Origins: Bowlby's Break with Psychoanalysis
Attachment theory did not begin with a theory about personality types or relationship patterns. It began with a clinical observation: children who experienced prolonged separation from their caregivers suffered in ways that existing psychological theory could not adequately explain.
John Bowlby was a British psychiatrist and psychoanalyst who, in the early 1940s, began studying juvenile delinquents and homeless children at the London Child Guidance Clinic. His 1944 paper "Forty-Four Juvenile Thieves: Their Characters and Home Life," published in the International Journal of Psycho-Analysis, documented a striking pattern — a specific character type he called "affectionless psychopathy" appeared almost exclusively in children who had experienced prolonged maternal separation in early childhood. The observation was clinical and provisional, but it planted the question that would drive the next four decades of his work: what does early separation from a caregiver actually do to a child, and through what mechanism?
Bowlby's answer required breaking with psychoanalytic orthodoxy. The dominant psychoanalytic account of the child's bond to its mother was drive theory — the infant becomes attached to the mother because she satisfies hunger, the primary drive. Bowlby found this account inadequate on both theoretical and empirical grounds. Drawing on ethology — the biological study of animal behavior — particularly Konrad Lorenz's work on imprinting in ducklings and Harry Harlow's experiments with infant rhesus monkeys (who preferred a cloth surrogate mother to a wire surrogate that provided milk, demonstrating that contact comfort rather than feeding drives attachment), Bowlby proposed something more radical: that the attachment behavioral system was a primary motivational system in its own right, shaped by natural selection because infants who maintained proximity to a protective caregiver were more likely to survive. Attachment was not a derivative of feeding; it was an evolved system for proximity regulation under threat.
He presented the theoretical foundations of this argument to the British Psychoanalytical Society in three papers between 1958 and 1960: The Nature of the Child's Tie to His Mother (1958), Separation Anxiety (1959), and Grief and Mourning in Infancy and Early Childhood (1960). The Society's reception was cool — leading figures including Anna Freud and Melanie Klein objected that Bowlby was replacing the inner world with behaviorism. The controversy was substantive: Bowlby was arguing that observable proximity-seeking behavior and its biological substrate were the right level of analysis, not unconscious fantasy or the vicissitudes of drive. The psychoanalytic establishment thought he had abandoned the essential subject matter of psychoanalysis.
The full theoretical statement came in the three-volume Attachment and Loss trilogy: Attachment (1969), Separation: Anxiety and Anger (1973), and Loss: Sadness and Depression (1980). The trilogy integrated ethology, control systems theory, cognitive science, and evolutionary biology into a comprehensive account of how the attachment behavioral system develops, what happens when it is activated, and what the consequences of disrupted caregiving are across the lifespan. The concept of internal working models — cognitive and emotional representations of the self in relation to attachment figures, built from early caregiving experience and carried forward into subsequent relationships — provided the mechanism linking early experience to later functioning.
Ainsworth and the Strange Situation
While Bowlby was developing the theoretical framework, Mary Ainsworth was generating the empirical evidence base. She had worked with Bowlby at the Tavistock Clinic in London in the early 1950s before moving to Uganda with her husband in 1954, where she conducted the first systematic observational study of mother-infant attachment in a naturalistic setting. The Uganda study (published as Infancy in Uganda in 1967) documented that infants displayed clearly differentiated attachment behaviors — using the mother as a secure base for exploration, showing distress at separation, using her as a haven of safety under threat — and that the quality of these behaviors varied across dyads in ways that seemed related to the mother's sensitivity and responsiveness. Some infants were securely organized; others were not. The individual differences were real and they were not random.
Ainsworth moved to Johns Hopkins University in 1958 and began the Baltimore study in 1963 — a longitudinal home observation study of 26 mother-infant pairs, visited regularly through the infants' first year. As a laboratory complement to the naturalistic observations, she developed the Strange Situation Procedure: a structured 20-minute protocol in which a mother, her infant, and a stranger are introduced, separated, and reunited across eight brief episodes. The design logic was specific — laboratory stress (an unfamiliar room, a stranger, brief separations) was introduced deliberately to activate the attachment system, and the infant's behavior during the reunion episodes was the key observational target. What the infant did when the mother returned, not merely during separation, was what distinguished organized attachment patterns from each other.
The Strange Situation findings, published with colleagues as Patterns of Attachment (Ainsworth, Blehar, Waters & Wall, 1978), identified three organized attachment patterns:
Secure (B): Infants explored freely using the mother as a secure base, showed distress at separation but were easily soothed by reunion, and returned to exploration afterward. In Ainsworth's Baltimore sample, approximately 65–70% of infants were classified secure.
Insecure-Avoidant (A): Infants showed little distress at separation and avoided or minimized contact with the mother at reunion, continuing to focus on the environment. Approximately 20–25% of infants in Ainsworth's sample.
Insecure-Resistant/Ambivalent (C): Infants were highly distressed before and during separation and were not soothed by reunion — mixing contact-seeking with angry resistance. Approximately 10–15% of Ainsworth's sample.
The critical observational finding linking the Strange Situation to home behavior was that secure infants in the laboratory had mothers who were more consistently sensitive and responsive during naturalistic home observations across the first year. The Strange Situation was not measuring something arbitrary — it was measuring the distillation of months of actual caregiving experience into an organized behavioral strategy.
Main, Solomon, and the Fourth Category
Ainsworth's three-category system was widely adopted through the late 1970s and early 1980s. But as researchers applied the Strange Situation to higher-risk populations — infants from maltreating environments, infants with mothers who had experienced unresolved trauma — a problem emerged: a substantial subset of infants could not be cleanly classified as A, B, or C. Their behavior in the Strange Situation was disorganized — they showed conflicting actions in the same moment (approaching while simultaneously turning away, freezing mid-movement, showing stereotyped repetitive motions, or appearing disoriented in the presence of the caregiver). The three-category system required forcing these infants into whichever category seemed the least bad fit.
Mary Main, one of Ainsworth's doctoral students at Johns Hopkins, working with Judith Solomon, developed the criteria for a fourth classification: disorganized/disoriented (D). The defining observation was that these infants appeared to lack a coherent behavioral strategy for managing separation stress. Main's theoretical account, developed with Erik Hesse, proposed that disorganized attachment develops when the caregiver is simultaneously the source of fear and the expected haven of safety — most commonly when the caregiver displays frightening, frightened, or dissociative behavior toward the infant. In such cases, the attachment system faces an unresolvable conflict: the figure who should provide safety is the source of alarm, and no organized strategy can resolve this contradiction. The result is behavioral disorganization and disorientation.
Main and Solomon first described the disorganized category in 1986, with fuller specification in 1990. Its appearance reorganized understanding of high-risk developmental trajectories — disorganized infant attachment is now one of the more robust predictors of subsequent psychopathology, including dissociation, borderline personality features, and anxiety disorders, particularly in samples with histories of early maltreatment.
The Adult Attachment Interview
Main's most consequential methodological contribution was not the disorganized category but the Adult Attachment Interview (AAI), developed with Carol George and Nancy Kaplan in 1985. The AAI is a semi-structured interview that asks adults about their childhood attachment relationships with parents — to describe each parent in a few words, to provide specific memories supporting those descriptions, and to reflect on how early experiences may have influenced their current personality and relationships.
The scoring of the AAI focuses not on what people report about their childhoods but on how they narrate them. The key construct is discourse coherence — the degree to which the narrative is organized, internally consistent, and plausibly connected to evidence. A person who describes their childhood as happy but cannot provide any supporting memories, or who becomes confused and incoherent when discussing past loss or trauma, receives a different classification than someone who describes a difficult childhood with clarity, reflective distance, and narrative coherence. The four AAI classifications — Secure-Autonomous (F), Dismissing (Ds), Preoccupied (E), and Unresolved/Disorganized (U) — correspond to adult states of mind regarding attachment, not to what happened in childhood.
The AAI's validation came through its predictive relationship with the Strange Situation. Van IJzendoorn's 1995 meta-analysis of 18 studies (N=854) found that a parent's AAI classification predicted their infant's Strange Situation classification at approximately 75% concordance for secure versus insecure. Critically, the AAI could be administered prenatally — before the infant had any experiences with the parent — and still predicted the infant's attachment organization months later. The interview was capturing the parent's representational state of mind, which then influenced caregiving sensitivity, which then organized the infant's attachment. The mechanism was not magic; it was caregiving behavior. But the correspondence between how a parent narrated their own childhood and how their child would behave in a laboratory separation procedure was striking enough to establish the AAI as the gold standard for assessing adult attachment from a representational perspective.
Adult Attachment and the Self-Report Tradition
The AAI's clinical rigor came with a practical cost: it requires a trained coder, takes approximately one hour to administer, and produces classifications through a specialized coding system that takes weeks of training to learn. For large-scale research, a different approach was needed.
Cindy Hazan and Philip Shaver's 1987 paper in the Journal of Personality and Social Psychology — "Romantic Love Conceptualized as an Attachment Process" — initiated the self-report adult attachment tradition. Hazan and Shaver proposed that romantic love in adults is governed by the same attachment behavioral system that organized infant-caregiver proximity. They adapted Ainsworth's three infant categories into three paragraph-length descriptions of how people typically feel in romantic relationships and asked adults to self-identify. The resulting three-category self-report measure was crude by later standards — a forced-choice single-item instrument — but its correlates were striking: secure adults reported more satisfying relationships, more positive models of self and partner, and more coherent narratives about childhood than avoidant or anxious adults.
The theoretical formalization of the four-category model came from Kim Bartholomew and Leonard Horowitz's 1991 paper, which introduced the 2×2 structure of positive/negative models of self and others, and from the subsequent work of Kelly Brennan, Catherine Clark, and Philip Shaver, whose 1998 factor analysis of 482 attachment-related items identified the two continuous dimensions — anxiety and avoidance — that now underlie most adult attachment measurement, including the ECR-R used on this site.
The convergence of the clinical-interview tradition (AAI) and the self-report tradition (ECR-R and its predecessors) has not been seamless. The two instruments assess somewhat different things — the AAI measures discourse coherence about the past; the ECR-R measures current feelings about romantic relationships — and their cross-classification agreement is lower than researchers initially hoped. This has generated ongoing debate about whether adult attachment is best understood as a narrative competency, a relational schema, or a pattern of affect regulation. The debate is unresolved, and both traditions remain active. What they share is the core theoretical claim: that early caregiving experiences become organized into internal representations that shape how proximity, support, and intimacy are approached in adult life.
For the current state of the empirical record — including cross-cultural evidence, stability research, and the clinical applications of attachment theory — see the research page. For the four adult attachment styles as they are currently understood, see Secure, Anxious-Preoccupied, Dismissive-Avoidant, and Fearful-Avoidant.