The Four Horsemen of Relationship Failure — Gottman's Warning Signs
Last reviewed by the Men Women Psychology editorial team.
The evidence
What the research actually shows
Working from his observational lab, John Gottman and Robert Levenson (1992) found that the way couples argue predicts later divorce with notable accuracy. They isolated four corrosive behaviors Gottman later nicknamed the 'four horsemen': criticism (attacking character rather than raising a complaint), contempt (mockery, sarcasm, eye-rolling, an air of superiority), defensiveness (counter-attacking or playing the victim instead of taking any responsibility), and stonewalling (shutting down and withdrawing from the interaction).
Across this body of work, contempt stands out. Gottman and Silver (1999) describe it as the most poisonous of the four and the strongest single predictor of divorce, because it communicates disgust and looks down on a partner rather than merely disagreeing with them. Research suggests contempt does measurable harm — couples who show it tend to report worse outcomes over time. None of the four is unique to one sex; both partners can deploy any of them.
There is one modest average pattern worth noting. Christensen and Heavey (1990) documented the demand-withdraw dynamic, in which one partner presses an issue while the other retreats. On average, women are somewhat more likely to occupy the demanding role and men the withdrawing or stonewalling role, but this reverses depending on who wants change, and the overlap between the sexes is large rather than absolute.
The mechanism
Why this happens
The horsemen tend to escalate in sequence. A recurring complaint that never gets resolved can harden into global criticism ('you always,' 'you never'), which invites defensiveness rather than repair. Over time, unaddressed criticism can curdle into contempt — a settled sense that the other person is beneath respect. Stonewalling often arrives last, as one partner, overwhelmed, simply stops engaging.
Stonewalling in particular is frequently driven by what Gottman calls flooding: when physiological arousal spikes during conflict, heart rate climbs and the capacity to listen or think clearly drops. Withdrawing can be an attempt to self-soothe rather than an act of indifference. This response is common in both sexes, though some research suggests men may flood somewhat more easily on average, which can feed the withdrawing role in the demand-withdraw cycle.
Contempt usually grows from long-stored resentment and a habit of scanning for a partner's faults rather than their strengths. When admiration and fondness erode, small irritations get reinterpreted as evidence of a bad character. The mechanism is less about a single fight and more about an accumulating climate of negativity that reshapes how each partner sees the other.
In practice
What this looks like in real life
A complaint sounds like 'I felt alone when you didn't call'; criticism sounds like 'you're so selfish, you never think about anyone but yourself.' The first targets a behavior in a moment; the second indicts the whole person, and tends to put the listener on the defensive before the real issue is even discussed.
Contempt is often nonverbal — the eye-roll, the sneer, the sarcastic 'oh, nice job.' A partner on the receiving end frequently feels not just criticized but diminished. This is the pattern most worth catching early, because it signals that respect itself is eroding.
Stonewalling can look like one partner going silent, looking away, busying themselves with a phone, or leaving the room mid-argument. From the outside it reads as cold or dismissive; internally the person is often overwhelmed and shutting down to cope rather than trying to punish.
Myth vs. evidence
What most people get wrong about this
The common misreading is that conflict itself is the danger. It is not — Gottman's research finds that happy and unhappy couples argue about the same things and argue often. What separates them is how they argue: the presence of the horsemen, not the presence of disagreement. Couples who never fight are not necessarily healthier; sometimes they have simply stopped engaging.
Another myth is that these patterns map neatly onto gender — that women nag and men withdraw. The behaviors are human, not male or female, and any partner can fall into any of the four. The demand-withdraw tendency is a soft average with heavy overlap, and it flips readily depending on which partner is seeking change in a given conversation.
Why it matters
What this means for relationships
Each horseman has an antidote that can be practiced. Criticism is countered by a gentle start-up that names a feeling and a specific need; defensiveness by taking responsibility for even a small part of the problem; contempt by deliberately building a culture of appreciation and fondness; and stonewalling by recognizing flooding and taking a genuine break to calm down before returning to the conversation.
The presence of the horsemen is a signal to change how you communicate, not a verdict that a relationship is doomed. Many couples learn to spot these patterns in themselves, interrupt them, and rebuild. Catching contempt early matters most, because rebuilding everyday respect and admiration is the foundation everything else rests on.
Where it varies
The nuance
These patterns describe couples on average, and individuals vary enormously. Janet Hyde's gender similarities hypothesis (2005) reminds us that on most psychological measures men and women are far more alike than different, so the mild demand-withdraw skew should not be read as a fixed rule about either sex.
Attachment style, stress, culture, and what each partner learned about conflict growing up shape which horsemen show up and how strongly. A securely attached person tends to repair more readily; someone who learned that conflict is dangerous may withdraw faster. The framework is a useful map, not a diagnosis of any one relationship.
Questions people ask about this
What are the four horsemen?
They are four destructive conflict behaviors identified by John Gottman: criticism (attacking character), contempt (mockery and superiority), defensiveness (counter-attacking or playing victim), and stonewalling (shutting down and withdrawing). When they become habitual, research suggests they predict relationship breakdown.
Which horseman is the most damaging?
Gottman's research points to contempt as the most corrosive and the strongest single predictor of divorce. Unlike ordinary disagreement, contempt communicates disgust and superiority, signalling that respect itself is eroding. It tends to do the deepest damage over time.
Is stonewalling the same as abuse?
Not usually. Stonewalling is often driven by emotional flooding — feeling overwhelmed and shutting down to self-soothe, not to control. It still harms connection and warrants attention. Deliberate, controlling silent treatment is a separate concern; context and pattern matter when telling them apart.
Do men and women show different horsemen?
On average, men are somewhat more likely to stonewall and women to take the demanding role in a demand-withdraw cycle, but this reverses depending on who wants change. The overlap is large, and any partner can fall into any of the four patterns.
Can a relationship recover from the four horsemen?
Often, yes. Their presence signals a need to change how you communicate, not an automatic ending. Each horseman has a practiced antidote, and many couples learn to catch and interrupt these patterns. Recovery tends to be hardest, though not impossible, once contempt is entrenched.
Does arguing mean a relationship is failing?
No. Gottman's research finds happy and unhappy couples argue about similar things and argue often. What distinguishes them is how they argue — whether the horsemen are present — not whether they disagree. Conflict handled with respect can actually strengthen a bond.
Research sources
These references point to the published research and established frameworks behind this page. They are provided for further reading; see our research methodology for how sources are selected.
- Gottman, J. M., & Levenson, R. W. (1992). Marital processes predictive of later dissolution. Journal of Personality and Social Psychology, 63(2), 221–233.
- Gottman, J. M., & Silver, N. (1999). The Seven Principles for Making Marriage Work. Crown.
- Gottman, J. M. (2011). The Science of Trust: Emotional Attunement for Couples. W. W. Norton.
- Christensen, A., & Heavey, C. L. (1990). Gender and social structure in the demand/withdraw pattern of marital conflict. Journal of Personality and Social Psychology, 59(1), 73–81.
- Hyde, J. S. (2005). The gender similarities hypothesis. American Psychologist, 60(6), 581–592.