Nobody wakes up and decides to wreck something they wanted. The experience is much more ordinary than that: you were fine, and then you were picking a fight about nothing, or you were fine, and then you had quietly gone somewhere they could not reach. Afterwards it is obvious. During, it felt like a reasonable response to something they did.
That gap — obvious afterwards, invisible during — is the actual subject of this page. Not why you do it, which turns out to be fairly simple, but how to catch it while it is happening, which is the only point at which anything can change.
What is usually going on
You leave before you can be left. The avoidant version. It does not look like leaving; it looks like clarity. Suddenly you can see all the ways this was never going to work, and the evidence is compelling, and you feel calm. What preceded it, almost always, was a good stretch — a trip that went well, a conversation that went further than usual, the point where the relationship became a thing you could lose. The faults you found are usually genuine faults. They were also available last month, when they did not matter.
You detonate something to see whether they stay. The anxious version, and the one people are more likely to recognise as sabotage because it is loud. A fight picked over something small. An accusation you do not believe. Leaving the room, not answering, saying the thing you know does damage. Underneath it is a question — are you going to go too — and the fight is the cheapest way to get it answered immediately. It works, in the sense that they usually stay and you feel briefly better. It also spends something, and the account is not infinite.
You cannot hold the good opinion. A quieter mechanism than either. Being liked by someone whose judgement you respect creates a discrepancy you have to resolve: either they are wrong about you, or you are wrong about you, and the first is easier. So you show them the worst version early, as a kind of correction, and you experience their reaction as confirmation rather than as consequence. This one is not really about attachment. It is about self-worth, and it responds to different things.
It is not sabotage at all — you knew and did not want to say. Occasionally the thing labelled self-sabotage is an accurate read arriving in a cowardly form. You did not want the relationship, you could not say so, and you made yourself impossible until someone else ended it. This is worth telling apart from the rest, because the others need practice and this one needs a sentence.
The timing is the tell
If you only take one thing from this page, take this. The content of the crisis is not diagnostic — it is always something plausible, because your mind is good at its job and would not hand you a bad reason. The timing is diagnostic.
So look at the last three relationships and mark two dates each: when it was going best, and when it went wrong. If the second date is consistently within a few weeks of the first, you are not having a run of bad luck with people who reveal themselves at the same point. You are watching a mechanism with a trigger, and the trigger is the going-well.
Then look at the depth rather than the duration. Some people end things at four months every time; others survive years and detonate at the point of moving in, or meeting the family, or the first time someone says something unguarded. The depth at which it fires is stable, and it is a much better description of the pattern than any story about the individual people.
What to do the next time it starts
- Treat the sudden clarity as a symptom. When certainty arrives all at once about something you were content with a fortnight ago, that speed is the signal. Real conclusions accumulate. This one appeared, and things that appear are usually doing a job.
- Impose a two-week rule and tell no one. Not a decision to stay — a decision not to act for fourteen days. Watch whether the case against them holds up when you are not fuelled. Most of the time it thins out considerably, and the few times it does not, you have lost nothing but a fortnight.
- Say the unarmed sentence. "Things are good and I've noticed I'm looking for problems." Or, for the other kind, "I wanted to know you'd stay and I went about it badly." Both are excruciating and both end the episode faster than anything else available, because they remove the cover the behaviour was operating under.
- Do not confess the whole pattern at once. A long explanation of your psychology in the middle of a crisis is another way of not being in the room. One sentence about what is happening now beats a complete theory.
- Write down what was good, while it is good. A short list, made during the good stretch, read during the bad one. The mechanism works by rewriting history, and a contemporaneous record is remarkably hard to argue with.
What it depends on
The two main versions want opposite things. If you leave, the practice is staying — the two weeks, the ninety seconds, the tolerance of a discomfort that turns out to be temporary rather than meaningful. If you detonate, the practice is asking plainly instead of testing, and then sitting through the gap between asking and being answered without escalating to get a faster reply.
Doing the wrong one actively hurts: telling a person who tests to sit still makes them wait in silence and then explode larger, and telling a person who exits to say what they need turns an avoidant retreat into an argument that ends things sooner. Which one you are is not reliably visible from inside — and it is worth knowing that these are patterns rather than verdicts, since attachment measures do move, slowly, in the direction of security. The test reads both dials, which is the difference between knowing you sabotage things and knowing what to do about it.