What the research on lasting couples actually found

Relationship advice is unusually confident for a field whose evidence base is mostly observational. Some of it, though, has held up across independent groups and decades, and it is worth separating from the rest.

The ratio finding

The most widely cited result from laboratory observation of couples is that stable relationships show a substantially higher proportion of positive to negative interactions during conflict than unstable ones.

The specific figure that circulates — five to one — comes from this work and should be treated as illustrative rather than as a target. The robust part is directional: the balance matters more than the presence of negativity.

Which is the first useful correction to popular advice. Couples who last are not the ones who avoid conflict. They are the ones whose conflict sits inside a larger volume of ordinary warmth.

The four behaviours that predict trouble

The best-known contribution from this literature is a set of communication patterns associated with poorer outcomes.

Criticism — framing a complaint as a statement about the person's character rather than about the behaviour.

Contempt — expressions of superiority, including sarcasm, mockery and eye-rolling. This is consistently identified as the strongest single predictor.

Defensiveness — responding to a complaint with counter-complaint or with justification rather than with acknowledgement.

Withdrawal — shutting down and disengaging during conflict, sometimes described as stonewalling.

The prediction accuracy claimed in some of this work has been criticised on statistical grounds, largely because the models were fitted and tested on overlapping samples. That criticism is fair and it does not eliminate the finding that these patterns are more common in distressed couples.

The practical value does not depend on prediction accuracy anyway. Contempt is corrosive whether or not a model can forecast divorce from it.

The bids finding

Less famous and possibly more useful.

Observational work has described small attempts at connection throughout a day — a comment about something out of the window, a question, a touch — and how the partner responds.

Couples who remained together responded to a substantially higher proportion of these than couples who separated.

This is the part of the research that translates most directly, because it concerns behaviour available continuously rather than during arguments.

Most people are capable of responding to a remark about the weather. Very few think of it as relationship maintenance.

Perpetual problems

A finding that surprises people.

A substantial proportion of the disagreements in long relationships are never resolved. They concern differences in temperament, family, tidiness, sociability, money and pace, and they recur for decades.

The distinction that matters is not whether such problems are solved but whether the couple can discuss them without contempt.

Which reframes the objective. The aim is not agreement on everything but a workable relationship to the disagreements that will not go away.

What the research does not support

Worth stating plainly, because the same literature is used to justify claims it does not make.

It does not support the idea that any specific technique reliably repairs a distressed relationship. Couples therapy has reasonable evidence behind particular approaches, and the effects are moderate rather than transformative.

It does not support strong claims about compatibility being detectable in advance. Prediction of relationship satisfaction from pre-relationship characteristics is weak in the research that has looked, including large studies using machine learning on extensive questionnaire data.

What predicts satisfaction reasonably well is how each person currently feels about the relationship, which is informative and not very useful for matching.

The limitations worth knowing

Most of this work is observational, so causal direction is uncertain. Contempt may erode a relationship, or an already-eroding relationship may produce contempt, and both are probably true.

Samples have been disproportionately from wealthy Western countries, and mostly heterosexual, which limits generalisation more than the confident popular presentation suggests.

And the effect sizes, where reported, are moderate. These are tendencies across groups, not rules that determine individual outcomes.

The short version

Keep the ordinary warmth high. Complain about behaviour rather than character. Do not express contempt, ever, even when it would be satisfying. Respond when your partner tries to connect about something trivial.

None of that is novel, and it has more behind it than most of what is written about relationships.

General information about relationships, not psychological or medical advice. Consult a qualified professional about your own circumstances.