It starts somewhere new — the washing up, a text that went unanswered, a comment about your job. Ten minutes later you are in the identical row you had last month, saying lines you could have written in advance, arriving at the same bad ending at the same speed.
A fight that repeats with that much precision is not a communication failure. Communication failures are messy and different every time. This is reproducible, and reproducible things have a structure.
Eric Berne called these games
Not games in the sense of fun, or even of deliberate manipulation. In Transactional Analysis — Berne's model of what actually goes on between people — a game is a sequence of moves two people run without either of them choosing to, which reliably ends in both feeling something familiar and bad. The familiarity is the point. The payoff of a game is not winning — it is confirming something you already believed.
That is why they repeat. If some part of you holds "nobody ever really listens to me", a conversation that ends with you unheard is not a failure. It is a result. The evidence arrived, the belief held, and the world stayed predictable — which is worth more to us than we would like to admit.
The three seats
Stephen Karpman noticed the roles are almost always three, and that people rotate between them mid-argument:
- Rescuer — helps where help wasn't asked for, then resents it. "I do everything around here."
- Persecutor — sets the standard and prosecutes the failure to meet it. "Why is it always me who has to say something?"
- Victim — is done to, and cannot act. "Nothing I do is ever enough."
The switch is what makes it feel like whiplash. A Rescuer who has given too much flips to Persecutor. A Persecutor who is challenged drops to Victim. That mid-argument lurch, where it suddenly stops being about the washing up — that is the switch, and it is the moment the game pays out.
How to spot yours: ignore the topic entirely and look at the ending. If you can predict the last five minutes before the first five have finished, you are in a game. The topic is interchangeable; the ending is the thing being produced.
You cannot win it, and you can leave it
Games need both seats filled. That is genuinely good news, because it means you can end one unilaterally — not by winning the argument, but by declining your role.
Concretely: stop supplying the next move. If you are usually the one who offers solutions, ask what they actually want instead of solving. If you are usually the one who escalates, say the plain version — "I notice we have had this exact conversation before" — and then stop talking. Naming the pattern out loud is the single most reliable interrupt, because a game only runs while both people pretend it is about the washing up.
It will feel worse before it feels better. The other person will usually push harder for a beat, because the familiar ending is not arriving. That escalation is the strongest evidence you have correctly identified the game.
Where this sits in the bigger picture
New to any of this? Start with what Transactional Analysis is. The full version, including the classic named games and how the switch works, is in the guide to the drama triangle. Most people have a seat they take first — and it is usually not the one they would guess.
Which seat do you take?
The three-minute egogram maps the five voices you speak from under pressure — the ones that decide which seat you reach for before you have thought about it. Free, no account.
Take the free quiz →