The Four Attachment Styles, Honestly

The first version of attachment theory was a 1951 report commissioned by the World Health Organization on the mental health of European children orphaned, displaced, or institutionalised by the Second World War. The author was a London child psychiatrist named John Bowlby, who had spent the 1930s working with delinquent and emotionally disturbed children at the London Child Guidance Clinic. The report — titled "Maternal Care and Mental Health" — argued, against the prevailing psychoanalytic and behaviourist orthodoxies, that the warmth and continuity of a child's relationship with one primary caregiver was a biological need on the order of food and sleep, and that disrupting it produced lasting damage. The book sold around 400,000 copies in its first English edition, was translated into a dozen languages, and supplied the policy basis for shutting down many of Europe's large foundling institutions through the 1950s.

Bowlby spent the next two decades turning that argument into a testable theory, drawing on Konrad Lorenz's ethology, on the cybernetics of feedback systems, and on his own clinical observations. The result, the three-volume Attachment and Loss published between 1969 and 1980, proposed that a human infant comes equipped with a set of inborn behaviours — crying, clinging, smiling, following — designed by evolution to maintain proximity to a caregiver, and that these behaviours, integrated by the second year of life into what Bowlby called an internal working model, form the template for all later close relationships. The theory's empirical engine, however, came not from Bowlby but from his collaborator Mary Ainsworth, who designed the Strange Situation Procedure in Baltimore in the late 1960s.

The procedure is short, technical, and easily misdescribed. A mother and her roughly 12-month-old infant enter an unfamiliar lab room with toys. Over a 21-minute sequence of seven episodes, a stranger enters, the mother leaves twice, and the mother returns twice. Trained coders rate the infant's behaviour during the reunion episodes — not the separation, the reunion — and assign one of three classifications. Type B, secure: the infant uses the mother as a base, protests at her leaving, and is comforted at her return. Type A, insecure-avoidant: the infant ignores the mother on return. Type C, insecure-ambivalent or resistant: the infant seeks contact but cannot be soothed, alternating clinging with anger. In a later 1986 paper, Ainsworth's student Mary Main added a fourth category, Type D — disorganized — for infants who showed contradictory or freezing behaviours associated with frightened or abusive caregivers. That fourth category is the source of the now-standard "four attachment styles" framing.

The leap from infants in Baltimore to adults swiping on Hinge happened in a single 1987 paper, published in the Journal of Personality and Social Psychology by Cindy Hazan and Phillip Shaver, then at the University of Denver. Hazan and Shaver translated Ainsworth's three infant categories into three short paragraph descriptions of romantic-relationship style, ran a self-report version in a Rocky Mountain News reader survey, and reported a roughly 56–24–20 split among adults — secure, avoidant, anxious. The paper has been cited tens of thousands of times and is the entire conceptual basis for almost every "what's your attachment style?" quiz online. What it is not is a measurement of the same thing the Strange Situation measures. The infant procedure is observational, takes a trained coder, and assesses a relationship; the adult quiz is a self-report and assesses a self-concept. The correlation between an adult's self-reported attachment style and their measured infant classification two decades earlier, in the longitudinal studies that have run the comparison, is around 0.3 — better than chance, far below the popular narrative of stable lifelong type.

Reliability is the harder problem. When the Strange Situation is repeated on the same infant-caregiver pair within a few weeks, classifications agree most of the time only when the family's social and economic background has not changed in between. When background variables shift — a parental job loss, a move, a new childcare arrangement — classifications change too. The British psychologist Michael Lamb and colleagues argued in the mid-1980s that this means the procedure measures the current state of the dyad's interaction more than any stable internal trait of the child, and that conclusion has held up. Cross-cultural challenges have been more pointed. Heidi Keller, the German developmental psychologist, has argued from fieldwork in over twenty countries — much of it in the Global South — that the universality assumptions baked into Ainsworth's procedure do not survive contact with cultures where infants are routinely cared for by multiple adults rather than a single mother figure. In Cameroonian Nso villages, for example, behaviours coded as "avoidant" in the Baltimore protocol reflect culturally normative independence training, not insecurity.

Patricia McKinsey Crittenden, an Ainsworth student herself, has spent the years since developing an alternative framework she calls the Dynamic-Maturational Model, which treats attachment patterns as adaptations to specific information-processing demands rather than fixed types. The four-styles-quiz version that circulates on social media bears roughly the same relationship to the underlying research as a horoscope does to astronomy. The original three categories were inferred from a 21-minute observation of a one-year-old. The fourth was added later for a small subset of cases. The romantic-relationship adaptation is a self-report typology, not a measurement of childhood. And the lifelong stability is, in the data, roughly the same as the test-retest reliability of an average personality questionnaire — informative, modest, and a long way from destiny.