Gottman’s Four Horsemen are four ways of arguing that the psychologist John Gottman found again and again in couples who were struggling: criticism, contempt, defensiveness and stonewalling. They are real, observable and worth learning to spot. What does not hold up is the famous claim built on top of them, that they let you predict divorce with more than 90% accuracy. And the most useful finding about them rarely gets mentioned: a short course teaching this material worked as well online as in person, and it worked without a therapist delivering it.
Here is what each horseman looks like in a real argument, what the research does and does not support, and what to do instead.
The Four Horsemen, plainly
Gottman borrowed the name from the four horsemen of the Apocalypse in the Book of Revelation. The Gottman Institute’s own descriptions are the clearest starting point.
| Horseman | What it is | What it can sound like |
|---|---|---|
| Criticism | Attacking your partner’s character, not a specific behaviour | “You never think about anyone but yourself.” |
| Contempt | Speaking from above: mockery, sarcasm, name-calling, eye-rolling | “Oh, well done. Did you work that out all by yourself?” |
| Defensiveness | Making excuses or playing the innocent victim so the complaint goes away | “I only forgot because you gave me ten other things to do.” |
| Stonewalling | Withdrawing and going unresponsive while the conversation is still happening | Silence, a blank face, suddenly looking at a phone |
Most people first recognise themselves in the gap between a complaint and a criticism. “I was worried when you didn’t call” is about a moment. “You’re so inconsiderate” is a verdict on a person. The first gives your partner something they can fix. The second only gives them something to defend, which is how the next horseman gets invited in.
If you have seen these called the “four horsemen of relationships” or the “four horsemen of the apocalypse in relationships,” it is the same framework.
What the research supports, and what it does not
This is where most articles on the subject go quiet, and it is the part that matters if you are deciding how seriously to take all this.
The behaviours are real
The four horsemen came out of years of watching couples argue in the lab, and researchers outside Gottman’s group now measure them too. A 2019 randomised trial for couples living with brain injury used a “Four Horsemen of the Apocalypse communication questionnaire” as an outcome, and couples who received the intervention improved on it compared with a waitlist. It was a small, specialised trial of 22 couples, so it shows the patterns can be measured and can shift, not much more. In a 2019 study of 130 couples, people with more borderline personality disorder symptoms “criticized more; their partners defended and stonewalled more”. The categories are concrete enough for independent teams to code and link to real outcomes.
The 90% divorce prediction is not reliable
Some version of this line appears on almost every page about the four horsemen: Gottman can predict which couples will divorce with more than 90% accuracy. The figures do come from real papers. A 1998 study of 130 newlywed couples reported predicting divorce or stability with 83% accuracy. A 2000 paper following 79 couples over 14 years reported a model that “predicted divorce with 93% accuracy”. That model did not rely on the horsemen alone. It also included marital satisfaction and whether the partners had thought about separating.
The problem is how those numbers were produced. The statistical models were built on couples whose outcomes were already known, then scored on how well they sorted those same couples. It is a little like writing an exam after seeing the answers. The honest test is crossvalidation: build the model on one group, then see how it does on a group it has never seen.
When Heyman and Smith Slep did that with archival data in 2001, naming Gottman’s 1998 study as their example, they found that “accuracy and predictive value drops precipitously during crossvalidation.” They also noted that these studies had not accounted for how common divorce actually is in the population, which changes what any accuracy figure means for a real couple. Their conclusion: results without crossvalidation “should be interpreted with extreme caution, no matter how impressive the initial results appear to be.”
Replication tells the same story. When another team tested the 1998 models in 85 young couples from at-risk backgrounds, “the major findings of Gottman et al. failed to replicate.”
None of this makes the four horsemen fake. It means the patterns go along with worse outcomes on average, and nobody can watch your argument and read your future from it. If you recognise all four in your relationship, you have learned something useful about how you fight. You have not received a forecast.
Is the Gottman Method “evidence-based”?
Partly, and less than its reputation suggests. The most current answer is a 2026 systematic review in the Journal of Marital and Family Therapy. It applied the criteria the American Psychological Association uses for empirically supported treatments to 40 studies and found that “no couple therapy models met the standards for strong empirical support.” Seven showed modest support, and Gottman couple therapy was one of them, alongside behavioural, cognitive-behavioural, emotionally focused, integrative behavioural, insight-oriented and strategic approaches. The limitations the reviewers hit most often were small samples and a lack of independent replication.
That describes the randomised evidence for Gottman-based work well. One trial randomly assigned 16 couples to ten sessions of Gottman couple therapy or a control group. An online Gottman education programme delivered over Zoom was tested with 72 couples. A programme for low-income couples experiencing situational violence, co-authored by Gottman himself, randomised 115 couples. The results point in a hopeful direction. They are not yet the large, independently replicated trials that would make any single couple therapy the proven one.
Contempt: why Gottman singled it out
Of the four, the Gottman Institute calls contempt “the single greatest predictor of divorce.” The reasoning is easy to feel. Criticism attacks what you did or who you are. Contempt comes, in the Institute’s words, “from a position of relative superiority,” and its target “is made to feel despised and worthless.” It is very hard to repair things with someone who seems to look down on you.
What that ranking rests on is thinner than the confidence suggests. The Institute’s page does not cite a study for it. In Gottman’s published papers contempt is a genuine signal, but it rarely stands out on its own. In the 14-year study, when the researchers used only the four horsemen to tell early divorcers from later ones, contempt mattered for both husbands and wives, but defensiveness separated the groups more strongly. That analysis covered just 22 divorced couples, and the authors noted its relatively low power. A 2025 study of 780 couples in Iran found contempt the most strongly connected of the four to the other conflict measures it tracked. That study looked at couples at a single point in time, so it cannot say what leads to what.
So take contempt seriously. It is the horseman most worth catching early. Just treat “the single greatest predictor” as a reasonable emphasis, not a measured ranking.
Stonewalling, briefly
Stonewalling is the horseman people ask about most, and it gets the most muddled treatment online. Most of the time it is a shutdown under overwhelm, not a chosen punishment: arousal climbs, nothing more can go in, and the person goes blank. Deliberate, patterned silence used to control someone is a different behaviour that happens to look the same from across the room. Telling the two apart matters, and our full guide to what stonewalling really is, and when it becomes something more serious walks through how.
The antidotes, as practice
The framework pairs each horseman with an antidote. One honest note first: these are the framework’s own prescriptions. They have not been tested one by one to show that each works on its own. Treat them as sensible practice, not proven technique.
| Instead of… | Try… | In practice |
|---|---|---|
| Criticism | A gentle start-up | Complain without blame. Say how you feel and what you need: “I felt left out tonight. Can we plan Saturday together?” |
| Contempt | A culture of appreciation and respect | Notice and say the small things you value, often, and outside of arguments |
| Defensiveness | Taking responsibility | Own even part of it: “You’re right that I didn’t call. I’m sorry.” |
| Stonewalling | Physiological self-soothing | Call a break, give it at least twenty minutes, calm your body, then come back |
The start-up deserves more attention than it gets. In a study of 124 newlywed couples, Gottman’s team found they could predict outcomes over six years “using just the first 3 minutes” of a conflict discussion. Be as careful with that as with the 90% figure, though: the independent replication above found that women’s negative start-up, one of the 1998 study’s headline findings, did not predict whether couples stayed together. How you open a hard conversation is still one of the few parts of it you fully control. For the stonewalling row, our guide to de-escalating a heated conversation explains why the break has to be spent calming down, not rehearsing your next point.
One more finding is worth holding onto, and it comes from outside the Gottman programme. In a trial of 134 distressed couples followed for five years, relationship outcomes were “more consistently linked with constructive communication than with destructive communication.” Getting rid of the horsemen matters. Building up the good side of how you talk may matter even more.
What has evidence if you are reading this instead of booking couples therapy
Most people who look up the four horsemen are not in a therapist’s office. They are trying to make sense of an argument from last night. So the most relevant question is whether learning this material on your own does anything.
The best Gottman-specific answer is a 2024 Norwegian study of the 12-hour Gottman Seven Principles course. It included 490 participants and a 242-person control group, measuring relationship quality before, after and six months later. The course “improves couple relationships and is equally effective whether delivered in person or online.” It also made no difference whether the people leading it had a clinical background, “suggesting that the material itself is sufficient.” One important caveat: this was not a randomised trial. It used propensity score matching, a strong way to compare similar groups that can only adjust for the differences researchers measured.
The strongest randomised evidence for largely self-directed couples work comes from a different school. In a 2016 trial of 300 couples, an 8-hour online programme adapted from integrative behavioural couple therapy improved relationship satisfaction compared with a waitlist (Cohen’s d = 0.69), and it improved partners’ depressive and anxious symptoms too. It included four 15-minute calls with staff, so it was not entirely solo. A 2025 meta-analysis of digital couples programmes found a moderate pooled effect on relationship satisfaction (Hedges’ g = 0.42 across six trials), with wide variation between programmes.
Taken together, the message is encouraging and specific. Structured material about how couples argue and repair can help, and it does not seem to need a clinician in the room. That is also a fair reason a conversational format can be useful for this particular material: talking through what happened, noticing which pattern showed up, planning a gentler start-up for next time. aidx.ai offers AI coaching and therapy that can be used that way. It is not couples therapy, and nothing here suggests AI replaces it. And if an argument involves fear, control or your safety, that is not a communication problem. The right next step is a domestic abuse service, not a better technique.
If the same fights keep coming back whatever you try, it is often worth looking at the older pattern underneath them. Our guide to breaking the cycle of repeating relationship patterns is a good place to start.
Common questions
What are the criticisms of the Gottman method?
There are four serious ones. First, the famous divorce-prediction figures came from models tested on the same couples they were built from, and accuracy dropped sharply under crossvalidation. Second, when an independent team tested the 1998 models in a different group of couples, the major findings failed to replicate. Third, researchers challenged the 1998 study’s big recommendations, including dropping active-listening training from couples work, citing nonrandom selection of couples and “the use of correlational data to make strong causal inferences”. Fourth, a 2026 review rated Gottman couple therapy as having modest, not strong, empirical support, held back by small samples and little independent replication. The same review found no couple therapy that met the bar for strong support, so this is not a reason to single Gottman out. It is a reason to be wary of anyone selling any approach as proven.
What are the four behaviors that cause 90% of all divorces?
No four behaviours have been shown to cause 90% of divorces. The phrase garbles Gottman’s claim that criticism, contempt, defensiveness and stonewalling can predict divorce with more than 90% accuracy. Prediction is not causation, and those accuracy figures came from models scored on the same couples they were built from. When such models were checked properly, accuracy dropped “precipitously”. The four behaviours are warning signs worth working on, not a formula for divorce.
What are the 7 principles of the Gottman method?
The seven principles come from Gottman’s book The Seven Principles for Making Marriage Work and the course based on it: enhance your love maps (know your partner’s inner world); nurture your fondness and admiration; turn toward each other instead of away; let your partner influence you; solve your solvable problems; overcome gridlock; and create shared meaning. This is the course that improved relationship quality in the Norwegian study above.
What are the Gottman Four Horsemen and their antidotes?
Criticism is answered with a gentle start-up, contempt with a culture of appreciation and respect, defensiveness with taking responsibility, and stonewalling with physiological self-soothing, which means a real break of at least twenty minutes to calm down before you continue. These pairings are the framework’s own prescriptions and have not been tested one by one.
What is the Four Horsemen?
In relationships, the Four Horsemen are Gottman’s name for four destructive conflict patterns: criticism, contempt, defensiveness and stonewalling. The original Four Horsemen of the Apocalypse come from the Book of Revelation, and Gottman borrowed the name as a metaphor.
Which of the four horsemen is the worst?
The Gottman Institute says contempt. It is the most corrosive to feel on the receiving end, and it was the most central of the four in a large 2025 study. The published prediction data are less clear-cut, though: in Gottman’s own 14-year study, defensiveness did at least as much work as contempt. In practice, the one worth working on first is the one you reach for most.
How many times has Gottman been married?
At least twice. A 2016 profile in the University of Wisconsin alumni magazine notes that he had been married and divorced before he met the clinical psychologist Julie Schwartz, who became his wife and long-time collaborator, Julie Schwartz Gottman. Some sources say his current marriage is his third. We could not confirm the exact number from a reliable source.
Can a relationship recover from the four horsemen?
Yes, often. The patterns are changeable. Scores on a four horsemen questionnaire improved in a randomised couples trial, and a short course in this material improved relationship quality, including for people who took it online. Recognising the pattern is a first step, not a verdict.
The part worth keeping
Strip away the prophecy and the four horsemen are still one of the most practical maps of a bad argument anyone has drawn. Complain about the behaviour, not the person. Watch for contempt. Own your part. Take a real break before you shut down. None of that needs a 90% statistic behind it to be worth doing, and the evidence suggests you can learn it from good material, at your own kitchen table.
Last reviewed: September 2026
This article is general information about relationship communication, not professional advice, diagnosis or treatment. If conflict in your relationship involves fear, control or your safety, contact a domestic abuse service: the National Domestic Abuse Helpline (UK) on 0808 2000 247, or the National Domestic Violence Hotline (US) on 1-800-799-7233. If you are in immediate danger, call your local emergency number.
References
- The Gottman Institute. The Four Horsemen: criticism, contempt, defensiveness, and stonewalling. gottman.com.
- The Gottman Institute. The Four Horsemen: the antidotes. gottman.com.
- Backhaus S, Neumann D, Parrott D, Hammond FM, Brownson C, Malec J. Investigation of a new couples intervention for individuals with brain injury: a randomized controlled trial. Arch Phys Med Rehabil. 2019;100(2):195-204.
- Beeney JE, Hallquist MN, Scott LN, et al. The emotional bank account and the four horsemen of the apocalypse in romantic relationships of people with borderline personality disorder: a dyadic observational study. Clin Psychol Sci. 2019;7(5):1063-1077.
- Gottman JM, Coan J, Carrere S, Swanson C. Predicting marital happiness and stability from newlywed interactions. J Marriage Fam. 1998;60(1):5-22.
- Gottman JM, Levenson RW. The timing of divorce: predicting when a couple will divorce over a 14-year period. J Marriage Fam. 2000;62(3):737-745.
- Heyman RE, Smith Slep AM. The hazards of predicting divorce without crossvalidation. J Marriage Fam. 2001;63(2):473-479.
- Kim HK, Capaldi DM, Crosby L. Generalizability of Gottman and colleagues’ affective process models of couples’ relationship outcomes. J Marriage Fam. 2007;69(1):55-72.
- Stanley SM, Bradbury TN, Markman HJ. Structural flaws in the bridge from basic research on marriage to interventions for couples. J Marriage Fam. 2000;62(1):256-264.
- Jackson JB, Codecá L, Miller RB, Simpson JE, O’Leary AM, Renner SM. Empirically supported couple therapy models for couple relationship distress: a replication and update. J Marital Fam Ther. 2026;52(4):e70162.
- Davoodvandi M, Navabi Nejad S, Farzad V. Examining the effectiveness of Gottman couple therapy on improving marital adjustment and couples’ intimacy. Iran J Psychiatry. 2018;13(2):135-141.
- Deylami N, Hassan SA, Alareqe NA, Zainudin ZN. Evaluation of an online Gottman’s psychoeducational intervention to improve marital communication among Iranian couples. Int J Environ Res Public Health. 2021;18(17):8945.
- Cleary Bradley RP, Gottman JM. Reducing situational violence in low-income couples by fostering healthy relationships. J Marital Fam Ther. 2012;38 Suppl 1:187-198.
- Hashemi B, Zarghami MH, Nodehi D. Understanding marital conflict: a network analysis of Gottman’s four horsemen and attitude toward the past and demographic variables in Iranian couples. J Sex Marital Ther. 2025;51(8):985-997.
- Carrère S, Gottman JM. Predicting divorce among newlyweds from the first three minutes of a marital conflict discussion. Fam Process. 1999;38(3):293-301.
- Baucom KJW, Baucom BR, Christensen A. Changes in dyadic communication during and after integrative and traditional behavioral couple therapy. Behav Res Ther. 2015;65:18-28.
- Zahl-Olsen R, Thuen F, Bertelsen TB. The effectiveness of the in-person and online Gottman Seven Principles Couple Enhancement Program: a propensity score matching design. J Marital Fam Ther. 2024;50(4):882-898.
- Doss BD, Cicila LN, Georgia EJ, et al. A randomized controlled trial of the web-based OurRelationship program: effects on relationship and individual functioning. J Consult Clin Psychol. 2016;84(4):285-296.
- Kernová L, Halamová J, Deriglazov D. Effectiveness of digital interventions on relationship satisfaction among couples: a systematic review and meta-analysis. BMC Psychol. 2025;13(1):1069.
- Gottman Institute. The Seven Principles for Making Marriage Work: couples’ guide. learning.gottman.com.
