The Four Horsemen of Conflict, and What Replaces Them
PsyCove Editorial · 8 min read · Updated 2026-10-28
- The four horsemen are observable behaviors rather than personality traits, which is why they can be changed.
- Contempt, the one that places a partner below the other, carries more weight than the other three.
- The direction of the finding replicated across follow-ups; the 90-percent accuracy figures did not.
- Each horseman has a specific replacement behavior, and the replacement is a skill you can rehearse.
- Ending conflict is not the goal. Repairing after it is what the longitudinal studies actually track.
What the four horsemen actually are
The list did not start as a list. Gottman's lab filmed couples in an apartment laboratory, coded their faces, voices and physiology second by second, then followed the same couples for years to see which pairs stayed together. Gottman and Levenson (1992) reported that patterns visible in a single argument discriminated couples who later separated from those who did not. The four behaviors below are what those codings kept turning up.
The framework is far easier to misuse than to use. It describes a statistical tendency in a small number of couples observed in one laboratory, and it predicts nothing about any individual relationship. Two people who fight hard, say what they think and repair within the hour will be mislabeled by this list; two people who never fight and quietly withdraw will be missed by it entirely. Read it as a description of four specific behaviors in a conversation, not as a score — and treat any verdict about a relationship drawn from the list alone as wrong, because those two cases are the ones the research design cannot see.
- Criticism: an attack on the person rather than the request, the difference between 'you always leave this to me' and 'I need help before I start cooking'.
- Contempt: mockery, sarcasm, eye-rolling, or a tone that puts one partner beneath the other. Gottman's analyses singled this one out.
- Defensiveness: counter-attack or victim stance in reply to a complaint, which stops the complaint from ever being heard.
- Stonewalling: withdrawal, going blank, leaving the room, ending the conversation without ending the problem.
Why contempt carries the most weight
Of the four, contempt is the one Gottman's writing treats as the strongest single signal. The reason is structural rather than moral. Criticism targets a behavior and can be answered with a behavior; contempt targets the person's worth and has no answer. Once one partner has been placed below the other, the conversation stops being about the dishes.
The 1992 result was narrower than its popular version. Negative affect reciprocity, the tendency of each partner to match the other's negativity, discriminated couples who later separated better than the total amount of negativity did. Two couples can produce the same count of harsh moments and end up in different places, because what differs is whether each harsh moment is met with escalation or with a de-escalating move.
Stonewalling sits at the far end of the same mechanism. When physiological arousal passes a threshold, people stop processing what they hear; Gottman called it flooding. The practical consequence is that the partner who goes blank is often not refusing to listen but has temporarily stopped being able to.
What the research supports, and where it stops
Two things have held up. The first is direction: negative reciprocity and the four behaviors predict poorer outcomes across follow-ups, including the 14-year divorce-timing analysis Gottman and Levenson published in 2000. The second is mechanism: observable behavior in a conflict conversation carries information that self-report questionnaires do not.
Two things have not held up as cleanly. The samples were small, with the 1992 study following dozens of couples rather than thousands, and the accuracy figures above 90 percent that circulate from the popular books come from those same small samples. Independent labs have reproduced the direction of the effect without reproducing the arithmetic. The honest summary is that the behavior matters and the precision is softer than the marketing.
A third limit is design. These were mostly married heterosexual couples in one region of the United States, observed doing a laboratory task. Whether the thresholds transfer to dating couples, to couples who never marry, or to relationships in other countries is genuinely open, and treating the list as universal is the exact mistake the cross-cultural literature warns about.
The four antidotes, and what they are for
Each horseman has a counter-behavior, and the counter-behavior is something you rehearse rather than a feeling you summon. None of the four requires the other person to change first, which is the entire point: a couple's trajectory is set less by who is right than by whether either person can make a de-escalating move while still upset.
- Against criticism, a gentle start-up: situation, then need, then request, in that order and in one breath.
- Against contempt, a described appreciation: name one specific thing the other person did this week that you valued. Contempt is a habit of attention, and habits of attention get replaced, not suppressed.
- Against defensiveness, partial responsibility first: 'you are right that I was late' before the explanation, because the explanation only lands after the complaint has been acknowledged.
- Against stonewalling, a scheduled break with a return time: twenty minutes, agreed out loud, with an hour named to resume. Without the return time, a break is just a withdrawal.
Repair matters more than the absence of conflict
The longitudinal work keeps pointing at one variable, and it is not conflict frequency. Reis and colleagues (2004) organized the intimacy literature around perceived partner responsiveness, the sense that the other person understands, validates and cares about what you just said. Overall, Fletcher and Simpson (2010) found that partner support aimed at self-improvement predicted better relationship quality over time. Neither construct requires a conflict-free relationship. Both require a partner who comes back.
This is also where the horsemen framing misleads. A couple that never raises anything looks free of the four behaviors while scoring badly on the variable that predicts staying together. Jehn's (1995) distinction helps: task conflict, disagreement about what to do, is often productive, while relationship conflict, disagreement about who the other person is, is what corrodes. Suppressing both is not the same as resolving one.
Where the framework breaks down
Three situations need a warning before any of the above applies:
- If one partner is afraid of the other, the four behaviors are not the problem and communication exercises are not the solution. Safety comes before technique.
- If the pattern includes coercion, monitoring or threats, this is not a conflict-style issue and should not be worked on as one.
- If the 'horseman' has become a word you use about your partner rather than about an argument, the label is now doing the damage. Calling someone contemptuous is itself a contemptuous move.
A five-minute check you can actually run
Take the last argument that never resolved. Write three lines: what you said, what you actually wanted, and what you did when the other person pushed back. Most people find their own horseman in the third line, and it is usually not the one they expected.
Then write the replacement sentence, situation, need and request, and say it out loud once. The rehearsal matters more than the insight, because the moment you need that sentence is precisely the moment you will not be able to compose it.
The goal is not a fight-free relationship. It is a relationship in which the fight has an exit.
Frequently asked questions
Are the four horsemen a sign a relationship will fail?
No. They are behaviors observed during conflict, and behaviors are the most changeable thing in a relationship. What the follow-up studies track is the pattern over time, not a single argument. A couple that uses all four occasionally and repairs well is in a different position from a couple where contempt is the default tone.
Is criticism ever appropriate?
A complaint is not criticism. A complaint names a specific behavior and its effect: the bins were left out and I had to bring them in twice this week. Criticism generalizes to character: you never think about anyone but yourself. The first can be answered. The second has to be defended against, which is why it produces the other three.
What if my partner stonewalls and will not talk at all?
Stop trying to finish the conversation while arousal is high. Name a specific time to resume, put it somewhere visible, and leave. If the return never happens, the problem is no longer stonewalling but avoidance of the topic, and that is worth naming directly rather than re-litigating the original argument.
Does this apply to dating couples, or only to married ones?
The evidence base is mostly married couples, many of them recruited in one region of the United States, so the honest answer is that it transfers as a hypothesis rather than as a finding. The mechanism, matching negativity with negativity, is likely general; the specific thresholds are not established for other groups.
Can a couple recover after contempt?
Yes, and the recovery is behavioral: the eye-roll stops being the reflex, and something specific gets named out loud instead. If contempt is accompanied by fear, coercion or threats, this is not a communication problem and a couples exercise will not fix it. Contact a domestic abuse service, and if distress is persistent, a qualified professional.
References
- 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.
- 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.
- Gottman, J. M., & Silver, N. (1999). The Seven Principles for Making Marriage Work. New York: Crown Publishers.
- Reis, H. T., Clark, M. S., & Holmes, J. G. (2004). Perceived partner responsiveness as an organizing construct in the study of intimacy and closeness. In D. J. Mashek & A. Aron (Eds.), Handbook of Closeness and Intimacy. Mahwah, NJ: Erlbaum.
- Overall, N. C., Fletcher, G. J. O., & Simpson, J. A. (2010). Helping each other grow: Romantic partner support, self-improvement, and relationship quality. Personality and Social Psychology Bulletin, 36(11), 1496–1513.
- Jehn, K. A. (1995). A multimethod examination of the benefits and detriments of intragroup conflict. Administrative Science Quarterly, 40(2), 256–282.