Topics / Relationships

What predicts a relationship lasting

Published November 20, 2026. 4 sources at the foot of the page.

What is a conflict pattern?

A conflict pattern is a habit of arguing that a couple falls back into whenever the same disagreement comes around. The words change from one time to the next and the shape does not.

Where you might meet it

Gottman and Levenson (1992) brought married couples into a laboratory and asked each pair to talk through a disagreement they were actually having, for about fifteen minutes, while cameras ran.

Trained coders then went through the recordings second by second and scored what each face and each voice was doing. The couples also wore sensors, so heart rate and other body measures were recorded alongside the talk.

After that the researchers waited. Gottman and Levenson (1992) returned to their couples four years later, and Gottman and Levenson (2000) followed a set of couples across fourteen years, recording who was still married and who was not.

That is the whole design: a short recorded disagreement, a detailed score of the behavior in it, and a follow-up years afterwards. Everything else in this course is an argument about what a design like that can and cannot tell anybody.

The four patterns that predicted dissolution

Of everything the coders scored, four conflict patterns did most of the work. Gottman called them the four horsemen, and that name has traveled a good deal further than the research behind it.

Criticism

Criticism goes at the person rather than at the problem. “The bin is full again” is a complaint; “you never think about anybody but yourself” is a criticism, because it makes a character out of a chore.

Contempt

Contempt puts one partner above the other: mockery, sneering, name calling, an eye roll delivered so that it will be seen. Gottman (1994) reports contempt as the strongest of the four, the pattern whose presence most separated the couples who later split from the couples who did not.

Defensiveness

Defensiveness answers a complaint with innocence or with a complaint of its own. Nothing lands, because every sentence is a reason the speaker is not the problem.

Stonewalling

Stonewalling is the partner who stops responding: no nod, no answer, eyes somewhere else. Gottman and Levenson found it more often in the husbands they recorded, and often in a body that was already racing.

The question is not whether a couple argues but which of these four patterns keeps showing up, and whether warmth outnumbers heat while they do.

The balance inside a disagreement

Gottman (1994) reports that what separated the stable couples was not the absence of conflict but the balance inside it: roughly five positive moments for every negative one while they disagreed, against something much closer to one to one among the couples who later separated.

Positive here means whatever the coders could see and hear: a joke that lands, an agreement, a question asked in a soft voice, a hand on an arm, a moment of taking the other person seriously.

The balance is a description of the recordings rather than a target anybody set. It says that in these samples the couples who lasted were not the calm ones, they were the ones who stayed warm while they argued.

Gottman (1994) also describes loud couples who stayed together for years and quiet couples who separated. Volume is not what the coding counted.

The most repeated number in this literature

The claim that this work predicts divorce with about 90 percent accuracy has been printed in newspapers, in books and on training slides for thirty years. It is the wrong way to read the studies, and the reason is not a technicality.

In these studies the statistical model was built using the same couples whose outcomes it then reported. Which behaviors went into the model was decided after the outcomes were known, and the weights were fitted to that one set of couples.

A number produced that way describes how well a model can be made to fit data already in hand. It is not a forecast, because nothing was forecast: no couple outside the set was sorted in advance and then checked against what happened.

Heyman and Slep (2001) made the point and tested it. Taking prediction equations of this kind and applying them to a separate sample of couples, they report that the accuracy falls a long way, which is what happens when a model fitted to one group of people meets another group.

Gottman and Levenson (2000) report accuracy figures of their own, and commentators since have argued about all of them for the same reason. A model that has already been shown the answers is not predicting them.

None of this makes the four patterns imaginary. It makes the headline number a claim about arithmetic rather than a claim about anybody’s marriage.

What these samples can carry

The couples Gottman and Levenson recorded were mostly white, married, heterosexual and living around one American city; the fourteen year follow-up drew on couples recruited in the Seattle area. Each study worked with a few dozen couples rather than thousands.

That is a narrow slice of the world. Nothing in it speaks for unmarried partners, for same sex couples, for couples outside the United States, or for the many people who would never volunteer to argue in front of a camera.

The coding also captures about fifteen minutes of one conversation. A disagreement staged for a laboratory is not every disagreement a couple has, and one recording is not a marriage.

The antidotes are recommendations

Gottman (1994) pairs each of the four patterns with something he suggests in its place: raising a complaint gently rather than as a verdict, building fondness and respect against contempt, taking a share of the responsibility instead of defending, and asking for a pause when the body is racing rather than going silent.

Those are his recommendations, and this course reports them as recommendations. They are not what the prediction studies measured, and nothing here promises an outcome for anybody who follows them.

The honest summary is small and still worth having. Four patterns are visible in how people argue, one of them stood out from the others, warmth during conflict went with staying together in these samples, and the famous number is not what it is usually said to be.

Sources

  1. Gottman, J. M., & Levenson, R. W. (1992). Marital processes predictive of later dissolution: Behavior, physiology, and health. Journal of Personality and Social Psychology, 63(2), 221-233.
  2. Gottman, J. M., & Levenson, R. W. (2000). The timing of divorce: Predicting when a couple will divorce over a 14-year period. Journal of Marriage and Family, 62(3), 737-745.
  3. Gottman, J. M. (1994). What Predicts Divorce? The Relationship Between Marital Processes and Marital Outcomes. Erlbaum.
  4. Heyman, R. E., & Slep, A. M. S. (2001). The hazards of predicting divorce without crossvalidation. Journal of Marriage and Family, 63(2), 473-479.

Practice it on Learn

Learn teaches relationship conflict in short games and lessons, with the same research as this page.

Practice relationship conflictSee what Learn is

Your first lesson is free and needs no account. For adults 18 and over.

More topics