What happens in your body mid-fight

Reviewed for evidence accuracy · July 2026

Here's the short version, in case you're reading this with your jaw tight and one ear on the other room.

What's happening to you mid-fight is physical. Past roughly 100 beats a minute (closer to 85 if you're very fit), your body has classified this conversation as an emergency, and the parts of your brain that listen, take another person's perspective, and find creative solutions start running at a fraction of capacity. This is measured, not metaphorical. It's called flooding, and above that line you cannot actually hear your partner, no matter how much you love them or how hard you're trying.

Which means the useful move, the one nobody wants, is to stop talking. And then the break only works if you do the right thing with it, which is almost never what people do.

The rest of this is why, and what the right thing is.

What flooding actually does to you

In the Love Lab at the University of Washington, couples came in, got wired to heart-rate sensors, and were asked to have a real disagreement while researchers coded the interaction second by second. The physiological data turned out to be one of the more durable things the field has produced. Gottman named the threshold state diffuse physiological arousal, and you'd recognize it from the inside without any of the vocabulary: your partner says a sentence with three parts and you catch only the last one, and it's the one that sounded like an accusation. Everything they do reads as hostile. The same three rebuttals loop in your head like they're on a track.

That's not you being difficult or dramatic. It's a threat system doing its job with the wrong information. It can't tell the difference between a bear and the person you have a mortgage with.

It's also why the good phrasing you learned doesn't show up when you need it. There's a real argument in the research about whether communication skills survive contact with a live fight, and the skeptics have a case: scripts tend to evaporate at exactly the arousal level that made you need them. Your body gets there before your training does. That isn't a character flaw, and it's the whole reason the break has to come first.

The other way out of the window

There's a useful frame here, borrowed from Dan Siegel, called the window of tolerance: a band of arousal within which you can feel something strongly and still think. Flooding is leaving that window through the ceiling.

You can also leave it through the floor. Some people, faced with the same escalation, go quiet, blank, and still. Nothing comes out. Gottman's lab called that stonewalling, and it looks, from the other side of the room, exactly like not caring, which is why it's the single most inflammatory thing a person can do while feeling nothing they'd recognize as calm. In much of the process research it's the low side of the same arousal problem.

I want to be careful here, because I don't know your partner and neither does any study. But if you've ever concluded that the person who shuts down in a fight is indifferent to it, the physiology is at least worth holding as a possibility. And if you're the one who goes blank, you already know that "say something" is the least achievable request in the English language when you're in it.

Why "let's just be rational" makes it worse

This is the part that surprises people, and it's the reason so many well-intentioned fights get worse right when someone starts trying to fix them.

Introducing logic to a flooded person means introducing more input into a threat-detection system, and it processes each new input as further threat. Your carefully sequenced account of who said what at your sister's wedding is not information to them. It's an incoming attack, delivered in a calm voice, which somehow makes it worse. Your accuracy is irrelevant. You can be completely right about the facts and still be pouring petrol.

The same goes for the reasonable-sounding requests. Calm down. Be logical. Let's just talk about this. All of them land as pressure on a system that already reads pressure as danger. So do the softer moves: a joke, a hand on the arm, a bid to reconnect. Those work beautifully in a body that feels safe, and they backfire in one that doesn't. Save them.

Twenty minutes, and what you do with them

Heart rate comes back down in roughly twenty minutes. There's no faster route, no breathing trick that beats it, no version where you're a big enough person to skip the wait.

But the twenty minutes is a floor, not a spell. It only counts if you actually let your body settle, and here's the thing the folk wisdom leaves out: replaying the fight resets the clock. Rehearsing your comeback keeps the alarm live. You can sit in the car for forty-five minutes, arriving at a devastating and completely just summary of your partner's character, and come back inside more flooded than when you left.

So the content of the break is the intervention. Not the duration. That single sentence is the most practically useful thing in this essay, and it's why "take a break" fails as advice for most people.

There's a study of Gottman's where couples in the middle of a conflict discussion were interrupted for about half an hour and told to read magazines. Just magazines. When they came back, their heart rates were lower and the conversation went visibly better. The magazines weren't therapeutic. They were absorbing, which is the entire point.

What ruminating looks like from the inside, since it rarely announces itself:

• Global, absolute language. "Always." "Never." "Why does this keep happening." "What's wrong with us."
• Case-building. Marshalling evidence. Scripting the line you'll open with. Keeping score. The phrase "and another thing," said to no one.
• Recruiting a third party. Drafting the text to a friend about how wrong they were. You are not looking for advice. You are looking for a verdict.

What works instead is boring and physical. Walk somewhere. Put on music loud enough to occupy you. Do something with your hands that requires enough attention to crowd out the narration. Active distraction beats sitting with your feelings, and it isn't close. Empty time is rumination time, and "just don't think about it" backfires: suppression rebounds, and the thought comes back with interest.

One more thing about the break, and it's what makes it a break rather than a door slamming. Announce it, and name a return time. "I need twenty minutes, and then I want to finish this." A break that's announced is self-regulation. A break that isn't is an exit, and to a partner who chases when they're frightened, it lands as abandonment, which is its own escalation.

Be ready for it not to land anyway. In the research, a repair attempt (anything that reaches across the escalation, including "can we stop") only works if the other person receives it, and a flooded person is exactly the person least able to hear goodwill in it. Make it anyway. Then go. You're not asking for permission to regulate, and your own arousal is one of the very few things in a fight that is entirely yours to move.

The bridge back

Somewhere in that twenty minutes, if it's going well, you get an opening: enough room to think about the fight without falling back into it.

The move that works there is a strange one. Instead of asking why (why does this always happen, why am I like this, why did they do that), get concrete, and get some distance. What was the moment it turned? Not the theme of the fight, the moment. And watch it as though it happened to someone else, on a screen, from a few rows back. That shift is well studied. It lowers reactivity without the rebound you get from shoving the whole thing down, and it puts you in the right mode for the conversation you're about to have.

Abstract self-questioning is the maladaptive mode. It feels like insight. It's usually rumination with a better vocabulary.

Coming back in

Two things have to be true before you re-enter, and most people check only one.

The clock has to have run, and something in you has to have actually settled. Time alone doesn't do it, for all the reasons above. Self-report alone doesn't do it either, because people under-report their own arousal with impressive consistency. The honest check isn't "am I calm." It's "can I still be curious about what was going on for them?" If the answer is no, you're not back yet, and the right move is more break, with a different distraction, and no shame attached to it.

When you do re-enter, the opening matters more than seems fair. Gottman's process data are consistent about this: how a conversation starts shapes how it ends, and the first minutes carry weight out of all proportion to their length. So don't lead with the case you built. Lead with what you felt and what you need, in that order, about a specific thing that specifically happened.

Then the repair itself, which is smaller than people expect. Each of you says what you felt, without justifying it and without commenting on the other's feelings. Each of you describes what it was like from where you were standing, and then finds one part of the other's version that makes sense given where they were standing. That last bit is the whole game, and it's where everyone balks, because under stress, validating your partner feels like losing. It isn't. Validation says your reaction makes sense from inside your frame. It does not say you were right, it does not concede the facts, and it is not an apology. "Given how that landed on you, of course you were hurt" costs you nothing you were actually defending.

Then each of you names one thing you did that made it worse. Then you agree on one small thing to do differently. That's it. That's the repair.

There's a further move, an old clinical one, where you find the confiding sentence sitting underneath the attacking sentence. "You never listen to me" is almost always carrying something closer to "I felt invisible in that moment and it scared me." Nobody can produce that sentence while flooded, which is precisely the problem, and it deserves its own piece rather than a paragraph here.

And do not try to solve the underlying issue. In the observational research, about 69% of a couple's conflicts are perpetual, rooted in durable differences in personality and need, and here's the part that should be more famous than it is: happy couples have the same proportion of unsolvable problems as couples heading for divorce. Same share. What differs is whether they're in dialogue with it or in gridlock. So repair the rupture. The problem will still be there next month, and that isn't a failure of the relationship.

Where this stops being true

Now the part most articles about flooding leave out, and we're not going to.

The 100 BPM threshold comes mostly from research on straight, married couples, and the arousal model isn't universal. In unhappy same-sex couples, the relationship between arousal and distress can run the other way, with low arousal signalling trouble. Interoception (your ability to read your own internal state at all) varies enormously between people, which makes self-report shakier than the tidy advice admits, and it's especially shaky for neurodivergent people, whose bodies were never the ones being measured. Inferring somebody's arousal from their text messages is exactly where machines are documented to fail, sarcasm and cultural expression being what they are.

The wider version of that caveat is worth saying plainly, because it applies to nearly everything you'll read on this subject: the foundational couples research is overwhelmingly heterosexual, married, often white, and mid-twentieth-century. How well it transfers to dating, to situationships, to queer relationships, to non-monogamous ones, or across cultures is genuinely under-studied. The arousal mechanism is about as solid as this field gets. The instrument for detecting it in yourself is not, and the population it was calibrated on may not be yours.

Two more honesty notes, both about the temptation to overclaim.

That same research programme produced the famous statistic about predicting divorce with better than 90% accuracy. It didn't survive. Heyman's 2001 cross-validation showed the models had been fitted to outcomes already known and never tested on fresh couples; when he built a comparable equation and validated it properly, accuracy fell to around 29%. You'll notice that figure appears nowhere else in this piece. The descriptive science held up. The fortune-telling didn't, and we think you're owed the difference.

And if you've absorbed the nervous-system vocabulary that's everywhere right now (regulated, dysregulated, vagal, shutdown), you should know that its most popular theoretical scaffolding is currently in a serious fight. In February 2026 a large group of researchers argued in print that polyvagal theory's neurophysiological foundations are untenable; Porges answered them in the same issue, and the dispute is live. The experiential language of regulation and co-regulation still describes something real, and we use it. The wiring diagram underneath it is not something we'll stake a claim on, and neither should anyone selling you a course. That's a piece of its own, and we'll write it.

If what you're feeling isn't really about the fight

Everything above assumes two people who are safe with each other and having a bad time. That assumption is doing a lot of work, and it isn't always true.

If you're afraid of your partner, if you find yourself managing your own behaviour to avoid their reaction, if the break isn't a break because there's no version of leaving the room that's safe, then this is a different situation and none of the advice above applies to it. Repair logic can make things worse there, and calling it conflict is a category error. It deserves a person, not a page.

The unglamorous part is the part that works

One more limit, and it's the one that decided the shape of this essay. Almost all of the work above studied two people, both willing, in a room with a researcher. The situation you're actually in (one person, flooded, alone, at eleven at night, with the other one behind a door) is the one the field has the least to say about. We build for that moment, and we try to be careful about what we claim, because the science is unusually clear about one thing here and unusually humble about the rest.

What's clear: you can't think while you're flooded, and neither can they, and the break isn't a failure of nerve. It's the only thing that makes the next twenty minutes different from the last twenty. It's a walk around the block and some loud music, and then coming back through the door and saying the plain thing first.

That's not the interesting part of the essay. It's the part that works.

✦Keep reading

• Three things worth practicing before the hard conversation
• The science we build on
• Can one person change a relationship alone?