Attachment Theory Explained: From Bowlby and Ainsworth to Adult Attachment Styles

In the middle of the twentieth century there were two respectable explanations for why a baby wants its mother, and both treated the wanting as derivative. The psychoanalytic account made the bond a by-product of feeding. The behaviorist account made it learned dependency, reinforced by adults who picked the child up when it cried, which led to the advice, given seriously, that responding to crying would produce a clingy child.

John Bowlby, a British psychiatrist working with separated and institutionalized children, thought both accounts had the causation backward. His proposal was that attachment is a primary behavioral system in its own right, as basic as feeding, shaped by natural selection because an infant who stayed near a protective adult survived and one who wandered off did not. Crying, clinging, following and searching are not weakness or manipulation. They are the system doing its job.

That reframing is the whole foundation, and it had an immediate practical consequence. If separation distress is adaptive rather than spoiled, then a child protesting a separation is behaving correctly, and the question shifts from how to train the behavior out to what the child's environment is teaching them to expect. Bowlby's work also laid the ground for the idea of a secure base: a caregiver whose reliable presence is what makes exploration possible rather than what prevents it.

Bowlby had a theory. Mary Ainsworth made it an empirical science, and she did it by watching. In the mid-1950s she spent months observing families in Uganda, visiting homes and recording how infants used their mothers while exploring, which gave her the raw material for her 1967 book on infancy there. She then ran a longitudinal study in Baltimore, with extended home visits across a baby's first year, building a detailed picture of how individual pairs actually behaved together over time.

The Strange Situation came out of that fieldwork rather than preceding it. Ainsworth needed a short, repeatable procedure that would reliably activate the attachment system, and she built one: a sequence of brief episodes in an unfamiliar room in which a caregiver leaves and returns while the infant has toys to explore and an unfamiliar adult present. The published 1970 paper with Silvia Bell reported the core finding that gives the procedure its logic, that the mother's presence encouraged exploratory behavior and her absence depressed exploration and heightened attachment behavior.

The crucial design choice is that the classification comes from the reunion, not the separation. How upset a baby gets when the caregiver leaves says more about temperament than about the relationship. What the baby does when the caregiver comes back, whether comfort is sought, received and used to get back to playing, is the part that distinguishes the patterns. Ainsworth and colleagues set out three: secure, avoidant and resistant, which is also called ambivalent.

The bridge from infant behavior to adult relationships is a concept Bowlby called the internal working model, and the jargon hides a simple idea. Out of thousands of repetitions of needing something and finding out what happens, a person builds a working expectation: whether people come when called, and whether they themselves are the kind of person people come for.

It is called a working model because it is a tool for prediction rather than a memory. You do not consult it. It arrives as the obvious interpretation of an ambiguous event: a partner is quiet, and one person's immediate reading is that something is wrong while another's is that the evening is pleasant. Both are running a model, and neither experiences it as an interpretation.

That is also why the models are self-confirming. Expecting people to leave produces behavior that strains relationships; expecting closeness to come with demands produces distance that reduces what is on offer. The model collects the evidence it generated. Breaking that circuit is most of what therapy aimed at attachment is actually doing.

For years a minority of infants in Strange Situation studies could not be sorted into the three categories cleanly. Their reunion behavior contained contradictions: approaching with the head averted, freezing mid-movement, a dazed expression, starting toward the caregiver and then stopping or going somewhere else.

Mary Main and Judith Solomon addressed this in 1990 with a published set of indices for identifying disorganized and disoriented behavior, so other researchers could code the pattern consistently. The reading is that these infants have no workable strategy available, which tends to be the case when the person who would normally provide comfort is also the source of alarm.

In adult self-report measures the nearest relative of this pattern is the fearful-avoidant style, high on both attachment anxiety and avoidance. The two should not be treated as the same score: one is an observed classification in infancy and the other is a questionnaire result in adulthood, and the relationship between them is conceptual.

For three decades attachment was developmental psychology. Then in 1987 Cindy Hazan and Phillip Shaver published a paper arguing that adult romantic love works like an attachment bond, and translated the infant categories into three short self-descriptions that adults could pick between. The design was simple to the point of being crude, and it opened an enormous literature, because suddenly the framework could be studied in anyone.

Kim Bartholomew and Leonard Horowitz refined it in 1991. Their four-category model splits the avoidant group in two, on the grounds that a person who dismisses closeness and a person who fears it are running different programs, and organizes the four around two underlying judgments: the model of self and the model of others. That gives secure, preoccupied, dismissing and fearful.

Modern adult measurement went further and dropped the categories as the primary unit. The standard approach scores two continuous dimensions, attachment anxiety and attachment avoidance, and treats the four styles as regions on that map rather than as types a person belongs to. Our own test reports a leading pattern because that is what readers can use, and the scores underneath it are the real output.

Three things stand up particularly well. First, caregiving quality is related to infant security in the direction the theory predicts, and the relationship is causal enough to be moved: a meta-analysis of 70 studies by Marian Bakermans-Kranenburg and colleagues found that randomized interventions reduced insensitive parenting with an effect of about 0.33 and reduced infant attachment insecurity with an effect of about 0.20, and that the programs which improved sensitivity were the ones that improved security.

Second, the adult dimensions do real predictive work. Attachment anxiety and avoidance are reliably associated with how couples handle conflict, how they seek and give support, how satisfied they report being, and how they regulate distress. Few frameworks in popular psychology have anything like this track record.

Third, patterns travel across generations, modestly. The 2016 meta-analysis by Marije Verhage and colleagues, covering 95 samples, found a correlation of about .31 between a caregiver's secure-autonomous state of mind and their infant's security, notably smaller than the figure reported two decades earlier, and concluded that caregiver sensitivity does not fully account for the link. The gap between the two, which researchers call the transmission gap, is still open.

The most common overreach is causal. The popular version says your adult attachment style was set by your infancy, and the longitudinal evidence does not support that. Chris Fraley and Glenn Roisman, reviewing the field in 2019, concluded that links between early caregiving and adult attachment styles are weak and inconsistent across measurement domains, that styles are more malleable in childhood and adolescence than in adulthood, and that early experiences do not fully determine adult outcomes.

The second overreach is categorical. Styles get used as four personality types, with a full behavior profile attached to each, which is a long way past a two-dimensional score with a lot of people sitting near the middle. The third is diagnostic: attachment style is not a diagnosis, it does not appear as a condition in any diagnostic manual, and insecure attachment is not a disorder.

The fourth is directional. The framework was designed to describe your own experience of closeness, and it works poorly as a tool for labeling a partner, an ex or a parent. A style you assign to somebody else is a theory about them built from their behavior toward you, which is exactly the information least likely to be impartial.

The table at the end of this page sets the two apart, because almost all of the confusion about attachment theory comes from the right-hand column being presented in the voice of the left.

None of the right-hand claims are necessary for the theory to be useful. The supported version is already the more interesting one.

The free Psychology.com Attachment Style Quiz is built on the two adult dimensions, with twenty questions written in the spirit of the widely used Experiences in Close Relationships framework. It asks how you generally think and feel in close relationships, and the result screen names the pattern your answers lean toward along with how close your other scores came. An optional in-depth report is available afterward for readers who want a longer write-up. It is a self-report questionnaire for education and reflection, not a clinical or diagnostic instrument, and it cannot reconstruct what happened in your childhood.

From here, the attachment styles index compares all four adult patterns, the secure attachment and insecure attachment pages cover each end of the dimensions, earned secure attachment covers what the research says about change, and secure attachment in infants covers the developmental side for readers who came looking for the Strange Situation rather than for themselves.