You already know the advice. Use "I" statements. Listen without interrupting. Don't make it about their character. It's decent advice, and it tends to desert you at the exact moment you reach for it, because almost nobody tells you the condition it runs on.
Communication skills only work in a body that can still hear. Above roughly 100 beats a minute, the parts of the brain that listen and take another person's perspective go partly offline, and a well-phrased sentence offered at that moment gets processed as one more thing coming at you. So every technique below is downstream of one question: can you still be curious? If the answer is no, the skill isn't the move. The break is.
That's the gate. Here's what sits behind it.
The gate: can you still be curious?
Gottman's lab wired couples to heart-rate monitors and watched what happens to a person mid-fight. Somewhere around 100 beats a minute (lower if you're athletic), the body stops treating the argument as an argument and starts treating it as a threat. Hearing narrows. The systems that listen, empathize and solve problems go partly offline. Above that line a person cannot take in what their partner is saying. Not won't. Can't.
Two things follow, and they're why this page opens the way it does.
The first: reasoning with someone in that state makes it worse. Laying out the facts, asking them to calm down, suggesting a better way to phrase it. A threat-detection system reads every new input as more threat, and you've just supplied one.
The second is the part the skeptics of this whole field have right, and we'd rather say it ourselves than have you discover it mid-fight. There's a long-running argument in the research about whether teaching couples communication skills changes anything, or whether the skills simply evaporate under flooding while only deeper shifts hold. The evidence leans toward the skeptics on skills taught alone. If your plan for a hard conversation is a script, the script is the first thing you lose access to when you need it, which is why anyone selling you phrases without mentioning your nervous system is selling you the easy half.
Calm, meanwhile, is a lot to ask of someone who's just been hurt, and it's the wrong bar anyway. Curiosity is the workable one. Can you still genuinely wonder what's going on for the other person, as opposed to rhetorically wondering it on the way to a verdict? If yes, you're in range, and the three moves below are available to you. If no, the honest next step is twenty minutes and a real distraction: a walk, music, something absorbing enough to occupy you. Stress chemistry needs roughly that long to clear and it only clears if you stop feeding it. Twenty minutes of composing your rebuttal isn't a break. It's a rehearsal, and it resets the clock.
The lab version of this is almost comically plain. Gottman interrupted couples mid-conflict and had them read magazines for half an hour. They came back with lower heart rates and had a better conversation than the couples who pushed straight on through.
The long version of that is next door: what happens in your body mid-fight.
1. Open like the opening is the whole thing
Gottman's process research on how conflict conversations unfold found that the first few minutes carry unusual weight. How a conversation opens shapes where it lands, and a harsh opening is very hard to walk back. (You may have seen the famous claim from that lab that it can predict divorce with 90-odd percent accuracy. It can't. When another researcher built a model that looked just as accurate and then tested it on a fresh sample, the accuracy fell to around 29%. The descriptive findings from that work are solid and we build on them. The prediction percentages aren't, and you won't catch us repeating them.)
Which makes the first sentence the highest-leverage sentence you have. Most of us spend it on a verdict.
"You never listen to me" is a verdict. "I felt invisible at dinner and I want to understand what happened" is a complaint with a person still inside it. Same grievance, different opening move, wildly different odds.
The usual instruction here is a formula: I feel X when Y, and I need Z. The formula is scaffolding, and it has two failure modes worth knowing. It can be weaponized, which anyone who has been on the receiving end of "when you leave the dishes, I feel disrespected" will recognize as an accusation with better grammar. And it's simply the wrong instrument for plenty of people. If you're autistic, or facts-first, or you find that naming a feeling on demand returns nothing but static, then the honest opening might be a plain literal request, or a correction of something that was factually wrong. That works. What has to be absent is the character verdict. The rest is style.
2. Ask one question, then leave the silence alone
Where the first and third moves rest on measured research, this one mostly rests on clinical practice, and we'll say so rather than dress it up:
What the research does describe is the mechanics of feeling heard, and it's fairly consistent. People report feeling heard when what comes back reflects the feeling and the need underneath what they said, and when no rebuttal arrives on its heels. They report feeling more attacked by four specific things: the word "but," being corrected on the facts, being problem-solved at before they've finished, and any defense that shows up before an acknowledgment.
Look at that list. It's almost entirely made of things you do while you're waiting for your turn.
So the one-question discipline is mainly a device for not doing them for thirty seconds. Ask one real question ("what was the worst part of that for you?" beats most of them) and then don't fill the pause with your interpretation of the answer. The silence is doing work.
One thing worth knowing while you're in there: validation isn't agreement. "I can see how that landed as dismissive" concedes nothing you were actually defending. Under stress it feels like surrender, which is why so few people manage it, and why it moves so much when someone does.
3. Keep the pattern and the person apart
The third move is the oldest and best-credentialed of the three. In Gottman's Four Horsemen, what separates criticism from complaint is a single move: a complaint is about a behavior, criticism is about a character. "You forgot to call the plumber" is a complaint. "You never follow through on anything" is criticism, and it reliably buys defensiveness in return, because there's nothing else available to do with it.
It's worth knowing where that road goes. Criticism's next-door neighbor is contempt, the one where the character verdict acquires an eye-roll, and contempt is the single most corrosive thing the research has found people do to each other. Every school in this field agrees on that, which is rare enough to be worth listening to.
The same distinction rescues the fight that won't die. In that same body of observational work, roughly 69% of what couples fight about turns out to be perpetual: it recurs for years, rooted in stable differences of personality, values and need. The part that gets dropped from the pull-quote is the part that matters. Thriving couples carry the same share of perpetual problems as the couples who split up. The difference isn't how many unsolvable things they've got. It's whether they keep a live conversation going with theirs or grind into gridlock.
So "we keep having this argument about money" isn't evidence of failure. It may be a fair description of the permanent shape of the thing, and the goal quietly shifts from solving it to handling it without damage.
A word about whose relationships this research watched
Nearly everything above comes from decades of observational work with married couples who were mostly straight, mostly white, and mostly filmed in the last century. That's the sample the field was built on. How cleanly it transfers to dating, to queer relationships, to non-monogamous ones, to relationships where one or both people are neurodivergent, is contested and genuinely under-studied. We'd rather hand you that than quietly imply the lab was watching you.
Our read is that the physiology travels furthest, because a nervous system under threat behaves like a nervous system under threat. The interpretive moves travel less far. Take the three above as a strong starting hypothesis about your relationship rather than a law about it, and let what actually happens in your house outrank the study.
If you're the only one trying
Here's where a page like this normally tells you that if you start with yourself, the relationship follows. We're not going to, because the best evidence we have says something more complicated, and you deserve the version we'd want if we were the ones reading.
The OurRelationship program has been tested in two forms: a couple version, with both partners in it, and an individual version with the same content and one partner working alone. The solo version improved the person: their wellbeing and their day-to-day functioning both went up. It did not move relationship satisfaction. Insight delivered to one person improved that person, and stopped there.
What did move relationships, in the research where anything moved them, was behavior. Finkel and colleagues had people spend a few minutes, a few times a year, writing about their fights from a neutral third party's point of view. One person, alone, doing something specific. Marital quality held up over the following year where it otherwise decayed. And CRAFT, a program that coaches one person in how to respond differently to a partner who refuses to come in, has shifted even resistant partners and the relationship around them. CRAFT was built for substance-use problems rather than ordinary conflict, so reading it across to your Tuesday-night argument is our extrapolation and not a proven finding. We'd rather flag that than lean on it.
The through-line is worth having anyway: understanding doesn't transmit between two people, and behavior does. A relationship is a feedback loop, and you hold half of the inputs.
Which turns this whole page into one practical instruction. If you're doing this alone, don't end with a realization. End with a behavior. One different thing you'll actually do or say next time, small enough that you'll really do it.
Our read of that (that changing your half is a genuine lever on the relationship, and not only on you) is a bet, and this is how we label our bets: What it does to your half of the loop isn't a bet at all. That part is the lever you have, and it's a real one.
One boundary on all of this, and it isn't fine print. Everything above assumes two people who are, underneath it, on the same side. If you're afraid of your partner, if you're managing your own behavior to keep someone from turning, or if "change your half" is functioning as a way to hold you responsible for someone else's cruelty, then this page is the wrong tool and the frame itself can do harm. That situation has its own kind of help, and reaching for it isn't something you have to earn by trying harder here first.
Here's the gap this page keeps pointing at. Every framework above assumes both partners are present, both are willing, and both are calm enough to do the work. One person, alone, mid-fight, at eleven at night, with a partner who isn't coming to the table: that's the moment the whole field has no scalable answer for, and it's the moment we're building for. The tool isn't open yet, and we're not going to pretend the in-the-moment version of any of this is settled science. Whether coaching survives compression into the middle of a live fight is exactly the question we're building against. In the meantime the method is here, whole, given away. It's more use to you tonight than a waitlist is.
Sources
- Gottman, J. M., & Levenson, R. W. (1992). Marital processes predictive of later dissolution: behavior, physiology, and health. Journal of Personality and Social Psychology. The flooding and physiological-arousal findings.
- Carrère, S., & Gottman, J. M. (1999). Predicting divorce among newlyweds from the first three minutes of a marital conflict discussion. Family Process. The start-up finding, which is descriptive and holds; the prediction claims built on this program of research are the part we don't use.
- Heyman, R. E. (2001). The hazards of predicting divorce without cross-validation. The critique of those prediction claims: a model that looked similarly accurate fell to roughly 29% when tested on a fresh sample.
- Gottman, J. M., & Silver, N. (1999). The Seven Principles for Making Marriage Work. Criticism versus complaint (the Four Horsemen), the 100-beat threshold, and the perpetual-problem share.
- Doss, B. D., et al. (2016). A randomized controlled trial of the web-based OurRelationship program. Journal of Consulting and Clinical Psychology. The individual-format trial of the same program is the source of the solo finding above.
- Finkel, E. J., Slotter, E. B., Luchies, L. B., Walton, G. M., & Gross, J. J. (2013). A brief intervention to promote conflict reappraisal preserves marital quality over time. Psychological Science.
- Miller, W. R., Meyers, R. J., & Tonigan, J. S. (1999). Engaging the unmotivated in treatment for alcohol problems: a comparison of three strategies. Journal of Consulting and Clinical Psychology. The CRAFT engagement trial.
Where the science is open, we've said so in the body rather than in a footnote.
Something here off, broken, or not sitting right? Tell us — a person reads it.
People and books appear here because we’d recommend them with nothing to gain. Some may partner with us over time; no one is listed because of it.