Transactions and Crossed Transactions

The problem it solves

A plain question about a schedule will regularly come back as something other than an answer. Somebody asks what time the release window closes and gets back "why is it always me who has to remember the schedule?", which is a common enough exchange that most people have been on both ends of it.

Neither side of that is unreasonable, since the question was a question and the colleague has probably been asked the same thing six times that week and is fed up. The conversation is no longer about the release window, though, and asked afterwards what happened, one of them would say they got their head bitten off and the other would say that nobody ever appreciates anything.

What is interesting is how mechanical it feels, because the same shape happens between people who like each other, on subjects neither of them cares about, several times a day. It's too regular to be a matter of personality, which suggests there is something structural underneath it worth looking at.

The idea

Berne's unit of analysis was the transaction, meaning one exchange, something sent and something sent back. His useful move was to say that both halves have an address, so a remark leaves one state in you and arrives at a particular state in the other person, and their reply leaves one of their states and arrives at one of yours.

Draw those as arrows and two things become visible.

Enlarge: One opening, two possible replies. The reply is where the exchange is decided.

Two columns of three stacked boxes, one per person, labelled Parent, Adult and Child. A solid arrow runs from the first person's Adult to the second person's Adult: the question as sent. Two possible replies come back. One returns Adult to Adult, parallel, and the exchange continues. The other runs from the second person's Parent to the first person's Child, cutting across, and the exchange stops.

One opening, two possible replies. The reply is where the exchange is decided.Close

One opening, two possible replies. The reply is where the exchange is decided.

A parallel transaction is one where the reply comes from the state that was addressed and returns to the state that sent it. Adult asks for a date and Adult supplies one, Parent instructs and Child complies, Child complains about the unfairness of it all and Parent soothes. In each of those there is no obstruction anywhere in the exchange, and Berne's observation was that a parallel exchange can continue indefinitely.

A crossed transaction is one where the reply comes from somewhere else, or lands somewhere else. Adult asks what time the window closes and Child answers a question about fairness that was not asked, so the original exchange stops and whatever happens next is about the crossing.

Two things follow that are easy to miss and are most of the value here.

Parallel does not mean healthy. The model describes flow rather than quality, and an instruction answered by compliance is textbook parallel and can run for a decade, which is precisely how a team that has stopped pushing back gets built, one frictionless exchange at a time. Smoothness tells you the pattern is unobstructed, and tells you very little about whether the pattern is worth having.

Crossing on purpose is the tool. If a pattern is running and you want it to stop, the way out is a reply that declines the address, and there isn't really another one. This is the single most useful thing on this page, and it works best with no commentary attached at all.

How to use it

There are three questions, and they go in this order.

Where did that arrive? The thing to look at is what the reply it seems to want would look like, rather than what they actually said, because a remark that needs you to defend yourself in order to work was aimed at somewhere you might defend yourself from.

Where did my answer come from? This is the one worth practising, because it's the half you control, and it is also where you will notice you have been running the same pattern for months.

Do I want this to continue? Sometimes the answer is yes, because a great deal of ordinary work (most of a working day, in fact) runs on parallel exchanges and should, but if you have had the same conversation with the same person four times the answer is no, and the mechanism for stopping it isn't saying so.

The deliberate cross, in practice, looks like this. An analyst arrives most weeks saying a piece of work is impossible and she is out of her depth, and warm reassurance is the parallel reply, because it answers the opening exactly where it was aimed. She does the work well and is back next Tuesday with the same opening, since no part of the exchange asked for anything different. The cross is "what specifically is blocked, and what did you try", which is less warm, harder to say, and ends the loop rather than servicing it.

The corresponding discipline is not to narrate any of this. "You just crossed that" is a Parent move with a textbook behind it, and it converts a technique that works into an argument about whether you're being clever at somebody.

Where it breaks down

Every concept here has one. It is the section most summaries leave out.

It reads written communication badly, and that is where people mostly use it. Berne was watching faces, and tone, pace, posture and timing carried most of the signal, all of which go missing in a channel or an email. What remains is word choice, the weakest cue of the set, and somebody drafting a terse reply on a phone between meetings produces text indistinguishable from somebody being magisterial, which makes the confident retrospective mapping of a three-day thread close to worthless.

Two people, in a room of nine. Most workplace communication is not a dyad, and the notation has no place to put an audience. A person answering their manager in a weekly forum is doing two things at once, answering and being seen to answer, and the second task changes the reply in ways the diagram can't represent, which makes this a real limitation rather than a misapplication.

Any exchange can be mapped afterwards, which is why a completed mapping is not evidence of anything. Give two people the vocabulary and they can assign vectors to a quarter's worth of messages with complete conviction and no way to be wrong (the conviction is the part to distrust). The satisfaction of a tidy diagram is not evidence that the diagram is true.

Crossing can be a way of dodging something that deserved an answer. The flat factual reply that ends an exchange is sometimes skill and sometimes a manager declining to hear a complaint that was badly delivered but real. The model has no way to tell those apart, and it flatters the person doing the crossing.

In the wrong hands it becomes a way of scoring somebody's replies. The notation describes an exchange, and it takes one short step to turn it on the person: telling a colleague which state their answer came from is a judgement wearing the clothes of an observation, and there is no reply to it, because disagreeing looks like proof. This is worth a great deal more applied to a message you are about to send than to one you have just received.

The evidence sits where it sits for all of this theory. The most thorough effectiveness review available pools forty-one trials and finds moderate to large effects for therapy built on transactional analysis, which is a real result and is also the ceiling. It says nothing about whether exchanges have addresses at all, because outcome studies can't isolate the component that did the work.

In one line

Notice where the reply came from, and remember that a conversation which runs smoothly forever is running smoothly because nothing is stopping it.