Attribution asymmetry

Why do people explain their own failure differently from an identical failure by someone else?

You are late to a meeting because the train stopped in a tunnel. A colleague is late to the same meeting, and the thought that arrives unbidden is not her train stopped too but she is disorganised. The behaviour is identical; the explanation is not. What is doing that work? It is tempting to answer "vanity", and vanity is part of it — but only part, and starting there hides something more interesting.

Begin with what an explanation of behaviour has to choose between. Roughly, we can locate a cause inside the person — their character, ability, intent — or outside them, in the situation they were standing in. Psychologists call these dispositional and situational attributions. Neither is inherently correct; the question is only whether we weight them well.

We do not. The best-known demonstration is Jones and Harris's 1967 study: participants read an essay defending Fidel Castro and were asked what the writer really believed. Even when told plainly that the writer had been assigned that position and had no choice, participants still inferred the writer was somewhat pro-Castro. The situation was handed to them explicitly, and they discounted it anyway. This bias toward the person over the situation is the fundamental attribution error, or, in the more careful name, correspondence bias.

Now the natural next step — and here we should slow down, because the textbook version of it is shakier than its fame suggests. The classic claim, from Jones and Nisbett in 1971, is the actor-observer asymmetry: I explain my behaviour situationally and yours dispositionally. Elegant, memorable, and when Bertram Malle assembled the accumulated studies in a 2006 meta-analysis, the overall effect was close to zero. It appears under some conditions — notably for negative events — and reverses or vanishes under others. So the general form of the asymmetry is not established; do not lean on it.

What does hold up well is narrower and slightly less flattering: the self-serving bias. Success we attribute to ourselves; failure we attribute to circumstance. This survives meta-analysis robustly, across ages and cultures. So the asymmetry in the opening scene is real, but it is organised around outcome valence rather than simply around self versus other.

Is that just self-flattery, then? Partly — but consider a purely informational account, which needs no motive at all. What do you know about your own lateness? You know the tunnel, the alarm that did fire, the ten previous mornings you were early. You have a rich situational history. What do you know about your colleague's? You have one datum: she walked in late. Behaviour is vivid and situations are invisible, and the person is what your eye is actually pointed at. Attributing to the person is what an observer with only behavioural data should do.

That account makes a testable prediction, and Storms tested it in 1973 by showing people videotapes of a conversation filmed from the other party's viewpoint. Seeing yourself from outside made your own behaviour look more dispositional. Move the camera, move the attribution. Whatever else is going on, perspective is doing real work.

One more constraint before we settle. Correspondence bias is not a human universal in the same strength everywhere: studies with East Asian participants — Miller's developmental work in 1984, Morris and Peng's on the same events reported in Chinese- and English-language newspapers in 1994 — find noticeably greater weight given to situation and context. So the "error" is partly a habit of attention that a culture can train.

Three refinements hold the picture together.

First, "error" is a slightly awkward label. In any single case we cannot say the person-explanation was wrong — sometimes the colleague really is disorganised. The finding is about systematic over-weighting across many cases, not about any one judgement.

Second, the motivational and informational accounts are not rivals to be settled. Self-serving bias plainly has a motivational component, and equally plainly you do have more situational data about yourself. Both contribute, and separating their shares in any given episode is not something the evidence supports doing confidently.

Third, be careful about which claim you carry away. "People explain themselves situationally and others dispositionally" is the version most often repeated and the one meta-analysis has substantially undercut. "People take credit for successes and blame circumstances for failures" is the sturdier finding. If you want one sentence, take the second.

• How correspondence bias interacts with power and status in workplaces and courts.
• Whether attributional retraining actually changes behaviour, or only self-report.

Every term the collection defines is gathered in the glossary.