior” to refer to both the behavioral system and to the behavioral components thereof—a usage that may occasion confusion among readers unaccustomed to the concept of behavioral systems. We have attempted to use the plural term, “attachment behaviors” to refer to the action components that serve the behavioral system, while reserving the singular term “attachment behavior” or the somewhat clumsy term “attachment behavioral system” to refer to the system.

Biological Function

The biological function of a behavioral system is to be distinguished from the causes of the behavioral system’s having been activated. It is an outcome of the behavioral system’s having been activated, but whereas there may be more than one predictable outcome, the biological function of the system is defined as that predictable outcome that afforded a certain survival advantage in the “environment of evolutionary adaptedness”—the original environment in which the system first emerged as a more or less environmentally stable system, and to which it may be said to be adapted in the evolutionary sense. Biological programing continues to bias members of the species to behave in the ways that gave survival advantage in this original environment. The biological function of the behavioral system may or may not give special survival advantage in one or another of the various environments in which populations now live, but unless changes in the average expectable environment render the behavioral system a liability, it will be maintained in the repertoire of the species.

Bowlby (1969) proposed that the biological function of the attachment system is protection, and he suggested that it was most specifically protection from predators in the environment of evolutionary adaptedness. Indeed, field studies of other species suggest that infants who get out of proximity to their mothers are very likely to become victims of predation. He argued, however, that even in the present-day environment of Western society a child is much more vulnerable to disaster (for example, to becoming a victim of a traffic accident) if alone rather than accompanied by a responsible adult (Bowlby, 1973). Indeed, he noted that even adults of any society tend to be less vulnerable to mishap if with a companion than when alone. Therefore, he felt comfortable about specifying protection as continuing to be the biological function of attachment behavior and its reciprocal parental behavior.

The implication is that the reciprocal behaviors of child and parent (Hinde, 1976a, 1976b, would term these “complementary” behaviors) are adapted to each other in an evolutionary sense. Thus, a child’s attachment behavior is adapted to an environment containing a figure—the mother figure—who is both accessible to him and responsive to his behavioral cues. To the extent that the environment of rearing approximates the environment to which an infant’s behaviors are phylogenetically adapted, his social development will follow a normal course. To the extent that the environment of rearing departs from the environment to which his behaviors are adapted, developmental anomalies may occur. Thus, for example, an infant reared for a long period, from early infancy onward, in an institutional environment in which he has so little consistent interaction with any one potential attachment figure that he fails to form an attachment may, when subsequently fostered and thus given an opportunity to attach himself, be unable to attach himself to anyone (e.g., Goldfarb, 1943; Provence & Lipton, 1962.)

The foregoing example raises an important point for attachment theory—namely, that just as an infant is predisposed to exhibit attachment behavior under appropriate circumstances, he is predisposed to form an attachment to a specific figure or figures. The predictable outcome of both the activation of the attachment behavioral system and attachment as a bond is the maintenance of a degree of proximity to the attachment figure(s); and similarly, in each case, the biological function is protection. We discuss attachment as bond and its relation to attachment behavior later in this chapter. Here we merely wish to point out that it is under very unusual circumstances that an infant or young child encounters conditions such that his attachment behavior does not result in the formation of an attachment. Although, as noted above, institution-reared infants may not become attached to anyone, most family-reared infants do become attached, even to unresponsive or punitive mother figures.

Goal-Corrected Behavior

Species-characteristic behavior systems may consist of fixed-action patterns that operate more or less independently of environmental feedback or that may at least have some fixed-action components in the system. Bowlby’s general theory of behavior specified, however, that species-characteristic behaviors may also be flexible and goal directed. Here he draws upon control-systems theory. A control system is a machine that may be described as operating purposively. The “goal” is built into the device by the men who program it, or “set” it. Feedback is the essential mechanism through which the machine achieves its goal. There is a mechanism for receiving “input” and one for effecting “output.” The results of the output are fed back through the receptor mechanism to affect further output in accordance with the way the device is programed.

The simplest kind of control system is a regulator—for example, a thermostat. The purpose is to maintain the temperature of a room at a level at which the thermostat is set—the specific “set-goal” of the device. (One may change the set-goal by changing the thermostat to another level.) When the receptor mechanism receives information that the room temperature has dropped below the level of the set-goal, it turns on the heating system through its effector mechanism; when information is received that the temperature has reached (or slightly surpassed) the set-goal level, it turns the heating system off. Many of the physiological systems operate homeostatically in essentially the same way as a regulator.

A more complex kind of control system is a servomechanism, such as power steering. In such a system the “setting” is continually changed by the human operator, and the system acts to bring performance into accord with the setting at each change. Another example is the action of the antimissile missile. Here the instructions are built into the machine in the course of its manufacture; its set-goal is the interception of another missile. Its effector system alters the speed and direction of its movement in accordance with feedback from its receptor mechanism, which monitors not only the distance and direction of the other missile but also the way in which the discrepancy between their relative positions changes as a result of their movements relative to each other. The set-goal and action of the missile is like that of the peregrine falcon that “stoops” to intercept another bird in flight. The only substantial difference between the falcon and the antimissile missile is that the missile’s program was built into it by its manufacturers, whereas the falcon’s biological program results from natural selection. In the case of the falcon this programing provides the equipment that enables continuously changing visual input to guide the movements that control the course and speed of flight, so that the predictable outcome is the achievement of the set-goal—the interception of prey.

“Goal corrected” is the term that Bowlby (1969) suggests as preferable to “goal directed” to describe behavioral systems that are structured in terms of set-goals. He suggests that complex behavioral systems of this sort are characteristic of the human species—systems that may be described as purposive and flexible and yet that have a basis of biological programing. The attachment system provides an interesting example, because it has both the features of a simple regulator and the flexibility of a much more complex control system. The setting of the set-goal—that is, the degree of proximity to an attachment figure specified by the set-goal—differs from time to time depending on circumstances. When the set-goal is set widely, a child may venture a substantial distance from his mother before the set-goal is exceeded, attachment behavior is activated, and the specified degree of proximity restored.

As suggested earlier, however, a variety of different conditions may activate attachment behavior, in addition to exceeding the distance (and time) away from the attachment figure that was specified by the “original” setting of the set-goal. Depending on the intensity with which such conditions may activate the attachment system, the set-goal may abruptly change its setting to specify the required degree of proximity more narrowly. Indeed, when the attachment system is activated to a high degree of intensity, the set-goal may be close bodily contact, and attachment behavior will not be terminated until this new set-goal has been achieved. Furthermore, there is substantial flexibility in the attachment behaviors that may be used for the achievement of the set-goal. The model of the simple regulator is approximated only when the attachment figure is stationary and inactive. The ways in which the attachment figure behaves influence the ways in which the child’s repertoire of attachment behaviors is deployed to achieve the current set-goal. Finally, although the “behavioral homeostasis” associated with the simple regulator model has general descriptive value, the attachment behavioral system is organized along much more complex lines. Overemphasis on the simple model has led many to assume that Bowlby’s attachment theory defines attachment behavior rigidly and exclusively in terms of seeking literal proximity—a conception that is inadequate even when describing the attachment and attachment behavior of a 1-year-old and that is clearly misleading when attempting to comprehend the behavior of the older child or adult.

Clearly Bowlby conceives of some very complex adult behavior stemming from species-characteristic behavioral systems. An example of this is parental behavior. In this case, however, there seems to be so much flexibility attributable to feedback from environmental conditions that the program followed by the system can only be perceived by stepping back from the details of behaviors in a given situation to look at the consistent pattern of behavior toward a common set-goal that is apparent across a variety of geographical and cultural environments.

Organization of Behavior

The behaviors classed together as serving a given behavioral system may be organized in different ways. The simplest mode of organization is chaining, in which the “output” of each link in the chain provides input to activate the next behavioral link—a mode of organization familiar to us through S—R psychology. Another more complex mode of organization, deemed by Bowlby to be more characteristic of most human behavior, is a hierarchical form of organization. One form of hierarchical organization is governed by a plan (Miller, Galanter, & Pribram, 1960.) In a plan, as Bowlby describes it, the overall structure of the behavior is governed by a set-goal, whereas the individual behavioral components for achieving the set-goal vary according to circumstances.

In the neonate the separate behaviors that may be classed together as attachment behavior because they promote proximity/contact with caregivers form a behavioral system whose components have minimal organization. Each behavioral component—for example, crying, sucking, smiling—has its distinctive conditions for activation and termination; and indeed, as Bowlby suggested, each might be viewed as a fixed-action pattern. About the middle of the first year of life, however, attachment behavior begins to become goal corrected and to be organized in accordance with plans although these may at first be very primitive.

As an example of a primitive plan of this sort, let us consider the case of the infant, engaged in exploratory play at some distance from his mother, who notices her get up and move away. Her movement may or may not have exceeded the limits of the proximity set-goal operative at the time, but the very fact that she takes the initiative in increasing the distance between them may arouse anxiety about her continuing accessibility, may narrow the limits of the set-goal, and may activate attachment behavior at a higher level of intensity. In such a case the baby may follow his mother with more urgency, seeking to establish closer proximity with her than before; he may signal to her by crying or calling, which may induce her to stop and wait for him or reverse direction and approach him; or he may do both. Even though this situation may evoke behavior no more complex than this, the baby may be viewed as having a primitive plan—namely, to get into closer proximity to his mother, and as having alternative behaviors available to him in terms of which he can implement his plan, choosing the one that best seems to suit his evaluation of the situation. Thus even a very simple plan has a set-goal and a choice of alternative behaviors, or perhaps a sequence of behaviors in terms of which the plan may be implemented and the set-goal achieved.

The Role of Cognitive Processes and Learning

It is clear that the organization of behavior in accordance with a plan involves cognitive processes and that these are far beyond the ability of the neonate. Only after considerable cognitive development has taken place does an infant become capable of plans. Although attachment theory cannot be identified as primarily a cognitive theory, Bowlby clearly conceives of the development of attachment as intertwined with cognitive development. Later in this chapter we mention some of the cognitive acquisitions that precede or coincide with important shifts in the course of the development of attachment. Here, however, we wish to make special mention of Bowlby’s (1969) concepts of “working models” and “cognitive maps,” which consist of inner representations of the attachment figure(s), the self, and the environment. Although it is obvious that such representational models become increasingly complex with experience, it is clearly necessary that some kind of simple representations of this sort be constructed before there may be hierarchical organization of behavior according to plans.

It is inconceivable that the way in which behavior systems characteristic of the human species operate would not be changed to a degree commensurate with the elaboration of representational models, and also with the further development of communication, especially the acquisition of language. Bowlby plainly indicates that this must be the case with the attachment system. Critics of attachment theory do not seem to have grasped the implications of either goal-corrected attachment behavior or hierarchical organization according to plans; on the contrary they seem to have paid attention only to the simple regulatory or homeostatic model, which Bowlby did discuss in detail in conjunction with presenting the concept of set-goal. Under certain circumstances and within a certain early age range, this model does indeed capture the main features of the regulation of attachment behavior. Bowlby would agree with his critics that literal proximity specified in feet and yards is a very inadequate way of delineating the set-goal of the attachment system in the case of the older child or adult. Even for an infant this model yields an oversimplified picture.

Bowlby (1973) emphasizes the importance of the infant’s confidence in his mother’s accessibility and responsiveness. If in the course of his experience in interaction with his mother he has built up expectations that she is generally accessible to him and responsive to his signals and communications, this provides an important “modifier” to his proximity set-goal under ordinary circumstances. If his experience has led him to distrust her accessibility or responsiveness, his set-goal for proximity may well be set more narrowly. In either case, circumstances—her behavior or the situation in general—may make her seem less accessible or responsive than usual, with effects on the literal distance implicit in a proximity set-goal. (Carr, Dabbs, and Carr, 1975, have demonstrated this point by comparing the effects of the mother’s facing or facing away from the child.) Simple expectations regarding the mother’s accessibility and responsiveness, as they differ with circumstance, are incorporated into the representational model a child constructs of his mother figure.

As the representational model of his attachment figure becomes consolidated and elaborated in the course of experience, the child becomes able to sustain his relationship with that figure over increasingly longer periods of absence and without significant distress—provided that the separations are agreed to willingly and the reasons for them understood. Under such circumstances the older child or adult may employ distant modes of interaction to reaffirm the accessibility and responsiveness of the attachment figure. Telephone calls, letters, or tapes may help to ameliorate absence; photographs and keepsakes help to bolster the symbolic representation of the absent figure. (Robertson and Robertson, 1971, reported deliberate use of such symbolic modes in supporting the ability to withstand separation of children even in the second and third years of life.) Our language usage offers testimony that proximity/contact is often conceived at the representational level. We talk about “feeling close” to some one, “keeping in touch,” and “keeping in contact.”

Nevertheless, inner representations cannot entirely supplant literal proximity and contact, nor can they provide more than minimal comfort in the case of inexplicable and/or permanent loss of an attachment figure—neither for a young child nor for a mature adult. When people are attached to another, they want to be with their loved one. They may be content for a while to be apart in the pursuit of other interests and activities, but the attachment is not worthy of the name if they do not want to spend a substantial amount of time with their attachment figures—that is to say, in proximity and interaction with them. Indeed, even an older child or adult will sometimes want to be in close bodily contact with a loved one, and certainly this will be the case when attachment behavior is intensely activated—say, by disaster, intense anxiety, or severe illness.

Interplay Among Behavioral Systems

Let us return to a consideration of attachment behavior as one of several behavioral systems that may be activated at a greater or lesser degree of intensity in any given situation. What happens when two or more systems are activated simultaneously? If one is very much more intensely activated than the others, that system determines the resulting overt behavior, and neither the observer nor the “behaver” may discern any conflict. If two systems are activated at more nearly equal levels of intensity, the more strongly activated may nevertheless determine the behavioral outcome, and the less intensely activated system may be represented only in terms of behavioral fragments, or perhaps identified in terms of the behavior that swings into action when the dominant system is terminated. An example is the behavior of a bird at a window feeding tray when a person comes to the window to observe the bird. In such a situation there is likely to be conflict for the bird between tendencies to feed and to flee. If feeding behavior is activated more strongly than flight, the bird will remain, but it may well manifest its conflict by interspersing feeding behavior with incipient “take-off” movements, which ethologists term “intention movements.” These movements are overt manifestations of the activation of the flight system, even though the bird continues to feed intermittently without actually flying away. If, on the other hand, the flight system is activated more strongly than the feeding system, the bird will fly away, but the fact that the feeding system is still at a significant level of activation will be shown if, as often happens, he soon returns to the feeding tray. And if the human observer is tactful enough to withdraw somewhat, it is likely that the flight system’s level of activation will be reduced to the extent that the level of activation of feeding behavior becomes relatively stronger and the bird will remain to feed. This kind of conflict with similar behavioral solutions may be seen in the responses of 1-year-olds to the stranger in the strange situation, and is reported and discussed in later chapters.

When the two competing behavioral systems are more nearly equal in level of activation, it is likely that both will be represented in overt behavior in one way or another. One way in which both might be represented is in alternate behaviors. Thus the bird, in our previous example, might alternate between flights away from the feeding tray and returns to peck a few grains before flying away again. Or in our strange situation, a 1-year-old child, conflicted between friendly approach to a pleasant but unfamiliar adult and a tendency to avoid her because she is unfamiliar, may approach the stranger but then immediately withdraw (usually returning to the mother), only to pause for a moment and then approach the stranger again, perhaps repeating this sequence several times.

Another way in which both competing systems may express themselves in overt behavior is in some kind of combination. Coy behavior represents such a combination. A person—child or adult—both attracted to another person and wary of him/her, may simultaneously smile and look away, the smile serving an affiliative or sociable system and the look away serving a wary/fearful system. Sometimes a behavior, not activated intensely enough to override another behavioral system that blocks its expression, may be redirected toward a goal object other than toward the one that elicited it. Thus a person whose aggressive/angry behavior is activated by the actions of another of whom he is also afraid or fears to offend may “redirect” aggressive behavior toward a third person or toward an inanimate object—an outcome referred to by psychoanalysts as “displacement.”

Even when there is no substantial degree of conflict between systems—that is, when one system is activated so strongly as to clearly override another—our understanding of the organization of behavior is greatly enhanced if we view the operation of one behavior system in the context of other systems. Thus, to comprehend how 1-year-olds manifest attachment behavior in the strange situation, we must trace through, episode by episode, the interplay among attachment behavior, wary/fearful behavior, exploratory behavior, and in some episodes, sociable (or affiliative) behavior directed toward the stranger. The training that most of us have received does not make it easy to conceptualize the interplay of as many as four complex systems, let alone to take into account the complex conditions that determine the level of activation of each of them. Bischof (1975) provides a control-systems model that illustrates the interplay among the four systems that are of most concern to us in strange-situation research. Bischof would be the first to agree that even his complex model represents an oversimplification of the complexities of real-life behavior. Nevertheless, we believe it to be a fine contribution toward an understanding of how intraorganismic and environmental conditions operate to determine which of four behavior systems will be activated most intensely and thus will control behavioral output. The model is not complex enough, however, to handle the manifestations of conflict behavior described earlier in this section.

Behaviors May Serve More Than One System

In each species there may be a few specific behaviors that are unique to one and only one behavioral system. Examples of this are difficult to find in the human species. Looking, for example, may serve a wide variety of behavioral systems, perhaps from earliest infancy onward. One looks at a novel object, and this serves the exploratory system. One seeks eye contact with an attachment figure, or at least monitors his/her whereabouts with an occasional glance. One glares at an antagonist toward whom one feels animosity. One may give a good, long look at a novel object, person, or situation that arouses wariness/fear, before either putting it at a distance, or “cutting off” the stimulus by looking away, or deciding that the object is more interesting than frightening and approaching it. Approach behavior itself may serve more than one system, as Tracy, Lamb, and Ainsworth (1976) have argued. Locomotor approach can serve the attachment system when the individual seeks proximity to an attachment figure. It can also serve exploration, food seeking, affiliation with figures other than an attachment figure, play, anger/aggression, and probably other systems as well. Furthermore, behaviors that in an early stage of development were especially linked with one behavioral system may at later stages occur, if only in fragmentary form, to serve either the same system or other systems. Thus, for example, behaviors displayed by an infant toward his mother may occur also in the adult as part of courtship/mating behavior. Thus in some species of birds, begging for food may be an integral feature of courtship—and human equivalents are not difficult to identify.

Bowlby (1969), in his chapter on “Beginnings,” enumerated various forms of behavior that “mediate attachment”—that is to say, specific behaviors that promote proximity, contact, and interaction with other persons and thus play a significant role in the development of attachment to one or a few such persons. We may identify these as “attachment behaviors,” because they clearly serve the attachment-behavior system, or as “precursor attachment behaviors” as Ainsworth (1972) did, because they are part of the equipment of the neonate and/or very young infant before he has become attached to anyone. There is nothing in attachment theory to imply that these behaviors serve the attachment system exclusively, even in early infancy. In his next chapter Bowlby listed a number of behaviors suggested by Ainsworth (1967) to be differentially displayed by an infant during his first year toward a particular figure toward whom he is, or is becoming, attached. Bowlby implied that these were useful indicators of the process of focusing on a specific figure. Some of them may also prove useful as criteria for describing an infant as having become attached to a particular figure. It was not intended by Bowlby and Ainsworth to imply: (1) that behaviors displayed differentially during an early phase of development necessarily continue throughout childhood and into adulthood to be displayed differentially to attachment figures; (2) that this list constitutes an adequate roster of behaviors that serve the attachment system during the second year of life and later; or (3) that these behaviors serve the attachment system exclusively. Indeed, as the organization of the attachment system becomes elaborated in the course of development, and as more and more forms of behavior become employed as alternative means of implementing the plans pertinent to interaction with attachment figures, it seems less and less useful to attempt an enumeration of attachment behaviors. Increasingly, the organization and patterning of behaviors become the focus of interest.

Attachment and Attachment Behaviors

Here we are concerned with the distinction between attachment as a bond, tie, or enduring relationship between a young child and his mother and attachment behaviors through which such a bond first becomes formed and that later serve to mediate the relationship. In developing attachment theory, Bowlby (1969) devoted much attention to attachment behavior as a behavior system, in the course of which he also discussed the specific behaviors that serve that system in infancy and early childhood. He devoted relatively little attention to an exposition of the relation between such behaviors and attachment as a bond. Indeed, we can assume that he considered it self-evident that the way in which the attachment-behavioral system became internally organized in relationship to a specific figure itself constituted the bond or attachment to that figure. Some readers, however, working within the framework of other paradigms, failed to grasp the organizational implications of the concept of a behavioral system, and concluded that attachment and overt attachment behavior were identical. Such a conclusion led to a variety of theoretical misconceptions: for example, that attachment has disappeared if attachment behavior, including separation distress, is no longer overtly manifested; that the intensity with which a child shows attachment behavior in a given situation may be taken as an index of the strength of his attachment; or that attachment consists in nothing more than the contingencies of the interaction between a child and his mother.

We have attempted to deal with the distinction between attachment and attachment behavior elsewhere (e.g., Ainsworth, 1969, 1972), and we return to this issue later in this volume, after presenting our findings. Here, however, we should like to remind the reader that Bowlby’s attachment theory came about through his efforts to account for the response of a young child to a major separation from his mother and to reunion with her afterwards (Bowlby, 1969, preface). Therefore, it seems appropriate here to review a few of the phenomena that it make it necessary to assume the existence of a bond between a child and his mother that, once formed, continues despite separation, independent of either overt manifestations of attachment behavior or the contingencies implicit in ongoing mother-child interaction. First, it is necessary to distinguish between brief separations of minutes (or even hours) that take place in the familiar home environment and about which a child will have formed a system of expectations and an involuntary separation lasting for days, weeks, or months, during which a child may be cared for by unfamiliar persons in an unfamiliar environment. It is the latter that we have termed “major” or “definitive” separations, to distinguish them from brief “everyday” separations in a familiar environment.

A child’s initial response to a major separation—either at the moment of parting or later when his expectations of a prompt reunion are violated—is to greatly intensify attachment behavior, protesting the separation and trying by all means at his disposal to regain proximity/contact with his attachment figure. This protest is usually more than momentary, but how long it lasts and how intense it is depend on a variety of circumstances. As separation continues, however, the child’s attachment behavior becomes either muted or more intermittently manifested, and eventually it may drop out altogether. If one were guided entirely by his overt behavior, one would say that he is no longer attached; but that the bond endures, despite absence of attachment behavior directed toward the absent figure, is vividly demonstrated in most children when reunited with the attachment figure. Whether with or without some delay, attachment behavior is activated at a high level of intensity—much higher than that characteristic of the child before separation. Were attachment identical with attachment behavior, one would be forced to conclude that separation first strengthens the bond, then weakens it, and finally destroys it. If one holds that the bond has altogether disappeared, it then becomes impossible to account for the fact that it reconstitutes itself so quickly after reunion. It seems to us more reasonable to view the bond as enduring despite the vicissitudes of attachment behavior.

If during separation from his mother a child is fortunate enough to be cared for by a substitute figure who plays a thorough maternal role, separation distress may be greatly alleviated, and the child may come to direct attachment behavior toward the substitute figure. Nevertheless such sensitive foster care does not diminish a child’s attachment to his own mother figure; on the contrary it facilitates rather than hampers the prompt reestablishment of normal relations with her upon reunion (Robertson & Robertson, 1971).

To be sure, there may be some delay in the reemergence of attachment behavior after a long period of separation, especially if separation was experienced in a depriving environment without adequate substitute mothering—and this delay is associated with the length and extent of disappearance of overt attachment behavior during the separation itself. Upon reunion the child may seem not to recognize his mother, or he may reject her advances, or he may seem merely to be uninterested in proximity to or contact with her. It is noteworthy that such behavior is not displayed to the father or to other familiar figures. Robertson and Bowlby (1952) identified such a response as “detachment” and attributed it to repression. The implication was that the bond—the attachment—had not disappeared but was still somehow internally represented, even though attachment behavior was absent. In support of the view that attachment as bond had not been lost are the many observations of children whose “detachment” suddenly gives way to intense attachment behavior—following the mother wherever she goes, showing distress when she is out of sight for a moment, and wanting close bodily contact much more frequently and intensely than was characteristic of them in the preseparation period. Given the sudden and dramatic shift between detached behavior and very intense attachment behavior, it is difficult to attribute the change to a process of relearning.

Whereas responses to separation and reunion especially highlight the distinction between attachment and attachment behavior, there are other more ordinary sources of evidence. The presence or absence of overt attachment behavior and the intensity with which it is manifested clearly depend on situational factors. For example, a child is more likely to manifest attachment behavior when he is hungry, tired, or ill than when he is fresh, fed, and in good health. It is difficult to conceive that his bond to his mother varies in strength from day to day or from moment to moment, even though the intensity of activation of attachment behavior so varies.

Emotion and Affect in Attachment Theory

In his general control-systems theory of behavior, Bowlby (1969) identified affect and emotion as “appraising processes.” Sensory input, whether conveying information about the state of the organism or about conditions in the environment, must be appraised or interpreted in order to be useful. Feelings (i.e., both affect and emotion) serve as appraising processes although not all appraising processes are felt (i.e., conscious). In the course of appraisal, input is compared to internal “set-points,” and certain behaviors are selected in preference to others as a consequence of this comparison. In this sense, feelings—whether “positive” or “negative,” pleasant or unpleasant—are focal in the control of behavior.

It was not until his 1973 volume, however, that Bowlby expanded on the role of feelings, giving particular attention to security, fear, anxiety, and anger. Let us briefly consider some important features of his argument. In the course of evolution each species develops a bias to respond with fear to certain “natural clues to an increased risk of danger.” It is of survival advantage for the individual to respond with avoidance, flight, or some other comparable form of behavior to situations that signal an increased risk of danger, without having had to learn through experience how to assess such risk. Among such natural clues to danger for the human species, he listed strangeness (unfamiliarity), sudden change of stimulation, rapid approach, height, and being alone. He particularly emphasized the tendency to respond especially strongly to compound situations in which two or more natural clues are simultaneously present. Although other clues to danger may be learned as derivatives of natural clues, through observation of the behavior of others or in more sophisticated risk-assessing processes, and although through experience a person’s fear may be reduced when natural clues to danger occur in now-familiar situations in which no risk has been encountered, these natural clues to danger nevertheless tend to continue to be appraised in terms of fear. Even a sophisticated adult is likely to experience fear in a compound situation, such as being alone in an unfamiliar environment in which illumination is suddenly reduced and strange noises are heard.

Fear behavior and attachment behavior are often activated at the same time by the same set of circumstances. When a young child is alarmed by one of the clues to increased risk of danger, whether natural or learned, he tends to seek increased proximity to an attachment figure. Should the attachment figure be inaccessible to him, either through absence or through an expectation of unresponsiveness built up through experience, he faces an especially frightening compound situation. Both components of such a situation are frightening, and the term fear may be applied to the appraisal of both. Bowlby presents a military analogy. The safety of an army in the field depends both on its defense against attack and on maintaining a line of communications with its base. Should the field commander judge that retreat is the best tactic, it is essential that the base be available to him, that he not be cut off from it, and that the commander in charge of the base be trusted to maintain the base and the support implicit in it. By analogy, the young child may be afraid of the threat implicit in the clues to danger he perceives in a situation, but he may also be afraid if he doubts the accessibility of his “base”—his attachment figure. Bowlby suggests that “alarm” be used for the former class of fear and “anxiety” for latter. This brings us squarely face to face with the issue of separation anxiety.

Bowlby emphasizes how crucial it is in a potentially fear-arousing situation to be with a trusted companion, for with such a companion fear of all kinds of situation diminishes, whereas when alone fear is magnified. Attachment figures are one’s most trusted companions. We all fear separation from attachment figures, but “separation” cannot be defined simply as a matter of absence of such a figure. What is crucial is the availability of the figure. It is when a figure is perceived as having become inaccessible and unresponsive, that separation distress (grief) occurs, and the anticipation of the possible occurrence of such a situation arouses anxiety.

Whereas a young infant is more likely to cry when he is alone than when he is in proximity or contact with his mother and his crying is most likely to be terminated promptly if his mother picks him up (Bell & Ainsworth, 1972), an older infant is likely to begin to form expectations and to experience anxiety relevant to his mother’s departure and/or absence. Thus, at some time in the second half of his first year, he begins to experience anxiety when his mother leaves the room, and may manifest this by crying or, after locomotion develops, by attempting to follow her.

Infants differ, however, in the consistency with which they exhibit distress in brief, everyday separations. It seems to us reasonable to suppose that there are concomitant differences in expectations. An infant who has experienced his mother as fairly consistently accessible to him and as responsive to his signals and communications may well expect her to continue to be an accessible and responsive person despite the fact that she has departed; and if she is absent for but a short time, his expectations are not violated. (This, of course, presupposes that the infant in question has developed a concept of his mother as a “permanent object” as Piaget (1937) used the term, but also that he has developed a “working model” of his mother as available to him in Bowlby’s sense of these terms.) On the other hand, an infant whose experience in interaction with his mother has not given him reason to expect her to be accessible to him when out of sight or responsive to his signals is more likely to experience anxiety even in little everyday separation, as Stayton and Ainsworth (1973) have shown. Such an infant may be identified as anxiously attached to his mother, and Bowlby (1973) elaborates the theme of anxious attachment, both in terms of the kinds of experience that may contribute to it, not only in infancy but also in later years, and in terms of the ways in which anxious attachment may affect later behavior.

The opposite of feeling afraid (whether alarmed or anxious) is feeling secure—or, according to the Oxford Dictionary, feeling “untroubled by fear or apprehension.” When an infant or young child is with an attachment figure, he is likely to be untroubled by fear or apprehension, unless he is troubled by his expectations that he/she may become inaccessible at any moment and/or fail to be responsive to his needs and wishes. Thus the mere physical presence of an attachment figure is not necessarily enough to promote a feeling of security, although it very frequently seems to do so. One could expect that the older the child and the better articulated his representational model of the attachment figure, the less likely that the mere physical presence of the figure would be enough to provide a secure or untroubled state; whereas in the case of an infant whose expectations and representational models are still in an early formative stage, it is perhaps not surprising that he appears to be secure in his mother’s presence, until her actions or some other aspect of the situation activate his anxieties.

Just as when an infant feels afraid, his attachment behavior is likely to be activated (as well as fear behavior), likewise when he feels secure, his attachment behavior may be at a low level of activation. This accounts for the phenomenon that we have termed “using the mother as a secure base from which to explore.” When the attachment behavioral system is activated at low intensity, the situation is open for the exploratory system to be activated at a higher level by novel features of the environment. It seems of obvious survival advantage in evolutionary terms for a species with as long and as vulnerable a period of infancy as that characteristic of humans to have developed an interlocking between the attachment system, whose function is protection, and exploratory (and also affiliative) behavior, which promotes learning to know and to deal with features of the environment (including persons other than attachment figures.) This interlocking permits a situation in which an infant or young child is prompted by intriguing objects to move away from his “secure base” to explore them, and yet tends to prevent him from straying too far away or from remaining away for too long a time; and the reciprocal maternal-behavioral system provides a fail-safe mechanism, for “retrieving” behavior will occur if the child does in fact go too far or stay away too long. The interlocking between systems of this sort has led some to propose that the biological function of attachment behavior is (or should include) providing an opportunity for learning. Bowlby (1969) obviously gives first place to the protective function and indeed might well have said explicitly that the biological function of exploratory behavior is learning about the environment, whereas the protective function of attachment behavior and reciprocal maternal behavior makes this possible. Obviously the functions of both systems are of crucial importance.

After this divergence from the theme of feelings as appraisal processes, let us return to anger. Bowlby (1973) reminded his readers about the literature on responses to separation that makes it clear that anger is engendered by separation or a threat of separation, and that this anger is particularly likely to be manifested at the time of reunion. The separation literature to which he referred, however, dealt with “major” or “definitive” separations in which a child was separated from attachment figures for a period of days, weeks, or months and was usually also removed to an unfamiliar environment. Perhaps separations of but a few minutes, whether in a familiar or unfamiliar environment, do not so consistently arouse angry feelings as do major separation experiences. Attachment-relevant anger is activated under conditions other than separation, however. If attachment behavior is activated at high intensity but not terminated by an appropriate response by the attachment figure, anger is very likely to ensue—whether the reasons for the nontermination are the absence of the figure (as in the case of separation) or its chronic tendency to be unresponsive.

This brief discussion of the affective implications of attachment has dealt with some of the most obvious aspects of affective involvement, but is far from complete. Both Bowlby (1969, 1973) and Ainsworth (e.g., 1972) have emphasized the notion that attachments imply strong affect—not only security, anxiety, fear, and anger, but also love, grief, jealousy and indeed the whole spectrum of emotions and feelings.

The Development of Child—Mother Attachment

Because this volume is not primarily devoted to the development of a child’s attachment to his mother figure, here we merely summarize what has been published in more detail elsewhere about the course of such development (Ainsworth, 1967, 1972; Bowlby, 1969). In 1972 we distinguished four phases of development of child—mother attachment; these correspond to Bowlby’s four phases, but with somewhat different titles. Three of these occur in the first year of life: (1) the initial preattachment phase; (2) the phase of attachment-in-the-making; and (3) the phase of clear-cut attachment. The 1-year-olds, to whom most of this volume is devoted, may be assumed to have reached Phase 3, and hence this phase will be considered more fully than either of the two earlier phases. A final phase was initially identified by Bowlby (1969) as: (4) the phase of goal-corrected partnership, which, he suggested, did not begin until about the end of the third year of life, or perhaps later. It is therefore only the 4-year-olds, and possibly some of the 3-year-olds discussed in Chapter 10, who are likely to have reached this final phase of development.

1. The Initial Preattachment Phase. Bowlby (1969) called this the phase of “orientation and signals without discrimination of figure.” It begins at birth and continues for a few weeks. From the beginning the baby is more “tuned in” to stimuli within certain ranges than to others, and it seems likely that the stimuli to which he is most responsive come from people. At first, however, he does not discriminate one person from another, and hence responds to his mother figure (i.e., his principal caregiver) in much the same way as he responds to other persons.

The infant can orient toward anyone who comes into close enough proximity, directing his gaze toward that person and tracking the latter’s movements with his eyes. He is equipped with a repertoire of signaling behaviors—for example, crying, which is present from birth onwards, and smiling and noncrying vocalizations, which soon emerge. These signals serve to induce other people to approach him and perhaps to pick him up, thus promoting proximity and contact; hence they are classed as attachment behaviors. In addition, the infant is equipped with a few behaviors through which he himself can actively seek or maintain closer contact—for example, rooting, sucking, grasping, and postural adjustment when held. (Rooting and sucking obviously serve the food-seeking system as well as the attachment system, and indeed in bottle-fed babies, they tend to become splintered apart from the attachment system.) When the baby is not in actual contact with a caregiver, however, he can rely only on his signaling behaviors to promote proximity/contact—a state of affairs that persists throughout this phase and the next one.

As mentioned earlier, Bowlby (1969) suggested that the original behavioral equipment of the neonate consists of fixed-action patterns and that these become organized together and linked to environmental stimulus situations in accordance with processes of learning that have become well known through S—R psychology. At the same time it is easy to consider the neonate’s fixed-action patterns as equivalent to Piaget’s (1936) reflex schemata and to account for their modification in Piagetian terms. In either case the infant, even during this first phase of development, begins to build up expectations (anticipations), although at first, as Piaget held, these are inextricably tied to his own sensorimotor schemata and do not extend to using one environmental clue as a basis for anticipating another environmental event.

Phase 1 may be said to come to an end when the baby is capable of discriminating among people and, in particular, of discriminating his mother figure from others. Because discrimination is learned much earlier through some modalities than through others, it is difficult to judge when Phase 1 has ended and Phase 2 begun. There is evidence that the mother can be discriminated very early through olfactory or somasthetic cues, whereas visual discrimination is relatively late in developing. Nevertheless, it is convenient to consider Phase 1 as continuing until the baby can fairly consistently discriminate his mother by means of visual cues, which tends to occur between 8 and 12 weeks of age.

2. The Phase of Attachment-in-the-Making. Bowlby termed this the phase of “orientation and signals directed towards one (or more) discriminated figure(s).” During this phase the baby not only can clearly discriminate unfamiliar from familiar figures, but also becomes able to discriminate between one familiar figure and another. He shows discrimination in the way he directs his various proximity-promoting (attachment) behaviors toward different figures, and these figures may also differ in how readily they can terminate an attachment behavior, such as crying. During this phase the baby’s repertoire of active attachment behaviors becomes expanded—for example, with the emergence of coordinated reaching. This phase of development roughly coincides with Piaget’s (1936) second and third stages of sensorimotor development, but here we shall not attempt to link cognitive development with the development of attachment, except to point out that the development of discrimination may be thought to involve Piaget’s processes of recognitory assimilation—or, for that matter, discrimination learning.

If simple preference of one figure over others is the criterion of attachment, then one could identify a baby as attached to a preferred figure in Phase 2. We prefer, however, to characterize a baby as incapable of attachment until Phase 3, during which he can take active initiative in seeking the proximity of an attachment figure.

3. The Phase of Clear-cut Attachment. Bowlby identified this as the phase of “maintenance of proximity to a discriminated figure by means of locomotion as well as signals.” As Bowlby’s label implies, the baby in this phase is very much more active than before in seeking and achieving proximity and contact with his discriminated (and preferred) figures on his own account, rather than relying as he did before on signaling behavior to bring them into proximity. Chief among his newly acquired behaviors is locomotion. Obviously locomotion can also serve other behavioral systems. But when a baby approaches a preferred figure, whether following a departing figure, greeting a returning figure, or merely seeking to be in closer proximity, we may infer that locomotion is serving the attachment system. A number of other active behaviors emerge that can be put into the service of the attachment-behavior system, including “active contact behaviors,” such as clambering up, embracing, burying the face in the body of the attachment figure, “scrambling” over the figure in an intimate exploration of face and body, and so on. Signaling behaviors continue to be emitted and may on occasion be intentional communications. Indeed language begins to develop during Phase 3.

Although the Phase-3 child is more active in seeking proximity/contact, clearly he does so only intermittently. He is active also in exploring his environment, manipulating the objects he discovers, and learning about their properties. The Phase-3 child is by no means focused constantly on his attachment figures, even though they may provide the secure background from which he moves out to familiarize himself with his world.

Bowlby (1969), using his control-systems model, pointed out that an infant’s behavior first becomes organized on a goal-corrected basis in Phase 3, and then gradually becomes hierarchically organized in terms of overall plans. To the extent that attachment behavior is so organized, certain of the attachment behaviors are to a greater or lesser extent interchangeable. In a given episode of activation, the set-goal of the attachment system may be set for a certain degree of proximity, but there may be a variety of alternative behaviors through which a child may attempt to approximate that set-goal. Thus the specificity of each form of attachment behavior becomes increasingly less important, whereas the set-goal and overall plan for accomplishing it grow increasingly significant. Furthermore, the characteristic way in which a child has learned to organize his behavior with reference to a specific attachment figure is of clearly greater importance than the intensity or frequency with which he manifests each of the behavioral components of the attachment system. It is our conviction that the onset of goal-corrected attachment behavior is an acceptable criterion of the onset of attachment. In offering this criterion, however, we do not mean to imply that attachment, once present ceases to develop; on the contrary there is much further development of attachment during Phase 3 and beyond. We shall not here go into descriptive detail about Phase-3 attachment behavior, for both Bowlby (1969) and we in this chapter have tended to cite our illustrative material from Phase-3 behavior.

Phase 3 commonly begins at some time during the second half of the first year, perhaps as early as 6 months in some cases, but more usually somewhat later. Its onset may be conceived as coincident with the onset of Piaget’s Stage 4 of sensorimotor development. The emergence of goal-corrected behavior may be conceived as coincident with the onset of the ability to distinguish between means and ends; and certainly hierarchical organization of behavior according to plans depends on means—ends distinctions and on achieving the ability for “true intention.” The notion of alternative means of achieving a set-goal that is implicit in plans has its parallel in Piaget’s concept of schemata becoming “mobile.” Furthermore, the achievement of at least a Stage-4 level of development of the concept of persons as having permanence—that is, as existing when not actually present to perception—seems to us (as well as to Schaffer & Emerson, 1964, and to Bowlby, 1969) a necessary condition for a child’s becoming attached to specific discriminated figures. In other words, our view of attachment implies a conception of the attachment figure as existing even when absent, as persistent in time and space, and as moving more or less predictably in a time—space continuum.

Despite the obvious connection between the concept of person permanence and separation distress, we are not convinced that the onset of crying when the mother leaves the room implies the acquisition of even a Piagetian Stage 4 concept of person permanence. Both Ainsworth (1967), in her study of Ganda babies, and Stayton, Ainsworth, and Main (1973), reporting on our longitudinal study of a sample of American babies, reported that crying when mother leaves the room occurs as early as 15 weeks. (In the latter study we were careful to eliminate episodes in which the baby was left alone or in which he had been just put down after having been held, because these were conditions likely to evoke crying from birth onwards.) We are inclined to believe that these very early instances of crying when mother leaves are an extension of the phenomenon, mentioned by Wolff (1969), of distress when a figure moves out of the infant’s visual field—an extension because in this case it is a discriminated figure disappearing at a substantial distance from the infant, implying both an extension of the visual field and the ability to visually discriminate among figures at a distance. There is no indication merely from the distress that the baby yet conceives of his mother as having existence after having disappeared from the visual field. For this, one would require, as Piaget suggested, search for the vanished person.

Nevertheless, even though instances of separation distress may occur before Phase 3 of the development of attachment (and before Stage 4 of the account by Piaget, 1936, of sensorimotor development), there is much evidence that separation distress is particularly likely to occur in Phase 3, even though it is clearly not inevitable in very brief separations either at home (Stayton, Ainsworth, & Main, 1973) or in the strange situation—as the findings reported in later chapters demonstrate. To us it is suggestive that it occurs fairly commonly at about the same time that locomotion and goal-corrected behavior first emerge. One could argue that a baby does not need to be attached to a specific figure or to organize his behavior on a goal-corrected basis until locomotion makes it possible to move away from his mother figure to explore the world. In any event it is a happy circumstance that these developmental acquisitions coincide—and as for crying and attempts to follow a mother who is disappearing or who has already disappeared, these acquisitions also have a survival function for the active, mobile child.

We have already mentioned expectations (anticipations) as beginning to be formed as early as Phase 1. It is clear that by the time an infant reaches Phase 3, these expectations become even more important. By this time, as Piaget (1936) points out, the child can begin to use one environmental event as a cue that another environmental event will follow. This implies that he can begin to anticipate his mother’s actions, insofar as these have a reasonable degree of consistency. Bowlby (1969) suggested that a baby in Phase 3, whose behavior has become goal corrected, is capable of taking into account in the plans through which he organizes his attachment behavior his expectations of how his mother is likely to act. That is to say he is capable of adjusting his plans to his mother’s expected behavior.

Phase 3 is conceived as continuing through the second and third years of life and thus obviously continues beyond the limits of Piaget’s Stage 4, spanning the rest of the sensorimotor period and comprehending at least the first portion of his preoperational period. This being so, it follows that attachment becomes increasingly a matter of inner representation of attachment figures and of the self in relation to them.

Bowlby emphasized that, although an infant’s attachment behavior and a mother’s reciprocal behavior are preadapted to each other in an evolutionary sense, the behavior of each partner is often dominated by other “antithetical” behavior systems. When an infant’s attachment behavior is activated, his mother may well be occupied with some activity antithetical to “maternal” behavior. Although the Phase-3 infant becomes increasingly capable of adjusting his plan for achieving the desired degree of proximity/contact with his mother in accordance with her current activity as interpreted in the light of the representational model of her that he has built up, there are limits to the success that his efforts are likely to meet, unless his mother abandons her plans in order to accommodate herself to his plan. The Phase-3 child is conceived as too “egocentric,” in Piaget’s (1924) sense, to be able to divine what his mother’s current plan might be and to act to change it so that it is in greater harmony with his own.

4. The Phase of a Goal-Corrected Partnership. To Bowlby (1969) the fundamental feature of the fourth and final phase of the development of child—mother attachment is the lessening of egocentricity to the point that the child is capable of seeing things from his mother’s point of view, and thus of being able to infer what feelings and motives, set-goals and plans might influence her behavior. To be sure, this increased understanding of his mother figure is far from perfect at first and develops only gradually. To the extent that a child has developed his representational model of his mother to include inferences of this sort, he is then able to more skillfully induce her to accommodate her plans to his, or at least to achieve some kind of mutually acceptable compromise. Bowlby suggested that when this point of development has been reached, mother and child develop a much more complex relationship, which he terms a “partnership.” That he termed it a “goal-corrected” partnership underlines the flexible, hierarchical organization of the child’s attachment behavior and of his mother’s reciprocal behavior that is implicit in the concept of “plans.” He surely did not mean to imply that goal-corrected behavior did not emerge until Phase 4, for he is explicit in pointing out that such behavior is characteristic of Phase 3 and serves to differentiate it from Phase-2 behavior.

Furthermore, as we have already implied, because of the development of communication and of the symbolic representations implicit in working models of self and of attachment figures, the kinds of interactions between a child and his attachment figures undergo much change. And as we have also previously implied, the forms of behavior through which the attachment system is mediated become much more varied, although they still feature, under certain circumstances, overt proximity/contact seeking.

Despite the increasing sophistication of the processes mediating a child’s attachment to his mother and others, and despite the fact that developmental changes continue, Bowlby did not conceive of such changes as involving processes different enough from those operating in Phase 4 to specify further phases of development. On the contrary, the processes implicit in Phase 4 were conceived as characteristic of mature attachments. Although Bowlby (1969, 1973) was specifically concerned with the attachment of a child to his mother figure, he conceived of attachments to other figures as approximating the same model—and he clearly stated that attachments continue throughout the entire life span. Attachment to parent figures may become attentuated as adulthood approaches and may become supplemented and to some extent supplanted by other attachments; but few if any adults cease to be influenced by their early attachments, or indeed cease at some level of awareness to be attached to their early attachment figures.