The sentence people say most often, after a controlling relationship finally becomes visible to those around it, is that it looked fine. Not that they had a bad feeling they talked themselves out of. Not that something was off and they let it go. It looked fine, and they are telling the truth.
That is the part worth sitting with, because the usual explanations do not account for it. We tend to assume the observers were careless, or too polite to ask, or that he was an unusually skilled performer. Sometimes that is true. But it does not explain the sheer consistency of the failure, and consistency at that scale is almost always structural rather than personal. Why coercive control is hard to spot is not a question about the attentiveness of the people watching. It is a question about what they were looking at.
They were looking at a system in equilibrium. What they could not see is who was paying for it.
THE PERSON HOLDING THE SYSTEM STEADY
Any system that stays stable in changing conditions has something in it doing the absorbing. Pressure arrives, and something flexes so that the overall state does not. In engineering terms, a regulator has to be able to meet at least as much variety as the system throws at it, otherwise the system destabilises and the instability becomes visible.
In a controlling relationship, the target becomes the regulator. She learns the moods and starts forecasting them. She adjusts the plans, manages the friends, softens the news, arrives early, explains before she is asked. Every unpredictable thing that could produce a scene gets absorbed by her before it reaches the surface. She becomes extremely good at this, because the cost of being bad at it is high and immediate.
The result is a relationship that presents as calm. What an observer sees is the output of an enormous amount of continuous labour, and the labour itself is invisible by design, because any part of it that showed would defeat its own purpose. The stability is not evidence that the system is healthy. It is evidence that someone is holding it steady, and the more effectively she does it, the less anyone can tell.
This produces an unpleasant inversion. The better the target is at absorbing, the more competent she appears and the more normal the relationship looks. The situation only becomes legible to others at the point where her capacity to absorb finally runs out, which is generally very late, and which is also the point at which she looks least reliable as a narrator.
ELAINE, AND THE COUPLE WHO PRESENTED WELL
COMPOSITE NOTE: Elaine is not a real person. She is a composite drawn from patterns that recur across many practitioners, and nothing here describes any individual case.
Elaine is a relationship counsellor with fourteen years of practice. A couple came to her for six sessions. He was open, self-critical, quick to acknowledge his part and visibly frustrated by his own temper. She was harder to reach: guarded, careful, giving short answers, and reluctant to name anything specific when invited to.
Elaine’s clinical read, formed early and reasonably, was that he was doing the work and she was withholding. The sessions began to organise themselves around drawing her out. Elaine noticed she felt more warmth toward him than toward her, registered that as her own countertransference, took it to supervision and adjusted for it. The dynamic did not change.
What Elaine had no way of seeing is that the woman in front of her went home with him afterwards. Every honest sentence spoken in that room carried a cost that would be collected later, out of sight, and possibly not that evening but at some unpredictable point over the following fortnight. She was not withholding. She was calculating, accurately, in an environment where candour was genuinely dangerous. Her guardedness was not resistance to the work. It was competence at survival, misread as resistance because the format had no way of representing it.
WHY THE FORMAT PRODUCED THE ERROR
Couples work rests on an assumption that holds in most cases and fails completely in this one: that both parties have roughly comparable power and a shared interest in the truth coming out. Where that assumption holds, the format is a good instrument. Where it does not, the format converts into something else entirely, because it creates a setting in which one party can speak freely and the other cannot, and then treats the difference between them as clinical data.
The person who can afford honesty appears insightful. The person who cannot appears defensive. A skilled practitioner working in good faith can be led, by the structure of the session itself, to an assessment that is precisely inverted. This is not a criticism of the practitioners involved. It is a description of what happens when an instrument is used outside the conditions it was designed for, which is a thing that can only be avoided by knowing that the conditions exist.
THE LAW HAD TO INVENT A NEW SHAPE IN ORDER TO SEE IT
The clearest evidence that this is a structural visibility problem rather than an attentiveness problem comes from the institutions that have wrestled with it hardest.
When controlling or coercive behaviour was criminalised in England and Wales under section 76 of the Serious Crime Act 2015, the offence had to be defined as behaviour occurring repeatedly or continuously. A course of conduct, in other words, rather than an act. The existing legal apparatus, built almost entirely around discrete incidents, was structurally incapable of holding the thing, and a new category had to be constructed for it.
Even then, the problem did not resolve. Research by Barlow and colleagues, examining how the offence was being handled in practice, found the pattern dissolving inside the recording systems themselves. Police records are built incident by incident, and an incident-shaped record cannot represent a pattern whose entire meaning lies in the accumulation. Each entry, taken alone, looks like very little. The relationship between the entries is the offence, and the relationship between the entries is exactly what the format discards.
The recorded-offence figures tell the same story from another angle. Recorded coercive control offences in England and Wales rose from around nine thousand in the year to March 2019 to nearly fifty thousand in the year to March 2025. That is not a five-fold increase in controlling behaviour. It is a five-fold increase in the ability to see it, which is a measure of institutional recognition and should never be presented as prevalence. The behaviour was always there. The instrument for registering it was not.
WHAT CHANGES WHEN YOU LOOK FOR THE SHAPE RATHER THAN THE EVENT
You stop waiting for an incident large enough to justify the concern.
If you are looking for a moment that would sound serious when described to a third party, you will be waiting a long time, and in many cases forever. There may be no such moment. The absence of one is characteristic, not exculpatory.
The absence of red flags stops being reassuring.
There is a paradox in this work that catches almost everyone. What people describe as red flags are usually visible precisely because they are not working yet. A pattern that has succeeded produces no flags at all, because the resistance it would have provoked has already been dissolved. Smoothness is not the opposite of the problem.
You reassess what guardedness means.
A person who gives careful, minimal, hedged answers about her own life is telling you something about the environment she returns to. That reading is not automatic and should not be applied indiscriminately. But it should be on the list of available interpretations, and at present it usually is not.
You get careful about who you put in a room together.
The single most consequential shift is procedural rather than perceptual. Once you know that the joint format can invert the picture, screening before joint work stops being an administrative nicety and becomes a clinical safeguard.
HOW TO LOOK
Ask about the shape of the account rather than its content. Can she give you a straight, unhedged description of one specific event? Does the account arrive pre-defended against objections nobody has raised?
Ask what happens after honesty. Not what happens during arguments, which she will have rationalised, but what the following days look like when she has said something true. The answer to that question is frequently the most diagnostic thing in an hour.
Ask what she has stopped doing. Range of movement over time is far more informative than any single event, because the erosion works by increments too small to notice individually. There is a reason philosophers use the heap of sand to illustrate this problem: no single grain makes a heap, and there is no moment at which a relationship visibly became abusive. Removed from the sequence, every individual step looks reasonable, which is why she cannot locate the moment either.
And take seriously the possibility that an outside reading beats a self-report here. The Gottman research on predicting relationship outcomes from structured observation is instructive on this point, with the caveat that its striking accuracy figures come largely from retroactive analysis on model-building samples and prospective performance is lower. The principle survives the caveat. Something watching from outside the system can see what the people inside it cannot, and in this particular situation the person inside has had her observational apparatus deliberately interfered with.
THE LENS, NOT THE FACTS
In 2019, the Court of Appeal quashed Sally Challen’s murder conviction. Almost nothing about the facts of the case had changed in the intervening eight years. What had changed was the availability of a frame capable of holding them, and with that frame in place the same evidence produced a different verdict.
That is the whole of it. There is no version of this work in which we look harder at the same picture and finally see it. The picture has been available the entire time. What determines whether anything appears in it is whether we brought an instrument capable of representing a pattern, or arrived, as almost everyone still does, looking for an incident.
