The strange situation, a twenty-minute test

Mary Ainsworth devised in the late 1960s a standardised sequence of about twenty minutes, in eight episodes, in which a child of around one year faces an unfamiliar room, an unfamiliar person, then two brief separations from their parent.

The procedure belongs to the theoretical framework proposed by John Bowlby, who described attachment as a behavioural system geared towards seeking proximity in moments of alarm. It aimed to observe that system under conditions comparable from one child to the next.

What is coded is not the reaction to separation, contrary to what is often reported, but conduct at reunion: how the child seeks or avoids contact and does, or does not, return to exploring.

Ainsworth distinguishes three configurations, secure, avoidant and ambivalent. Mary Main and Judith Solomon added in 1990 a fourth category, called disorganised, for brief and contradictory conduct the first three did not classify.

Distributions observed in Western samples place a majority of children in the secure category, with proportions that vary appreciably by country and sample.

It measures a relationship, not a trait of the child. The same child may be classified differently with two different adults, a result found in several studies and rarely mentioned when attachment vocabulary is used outside the field.

The classification is a research category, established from coding by trained observers. It was not designed as an individual diagnostic instrument, and its authors have said so repeatedly.

Coding requires lengthy training of observers, which limits sample sizes. Other instruments, such as the Attachment Q-sort, completed on the basis of home observations, have been developed to broaden measurement, with imperfect agreement with the laboratory procedure.

Cultural validity was discussed from the 1980s. Takahashi's work in Japan showed that separation had a different meaning depending on childcare habits, which shifts the distribution of categories without anything indicating a difference in the quality of the relationship.

A second criticism concerns stability over time: classifications change when family circumstances change, which defenders of the model regard as an expected property rather than a flaw.

Meta-analyses find associations between early classification and later indicators of social conduct, with small to moderate effect sizes.

Those associations permit no individual prediction. They describe average differences between groups, compatible with very varied individual trajectories, including ones opposite to the average trend. They are also hard to separate from the living conditions that accompany the initial classification and persist afterwards, which leaves uncertain the share attributable to the early relationship itself.