What marital satisfaction scales measure

The two most used scales in the literature are old. The Marital Adjustment Test, published by Harvey Locke and Karl Wallace in 1959, and the Dyadic Adjustment Scale, published by Graham Spanier in 1976, still serve as a reference or as the basis for abridged versions.

This longevity has an advantage, the comparability of results across several decades, and a drawback, the persistence of items designed in a social context very different from today's.

These scales combine two types of question. Overall evaluation questions, of the kind: on the whole, are you satisfied with your relationship. And frequency questions bearing on behaviours or disagreements: how often do you argue, confide in each other, consider separating.

This combination has been criticised. Mixing an evaluation with behaviours produces a composite score that can vary for opposite reasons: a couple who discuss their disagreements a great deal may obtain a lower score than a couple who avoid them.

All these measures rest on self-report. Yet marital satisfaction is a domain in which people do not answer freely: admitting that one's relationship is going badly engages one's self-image, including before an anonymous questionnaire.

The distributions obtained show it: answers cluster heavily on the favourable side of the scale, which reduces the ability to distinguish situations in that zone. A high score there is therefore far less informative than a low one.

A difficulty specific to this field is that the object studied comprises two people who do not answer the same thing. The correlation between the scores of two members of the same couple is positive but far from perfect.

Studies that question only one of the two, still numerous, therefore measure the satisfaction of a person and not the state of a relationship. That is a limit to keep in mind before any result concerning couples.

Three questions are enough. Were both members of the couple questioned. Does the scale mix evaluation and behaviours. And is the reported effect expressed in interpretable units, or only as a statistically significant difference.

This last point is most often neglected. On large samples, a minuscule gap becomes significant, which makes it neither perceptible nor useful to anyone.