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


Preface (2015)

Everett Waters, Inge Bretherton and Brian E. Vaughn

A preface generally recounts how the idea for a book evolved or how a project developed to the point of requiring a book-length presentation. The story behind Patterns of Attachment is exceptionally well-documented (e.g., Ainsworth & Bowlby, 1991; Bretherton, 2013, 1991; Karen, 1998; van Dijken, van der Veer, van IJzendoorn, & Kuipers, 1998). Nonetheless, it is worth retelling for new readers.

This preface is written with several important goals in mind. The first is to emphasize that Patterns of Attachment remains a core resource in attachment study and deserves a close reading. The second is to make the book, and the foundations of developmental attachment theory, more accessible by clarifying and updating points of theory and method that have been the seeds of misunderstandings, and at times, controversy. In addition, we have added new appendices that include the full text of the measures for secure base behavior at home and the four maternal sensitivity constructs used in the Baltimore longitudinal study. For economy, these important measures were only presented in outline in the original printing and in subsequent journal articles. However, after circulating for decades as mimeographed artifacts from Mary Ainsworth’s laboratory, it is high time they appear in full and in their appropriate context.

A Modern Classic

It is rarely clear from the outset that a scientific study will become a landmark. Moreover, it is never clear exactly where its greatest impact will fall. Patterns of Attachment appeared at a critical moment in the development of attachment study. Only the first volumes of Attachment and Loss (Bowlby, 1969, 1973) had appeared and, as theoretical works, it naturally made more predictions and raised more questions than they answered. Bowlby’s ethological approach seemed an advance over psychoanalytic and learning theory approaches to infant—mother relationships. Moreover, his emphasis on ethology, cognition, and control systems suggested deploying new strategies and tools. But would it work? Would difficult issues in social and emotional development yield to this new approach? Would it help resolve seemingly intractable issues regarding the importance of early experience? Patterns of Attachment was clearly reporting the kinds of progress necessary to bring attachment study to center stage. But would it attract enough new students to pursue all its implications for theory and practice?

More than three decades on, Patterns of Attachment has brought us a long way toward achieving these goals and has enriched developmental psychology and related fields beyond what Mary Ainsworth and her students could have imagined. In a sense, the Baltimore project was a logical extension of the Infancy in Uganda project (Ainsworth, 1967). It retained and built upon the observational descriptive methodology of the Uganda study. However, the observations were organized around a more formal schedule and research design that supported stronger tests of key attachment theory hypotheses. It was also an opportunity to consolidate insights from the Uganda observations into a more focused and formalized set of tools for quantifying maternal and infant behavior. Access to laboratory facilities also made it possible to coordinate the naturalistic observations with semi-structured laboratory observations in what became the Strange Situation Procedure (SSP). In turn, the SSP provided a window on facets of infant attachment behavior that were less salient in naturalistic settings.

The Baltimore project did much to validate and insure the good health of John Bowlby’s attachment theory. It also opened the door to translational research that has helped establish infant psychiatry as a distinct discipline and served as a template for early intervention strategies. What is critical now is that Patterns of Attachment remain available for students and young researchers. This is assured by its inclusion in the Psychology Press Classic Editions series.

Patterns of Attachment

Patterns of Attachment reports the methods and key results of Mary Ainsworth’s landmark Baltimore longitudinal study, in which she and her students observed infant—mother interaction and attachment behavior throughout the first year of life. Following upon her naturalistic home observations in Uganda, the Baltimore project yielded a wealth of enduring, benchmark results on the nature of the child’s tie to its primary caregiver and the importance of early experience. It also addressed a wide range of conceptual and methodological issues common to many developmental and longitudinal projects, especially issues of age appropriate assessment, quantifying behavior, and comprehending individual differences. In addition, Ainsworth and her students broke new ground, clarifying and defining new concepts, demonstrating the value of the ethological methods and insights about behavior, and deploying plain hard work to surmount obstacles to good measurement.

As in Infancy in Uganda, Ainsworth showed an exceptional “eye” for maternal and infant behavior and for the way meaning is conditioned by timing and context (Bretherton, 2003). Indeed, her conceptualization and lengthy naturalistic observations of maternal behavior are still unsurpassed in developmental research. Ultimately, the Baltimore observations provided decisive support for viewing attachment as a secure base relationship. They also revealed reciprocal links between proximity seeking, exploration, and sensitivity to physical, behavioral, and emotional context that neither psychoanalysis nor learning theory had highlighted or explained.

Attachment theory evolved from John Bowlby’s critique of psychoanalytic drive theory and his own clinical observations, supplemented by his knowledge of fields as diverse as primate ethology, control systems theory, and cognitive psychology. By the time he had written the first volume of his Attachment and Loss trilogy, Mary Ainsworth’s naturalistic observations in Uganda and Baltimore, and her theoretical and descriptive insights about maternal care and the secure base phenomenon, had become integral to attachment theory. This combination of theory and observation was logically compelling and presented by both Bowlby and Ainsworth with exceptional clarity. Nonetheless, their work might not have passed the test of time were it not for the Strange Situation Procedure (SSP) reported in Patterns of Attachment. Here was a structured, quantifiable, and reproducible assessment procedure that was much more economical than naturalistic observation.

Now, as we enter the fourth generation of attachment study, we have a rich and growing catalogue of interesting and well-validated approaches to measuring attachment-related behavior and representations from infancy to adulthood. Yet, each of them has roots in the SSP and the secure base concept. The Psychology Press Classic Editions series celebrates books that are widely recognized as enduring classics in Psychology. Patterns of Attachment has certainly endured. More importantly, its significance and influence continue to grow.

Impact and Endurance

It is difficult to separate the impact and endurance of a particular book from the good health of the field it represents. Qualitative and quantitative evidence confirm that, after more than 35 years, continuing interest in Patterns of Attachment reflects and contributes to the health of attachment study.

Twenty Studies

At the opening of the 21st century, developmental psychologist Wallace Dixon (2002) asked members of the Society for Research in Child Development to identify twenty studies (1950—2000) that had revolutionized child psychology. Alongside Piaget’s (1936/1952) The Origins of Intelligence, Vygotsky’s (1978) Mind in Society, and Chomsky’s (1957) Syntactic Structures, the top five included both Bowlby’s (1969) Attachment and Loss (Vol. 1) and Patterns of Attachment. In 2016 Dixon updated his survey, focusing again on the most recent 50 years and adding additional dimensions. Dixon’s (2016) judges ranked Patterns of Attachment among the Most Revolutionary (#1), the Most Important (#1), and Most Fascinating (#2) developmental psychology studies of the past 50 years. The only ranking in which it finished out of the running, so to speak, was Most Controversial. Although attachment study hardly seems controversial today, for many of us, acceptance feels like it has been a long time coming.


Although scholarly impact is a difficult phenomenon to measure, the citation indexes maintained by the Web of Science and Google Scholar are widely used for evaluating and comparing both journal impact and the impact of books and articles within a field of study. Both data bases report the extent to which individual works or authors are cited by other scientists in a given time period. Google Scholar bases its information on a broad (some would say indiscriminant) search of citations in books, articles, manuscripts, conference reports, etc., accessible through the Internet. In contrast, Web of Science limits its counts on a selection of source journals. It also provides useful analytic tools that help identify and interpret patterns of results. Although Google Scholar typically reports higher citation counts than Web of Science, this is a matter of focus rather than of validity or utility.

A Google Scholar search conducted March 2015 indicated that Patterns of Attachment has been cited almost 15,000 times, an astounding number for an empirical monograph rooted in behavior observations. This compares favorably with Piaget’s The Origins of Intelligence (10,849 citations in Google Scholar), which established, and for decades defined, the field of early cognitive development. Moreover, it surpasses Thomas and Chess’s (1977) classic, Temperament and Development (3,427 citations) which also focuses on patterns in infant development and was included in Dixon’s (2002) Twenty Studies that Revolutionized Child Psychology.

The narrower Web of Science data base identifies over 5,000 citations to Patterns of Attachment. More importantly, it provides a citation report showing the pattern of citations year by year. For most publications, citation count rises beginning 2—3 years after a book or article appears and then declines as a field moves on and interest in a particular work wanes. In marked contrast, interest in Patterns of Attachment has been constant, even increasing, across the three decades since it was first published (see Figure P.1). This too sets Patterns of Attachment apart from most observational and experimental research in developmental psychology. Of course the field has grown since 1978 and there are more attachment researchers and more journals publishing attachment-related research. However, these reflect, as much as account for, the enduring interest in and impact of this psychology landmark.

Figure P.1 Articles citing Patterns of Attachment 1978—2013 (Web of Science).

FIGURE P.1 Articles citing Patterns of Attachment 1978—2013 (Web of Science).

Patterns of Influence

It is also useful to examine how Patterns of Attachment has influenced and supported enduring and emerging themes in attachment study. One approach is to look for themes among the books and articles that have cited Patterns of Attachment. This could be done at different levels of detail and with different degrees of formal analysis. However, even an informal content analysis of titles citing Patterns of Attachment reveals that it has had broad impact and served as a catalyst for explorations in new directions. It also shows that attachment study has kept pace with new directions in psychology. Having identified all the books and articles that cited Patterns of Attachment between 1978 and 2013, we identified those that had been cited at least 100 times. We then used titles and abstracts to sort these 375 high impact items into categories. The twenty categories with the most members are listed alphabetically in Table P.1.

TABLE P.1 Categories of Articles Citing Patterns of Attachment 1978—2013 (Web of Science)

1. Attachment across cultures

11. Attachment and alternative care arrangements (daycare, adoption)

2. Attachment and social competence

12. Attachment-based interventions

3. Attachment to fathers and other non-maternal figures

13. Child maltreatment

4. Disorganized attachment

14. Effects of early experience

5. Emotion regulation

15. From sensorimotor to formal representations

6. Infant—mother interaction, maternal characteristics (e.g., depression)

16. Maternal sensitivity

7. Measurement alternatives and extensions

17. Neurobiology and genetics of attachment classifications

8. Patterns of attachment in adult relationships

18. Relevance of classification concepts across age

9. Relevance to at-risk and developmental psychopathology

19. Stability and change across age

10. Stimulus to advances in attachment and developmental theory

20. Validity in various age groups and populations

A number of these are enduring topics. However, even within familiar categories, the lines of research have evolved quite a bit since the early years of attachment study. Others, such as emotion regulation, neurobiology and genetics of attachment classifications, disorganized attachment, and perhaps even relevance to risk and developmental psychopathology (insofar as this was not yet a distinct field of study) could hardly have been imagined when Patterns of Attachment first appeared. This is an important indication of the book’s heuristic value and, the good health of attachment study today.

The Attachment Paradigm

John Bowlby liked to refer to attachment theory as a new paradigm, a new way of understanding the infant’s tie to primary caregivers. Paradigm can also refer to a community of theorists and researchers bound together by shared principles and methods (Kuhn, 1962/2012; Masterman, 1970). Thus, the attachment paradigm refers to both Bowlby—Ainsworth attachment theory and to the community that shares and contributes to their perspective, as distinct from the psychoanalytic and learning theory perspectives.

Paradigm can also refer to one or more prototypical problems or key techniques associated with a theoretical or methodological approach (Kuhn, 1962/2012; Masterman, 1970). As students become skilled in solving such problems or using a particular tool, they come to understand the practical meaning of key theoretical concepts. They also learn to recognize the contexts in which a theory or methodology is relevant. Eventually, with much experience across many trials, they acquire the expectations and fluency characteristic of experts. In addition to their use as a tool for discovery, the attachment—exploration balance and the SSP are very much the paradigm through which generations of students have learned to recognize attachment behavior and understand attachment theory. This, as much as any theoretical insight or empirical result, is why Patterns of Attachment endures.

The Secure Base Phenomenon

John Bowlby introduced attachment theory in a 1957 talk to the British Psychoanalytical Society. It was published the following year as “The nature of the child’s tie to his [sic] mother” (Bowlby, 1958). The key to the paper was a proposal for replacing psychoanalytic drive theory and cathectic bonding with a more tenable, empirically accessible motivation model. For this Bowlby turned to ethology and comparative psychology and the concept of instinctual behavioral responses. There was as yet no mention of exploratory behavior or the attachment—exploration balance, both of which are central to Patterns of Attachment and current attachment theory.

Bowlby (1958) identified as attachment-related five behaviors that have the predictable outcome of helping maintain proximity to the mother and contribute to maintaining her availability. These included smiling, crying, clinging, following, and sucking. Bowlby proposed that, over time, and with appropriate experience, these become integrated into an attachment behavioral system that is keenly sensitive to inner and environmental cues and context.

Bowlby’s use of concepts from ethology led some early critics to characterize his proposal as an instinct theory, with the now outdated connotation of behaviors that emerge early in development and are relatively inflexible (stereotyped) in both form and response to the environment. In response, Bowlby (1969) went to considerable lengths in Attachment and Loss (Vol. 1) to provide a more sophisticated understanding of the instinct concept as applied to behavioral systems. Criticisms of attachment theory as an instinct theory are simply uninformed. The ethological, evolutionary concept central to the theory is not the innateness of the attachment system but the importance of evolved biases in infant learning abilities which are part of our primate evolutionary heritage and critical to attachment development. It is not attachment that is inherited. It is the capacity to become attached and to construct (through interaction with an appropriate caregiving environment) a system for using one or a few figures as a secure base.

Designating a specific set of behaviors as attachment behaviors was also a source of misunderstanding, measurement problems, and criticism until Sroufe and Waters (1977), Hay (1980), and others clarified what Bowlby and Mary Ainsworth had known all along—that every behavior accessed by an attachment control system is available to other motivational—behavioral systems as well. There are no uniquely attachment-related behaviors. Thus, simply adding up all the instances of a particular behavior irrespective of context is unlikely to measure anything very interesting regarding attachment relationships.

Mary Ainsworth always said that her goal in Uganda was not to confirm Bowlby’s attachment theory but to see whether conceptualizing the child’s tie to its mother as a secure base relationship fit what mothers and babies actually do. In fact, she didn’t send Bowlby regular updates on the study or engage in much back and forth as it unfolded. Then, in 1958, he sent her his paper on the child’s tie to its mother. This prompted her to provide an overall picture of her observations, which Bowlby generously described as a “happy convergence” with his own conclusions. In the spring of 1959, Bowlby visited Mary Ainsworth in Baltimore. This gave them a chance to compare notes and for her to fill him in on some of the initial analyses of her Uganda data. Then in 1961, at Bowlby’s invitation, she began presenting her work at the Tavistock Seminars on Mother—Infant Interaction which he organized in London. Attachment theory evolved quite a bit between 1958 and 1969. Much of this evolution was motivated and guided by Mary Ainsworth’s emphasis on the secure base concept and her observations of proximity-seeking and exploratory behavior in Uganda and in the 26 mother—infant dyads she and her students studied in the Baltimore longitudinal study (1963—67).

Bowlby introduced the control systems concept in the first volume of his Attachment and Loss trilogy (Bowlby, 1969, Ch. 13). This was a significant step beyond the 1958 formulation, both in addressing the underlying mechanisms and the context sensitivity of attachment responses. In addition, rather than focusing exclusively on the conditions under which the attachment responses are switched on and off, the emphasis turned to the ongoing dynamic balance between exploration and proximity seeking. It was apparent in naturalistic observations spurred by Bowlby’s work that infants actively monitor their caregivers’ location and availability while exploring. That is, exploration was pivoting around or referenced to a secure base figure about whom information and expectations were continuously updated by an attachment control system. Indeed, if there were no exploration there would be no need for proximity seeking. The infant could, as in many species, simply stay on or within reach of its mother.

Most mammals are capable of seeking proximity/contact early in infancy. In addition, they all show extensive exploratory interest in their environments. Moreover, they all use both proximity-seeking and exploratory locomotion for a variety of purposes including information seeking and foraging (Hay, 1980). Yet, they do not all show the kind of lasting bonds, extended and extensive exploration, and parental care we associate with parent—offspring (not to mention adult—adult) human attachment relationships. The attachment control system coordinates these two systems, over time and in light of context, to serve two functions: (a) maintaining proximity to the primary caregiver, and (b) supporting exploration and learning. Although Bowlby (1969) emphasized the role of proximity in affording a degree of protection, he clearly recognized that both functions play significant roles in evolution and development.

Infants and children learn much more when exploring with a caregiver, who can scaffold and co-construct problem solutions and problem-solving skills, than they could exploring alone. Thus, rather than thinking of such behaviors as more (or less) attachment-related, it is useful to think of attachment as a system which, for extended intervals, takes control of proximity and exploratory systems and coordinates them in ways that promote both safety and the experience necessary to build a human nervous system (Waters, 2008). That is, attachment is not the behaviors but the higher level system that organizes them (Bowlby, 1969; Sroufe & Waters, 1977). This perspective is important for appreciating the roles of internal working models and for building computational models of the attachment—exploration balance (e.g., Bischof, 1975; Petters, 2006; Petters, Waters, & Sloman, 2011; Petters, Waters, & Schönbrot, in press, 2015).

The close link between attachment and exploration, which was so evident in the Uganda and Baltimore observations, found fuller theoretical development in the second volume of Bowlby’s trilogy (Bowlby, 1973), especially Chapter 21 (“Secure attachment and the growth of self-reliance”). Still, Bowlby felt that the iconic image of an infant retreating to its mother for safety or comfort was only half of the picture. Attachment theory required a new, equally evocative concept to encompass this and the mother’s role as base from which to explore. Mary Ainsworth had often spoken of the infant’s excursions out to explore and back to the mother as the secure base phenomenon. Problem solved—and thus the title of Bowlby’s 1988 collection of lectures and articles, A Secure Base, and his dedication, “to Mary D. S. Ainsworth who introduced the concept of a secure base.”

The Baltimore Project

The Baltimore study was not merely “a psychological study of the Strange Situation.” It advanced from Infancy in Uganda on a number of fronts. First of all, it shed the idea that strength and onset of attachment were measurable phenomena. It would be more productive for both theory and measurement to focus instead on: (a) the skill and confidence with which an infant used its mother as a secure base once attachment was clearly established, and (b) the infant’s expectations regarding the mother’s accessibility and responsiveness and her cooperation with its ongoing behavior.

The Baltimore project also benefited from Mary Ainsworth’s contact with John Bowlby as he developed the Attachment and Loss trilogy. In addition to using her as a sounding board for testing his own ideas, Bowlby repeatedly acknowledged the value he placed on her insights and detailed knowledge of maternal and infant behavior. And, of course, he credited her with the secure base formulation that captured the essence of the relationship so well. This collaboration gave the Baltimore project had a head start on the field when it came to testing hypotheses based on the most recent developments in attachment theory. This was particularly important when it came to insuring that home observations covered the full range of theoretically relevant behaviors. It also influenced the design of the SSP episodes.

Finally, it was easier in Baltimore than in Uganda to follow a structured recruitment plan and research design in which each mother—infant dyad could be observed at home for the same 16 hours during each quarter of the first year. Even today, with ready access to sound and video recording, the narrative records from these observations are unparalleled as longitudinal descriptions of development in the first year.

A Flair for Measurement

Patterns of Attachment established the SSP as the hallmark of infant attachment research. Nonetheless, the measures developed for assessing (1) maternal interactive behavior, and (2) the attachment—exploration balance at home were also important to the Baltimore project. The maternal interactive behavior scales, in particular, are unparalleled examples of translating detailed observations into workable quantitative measures.

Maternal Sensitivity

After more than 40 years, the Ainsworth Maternal Sensitivity Scales remain cornerstones of research on infant—mother interaction. As illustrated and discussed in the December 2013 special issue of Attachment and Human Development (Grossmann, Bretherton, Waters, & Grossmann, 2013) they are some of the most elegant behavior descriptions in psychology and convey a great deal about Mary Ainsworth’s skill as an observer and theorist. Moreover, as with the secure base concept and the attachment—exploration balance, the maternal caregiving and interaction constructs, (1) sensitivity to signals, (2) cooperation with ongoing behavior, (3) acceptance of age-related requirements, and (4) physical—psychological accessibility, have proven useful for theory and research strategy across the lifespan. Indeed, they would have to because the very idea of a secure base relationship entails someone using, and someone providing, secure base support. Neither can be defined without reference to the other—an example of a truly dyadic phenomenon.

In the past, the four caregiving and interaction scales have only been available in mimeograph or online. They deserve to be more accessible and more often studied in detail. Thus, they have been included in this reissue as Appendix IV. Each of the scales consists of an introductory essay that describes how the construct is conceptualized and how it plays out in the home environment. In addition, each essay includes thoughtful observations about how best to observe and the kinds of difficulties that arise in observing and interpreting behavior at this level of detail. As always in Ainsworth’s work, the emphasis is on observing the behavior, not rating a psychological construct, and when assigning scores, looking for a convergence of indications rather than placing heavy bets on individual acts.

Conceptualizing the four caregiving and interaction constructs, translating them into measures, and relating them to attachment outcomes in a truly modern and theoretically grounded way was one of the singular accomplishments of the Baltimore study. A key innovation here was basing the scale anchors on vignettes from the home observations and allowing more than one interactive vignette to anchor a given scale point. Keeping the scales close to actual behavior and finely attuned to context was a significant advance over scales based on stereotypes, informal observations, and arbitrary anchors. Importantly, the link between sensitivity constructs and infant security has stood the test of time in a number of longitudinal studies (see Grossmann et al., 2013). In addition, the development of the Maternal Behavior Q-set (MBQ) by Pederson and Moran (1995) has made it much easier to collect and quantify naturalistic observations of mother—infant interaction.

Recent research shows that there is considerable room to expand the Patterns of Attachment conceptualization of maternal sensitivity, especially at older ages, to give added weight to support for independence and for exploration. In recent research, Bernier, Matte-Gagne, Belanger, and Whipple (2014) have found that doing so substantially improves a model linking maternal AAI coherence to maternal sensitivity and child security. Similar work, extending current conceptualizations and measurement of secure base support to older ages (Crowell, Treboux, Gao, Fyffe, Pan, & Waters, 2002), and examining it in different cultures and family circumstances (e.g., Posada, Carbonell, Alzate, & Plata, 2004) is a promising direction for new research based on new ethological attachment studies (e.g., Grossmann et al., 2013).

Secure Base Behavior at Home

Bowlby’s conceptualization of the child’s relationship to its primary caregiver as a secure base relationship was a key departure from psychoanalysis. This opened the door to a control systems approach to motivation which accounted for the infant’s sensitivity to its environment and the apparent purposefulness of its excursions from and back to its mother, without resorting to unobservable libidinal drives or implausible schedules of reinforcement, or invoking some equally magical alternative. To establish this, Bowlby reviewed ample behavioral and neurophysiological evidence of control systems organizing and regulating the behavior of other species. Moreover, evolutionary theory provided a very plausible explanation for human infants’ ability to construct such a system. The challenge in the Baltimore study was to build upon the experience and results Ainsworth reported in Infancy in Uganda to quantify individual differences in infants’ secure base behavior.

The performance of a control system is ordinarily evaluated in terms of how closely its output tracks a set goal (see Bowlby, 1969; Waters & Deane, 1985). For example, the variance of room temperature around a thermostat setting of 72 degrees reflects how well a thermostatic control system accomplishes its goal of maintaining a stable, comfortable room temperature. Unfortunately, the set goal of the attachment control system is not as easily specified. In early formulations, Bowlby spoke of an infant maintaining a degree of proximity to its mother. At the same time, he recognized that this depended very much on contextual factors, recent events, the familiarity of the setting, whether the infant was tired or ill, etc. It also seemed to depend on the infant’s expectations regarding its mother’s availability and responsiveness. Thus, the problem Mary Ainsworth had to solve in the Baltimore study was to go beyond specific proximity-seeking and exploratory behaviors to capture how well an infant managed the attachment—exploration balance across time.

The solution was to use the scorer’s understanding of the secure base concept to match infants’ behavior to one of five levels of secure base use. These ranged from: (1) using the mother as a secure base and maintaining a smooth balance between proximity and exploration, to (5) patterns in which the attachment—exploration balance was not consistently maintained. As with the maternal caregiving and interaction scales, the measure was derived from transcripts of the actual Baltimore home observations. In addition, sub-classifications at each level described different patterns of secure base behavior that could be considered similarly effective or ineffective. Importantly, the anchors were not discrete behaviors. Instead, they were patterns of behavior as it was organized (or not) over time and context.

Mary Ainsworth was “not altogether satisfied” (Ainsworth, Blehar, Waters, & Wall, 1978, p. 241) with this approach. Nonetheless, it served well enough to anchor the SSP to behavior in a naturalistic setting, confirming its relevance to Bowlby’s theory. It also served well enough to validate SSP behavior profiles in terms of attachment security. Although largely replaced by the attachment Q-set (Waters & Deane, 1985) we have included the complete criteria for scoring the attachment—exploration balance at home as Appendix V as a reminder of the important role secure base behavior at home played in the Baltimore Study.

Despite its convenience and demonstrated usefulness, the SSP was never intended to foreclose further ethological study of secure base behavior. Moreover, now that research has established the relevance of the secure base concept beyond infancy, indeed well into adulthood (e.g., Waters & Cummings, 2000; Crowell et al., 2002), and with hand-held (and smaller) video recording equipment available and inexpensive, research on secure base behavior in naturalistic environments deserves high priority on the agenda of attachment study.

Origins of the Strange Situation

The Uganda observations were conducted in 1954 but, due to personal circumstances, were not published until 1967; observations for the Baltimore longitudinal sample were conducted in 1963—64. Quite a few studies examining brief mother—child separations predate Ainsworth’s Uganda and Baltimore studies and have been mentioned as precursors, if not models, for the SSP. For example, Shirley and Poyntz (1941) studied 101 boys and 98 girls aged 2—8-years-old who made semi-annual, day-long visits to a Harvard School of Public Health clinic as part of a larger longitudinal study of child development. Their approach was primarily descriptive and normative, with the goal of shedding light on children’s reactions to separation from their mothers. In most instances, a staff member picked the child up at home in the morning and drove them to the clinic for a full day of physical, medical, and psychological tests, along with lunch, play breaks, and a nap. In a quarter of the cases a parent brought the child to the clinic. A parent retrieved the child in the afternoon.

Shirley and Poyntz kept diary-type records of the child’s day at the clinic. In addition, at the end of the day, they obtained a verbal report of the initial separation and the trip to the clinic from the staff member who had transported the child. Subsequent analyses of specific behavior categories were based on these records. In addition, their report included descriptions of a number of cases, often from visits at several ages, to illustrate age trends and differences within age. The report also included comments on a wide range of behaviors that were not addressed in formal analyses. Their comments were on phenomena such as: (1) mother’s styles of negotiating the departure for the clinic, (2) low-keyedness during play-time (e.g., “killing time,” rejecting toys, sedentary play, absence of talking), even after obvious separation-related distress had abated, and (3) resuming crying upon mother’s return. Although observed in older children, such behaviors will be familiar to readers of Patterns of Attachment.

In a second report, based on the same sample, Shirley (1942) developed a weighted “adjustment assay” intended to reflect a child’s degree of separation-related distress independent of age and sex. She then identified 12 girls and 22 boys whose mothers met criteria for “over-protective” or “rejecting.” Based on a tabular (as opposed to statistical) analysis, Shirley and Poyntz (1941)concluded that:

A child’s level of adjustment depends little upon the extrinsic features of the day, and little even upon his health. It depends much more upon the wholesomeness of his upbringing in the home, and the security and confidence and affection given him by his parents. A secure and wholesomely loved child goes forth to meet a new experience in a spirit of adventure, and comes out triumphant in his encounters with new places, new materials, and new friends, old and young. A child that is over-sheltered or under-loved goes forth from home with misgivings and doubts, and gives an impression of inadequacy and immaturity in his encounter with new experiences that makes him unwelcome either in the society of adults or children. (p. 217)

Shirley and Poyntz seem to have had a very good sense for children’s separation-related responses and recognized the importance of the behavioral, emotional, and situational context when interpreting the meaning of behavior. Yet, their work hardly seems modern in comparison to the conceptualization, quantification, and analysis of both infant and maternal behavior in Patterns of Attachment.

Despite the value of the Shirley and Poyntz study, Ainsworth (personal communication) has said that she had in mind Jean Arsenian’s (1943) study of mothers and children in a Massachusetts Reformatory for Women when she decided to develop a separation—reunion procedure for the Baltimore study. Arsenian studied the 24 young children (11.2—30 months) in order to better understand the dynamics of childhood “security” through the medium of an unfamiliar room. However, as she notes, due to the circumstances, the children had only intermittent contact with their mothers, half of whom served as aides in the institution nursery, the remainder being assigned to work in other parts of the facility. Moreover, having a baby seems to have been a source of status among the inmates, and the mothers often traded on this by being characteristically over-protective and over-emotional (i.e., demonstrative) with their children. Arsenian also suggests that the children had less opportunity to explore than usual for children their ages and thus their reactions were “probably more intense than for a non-institutional group.”

The value of the study is limited by these unusual sample characteristics, arbitrary decisions underlying behavior coding, and the use of Lewinian field theory as an interpretive framework. Thus, aside from focusing on children in an unfamiliar room, Arsenian (1943) is unlikely to have influenced the particulars of the Baltimore study’s observations or the laboratory assessments. Nonetheless, her conclusions certainly parallel Bowlby’s and Ainsworth’s intuitions about the origins of infant security:

The extent to which the strange situation was made secure by the presence of the adult evidently varied with the dependence of the child and with the history of his previous relationship with the adult. For independent children, the “substitute” mothers were adequate sources of protection in the situation. Dependent children, on the contrary, were secure only in those instances where the adult who accompanied them was their own mother, whose affection and solicitude had been experienced constantly in the past. (p. 241)

This begins to capture the relational interpretation that is essential to Bowlby—Ainsworth attachment theory and must have resonated with Mary Ainsworth’s own thinking.

More important than either of these studies was John Bowlby’s invitation to participate in the 1961, 1963, and 1965 meetings of the Tavistock Seminars on Mother—Infant Interaction in London. These meetings afforded Mary Ainsworth a chance to present preliminary results from her Uganda observations and to keep abreast of Harry Harlow’s reports of his experimental studies on infant—mother interaction and attachment in rhesus macaques (see, Suomi, van der Horst, & van der Veer, 2008). Harlow was a talented experimenter and a keen observer. In addition he was careful and systematic in formulating hypotheses and interpreting results. It is clear that prior studies, particularly Arsenian’s, primed Mary Ainsworth to consider adding a standardized separation—reunion procedure to the Baltimore longitudinal study. However, it is equally clear that the SSP is not simply adapted from these studies. It was specifically designed to test hypotheses about the attachment—exploration balance and the secure base phenomenon.

Highlights and Issues

We usually consider a project fully reported once the key results have appeared in a series of journal articles. This may suffice even for high profile research that issues from especially active laboratories. However, this approach has diminished the impact of many important longitudinal projects, leaving us with a few key findings rather than a coherent picture and comprehensive evaluation. With few exceptions (e.g., Block, 1971; Sroufe, Egeland, Carlson, & Collins, 2005; and the contributors to Grossmann, Grossmann, & Waters, 2005) the organizers of major longitudinal studies have rarely found time or opportunity to portray the full sweep of their projects. Fortunately, this was not the fate of the Baltimore project.

Patterns of Attachment provides a much more coherent picture of the Baltimore project than would have emerged from journal articles alone. In addition, a substantial literature on the evolution of attachment theory provides useful reflections on the project’s context, rationale, and goals (e.g., Ainsworth & Bowlby, 1991; Bretherton, 2003). In addition to methods and results, the book also provides a concise yet detailed overview of attachment theory, and reviews and integrates previously published results. By giving full expression to the background and rationales for the design issues, measures, and key decisions that shaped the Baltimore study, Patterns of Attachment goes beyond documenting and teaching; it illustrates the ethological approach in developmental research and effectively mentors readers who are new to attachment study.

Normative Issues

The theory Bowlby outlined in 1958 and extended in Attachment and Loss (Vol. 1) addressed the nature of human infant—mother relationships in evolutionary/normative perspective before turning to the sources of individual differences. Similarly, after describing the SSP methodology, Patterns of Attachment turns first (Chapters 5 and 13) to normative issues. Here the focus is on behavior patterns illustrating the infant’s interest in its social and physical environments and the sensitivity of attachment and exploratory behavior to context. These patterns, which stand out in strong relief across SSP episodes, are exactly the details and complexities on which psychoanalytic theory and the operant theory of dependency had faltered.

The normative patterns of proximity seeking and contact maintaining toward mother and stranger across episodes (Chapter 5, Figures 2—9) clearly confirmed what was also evident in the home. One-year-olds were continuously monitoring a wide range of internal and environmental inputs in order to maintain what Bowlby had described as an “apparently purposeful” attachment—exploration balance that could only be plausibly explained with a motivational model at least as complex as a behavioral control system.

The SSP also provided normative data relevant to the interchangeability of attachment figures (mother vs. stranger) and the kinds of cues that activate and terminate attachment behaviors. Although such issues are not hotly contested today, establishing a sound normative picture of the attachment—exploration balance in different contexts helped ease the way for understanding and acceptance of the new paradigm.

Individual Differences

Although the normative issues addressed in the Baltimore project were central to attachment theory, Patterns of Attachment is best known for the secure, avoidant, and resistant/ambivalent classifications used to summarize individual differences across SSP episodes. These patterns reflect qualitative differences in the manner and effectiveness with which an infant organizes and maintains its secure base behavior with respect to a particular figure. That is, how well the attachment control system tracks its set goal of felt security across time and context.

Unfortunately, it can be difficult to evaluate how well a control system tracks its set goal. This is especially so if, instead of maintaining a single variable such as distance or temperature, the set goal is conceptualized as satisficing (Simon, 1956, p. 129; Sroufe & Waters, 1977) over several facets or variables. The task was all the more difficult in the Baltimore home observations because each home had a unique physical layout and the mothers’ behavior was unconstrained by specific instructions. In contrast, the SSP provided a consistent physical layout, the mothers’ behavior could be constrained somewhat without seeming out of place as it might at home, and the range of infant behavior was somewhat limited by the layout, toys, and sequence of episodes. Nonetheless, capturing the organization of behavior in the SSP, as opposed to simply quantifying discrete behaviors, presented a significant challenge. The strategy adopted for the Baltimore project was to identify behavior profiles associated with more or less effective secure base use over the episodes of the SSP and then to relate these to secure base behavior at home and to antecedent patterns of maternal care.

Even before the Baltimore project, Mary Ainsworth had a great deal of experience summarizing behavior in terms of patterns and classifications. As early as her Ph.D. thesis she noted that, in conceptualizing adult adjustment in terms of security, “it has become apparent that the pattern of adjustment (over domains) is more significant for the understanding of the individual than any single measurement, or any total score” (Salter, 1940, p. 13). She had also found classification a useful tool for organizing her Uganda observations. Pattern-based analysis had also played a significant role in her work as a psychodiagnostician at Baltimore’s Sheppard and Enoch Pratt Hospital in Baltimore between 1955 and 1961 (Ainsworth, 1983).

The Baltimore project was initiated in an era when many (perhaps most) psychologists viewed classifications systems with skepticism, associating them with unreliable assessments, illusory typologies, and armchair interpretations. However, this was not the spirit in which Ainsworth proposed her use of classificatory methods. Instead, she used classification as a tool for description, for representing order and organization in the overwhelming complexity (and volume) of her observational data (see Patterns of Attachment, pp. 55—59). She specifically eschewed the notion that attachment patterns explain behavior. Instead, when patterns were detected they became the phenomena to be explained, starting points for the next step in discovery. Despite some initial objections, significant links to secure base behavior at home and to key aspects of maternal care demonstrated the value of this approach.

The ABC’s of Attachment Patterns

Today, Patterns of Attachment is familiar first and foremost for the ABC classification system. Less well known (or remembered) is the initial reason for delving into individual differences. This is not to say that a classificatory system based primarily on reunion behavior was obvious. Only that the ABC classification system is rooted in issues other than individual differences per se.

Early Experience and Attachment Security

One of the key insights Bowlby valued in psychoanalysis was the idea that early experience plays an important role in later development. For developmental psychologists, this implied, among other things, the prediction that better quality maternal care would be associated with earlier attachment onset. Psychoanalysts, psychologists, pediatricians, and even baby care experts had suggested a wide range of maternal behaviors as likely to accelerate or delay attachment development. Sorting out the facets of early care that influenced attachment development was viewed as significant for both theory and practice.

The narrative records from each quarter’s home visits provided detailed information about a wide range of potential influences on attachment development. These ranged from breast vs. bottle feeding, schedule vs. ad lib feeding, various sleeping arrangements, frequency or duration of close bodily contact and face to face interaction, to interactive behaviors such as sensitivity to signals, cooperation with ongoing behavior, acceptance of the baby’s needs, and psychological and physical accessibility.

Separating the wheat from the chaff required a criterion for attachment onset with which maternal behaviors could be correlated. Unfortunately, it was already becoming clear that there was not going to be a well-validated criterion for attachment onset. Although there might be a fairly narrow window within which preference was established, this occurred very early and proved very sensitive to context. Moreover, this was more akin to bonding than to the kinds of attachment behaviors that Bowlby incorporated into his control systems model and that most experts had in mind as signs of strong or weak, secure or insecure, attachment. Similarly, developmental psychologists recognized that the ability to distinguish mother from others was a necessary precursor of attachment. However, it emerged long before the infant was capable of using her as a secure base. Similarly, responses to strangers (Spitz, 1965) and to separation from mother (e.g., Schaffer & Emerson, 1964) were suggested as signs of attachment onset. Such behaviors were too sensitive to context to serve as reliable criteria (Sroufe, Waters, & Matas, 1974). Other potential criteria for attachment onset seemed too closely tied to the mechanisms that control locomotor and sensori-motor development to reflect the impact of maternal behaviors.

In one of the key insights of early attachment study, Mary Ainsworth recognized that instead of sorting out maternal behaviors by correlating them with the age of attachment onset, she could instead correlate them with qualitative outcomes once attachment was clearly established. That is, instead of searching for maternal behaviors that led to earlier attachment onset, she would search for those that were most closely associated with a good outcome toward the end of infancy—where a good outcome refers to a criterion rooted in attachment theory, i.e., the attachment control system tracking its set goal smoothly and consistently.

Associations between SSP classifications and first and fourth quarter maternal behaviors, presented in Chapters 7 and 8 and in Tables 15—17, did a very nice job of identifying key maternal behaviors. The primary first quarter correlates mapped very well into four facets of maternal care assessed using behaviorally based rating scales, sensitivity to infant signals, cooperation with ongoing behavior, acceptance of the infant’s needs, and physical and psychological availability. Importantly, these are the variables that are most plausibly linked to acquiring expectations about the mother’s availability and responsiveness, and to the sense that she is “always there for me.” This result provided important support for Bowlby’s and Ainsworth’s view of attachment as a secure base relationship grounded on trust rather than the strength of a libidinal bond. It also supported the idea that, contrary to psychoanalysts’ expectations, negative attachment outcomes could arise in the absence of significant trauma.

Home Behavior and the Validity of the SSP

Obviously, it would be difficult to justify the SSP as a measure of individual differences in attachment security or secure base use if it were not significantly linked to secure base behavior at home. The key results on this point are summarized in Chapter 12 and Table 29, Classification of Strange Situation Behavior and Classification of Attachment—Exploration Balance Behavior at Home. Surprisingly, nearly two decades (and many studies using the SSP) passed before this key result was replicated (Vaughn & Waters, 1990). Van IJzendoorn, Vereijken, Bakermans-Kranenburg, and Riksen-Waldraven (2004) have summarized a number of subsequent studies validating the SSP against Attachment Q-set (AQS) assessments of secure base behavior at home. The results consistently show that secure base behavior at home is significantly related to secure vs. insecure classifications in the SSP. This is important evidence that the secure vs. insecure SSP classification is related to the security construct and to the secure base phenomenon observed at home.


The key findings from the Baltimore study have generally stood the test of time. However, as in other areas of science, the results in replication studies are not consistently as strong as in the original. A number of factors can influence replication results. These include: (a) familiarity with the SSP scoring and classification system, (b) observation skills and familiarity with the home observation scales, (c) the number and duration of home observations, (d) the samples on which replications are performed, and (e) the theoretical relevance of predicted correlates. Naturalistic observation is difficult and requires considerable experience with the behavior being observed. It is never as simple as reading scale definitions and starting to observe. Ideally, observers have access to training materials, pilot subjects, and mentors who can help build expertise in naturalistic observation before they begin collecting data.

Although training in SSP procedure and scoring has long been available through Alan Sroufe’s group at the University of Minnesota, nothing comparable has been developed to help researchers become expert observers of maternal and infant behavior at home. Waters and Deane’s (1985) Attachment Q-set for assessing infant secure base behavior and Pederson and Moran’s (1995) Maternal Behavior Q-set for assessing maternal sensitivity in naturalistic settings have been very useful in this regard. The items in these Q-sets provide considerable information about the level of detail at which the “action” takes place and the role context plays in the meaning of key behaviors. Nonetheless, a Q-set cannot fully replace working with an experienced mentor. A set of first-rate video recordings that effectively capture the experience of making 60—90 minute observations in a variety of contexts would significantly advance training opportunities. However, even with current video recording equipment, capturing, editing, and providing commentary across the full range of individual differences would be a significant undertaking.

It is a significant limitation that many researchers continue to base assessments on observations that are simply too few, too brief, and sample too few contexts to provide a representative sample of a caregiver’s typical behavior. Although it may not be necessary to obtain the full 12—16 hours of observation that were collected each quarter in the Baltimore study, single observations of brief interactions in one or two contexts (e.g. feeding and free play) are unlikely to provide reliable estimates of infants’ or mothers’ typical behavior. Waters (1978) illustrated the use of traditional psychometric methods for assessing the duration of observation intervals necessary to reliably assess individual differences in infant and maternal behaviors and there have been many relevant developments in reliability assessment since then. These methods should be used more often in designing observation schedules and evaluating the representativeness of observational data.

It is also important to take into account the diversity of samples used in current research. The Baltimore sample was very homogeneous. Moreover, very few of the mothers worked outside the home or made use of extensive substitute care during the first year. Today, substitute care arrangements are very common and very diverse with respect to frequency, duration, and quality. It is very useful that attachment research has branched out to include a wide range of cultural and at-risk samples. After all, cross-community, cross-cultural, and clinical relevance have always been central to the translational goals of attachment research. At the same time, attachment researchers have not always been attentive enough to the measurement implications of sample characteristics. Low correlations between facets of care and infant security in families very different from the Baltimore sample deserve more probing analysis than simply dismissing them as failures of replication. Sometimes a low correlation is the answer to an interesting new question.

Finally, the strength of replication results is going to be related to how tightly a prediction is tied to attachment theory in the first place. Researchers who are new to attachment study or not yet thoroughly versed in attachment theory have often designed studies that seem less tied to predictions from attachment theory than to the hypothesis that “all good things go together” (Waters, Corcoran, & Anafarta, 2005). When such findings are significant in an initial study it is likely that they are either due to spurious influences or have simply occurred by chance. In replication studies they are likely to be attenuated or to disappear altogether. This hardly counts as a replication failure. Indeed, it is useful that such results tend to fall away because they only represent noise in efforts to integrate and sythesize accumulating results.

The evolution of measurement instruments and methods plays an important role in driving science forward. When new instruments and methods become available, early adopters play a useful role by investigating their relevance in new domains. However, instrument-driven research is not a substitute for theory-driven hypothesis testing. Indeed, it often becomes little more than an empty search for statistically (if not theoretically) significant results. Attachment study is not immune to this. The problem with significant but theoretically tenuous correlations is that they eventually accumulate to the point that they cannot be interpreted within any sensible attachment theory framework. Sometimes even significant correlations can be too much of a good thing. Progress in attachment study depends on research that is theory-driven, not instrument-driven. With occasional lapses, attachment researchers have managed this rather well. This is important because theoretical reviews and meta-analyses depend on the quality of the underlying research. The Baltimore study remains a valuable model not only for its results but also for the level of expertise and effort underlying its results.

The Baltimore study entailed a level of craftsmanship that does not easily scale to large samples. At the same time, larger samples and more diverse samples open doors that the Baltimore study could not. In the end, the validity of the SSP rests broadly on its links to secure base behavior outside the laboratory, other theoretically based external correlates, and discriminant validity vis à vis alternative interpretations. Overall, the SSP has earned its reputation as a measure of an infant’s confidence in a particular caregiver’s accessibility and responsiveness and its ability to use her (or him) as a secure base from which to explore and as a haven of safety and comfort when required. Nonetheless, it is important to confirm that the SSP is providing a window on secure base behavior at home whenever it is first used in any sample markedly different from the healthy, home reared, one-year-olds observed in the Baltimore study. This step has been overlooked in far too many cross-cultural samples, studies of new classification schemes, and studies of the SSP beyond infancy. It should become standard procedure.

Of course, the results of such studies will not always be positive (e.g., Posada, 2006). The fact that the SSP cannot stand in place of home observations in every instance is problematic for individual studies. An SSP that is not correlated with relevant home behavior would likely have other interesting correlates. However, these correlates should not be interpreted as correlates of secure base behavior. Importantly, this would not present a challenge to attachment theory per se. Why? Because infant attachment theory is not a theory about the SSP. Both the evolutionary rationale, the concept of biases in infant learning abilities, and the hypotheses about origins in the early caregiving environment address the behavior of human infants in naturalistic settings. The theory stands or falls on these issues, not on the validity of a particular test. Inevitably, there will be some cultures, age groups, or special populations in which the SSP is not significantly correlated with secure base behavior at home. In these contexts we can always base our assessments on observations in the home and other naturalistic settings.

Disorganized Attachment

The ABC classification system was always intended to be open to extension to capture newly noticed behavior patterns and data from new populations. Even during the Baltimore project, Ainsworth and her students noticed reunion responses in the SSP that were not fully comprehended by the initial classification scheme. For example, the B4 pattern was not seen in the first 23 mother—infant dyads (subsample 1) of the Baltimore project.

By far the most influential addition to the ABC classification system has been Main and Solomon’s (1986, 1990) discovery of a group of infants who were initially designated unclassified and are now classified Disorganized (Group D). The hallmark of the D group is a diverse array of odd, fearful, disjointed, contradictory and seemingly inexplicable behavioral responses exhibited by infants to the caregiver in SSP reunion episodes. In the Baltimore study these behaviors were infrequent and too difficult to comprehend to suggest a new classification. However, such behaviors were quite common in studies of maltreated and high-risk samples and clearly warranted theoretical analysis, a distinct classification, and prospective studies of antecedents and sequelae (e.g., Carlson, Cicchetti, Barnett, & Braunwald, 1989). Much of this work is summarized in Lyons-Ruth and Jacobvitz’s (2008) Handbook of Attachment chapter and in Solomon and George’s (2011) volume devoted to disorganized attachment and caregiving.

Attachment Is a Relationship, Not a Trait

Attachment theory and the Baltimore project were designed to shed light on a child’s first and most lasting relationship. For almost four decades, discoveries from SSP research have guided the development of attachment theory and helped clarify the place of attachment in social and emotional development. The fact that SSP classifications with mother and father are often different makes the important point that the procedure primarily assesses individual relationships, not general relationship styles (see Sroufe, 1985). Avoidance and resistance are test behaviors observed primarily in the SSP reunion episodes. Such behaviors are rarely observed in non-test settings. Their value for assessment is based on links to the secure base organization of behavior at home, not discrete behaviors that seem phenotypically similar to avoidance or resistance in the SSP. Indeed, Sroufe, Fox, and Pancake (1983) have reported that infants who were avoidant or resistant in the SSP, were, paradoxically, overly dependent on their preschool teachers when they were observed at 47—60 months of age (p. 1625).

The idea that qualities of first relationships eventually become or moderate trait-like individual differences (e.g., Bowlby, 1988) is intriguing and finds some support in personal and clinical experience. However, there is little research support for generalizing from specific attachment-related behaviors to trait-like consistencies across context, behavioral domains, or age. Moreover, as Rutter (1995) has pointed out, it is not clear what kinds of processes could lead to such outcomes. Bowlby-Ainsworth attachment theory is built on the recognition that, even in infancy, attachment behavior is sensitive, adaptive, and coherent across context and age. The limitations of trait constructs became evident as soon as developmentalists recognized the meaning, complexity, and coherence of attachment behavior. A satisfactory descriptive/explanatory framework required an entirely new paradigm that drew concepts from cognitive psychology, ethology, control systems, and evolutionary theory.

Despite these caveats, it seems likely that the use of avoidant, resistant, disorganized, etc. as descriptors will persist in the attachment literature and in informal discussion. Although this is often convenient, it is important to keep in mind that these are merely labels. They should not be reified and their verbal associates are not a sound basis for drawing inferences or generalizations. Although humans are comfortable thinking in terms of traits and types, truly trait-like consistency is relatively uncommon. Moreover, as Wiggins (1997) has emphasized, traits label and summarize behavior. They do not explain it. If avoidance, resistance, and disorganized behavior were strongly trait-like across situations and age, the challenge would be to explain why. If we allow the charm of interesting labels to undermine clear thinking and problem formulations or to suggest magical explanations, we risk losing the key descriptive and theoretical insights underlying attachment theory.


John Bowlby’s and Mary Ainsworth’s developmental attachment theory is one of the landmark accomplishments of 20th century social and behavioral sciences. It has generated a wealth of empirically testable hypotheses and innovations in assessment methods. Research has supported the key hypotheses and these results have fared well in replications across a wide range of communities and cultures. Moreover, Ainsworth’s conceptualizations of maternal care and interaction and the secure base phenomenon have provided a useful framework for research on attachment relationships well beyond infancy (e.g., Ainsworth, 1989; Crowell et al., 2002; Waters & Waters, 2006). Although attachment theorists and researchers need to be vigilant about keeping the secure base concept at the center of theory and assessment, the theory remains a rich source of new insights about relationships and development. It is also beginning to realize Bowlby’s and Ainsworth’s goal of having significant impact on prevention and intervention (e.g., Berlin, Zeanah, & Lieberman, 2008; Atkinson & Goldberg, 2004; Marvin, Cooper, Hoffman, & Powell, 2002). We are delighted that Taylor & Francis have selected Patterns of Attachment for inclusion in the Classic Editions series. Patterns of Attachment is indeed a classic and deserves to remain widely available as a resource and a model for new generations of attachment researchers.

Everett Waters, Ph.D., Professor, Department of Psychology, State University of New York at Stony Brook

Inge Bretherton, Ph.D., Professor Emerita, Department of Human Development and Family Studies, University of Wisconsin-Madison

Brian E. Vaughn, Ph.D., Professor, Human Development and Family Studies, Auburn University


The authors wish to express their appreciation to Georgette Enriquez for help in designing this reissue. We also thank Professor Harriet Waters for several rounds of insightful comments that greatly improved drafts of this manuscript. This work was supported in part by the New York Attachment Consortium and the Center for Mental Health Promotion.


Ainsworth, M. D. S. (1967). Infancy in Uganda: Infant care and the growth of love. Baltimore: Johns Hopkins University Press.

Ainsworth, M. D. S. (1983). Mary D. Salter Ainsworth. In A. N. O’Connell & N. F. Russo (Eds). Models of achievement: Reflections of eminent women in psychology. New York: Columbia University Press (pp. 200—219).

Ainsworth, M. D. S. (1989). Attachments beyond infancy. American Psychologist, 44, 709—716.

Ainsworth, M. D. S., Blehar, M., Waters, E., & Wall, S. (1978). Patterns of attachment: A psychological study of the Strange Situation. Hillsdale, NJ: Lawrence Erlbaum.

Ainsworth, M. & Bowlby, J. (1991). An ethological approach to personality development. American Psychologist, 46, 333—341.

Arsenian, J. M. (1943). Young children in an insecure situation. Journal of Abnormal and Social Psychology, 38, 225—249.

Atkinson, L. & Goldberg, S. (Eds) (2004). Clinical applications of attachment. Mahwah, NJ: Lawrence Erlbaum.

Berlin, L. J., Zeanah, C. H., & Lieberman, A. F. (2008). Prevention and intervention programs for supporting early attachment security. In J. Cassidy & P. Shaver (Eds) Handbook of attachment (2nd ed.). New York: Guilford Press (pp. 745—761).

Bernier, A., Matte-Gagne, C., Belanger, M., & Whipple, N. (2014). Taking stock of two decades of attachment transmission gap: Broadening the assessment of maternal behavior. Child Development, 85, 1852—1865.

Bischof, N. (1975). A systems approach towards the functional connections of attachment and fear. Child Development, 46, 801—817.

Block, J. (1971). Lives through time. Hillsdale, NJ: Lawrence Erlbaum.

Bowlby, J. (1958). The nature of the child’s tie to his mother. International Journal of Psycho-Analysis, 39(5), 350—373.

Bowlby, J. (1969, 2nd ed. 1982). Attachment and loss (vol. 1), Attachment. New York: Basic Books.

Bowlby, J. (1973). Attachment and loss (vol. 2), Separation. New York: Basic Books.

Bowlby, J. (1988). A secure base: Parent—child attachment and healthy human development. New York: Basic Books

Bretherton, I. (1991). The roots and growing points of attachment theory. In C. M. Parkes, J. Stevenson-Hinde, & P. Marris (Eds) Attachment across the life cycle. London: Tavistock/Routledge (pp. 9—32).

Bretherton, I. (1992). The origins of attachment theory: John Bowlby and Mary Ainsworth. Developmental Psychology, 28, 759—775

Bretherton, I. (2003). Mary Ainsworth: Insightful observer and courageous theoretician. In G. A. Kimble & M. Wertheimer (Eds), Portraits of pioneers in psychology (vol. 5). Washington, D.C.: American Psychological Association.

Bretherton, I. (2013). Revisiting Mary Ainsworth’s conceptualization and assessments of maternal sensitivity—insensitivity. Attachment and Human Development, 15(5—6), 460—484.

Carlson, V., Cicchetti, D., Barnett, D., & Braunwald, K. (1989). Disorganized/disoriented attachment relationships in maltreated infants. Developmental Psychology, 25, 525—531.

Chomsky, N. (1957). Syntactic structures. The Hague: Mouton de Gruyter.

Crowell, J. A., Treboux, D., Gao, Y., Fyffe, C., Pan, H., & Waters, E. (2002). Assessing secure base behavior in adulthood: Development of a measure, links to adult attachment representations, and relations to couples’ communication and reports of relationships. Developmental Psychology, 38, 679—693.

Dixon, W. E. (2002). Twenty studies that revolutionized child psychology. Saddle River, NJ: Prentice-Hall.

Dixon, W. E., Jr. (2016). Twenty studies that revolutionized child psychology (2nd ed.). Saddle River, NJ: Pearson.

Grossmann, K. E., Bretherton, I., Waters, E., & Grossmann, K. (Guest Eds) (2013). Maternal sensitivity: Papers in honor of Mary Ainsworth’s 100th Year. Attachment and Human Development, 15(5—6).

Grossmann, K. E., Grossmann, K., & Waters, E. (2005). Attachment from infancy to maturity: The major longitudinal studies. New York: Guilford Press.

Hay, D. F. (1980). Multiple functions of proximity seeking behavior. Child Development, 51, 636—645.

Karen, R. (1998). Becoming attached. New York: Oxford University Press.

Kuhn, T. (1962/2012). The structure of scientific revolutions, 4th ed. Chicago: University of Chicago Press.

Lyons-Ruth, K. & Jacobvitz, D. (2008). Attachment disorganization: Genetic factors, parenting contexts, and developmental transformation from infancy to adulthood. In J. Cassidy & P. R. Shaver (Eds), Handbook of attachment theory and research, 2nd ed. New York: Guilford Press (pp. 666—697).

Main, M. & Solomon, J. (1986). Discovery of a new, insecure-disorganized/disoriented attachment pattern. In T. B. Brazelton & M. Yogman (Eds), Affective development in infancy. Norwood, New Jersey: Ablex (pp. 95—124).

Main, M., & Solomon, J. (1990). Procedures for identifying infants as disorganized/disoriented during the Ainsworth Strange Situation. In M. T. Greenberg, D. Cicchetti, & E. M. Cummings (Eds), Attachment in the preschool years. Chicago: University of Chicago Press (pp. 121—160).

Marvin, R., Cooper, G., Hoffman, K., & Powell, B. (2002). The circle of security project: Attachment-based intervention with caregiver—preschool child dyads. Attachment and Human Development, 1(4), 107—124.

Masterman, M. (1970). What is a paradigm? In I. Lakatos & A. Musgrave (Eds), Criticism and the growth of knowledge. Cambridge, UK: Cambridge University Press.

Pederson, D. & Moran, G. (1995). Appendix B: Maternal behavior Q-set. Monographs of the Society for Research in Child Development, 60(2—3), 247—254.

Petters, D. (2006). Implementing a theory of attachment: A simulation of the strange situation with autonomous agents. In Proceedings of the Seventh International Conference on Cognitive Modelling. Trieste: Edizioni Golardiche (pp. 226—231).

Petters, D., Waters, E., & Schönbrot, F. (in press, 2015). Modeling and simulating attachment H. Waters. Measuring attachment. New York: Guilford.

Petters, D., Waters, E. & Sloman, A. (2011). Modeling machines that can love: From Bowlby’s attachment control system to requirements for romantic robots. Emotion Researcher, 26(2), 5—7.

Piaget, J. (1936/1952). The origins of intelligence in children. London: Routledge and Kegan Paul.

Posada, G. (2006). Assessing attachment security at age three: Q-sort observations and the MacArthur Strange Situation adaptation. Social Development, 15(4), 644—658.

Posada, G., Carbonell, O. A., Alzate, G., & Plata, S. (2004). Through Colombian lenses: Ethnographic and conventional analyses of maternal care and their associations with secure base behavior. Developmental Psychology, 40(4), 508—518.

Rutter, M. (1995). Clinical implications of attachment concepts: Retrospect and prospect. Journal of Child Psychology and Psychiatry, 36, 549—571.

Salter, M. D. (1940). An evaluation of adjustment based upon the concept of security. University of Toronto Studies: Child Development Series No. 18. Toronto: University of Toronto Press.

Schaffer, H. R. & Emerson, P. E. (1964). The development of social attachment in infancy. Monographs of the Society for Research in Child Development, 29(3), serial no. 94.

Shirley, M. M. (1942). Children’s adjustments to a strange situation. Journal of Abnormal and Social Psychology, 37, 201—217.

Shirley, M. & Poyntz, L. (1941). The influence of separation from the mother on children’s emotional responses. Journal of Psychology, 12, 251—282.

Simon, H. A. (1956). Rational choice and the structure of the environment. Psychological Review, 63(2), 129—138. doi:10.1037/h0042769.

Solomon, J. & George, C. (2011). Disorganized attachment and caregiving. New York: Guilford.

Spitz, R. (1965). The first year of life. New York: International Universities Press.

Sroufe, L. A. (1985). Attachment classification from the perspective of infant—caregiver relationships and infant temperament. Child Development, 56, 1—14.

Sroufe, L. A., Egeland, B., Carlson, E., & Collins, W. A. (2005). The development of the person: The Minnesota study of risk and adaptation from birth to adulthood. New York: Guilford.

Sroufe, L. A., Fox, N., & Pancake, V. (1983). Attachment and dependency in developmental perspective. Child Development, 54(6), 1615—1627.

Sroufe, L. A. & Waters, E. (1977). Attachment as an organizational construct. Child Development, 48, 1184—1199.

Sroufe, L. A., Waters, E., & Matas, L. (1974). Contextual determinants of affective expression in infancy. In M. Lewis & L. Rosenblum (Eds), The origins of fear. New York: John Wiley & Sons.

Suomi, S. J., van der Horst, F. C. P., & van der Veer, R. (2008). Rigorous experiments on monkey love: An account of Harry F. Harlow’s role in the history of attachment theory. Integrative Psychological and Behavioral Science, 42, 354—369.

Thomas, A. & Chess, S. (1977). Temperament and development. Oxford: Brunner/Mazel.

van Dijken, S., van der Veer, R., van Ijzendoorn, M., & Kuipers, H. (1998). Bowlby before Bowlby: The sources of an intellectual departure in psychoanalysis and psychology. Journal of the History of the Behavioral Sciences, 34(3), 247—269.

van IJzendoorn, M. H., Vereijken, C. M. J. L., Bakermans-Kranenburg, M. J., & Riksen-Walraven, J. M. (2004). Assessing attachment security with the Attachment Q Sort: Meta-analytic evidence for the validity of the observer AQS. Child Development, 75(4), 1188—1213.

Vaughn, B. & Waters, E. (1990). Attachment behavior at home and in the laboratory: Q-sort observations and Strange Situation classifications of one-year-olds. Child Development, 61, 1965—1973.

Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. In M. Cole, V. John-Steiner, S Scribner, & E. Souberman (Eds) Cambridge, MA: Harvard University Press.

Waters, E. (1978). The reliability and stability of individual differences in infant—mother attachment. Child Development, 49, 483—494.

Waters, E. (2008). Live long and prosper: A note on attachment and evolution. Retrieved online August, 2014.

Waters, E., Corcoran, D., & Anafarta, M. (2005). Attachment, other relationships, and the theory that all good things go together. Human Development, 48, 80—84.

Waters, E. & Cummings, M. (2000). A secure base from which to explore relationships. Child Development, 71, 164—172.

Waters, E. & Deane, K. (1985). Defining and assessing individual differences in attachment relationships: Q-methodology and the organization of behavior in infancy and early childhood. In I. Bretherton & E. Waters (Eds), Monographs of the Society for Research in Child Development, 50(1—2), 41—65.

Waters, H. & Waters, E. (2006). The attachment working models concept: Among other things, we build script-like representations of secure base experiences. Attachment and Human Development, 8, 185—197.

Wiggins, J. S. (1997). In defense of traits. In R. Hogan, J. Johnson & S. R. Briggs (Eds), Handbook of personality psychology, San Diego: Academic Press. (pp. 97—117).

Preface (1978)

This book is about the attachment of infants to their mother figures. In it we focus on how infant behavior is patterned. We approach this patterning in two main ways. First, we examine the way in which a baby’s behavior is patterned when the attachment system is activated at varying levels of intensity through simple manipulations of his environment in a laboratory situation, which we have called the “strange situation.” When examining the baby’s responses to controlled environmental changes, we observe the way in which his or her attachment behavior interacts with other behavioral systems that are also activated at varying levels of intensity and that may either compete or conflict with attachment behavior or augment the intensity with which attachment behavior is manifested. Second, we identify certain important individual differences in the way in which behavior is patterned—both attachment behavior and behavior antithetical to it—and seek to understand how such differences may have arisen and how different patterns of attachment may influence development.

We undertook writing this book in order to present the information about infant—mother attachment that we had gained through the use of a standard laboratory situation and to compare the manifestations of attachment in that situation with manifestations of attachment observed at home. We also wished to review the findings of other investigations of attachment, especially those that are directly comparable with ours because of their use of our strange-situation procedure, and to compare their findings with ours, including the findings of investigations that studied children older than the 1-year-olds upon which our work focuses and those that are concerned with an infant’s attachment to figures other than the mother. We report much empirical detail, which will be of interest to all those who investigate a young child’s early interpersonal relations. The empirical detail leads, however, to a discussion of theoretical issues of major significance. Implicit in both the empirical findings and in the theoretical discussions are clues both to the understanding of developmental anomalies and to ways in which such anomalies might be prevented, assuming the feasibility of early intervention in families in which new babies are expected or have recently arrived. Therefore, we believe that this volume will be of interest not only to those concerned with theory and research into early social development, but also to diverse classes of persons concerned with the practical job of providing better infant care and facilitating optimal development in young children.

It seems suitable in this preface to introduce the reader to the strange situation and to describe how we happened to use it and why we judged the findings stemming from its use to be of sufficient significance to focus a book on them. The “strange situation” was the label assigned by Ainsworth and Wittig (1969) to a standardized laboratory procedure in which several episodes, in fixed order, were intended to activate and/or intensify infants’ attachment behavior. These episodes were designed to approximate situations that most infants commonly encounter in real life. The adjective “strange” denotes “unfamiliar,” rather than “odd” or “peculiar”; it was used because fear of the unfamiliar is commonly referred to as “fear of the strange” (e.g., Hebb, 1946). All of the instigations to attachment behavior used in the strange situation involved unfamiliarity.

The strange situation was originally devised in 1964 for use in conjunction with an intensive longitudinal study of the development of infant—mother attachment throughout the first year of life, a naturalistic study in which infants were observed in their familiar home environments. This study of 26 mother—infant pairs living in the Baltimore area had been preceded by a comparable but less intensive study of 28 dyads living in country villages in Uganda (Ainsworth, 1967). Despite many similarities between the two samples in regard to attachment behavior, three behavioral patterns that had been highlighted in the Ganda study emerged less strikingly in the American study: the use of the mother as a secure base from which to explore; distress in brief, everyday separations from the mother; and fear when encountering a stranger. Perhaps if stronger instigation were provided, the American babies might be induced to behave in much the same ways as had the Ganda infants. In the belief that these behaviors might be evoked more incisively in an unfamiliar situation than in the familiar home environment, the strange situation was devised.

First, let us consider the use by an infant of his mother as a secure base from which to explore the world. One of us (Salter, 1940) had long been interested in the hypothesis, originally formulated by Blatz,1 that a young child who had gained security in his relationship with his parents was emboldened thereby to strike out to explore the world, willing to risk the insecurity initially implicit in a learning situation because he could rely on his parents to be available, responsive, protective, and reassuring. If his adventure evoked undue anxiety, the child could easily return to “home base,” in the expectation that his parents would provide the reassurance he needed. If, on the other hand, his relationship with his parents was insecure, then he might not dare to leave them to explore, not trusting them to remain available to him if he left or to be responsive when he needed them. Lacking trust, he would stick close to his base, fearing to risk the anxiety implicit in exploration and learning. This hypothesis was confirmed in the Ganda study (Ainsworth, 1963, 1967). Infants who were judged to be securely attached to their mothers explored actively while their mothers conversed with the observers, and indeed they might well leave the room or even the house in order to extend their exploratory activities. Yet most of these same infants were acutely distressed and ceased exploration if it were the mother who left them. By contrast, infants who were judged to be anxiously attached tended to remain close to the mother, perhaps clinging to her and exploring little or not at all.

In the course of the longitudinal study of Baltimore infants, however, nearly all babies left their mothers to explore the familiar home environment (Ainsworth, Bell, & Stayton, 1971), whether or not they were judged to be secure in their attachments to their mothers (Stayton & Ainsworth, 1973). Perhaps individual differences could be discriminated in an unfamiliar environment that might hence be expected to provide stronger instigation to attachment behavior.2 Perhaps those who were anxiously attached to their mothers might be unwilling to explore when placed in an unfamiliar situation, whereas those who were securely attached would explore even a strange situation with the mother present.

Antedating our strange situation was Arsenian’s study (1943) of young children in an “insecure” situation and Harlow’s (1961) work with rhesus infants in an open-field situation. Both studies showed the effectiveness of the mother or mother surrogate in providing security for exploration. Subsequently, several studies of infants with and without their mothers in unfamiliar situations have provided clear-cut confirmation of the hypothesis that infants and young children tend to explore an unfamiliar environment in the mother’s presence, but slow down or cease exploration in her absence (e.g., Cox & Campbell, 1968; Rheingold, 1969), although infants will indeed leave their mothers on their own initiative in order to explore (Rheingold & Eckerman, 1970). The present study not only adds further evidence of these normative tendencies, but also throws light on individual differences in maintaining exploration under conditions that also activate attachment behavior.

Second, distress upon being separated from the mother has long been conceived as an indication that an infant has become attached to her (e.g., Schaffer & Emerson, 1964). Our longitudinal study of Baltimore infants showed, however, that the average baby did not consistently protest his mother’s departure in the familiar home environment (Stayton, Ainsworth, & Main, 1973). Indeed some babies, who, by other behavioral criteria, were clearly attached to their mothers, showed very infrequent separation distress. The same finding had been noted in the case of Ganda infants (Ainsworth, 1963, 1967), but nevertheless the latter more frequently protested separation in a familiar environment than did the Baltimore babies. On the other hand, it is well known that, once attached to a mother figure, infants and young children tend strongly to protest being separated against their will and placed in an unfamiliar environment for any substantial length of time (e.g., Bowlby, 1953; Heinicke & Westheimer, 1966; Schaffer & Callender, 1959; Yarrow, 1967). Therefore it was of interest to subject the infants in the longitudinal sample to very brief separation experiences in an unfamiliar environment in order to compare their responses with similar minor separations in the home environment. It was expected that most would be distressed by separation in the strange situation, even though they might be infrequently distressed by little separations at home.

Third, it was of interest to observe infants’ responses to a stranger in an unfamiliar environment. Although Spitz (e.g., 1965) maintained that fear of strangers (i.e., 8-month anxiety) was a milestone in normal development and a criterion that an infant had achieved “true object relations,” and although Ganda infants (Ainsworth, 1967) had been observed to be conspicuously afraid of strangers toward the end of the first year, the Baltimore babies did not consistently show such fear in the familiar environment of the home. Therefore it was of interest to see whether the context of an unfamiliar environment would heighten their fear of strangers.

The structure of the strange situation followed from these lines of hypothesis and interest. Exploratory behavior was to be observed both in the mother’s presence and in her absence. The infant’s response to a stranger was likewise to be observed both in the mother’s presence and in her absence. His response to his mother’s absence was to be seen both when he was alone and when he was left with a stranger. His response to his mother’s return after an absence was to be compared with his response to the return of the stranger after an absence. The episodes of the strange situation, which are described in detail in Chapter 2, followed from these considerations.

The 1-year-old, accompanied by his mother, was introduced to an unfamiliar but otherwise unalarming playroom where massive instigation to exploratory behavior was provided by a large array of toys. In the next episode, an adult stranger entered, who was tactful but nevertheless unfamiliar. Then came a brief separation episode in which the mother left the baby with the stranger. Then after an episode of reunion with the mother, there was a second separation in which the baby was first alone in the unfamiliar environment and then again with the stranger, who returned before the mother reentered. Because it was anticipated that experience in each episode would affect behavior in the next episode, the instigation to attachment behavior expected to be the weakest was placed at the beginning and that expected to be strongest toward the end. The expectations that these mild instigations would be cumulative in their effect were fulfilled.

It must be emphasized that the strange situation does not constitute an experiment in the literal meaning of this term. Different groups of subjects were not assigned to different treatments in order to ascertain the relative effect of these treatments on some dependent behavioral variable. Nor was it our intent to assess the relative effects of the different kinds of instigation upon intensity of attachment behavior—an intent that would have demanded control of order effects. On the contrary, the strange situation was designed as a controlled laboratory procedure in which individual differences among infants could be highlighted, precisely because they were exposed to the same situation with the same episodes in the same order.

The findings that have emerged from the use of this procedure have indeed highlighted individual differences in the way infants respond to an accumulation of instigations to attachment behavior. Different patterns of strange-situation behavior, we propose, indicate differences in the way infant—mother attachment has become organized. We have observed the same patterns in four separate samples of 1-year-olds, and other investigators who have used our techniques for the identification of patterns of attachment have confirmed our findings. Just because the procedure provides increasingly strong instigation to attachment behavior through its cumulative nature, one may observe in a relatively short span of time attachment behavior under conditions of activation from relatively weak to very strong. In the familiar home environment, occasions for strong activation of attachment behavior are infrequent, so that it requires many hours of observation to encompass a similar range, especially in the case of a healthy infant reared in a social environment that is sensitively responsive to him.

Nevertheless, in our longitudinal study that provided for approxmiately 72 hours of observation of each infant throughout the first year, it was possible to observe patterns of attachment and, further, to relate these to patterns of maternal behavior. For the sample of infants thus longitudinally observed, it was possible to examine continuities and discontinuities of specific behaviors between the home and laboratory environments; more important, these two sets of data enable one to perceive the patterning or organization of behaviors that reflects continuity of an attachment of a distinctive nature, despite discontinuities in specific behaviors.

Consequently, the findings reported in this volume go far beyond the specific issues that the strange situation was initially designed to investigate. They throw light upon qualitative differences in the nature of the attachment relationship itself, and, in conjunction with longitudinal data provided both by ourselves and by other investigators, they also yield hypotheses of how such qualitiative differences arose and how they exert an influence on subsequent development.

To anticipate a more detailed report of our findings, we can note that the episodes of the strange situation that made the most significant contribution to the identification of patterns of attachment were the reunion episodes—those in which the mother rejoined the baby after having been away for some minutes. This comes as a surprise to some who may have assumed that responses during the separation episodes—the episodes during which the instigation to attachment behavior might be assumed to be strongest—would be most significant. To us it was not surprising. The entire separation literature (cf., Ainsworth, 1962) suggests that the response to reunion after separation may well yield a clearer picture of the state of attachment than did the response to separation itself. After a relatively brief separation—lasting a few days or even a few weeks—it is common to observe a great intensification of attachment behavior upon reunion. The child seeks to be in close bodily contact with his attachment figure and also seeks to maintain close proximity over much longer periods than was previously characteristic of him. It seems that separation has shaken his trust in the mother’s accessibility and responsiveness, so that he scarcely dares to let her out of sight lest she disappear again. Furthermore, he may be more ambivalent toward her than previously. It seems that the angry feelings aroused during the separation, when he felt abandoned, are not altogether dissipated upon reunion, but mingle or alternate with his desire for renewed contact, so that he both rejects and seeks to be close to his attachment figure.

Furthermore, a child may respond to separation, especially to a long and depriving separation, with “detachment” behavior, which gives the impression that he is indifferent to the whereabouts and behavior of his attachment figure. In fact, however, detachment seems likely to be a product of intense conflict between attachment behavior activated at high levels of intensity and avoidant behavior evoked by the seeming rejection implicit in the failure of the attachment figure to respond to him during the separation. This detachment behavior, like angry rejecting behavior, is not likely to vanish immediately upon reunion. On the contrary, it may be strengthened by the high-intensity activation of attachment behavior occasioned by reunion. Consequently a child may seem not to recognize his mother or may seem indifferent to her for a period of time after reunion and before intensified attachment behavior overtly reasserts itself.

Although one might expect to find these various reunion behaviors—whether they be intensified attachment behavior, angry resistance, or avoidant detachment—to be less conspicuous and/or less prolonged after the brief separations implicit in our strange situation, nevertheless it seemed reasonable to us to be alert for responses, similar in kind if not in degree, in the reunion episodes. Furthermore, because the strange-situation separations were so brief, it makes sense to suppose that individual differences in reunion behaviors reflect characteristics of the infant’s attachment relationship to his mother—characteristics that were consolidated long before the strange situation was first encountered.

The final task of this preface is briefly to outline the structure of this volume. But before proceeding to that task, one further point is most suitably discussed here. The strange situation is admittedly somewhat stressful. Some have suggested that it is unjustifiably stressful. We must disagree. We would not have subjected over 100 infants to an unduly stressful procedure. We designed the situation to approximate the kind of experiences that an infant in our society commonly encounters in real life. All American mothers whom we have encountered do not hesitate to take their babies at least occasionally into unfamiliar environments—for example, to visit an adult friend unfamiliar to the baby or, less commonly, to take him to a day-care center, to a babysitter’s home, or to a play group. While they are in this unfamiliar (but not otherwise alarming) environment, the mother may leave her baby for a few minutes—either alone or with a stranger—whether to accompany her hostess to another room, to go to the telephone, or to visit the bathroom. The strange situation was modeled on such common real-life experiences.

None of the mothers in any of our four samples came to the laboratory without having been informed in detail of every step in the procedure, how we expected a range of babies to respond, and why we had designed the episodes in the way that we had. Nearly all mothers that we approached agreed to participate with their babies; only one did so with any apparent misgivings, and she was the one mother in our longitudinal sample who had a full-time job and whose baby had begun to react negatively to her daily departures and returns. We emphasized that any episode could be curtailed if a baby became unduly distressed, but it was we who nearly always initiated a curtailment, while the mother showed no concern.

After the strange situation was over, we always spent substantial time with the mother and baby, giving the mother an opportunity to discuss the baby’s reactions if she wished, but in any case offering an occasion for pleasant social interaction. In no case did we observe any continuing distress or any adverse effects attributable to the strange situation, and in the case of our longitudinal sample this was so in a follow-up visit three weeks later. Indeed we were soon convinced that we were far more concerned about the anxiety that might have been associated with the brief separation experiences implicit in the strange-situation procedure than were the parents—who had little or no compunction about imposing much longer separations on their babies, often under less than optimum conditions.

Nevertheless we acknowledge that the strange-situation procedure might not approximate common experiences of infants who are reared differently, whether in other societies or by atypical parents in our own society; and we cast no aspersions by our term “atypical,” for these may be highly sensitive parents who avoid all unnecessary occasions for separation. It seems entirely likely that Ainsworth’s (1967) Ganda infants and Konner’s (1972) Bushman babies could not have tolerated the strange situation. Recently Takahashi (personal communication) informed us that the Japanese mothers of her sample would not consent to leaving their babies alone in an unfamiliar situation, although they did not object to leaving them with a stranger. The strange situation surely should not be imposed on a baby whose parents are reluctant to cooperate, especially if they have reason to expect that he would be especially disturbed either by separation or by encountering a stranger. For all but a few infants in our middle-class society, however, we are convinced that there is no uncommon stress implicit in the strange-situation procedure, and we are even more convinced that the scientific yield of the strange-situation procedure has been great indeed.

Now let us introduce the reader to the rest of this volume. Chapter 1 deals with the theoretical background that underlies our research. It is necessary in order to follow our interpretations of the findings. Those who are thoroughly conversant with ethological—evolutionary attachment theory (e.g., Ainsworth, 1969, 1972; Bowlby, 1969, 1973) will perhaps find little new in Chapter 1 and may wish to speed on to later chapters.

Part II deals with method. Chapter 2 introduces the reader to our total sample of 106 infants and presents the strange-situation procedure in the kind of detail necessary if others are to replicate it. Chapter 3 presents the behavioral measures we used in our data reduction. There are three types of assessment: (1) frequency measures of an ordinary kind, which are used chiefly to deal with “discrete” behaviors (specific behaviors considered separately from other behaviors); (2) special scoring of interactive behaviors (“categorical” measures that assume a degree of equivalence among goal-corrected behaviors with a common set-goal, and that thus themselves take behavioral patterning into account); and (3) classification of infants according to the patterns of behavior they displayed. Although the frequency measures are almost self-explanatory, the reader will need to become familiar with the categorical measures and with the classificatory system in order to follow our presentation of findings with understanding and ease.

Part III is concerned with results, both of our own strange-situation research and that of others who have used the strange-situation procedure with little or no modification. Chapter 4 contains a descriptive account of behavior in each episode of the strange situation. This analysis is ethologically inspired. It seemed desirable to provide this detailed account of strange-situation behavior before reducing the data to more manipulable behavioral measures. This account is prerequisite to the analysis of the activation and termination of specific behaviors, of changes in behaviors as the activation of the attachment system becomes more intense, and of the ways in which different attachment behaviors are alternative to each other and hence interchangeable to some extent. Chapter 5 is a normative account of behavioral changes across episodes of the strange situation. This analysis, reported previously for a smaller sample (Ainsworth, Bell, & Stayton, 1971), deals with the variations across episodes of the various behavioral measures. In a sense, it summarizes the detailed episode-by-episode analysis of Chapter 4.

Chapters 6, 7, and 8 deal with individual differences in strange-situation behavior. Chapter 6 is devoted to a multiple discriminant function analysis, which examines the reliability of the classificatory system that is our primary method of identifying patterns of attachment. Among other things, this analysis ascertains the extent to which the specifications for classification actually contribute to discriminating one classificatory group from the others. Chapters 7 and 8 focus on individual differences in our longitudinal sample, comparing strange-situation patterns with behaviors manifested at home during both the first and fourth quarters of the first year. Chapter 7 compares infant behavior at home with behavior in the strange situation. This analysis is highly pertinent to the issue of the stability of both attachment behaviors and patterns of attachment over time and across situations. It is also essential to the interpretation of strange-situation patterns as indicative of qualitative differences in the infant—mother attachment relationship. Chapter 8 examines the relationship of maternal behavior at home to infant behavior in the strange situation—an analysis that throws light upon the influence of individual differences in maternal behavior on individual differences in the quality of the attachment of infant to mother.

Chapters 9 and 10 are review chapters. Chapter 9 deals with the findings of other investigations of the behavior of 1-year-olds in the strange situation, whereas Chapter 10 is concerned with the behavior of children between 2 and 4. These important chapters extend the scope of our research. In most instances the findings reported therein confirm and extend our findings, although some studies, especially some of those dealing with older children, suggest limitations. Other studies yield apparent discrepancies between their findings and ours that seem best explained in terms of the use of different methods of appraisal.

We then return again specifically to a consideration of individual differences. Chapter 11 examines the stability of patterns of attachment and attachment behavior shown when the strange situation is repeated after varying lapses of time. Chapter 12 considers individual differences in patterns of behavior as they are more finely reflected in subgroup differences, over and above the way in which they are reflected in differences among the three main classificatory groups that were the theme of many of the findings reported in Chapters 6 through 11. These subgroups are too small for one to be able to meaningfully assess the statistical significance of the differences among them. Hence the reader who is interested in the general thrust of our argument rather than in possibly suggestive detail may wish to skip on to Part IV.

In Part IV the findings reported in Part III are discussed in the light of both theoretical considerations and other relevant findings reported in the research literature. Chapter 13 focuses on the discussion of the normative findings, which may now be better understood after our consideration of individual differences. Chapter 14 considers individual differences in the light of diverse theoretical paradigms—evolutionary—ethological attachment theory (summarized in Chapter 1) and two paradigms stemming from social-learning theory. Here we attempt to deal with some recent criticisms of attachment research and of the concept of attachment. It seems obvious to us that these criticisms are attributable to divergent paradigms, leading to research asking different questions, and conducted with procedures different from ours. Insofar as it is possible to make a bridge between divergent paradigms, we believe that the findings reported in this volume provide a definitive reply to the kind of criticisms made to date. Finally, Chapter 15 provides an interpretation of the patterns of attachment that have emerged as the most significant set of findings of our research, along with a discussion of some of the ways in which they seem likely to influence early development.


Tied as it has been to our longitudinal research into the development of infant—mother attachment, the body of findings reported in this volume has taken many years to amass, and we are indebted to many who have played significant roles in this endeavor. Our first debt of gratitude is to the Foundations’ Fund for Research in Psychiatry, which in 1962—63 awarded the grant, 62—244, that made it possible for this research to be launched. Since then, the research has been supported by USPHS grant RO1 HD 01712 and by grants from the Office of Child Development, the Grant Foundation, and the Spencer Foundation; this support is gratefully acknowledged. The appointment of the first author to the Center for Advanced Studies at the University of Virginia during 1975—77 finally facilitated the preparation of this volume.

Special appreciation goes to Barbara Wittig, who helped devise the strange situation and who carried out many of the original observations of infants both in the strange situation and at home. We also thank George D. Allyn and Robert S. Marvin, who also participated in data collection with the longitudinal sample, Sample 1. We are indebted to Silvia Bell and Mary Main for their independent work in collecting and analyzing data with Samples 2 and 4. Indeed we are deeply grateful to Mary Main both for the special data analysis to which she has given us access for this book and for her theoretical contribution, which has done much to forward our understanding of infants who show an anxious, avoidant pattern of attachment. Among those who played an active part in collecting and scoring Sample 3 data are Donelda Stayton, Larry Schutz, Thomas Pentz, Natalie Hirsch, and Inge Bretherton. Inge deserves special thanks for her excellent analysis of the total sample data in regard to behavior directed toward strangers. We thank the various gracious women who played the unrewarding role of the stranger in the strange situation. We cannot enumerate all those who helped with the analysis of the home-visit data for Sample 1, but we wish to thank them, together with others who helped with the analysis of the strange-situation data—including Elaine Jacobs, Rob Woodson, and Mark Greenberg. We are grateful for the excellent film records that William Hamilton made of the babies in Sample 3, and also for those made by Robert Marvin for some of Sample 1.

We are deeply indebted to the 106 mothers who, with their babies, participated in the strange situation. Most of them were motivated by a desire to support a study that aimed to extend our knowledge of early social development. We trust that their efforts will, in due course, result in some useful guidelines for mothers of young infants to facilitate the establishment of secure and harmonious attachment relationships. As for the 1-year-olds—who will probably not remember—we trust that their adventure in the strange situation and in social interaction afterwards will have been on the whole enjoyable.

Our debt to John Bowlby is great and many faceted. Not only is his formulation of attachment theory focal to the interpretation of our findings, but also his advice and encouragement have been vital throughout the several stages of this long project. In addition, although none of them can be held accountable for the final form of this volume, we were very much helped by John Bowlby, as well as by Robert Hinde and Mary Main, who read all or parts of earlier drafts and made cogent suggestions.

Finally, we wish to express our deep appreciation to those whose independent work with the strange situation has contributed to the literature reviewed in Chapters 9 and 10: Joyce Brookhart, Dante Cicchetti, David Connell, Shirley Feldman, Ellen Hock, Margaret Ingham, Michael Lamb, Alicia Lieberman, Sue Londerville, Eleanor Maccoby, Leah Matas, Saul Rosenberg, Felicissima Serafica, Lisa Tomasini, and Bill Tolan, as well as others acknowledged earlier for their contributions—Silvia Bell, Mary Main, Robert Marvin, and Thomas Pentz. We are indeed grateful for their response to our inquiries, for the unpublished material that many of them provided, and for their care in editing drafts of our review of their work.






1 MDSA first heard William Blatz speak of a child using his parents as a secure base from which to venture forth to learn when she was a student in his course at the University of Toronto in 1934—35. It was not until 30 years later (Blatz, 1966) that he explicitly published his “security theory.”

2 It now seems likely to us that the Ganda infants, being more afraid of strangers than the Baltimore infants were, found even the familiar home environment more stressful because of the presence of the visitor-observers, and that this highlighted individual differences in their use of the mother as a secure base from which to explore.