Why the same arguments keep coming back

From the 1980s onwards, the American psychologist John Gottman filmed thousands of conflictual conversations in an apartment turned into a laboratory at the University of Washington. Coding these exchanges frame by frame produced a result nobody expected.

In his data, roughly two thirds of marital disagreements are never resolved. They bear on differences of rhythm, temperament, relationship to family or to order, and return in various forms over decades. He calls them perpetual disagreements.

That proportion changes how a recurring argument should be read. It stops being the sign of a problem badly handled and becomes the ordinary manifestation of a stable difference between two people, a difference no conversation will make disappear.

The practical consequence is unexpected: looking for the solution is sometimes the mistake itself. A perpetual disagreement is not settled; it is kept in a liveable state or it festers, and it is that fork in the road which separates the observed trajectories.

What Gottman associates with deterioration is not the frequency of disagreements but the way they unfold. Four behaviours recur: criticism, which targets the person rather than the fact, contempt, which belittles them, defensiveness, which sends the blame back, and withdrawal, that wall of silence where one stops answering.

One observation from the physiological measurements deserves flagging, because it contradicts common sense. The person who goes quiet is not indifferent: their readings indicate intense activation instead. Silence is saturation, not disinterest.

In couples who stay together, Gottman records about five positive exchanges for one negative exchange during a disagreement: a touch of humour, a sign of partial agreement, a question. Among those who separate, the ratio moves towards one to one.

This figure is often quoted outside its frame, as if it held for daily life. It bears on the moment of disagreement, precisely where these gestures are the most costly to produce.

Gottman announced very high prediction rates, which made his reputation. They have been contested, notably by Richard Heyman and Amy Slep in 2001: his models were often computed on the very data whose outcome was already known, which amounts to description after the fact rather than prediction.

The behaviours described, for their part, remain solid and replicated. What has to be abandoned is the idea of an individual prognosis: knowing that a behaviour is associated with increased risk in a population says nothing about what will become of any given couple.