The Drama Triangle
Three roles that turn a problem into a performance, and the switch that keeps it running.
At a glance
The problem it solves
Teams that have the same argument every few weeks are usually not arguing about the subject of the argument. Everyone involved can describe it, most of them can predict how it will go, and none of them can end it. Each round has a clear villain and a clear casualty, and those are usually not the same two people twice running. Each participant is behaving reasonably given how the situation looks to them, and the situation gets steadily worse.
The manager's version is quieter and more expensive. Somebody brings you a problem and you solve it, because you can and it's faster, so they bring you the next one. A year or two of that and you're doing two jobs, you're privately resentful, and the person you have been helping has visibly stopped trying to solve anything. At some point you will say something about it, and it will come out a good deal harder than you meant (it usually does).
Both of these look like problems about specific people, and they keep happening with different people.
The idea
Stephen Karpman, writing in 1968 about the plots of fairy tales, noticed that a very small number of parts kept recurring, and that the same parts turn up in conflicts which won't resolve. There are three of them.
Victim does not mean somebody who is being harmed, but somebody taking the position that this cannot be solved and is being done to them.
Persecutor is the one applying blame, pressure or control, nearly always while feeling completely justified, and usually in response to something real. Nobody takes this role believing themselves to be the villain of it.
Rescuer is the one stepping in uninvited to do something the other person could do, whose help quietly confirms that they could not. It's the role that looks like virtue, which is what makes it the hardest of the three to put down.
The qualification in the first of those is not a detail, and it is the first thing that disappears when this gets used. The roles apply where the position does not fit the facts, and nowhere else. Somebody genuinely being treated badly is not in Victim, and a colleague who does you a favour you asked for is not Rescuing. Drop that and what remains is a vocabulary for reframing anybody with a difficulty as performing one.
The mechanism, and the reason this is more than a way of labelling three people, is the switch.
Three labelled stations in a row, Rescuer, Persecutor and Victim, joined by arrows into a closed loop. Each arrow carries the condition that moves somebody on: the rescuer tires of carrying it, the one applying pressure is challenged, the one hard done by recruits a new helper. A line below the loop notes that the way out is not a fourth station but answering the request that was actually made.
The roles rotate, and they rotate within individuals. The person who has been carrying somebody for a year and a half finally says what he thinks in a planning meeting, and is now the one applying pressure. Challenged on it, he explains that nobody has ever supported him and he's on his own here, and is now the one hard done by. Somebody steps in to help, and the circuit is complete.
That rotation is what turns a disagreement into a fixture, and it is also the thing a static picture of three roles cannot show, which is why the diagram above is drawn as a circuit with the switch conditions written on it.
Karpman's original paper is five pages long, it's published on his own site, and it's linked in the references. Read it there rather than here.
How to use it
Take the roles off the people. Nobody is a Rescuer, and what people do is rescue, in particular situations, usually with the same one or two colleagues, and often at work but not at home, or the other way round. The moment a role becomes a description of a person it stops being observable and starts being an accusation.
Learn one test and apply it to yourself. Did they ask for this, could they do it themselves, and will I resent it later? One yes is worth noticing and two is an answer, and this single question set is most of the practical value on this page. It works because it's checkable, which little else here is, and because you already know the answers.
Note what the test does not do. Much of a manager's job is helping people with things they could do and should not be doing today. Fixing a junior's deployment problem in twenty minutes because you have a release on Thursday, explaining what you did, and going back to your own work is not Rescuing, since there is a reason beyond her competence, she's better off afterwards, and nothing is being banked for later.
Get out by answering the request that was actually made. The exit is not detachment and it is not a fourth position. When somebody brings you something they could handle, ask what they have already tried and what they want from you, which is one sentence. It ends more of these than any amount of pattern-naming, because it converts an implicit handover into an explicit decision, and the answer is sometimes "please go and talk to the API team", at which point doing so is a choice rather than a reflex.
Acey Choy proposed a positive counterpart to the three roles in 1990, which is vulnerability in place of helplessness, assertiveness in place of pressure, and caring in place of rescue. It's the standard answer to "so what do I do instead", and in practice it comes to much the same thing as the question above.
Do not say any of this out loud to anyone. "You're playing Victim" is pressure applied from above with a citation attached, and it's unanswerable, because objecting to it looks like the role. If you want a team to have some of this, teach the Rescuer test, apply it to yourself in front of them, and leave the other two words out of the shared vocabulary entirely.
Where it breaks down
It is far too easy to apply to somebody who is actually being harmed, and this is the serious one. The model's own definition requires the position to be unwarranted, and that qualification doesn't survive contact with real use. An engineer raising a formal complaint gets told she is stepping into Victim, and a tester who has flagged the same quality problem twice becomes "very Persecutor lately", after which the label does exactly the harm the original complaint was about. This failure runs downhill by default, from people with standing to people without it, which is what makes it something more than a misunderstanding.
It explains everything and predicts nothing. Three roles plus rotation between them will fit any conflict once it has happened. There is no outcome the model forbids, so no outcome tests it, and the fluency with which a past dispute can be mapped is not evidence of anything except that the categories are broad.
There is essentially no empirical literature on it. This is among the most widely taught ideas in transactional analysis and among the least studied, and a sympathetic 2010 survey of the field's research base is instructive mostly for what it does not contain. The best available review pools forty-one trials of transactional analysis therapy and finds moderate to large effects, and no part of that isolates the triangle, which appears in no arm of any trial as a thing being tested. It is a widely shared observation rather than a finding.
It hands a manager a respectable reason to withhold help. Because Rescuing looks like helping, the model can be used to reclassify ordinary support as enabling, and reluctance as discipline. It offers no way to tell the difference beyond the three questions, and those are questions about your own motives, so anybody inclined to answer them favourably will (and most people are).
As soon as the words are shared, they become insults. A team six weeks past the workshop has three new ways to dismiss somebody without engaging with what they said. Watch for conflict and raised concerns falling together, because that's suppression rather than resolution, and it is the most common outcome of teaching this well.
In one line
Nobody is a rescuer, the roles rotate, and the way out is asking what they have tried and what they want from you.
References
- Primary
Fairy Tales and Script Drama Analysis (opens in a new tab)
Five pages, published by the author on his own site, and the place this idea belongs. Go and read it rather than any summary, including this one, since it's shorter than most articles about it and considerably more careful about what it claims.
- Further
The Winner's Triangle (opens in a new tab)
Choy's proposal for what the three roles look like without the discounting in them, which is vulnerability in place of helplessness, assertiveness in place of pressure, and caring in place of rescue. The standard answer to "so what do I do instead."
- Further
Games People Play
The wider theory of games that Karpman's three roles compress into one diagram, including the payoff idea this page skips, which is that a repeating pattern is maintained by what each participant gets out of the ending.
- Critique
Scientific evidence base for transactional analysis in the year 2010 (opens in a new tab)
A sympathetic survey of what fifty years of research had established, and instructive here for an absence, since one of the most widely taught ideas in the field is also one of the least studied.
- Critique
The Effectiveness of Transactional Analysis Treatments and Their Predictors (opens in a new tab)
The strongest effectiveness evidence the field has, and worth reading for what it cannot do. Therapy built on this theory helps people, and no part of that isolates the triangle, which appears in no arm of any trial as a thing being tested.