How to ask, and when

Most people rehearse this conversation for weeks and then open it badly, usually at eleven at night, usually after a drink, usually starting with the sentence they promised themselves they would not start with. The rehearsal does not help because it is rehearsing an argument. What you want is a conversation, and those are opened differently.

Decide one thing first

Before you say anything, answer this for yourself: what do you want to happen afterwards?

There are only a few honest answers. You want to be reassured. You want the behaviour to stop, whatever it is. You want to know so you can decide. Or you have already decided and you want to say so.

These are four different conversations and they do not open the same way. The one that goes worst is the one where you have not picked — you ask a question shaped like you want reassurance, get it, and then are angrier than before, because reassurance was not actually what you came for.

You do not have to tell them which one it is. You do have to know.

Name the observation, not the conclusion

The single most useful move is to say what you noticed and stop there.

"You've been taking your phone into the other room to answer it for about a month, and I noticed" is a sentence about a fact. It is checkable, it is not an accusation, and there is no way to answer it except by talking about the thing that actually happened.

"Who is she" is a sentence about a conclusion. It has one answer if you are right and one answer if you are wrong, and both answers sound identical.

This is the same distinction the eighteen signs are built on: an observation is a thing that happened, and it almost always has an ordinary explanation available alongside the one you are afraid of. Leading with the observation leaves room for the ordinary explanation to be the true one. Leading with the conclusion does not, which is why the conversation becomes about the accusation rather than about the thing.

The four openings that reliably fail

The test. Asking something you already know the answer to, to see whether they lie. If they pass, you learn nothing you did not know. If they fail, you have started the most important conversation of your relationship from a position you have to admit to.

The ambush. Opening with evidence you have been holding for two weeks. It feels like strength. What it actually does is tell your partner that you spent a fortnight collecting rather than asking, and that becomes the subject.

The hypothetical. "What would you do if someone at work liked you?" Nobody answers the hypothetical. They answer the fact that you asked it, usually by going quiet.

The late-night version. Tiredness narrows both of you to your worst arguments. Whatever this conversation is, it is not urgent enough to have at midnight, and it is the one conversation where you genuinely get one clean attempt.

How to open it

Pick a time when neither of you is leaving in twenty minutes. Say that you want to talk about something and that it is not an ambush. Then say the observation, plainly, in one or two sentences, and stop talking.

The stopping is the hard part and it is the part that works. Silence after a plain statement is uncomfortable enough that it gets filled, and what fills it is usually more informative than anything you could have asked for.

If you want a form of words: "I've noticed [the thing]. It's been on my mind and I'd rather ask you than keep thinking about it." That sentence does not accuse, does not test, does not hypothesise, and cannot be answered with a yes or a no.

What the answer will actually look like

Prepare for it to be unsatisfying, because most of them are.

An honest partner who has done nothing is often defensive, because being asked is itself an accusation of a kind and it stings. A partner who has done something is sometimes calm, because they have had longer than you to prepare. Defensiveness is not evidence and composure is not innocence. Both readings are wrong often enough to be useless, which is exactly why this site does not turn behaviour into a likelihood.

What is informative is specificity. An ordinary explanation tends to come with detail attached without being asked — the name, the reason, the boring context — because there is nothing to manage. An explanation that stays general, and stays general on a second pass a week later, is worth noticing. Not concluding from. Noticing.

If it goes badly

Sometimes the conversation ends with nothing resolved and both of you further apart. That is a common outcome and it does not mean you asked wrong.

You can come back to it. A second conversation, days later and calmer, is more useful than an escalation on the same night. If you find yourself unable to leave the subject alone in the meantime, that is worth reading about separately — the interval where you know nothing is its own problem and it has its own handling.

And whatever happens, do not go to the phone. It is unlawful in most places to access someone's accounts without their permission, and it converts a conversation you might have recovered from into one you probably will not — the guide on that line covers what it actually costs.

Before you open it

If you have not yet worked out what you have actually noticed, as distinct from what you have been imagining, the eighteen questions are a way to separate the two. They produce a weighted count of your own observations and a suggested next step. They do not produce a verdict, and nothing you answer leaves your device.