The Attachment Style Assessment: Two Dimensions, Not Four Boxes

Why the dimensions matter more than the label

Attachment research moved from categories to dimensions decades ago, and the measurement literature is fairly settled on it: analyses of the underlying structure find two continuous dimensions rather than four discrete groups. The category system survives because it is easy to talk about, and because the four names describe recognisable patterns. But nothing in the data marks a boundary between anxious and secure — the transition is gradual, and someone at 55 on anxiety and someone at 45 have almost identical profiles and different labels.

This is the source of the most common complaint about these assessments, which is that the result changed between tests. Often nothing changed except which side of a midpoint a nearly-identical score landed on. Someone whose anxiety score moves from just above to just below the cut point gets a different category name and a different-sounding description from a shift well inside measurement noise.

It also explains why people recognise themselves in more than one description. A moderate score on both dimensions genuinely does contain elements of several patterns, and reading four vivid descriptions and finding pieces of yourself in three of them is the correct experience for that profile rather than a sign the test failed. The way out is to read the two numbers, and how far each sits from the middle, rather than the name attached to them.

What each dimension is picking up

Attachment anxiety is the dimension organised around availability. High scorers monitor the state of the relationship closely, read ambiguity as a signal of withdrawal, and need explicit reassurance more often than a low scorer finds intuitive. The behaviours it produces — checking, seeking contact, pressing for resolution during conflict — are attempts to restore certainty, which is why they intensify precisely when the other person pulls back.

Attachment avoidance is organised around dependence. High scorers are uncomfortable relying on someone else and uncomfortable being relied on, prefer to handle difficulty alone, and experience escalating closeness as pressure. The characteristic move under stress is distance — becoming busy, becoming reasonable, going quiet — which is not indifference, though it reliably reads as indifference to a partner high on anxiety.

The third thing worth measuring alongside them is whether you hold a working assumption that connection survives distance. That is what a secure profile amounts to in practice, and it is not simply the absence of the other two: it shows up as being able to be apart without the relationship feeling at risk, and being able to raise a problem without treating it as a threat to the whole thing. Assessments that report only anxiety and avoidance leave this implicit, and it is often the more legible thing to notice about yourself.

The pairing effect, and why your score is not entirely yours

The two dimensions interact badly in a specific combination, and it is common enough to be worth naming. One partner high on anxiety and one high on avoidance produce a loop in which each person's coping strategy is the other's trigger: the anxious partner seeks contact to restore certainty, the avoidant partner withdraws to reduce pressure, and both escalate for reasons the other cannot see. Neither pattern is dysfunctional alone. The interaction is what causes the damage.

This means an attachment score taken during a particular relationship is measuring the relationship as well as you. Anxiety scores rise with a partner who is inconsistent and fall with one who is reliable, and the movement is real rather than error — attachment is partly relationship-specific, and someone can be substantially more secure in one relationship than they were in another. Measuring during a stable period and during a difficult one gives two different, both accurate, readings.

It also means the useful comparison is often between two people rather than against a norm. Two partners each seeing both sets of scores tends to reframe the dynamic from a dispute about who is being unreasonable to a description of two strategies that happen to collide — which is more actionable than either person's private theory of the other, and it does not require anyone to be told they scored badly.

Where the two dimensions came from

The dimensional model was not the original one, and the history explains why four-box results are still everywhere. The first widely used adult attachment measure asked people to pick which of three paragraphs described them best, adapting infant attachment categories to romantic relationships (Hazan & Shaver, 1987). It was influential and it forced a choice between prewritten types. A four-category revision followed, organising the styles around whether your model of yourself and your model of others were positive or negative, which added a distinction between two kinds of avoidance — fearful and dismissing (Bartholomew & Horowitz, 1991).

The shift to two continuous dimensions came from a measurement analysis rather than a new theory. Applying item response theory to the existing self-report scales, Fraley, Waller and Brennan found the items were behaving as indicators of two continuous dimensions and, importantly, that most scales measured the secure end of each dimension poorly — they discriminated well among insecure people and badly among relatively secure ones (Fraley et al., 2000). The revised scale they built from that analysis addressed the imbalance, and it is the item set most well-constructed attachment questionnaires now derive from.

That is the technical reason a two-score result is preferred over a four-box label. The categories were a starting point that the measurement did not support: cutting two continuous dimensions at their midpoints throws away the distance information and makes anyone near the middle's label unstable. It is also why an older test may put you in a box while a newer one gives you two numbers, without either being broken.

How much attachment moves

More than the popular framing suggests. Attachment patterns are usually described as forming in childhood, which is accurate about their origin and misleading about their permanence. Longitudinal work finds meaningful change in adult attachment over periods of years, with the largest shifts following sustained relationship experiences that contradict the existing expectation — a reliable partnership lowering anxiety, a period of being depended on lowering avoidance.

Therapy also moves it, and the mechanism is not mysterious: both dimensions are built from expectations about what happens when you need someone, and expectations update when the evidence changes consistently enough. What does not appear to work is deciding to be secure. The change follows repeated experience rather than resolution, which is why the timescale is months and years rather than weeks.

How much it moves has been tested directly against two competing models. One holds that early experience lays down an enduring foundation that continues to exert influence however much the surface changes; the other holds that attachment is revised continuously, so recent experience progressively overwrites the old pattern and early influence fades. A longitudinal study tracking people repeatedly over time found the data favoured the enduring-foundation account: current attachment kept a persistent trace of the earlier pattern rather than being fully rewritten by recent relationships (Fraley et al., 2011). Both change and continuity are real, in other words, and the change is layered over something that stays.

For anyone taking the assessment more than once, that timescale is the one to respect. Retaking within weeks measures the state of your current relationship and how the last fortnight went. Retaking after a year, particularly across a change in circumstances, measures something closer to the thing the label claims to describe — and your own two results are a comparison that needs no reference sample to interpret.