Toxic is not a clinical term, which is why it is worth taking apart. It gets used for three very different situations, and the response appropriate to one is actively harmful applied to another. This guide separates them, covers the four behaviours that actually predict a relationship failing, and is explicit about the line where the advice stops being about improving a dynamic.
Three different things wearing one word
A poor match. Two people who are not cruel and who nonetheless make each other worse. Nobody is doing anything wrong and it is still not working. What this needs is honesty, and frequently an ending.
A damaging pattern. A dynamic neither person wants and both maintain, usually a loop where each response makes the next one more likely. This is what couples therapy is genuinely good at, and it improves when both people want it to.
Abuse. A pattern of control, intimidation or violence. This is not a shared problem to work on jointly, and treating it as one is dangerous: couples therapy is generally contraindicated where there is coercive control, because it gives an abusive partner information and a new arena.
Most writing on this subject moves between the three without noticing. Working out which one you are in is the first useful thing to do.
What actually predicts breakdown
Gottman’s observational research is the most useful body of work here, because it watched couples interact and then followed them, rather than asking them to describe themselves. [gottman-1994-what] Four behaviours predicted separation with unusual accuracy.
| Behaviour | What it sounds like | The antidote |
|---|---|---|
| Criticism | ”You never help” rather than “I need help tonight” | Describe the need, not the character |
| Contempt | Sarcasm, mockery, eye-rolling, superiority | Deliberate expressions of regard; this is the one that matters most |
| Defensiveness | Counter-attack, or wounded victimhood | Accept any part of the complaint that is fair |
| Stonewalling | Going silent, shutting down, leaving the room | Name it, take twenty minutes, come back |
Contempt is the strongest single predictor of separation. It is also the one people most reliably underestimate, because a raised voice feels like the serious thing and an eye-roll does not.
Stonewalling is worth defending slightly. It is often not indifference: the person doing it is frequently physiologically overwhelmed, with a racing heart, and has stopped being able to process anything. [gottman-1992-behavior] That does not make it harmless, and it does change the fix, which is a stated pause rather than pushing through. Our guide to stonewalling covers it in detail.
Conflict is not the problem
The most useful correction in this literature is that stable and unstable couples are not well separated by how much they argue. Some very stable couples argue a great deal.
What separates them is the balance of positive to negative interaction, and whether repair attempts get accepted. A repair attempt is anything that tries to de-escalate: a joke, an apology, a hand on an arm. In struggling couples these are made and not received, which is a different failure from not making them.
The practical consequence is that a relationship with no visible conflict is not automatically healthy. Quiet contempt outperforms loud argument as a warning sign.
Which of the three is this?
Tick what is true of the past few months. The aim is to work out which situation you are in, because that decides what would actually help.
0 of 6 ticked
Isolation, monitoring, control of money and fear together describe coercive control rather than a bad dynamic. That needs specialist support rather than joint therapy, and it is worth talking to someone outside the relationship.
The earlier items describe a damaging dynamic, which is what couples work is genuinely good at when both people want it to change.
None of the markers are present. Ordinary conflict without these is not what this page is about.
No relationship screener is published on this site. If low mood or anxiety has built up around this, the hub has instruments for both.
The line where this stops being about dynamics
Coercive control is a pattern of intimidation, isolation, monitoring and regulation of everyday life, and Stark’s account of it changed how the behaviour is understood: the harm is in the cumulative pattern of constraint rather than in any single incident. [stark-2007-coercive] It frequently occurs without physical violence.
That is why it takes so long to see from inside. Each incident, described alone, sounds explicable, and the person describing it often ends up sounding unreasonable. Our guides to coercive control, emotional manipulation and the grey rock method cover the specifics.
Two things are worth stating plainly. Couples therapy is generally not appropriate where coercive control is present. And leaving is the highest-risk period rather than the end of the danger, which is why specialist domestic abuse services exist, why they plan rather than instruct, and why “just leave” is poor advice offered confidently.
When to seek help
Speak to a doctor or a therapist if you have started doubting your own memory of events, if you are managing your behaviour continuously to avoid a reaction, or if you have withdrawn from people who used to know you well.
If any of the last three items in the check above apply, contact a specialist domestic abuse service rather than a general counselling service, and consider doing it from a device the other person does not have access to.
If you are in immediate danger, contact your local emergency services.
How MyFreud can help
When your account of events keeps getting overwritten, a record you wrote at the time is the thing that holds. Tracking mood alongside what happened builds that over weeks, and it is very often the point at which people stop wondering whether they are imagining it.