Red flags in a relationship — difficult, or dangerous?

One of the lists that ranks for this puts “physical, mental, or emotional abuse” at number four and “conflict avoidance” at number nine, in the same typeface, with the same bullet. A list that weighs those equally cannot answer the question you arrived with.

The genre has settled on ten. Ten red flags, ten signs of a toxic relationship, ten things to watch for — bulleted, equal length, equal weight. It is a format that cannot fail to be read and cannot possibly help, because the work you need doing is the sorting, and the format has already flattened it.

The flattening does damage in both directions, which is the part nobody writing these seems to consider. Someone in an ordinary rough patch reads sixteen items, ticks four, and concludes their relationship is toxic. Someone living under sustained control reads the same sixteen, ticks four, and concludes the same thing — and “toxic” is a word that implies two people, a shared dynamic, and something couples work could fix. The list quietly tells the reader in most danger that she has a communication problem.

What the people who measure this actually measure

There is a whole field here — forty years of it, indexed and free to find — and it does not work in traits at all. The instruments researchers and advocates trust do one or more of three things that no list does.

• They measure frequency and intensity over time. The Checklist of Controlling Behaviors runs to eighty-four items across ten areas — isolation, intimidation, threats, economic abuse, minimising and denying and so on — and it was validated on more than two thousand women staying in refuge. What it records is not whether a thing happened but how often and how hard.
• They measure what it is like to be there. The Women’s Experience with Battering scale was built out of interviews rather than out of a list of acts, and it asks about the person answering: their sense of vulnerability, their loss of power, their altered sense of themselves.
• They measure danger. The Danger Assessment is a validated instrument for estimating the risk of being killed by a partner. It exists, advocates use it, and nothing in the listicle genre acknowledges that such a thing is possible.

Three questions that do the work the list cannot

These are not a quiz and there is no score. They are the three axes the instruments measure on, in language you can actually apply to your own week.

Contempt, and the number you have seen everywhere

The one piece of research that travels through this whole subject is Gottman’s Four Horsemen, and it is worth having properly rather than as a half-remembered bullet.

Contempt is the one to take seriously. It is the behaviour most consistently found in couples who come apart, it is the hardest of the four to explain away, and unlike most of what gets called a red flag it is visible in a single evening.

Now the part almost nobody prints. You have probably met the claim that this work predicts divorce with around ninety-four per cent accuracy. That number is not a finding about your relationship, and three things about it are worth knowing.

• The study it is pinned to followed 73 couples for four years — 1983 to 1987 — and its own abstract says its central model “received preliminary support”. There is no accuracy figure in it at all.
• The accuracy was calculated backwards. The team already knew which couples had divorced, fed those outcomes in with the coded video, and derived the equation that best separated the groups. Fitting a model to known answers and then reporting how well it separates them is not the same act as predicting.
• Serious marital researchers said so in print. Stanley, Bradbury and Markman published a critique in the same journal concluding that the recommendations “exceed the scope and methods of the underlying research”.

None of which means contempt does not matter. It means contempt is a signal worth acting on and not a verdict that has been handed down about you.

Where the line actually is

The word the field uses is coercive control, and the reason it matters here is that it is not a more severe version of a bad relationship. It is a different thing with a different shape: isolation from people, surveillance, intimidation, degradation, and the regulation of ordinary life — money, clothes, food, sleep, where you go and when you come back. The researcher most associated with it describes the result as entrapment, and argues that the harm is the regime rather than any incident inside it.

Which produces the single most important sentence in this literature, from a review by Stark and Hester:

Coercive control is not a type of violence. Indeed, level of control predicts a range of negative outcomes heretofore associated with physical abuse, including post-separation violence and sexual assault.

Stark & Hester (2019), Violence Against Women

Read that twice. Control is not the mild end of the scale — it is the better predictor. And because the harm is a pattern rather than an event, there is, by design, no incident to point at. Somebody describing it on a forum put the mechanism better than the literature does: it starts so small, with tiny things that grow bigger and bigger until they take over your entire life. She said it took her thirteen years to name it.

Why people stay, which is not what you have been told

There was never a moment big enough

Every step is small relative to the one before it. That is not a failure of judgement; it is what a gradual pattern does to the person inside it, and it is exactly what you would expect if the harm is a regime rather than a series of events.

The good days are the mechanism, not the consolation

Unpredictable reward is stickier than reliable reward — the oldest finding in behavioural psychology, and the reason slot machines work. Applied here, intermittent kindness and remorse make the bond feel stronger than it would in a steady relationship, not weaker. It is also why “but he can be so lovely” is evidence, and not evidence against.

Self-blame, not sunk cost

The economists’ story is that people stay because of what they have already put in. What people actually write is different, and much sadder: they assumed it was them. That they were the difficult one, the one who could not make themselves understood, the one asking too much.

And the other error: if everything is a red flag, nothing is

There is a real cost on the other side, and a clinical psychologist has written a whole book about it. Her argument has two edges and the page needs both. The first is that therapy vocabulary has spread far enough that people now leave ordinary, imperfect, fixable relationships because a word fits loosely. The second is sharper: when “gaslighting”, “narcissist” and “abuse” are used for everything, they stop carrying weight — and the people they were coined for are believed less.

Her line about the flags themselves is the right note to end the diagnostic half on: they are not always deal-breakers. Sometimes they are just where the growth starts.

The case nobody writes about

Here is the gap. Every page on this subject assumes a wrongdoer. There is healthy, and there is your partner doing something to you, and nothing in between.

But a great many relationships that are struggling are not like that at all. Nobody is behaving badly. Two people want different things, or want them at different times, and neither has done anything you could put on a list. There is no villain, so nobody is writing for them — except divorce lawyers, who rank well on this phrase and whose business is the filing.

Where inbetween fits, and where it must not be used

The second thing first, because it matters more.

If what you have read here about coercive control is your situation, this app is not for you. The National Domestic Violence Hotline does not encourage anyone in an abusive relationship to seek counselling with their partner, and their reasoning applies to us at least as much: what you disclose can be used against you, the power difference is not something a fair process can even out, and anything that carries messages between two people is one more channel to be monitored through and punished by. Call the number at the top of the page. Do not use this.

Where inbetween is genuinely the right thing is the case in the section above — the one with no villain in it. You talk to it, they talk to it, and it is the only one who hears both sides, so you can say the unreasonable version of what you want without it landing on them. Nothing crosses unless you allow it. That is useful when two people cannot work out what the other actually wants. It is useless, and worse than useless, against somebody who is choosing this.

If you have decided it is the first kind, the trust page covers what comes next, and what a first therapy session is like is a plainer description than most.