Evidence status: mixed results Classic studies and reviews support a correspondence bias: people can infer attitudes or traits from behaviour even when they know it was constrained. The familiar claim that we explain ourselves through situations and others through personality is not a universal law. A 2006 meta-analysis of 173 studies found almost no overall actor–observer difference. This is a version 0.1 research draft, with evidence reviewed to August 2026.
How Everyday Psychology Takes Shape · Season Two: How Do We Read Other People? · Article 5
A shop assistant gives short answers, and we decide that he is cold. A colleague says nothing in a meeting, so she lacks initiative. A driver abruptly changes lanes: that person is simply irresponsible. Behaviour has a visible shape. Circumstances are distributed outside the frame, and personality becomes the easiest cause with which to complete the story.
Sometimes the story is right. Habits do influence behaviour across settings, and stable individual differences are legitimate subjects of psychology. The question is not whether personality exists. It is how much evidence is needed to move from one action to “this is the kind of person he is”.
Popular accounts often merge several concepts into one law. A careful treatment must separate correspondence bias, the fundamental attribution error, and the actor–observer asymmetry. They belong to the same research history, but they do not name one empirical result.
Inferring an author’s attitude when the position was assigned
In a classic 1967 study, Edward Jones and Victor Harris asked participants to read essays supporting or opposing Fidel Castro. Some readers were told that the author had freely chosen the position; others were told that the position had been assigned. Even in the assigned condition, readers inferred different underlying attitudes from the content of the essays.
This kind of result became known as correspondence bias: perceivers draw an overly strong correspondence between an observed act and an internal disposition, failing to use situational constraint fully. In a 1995 review, Daniel Gilbert and Patrick Malone described several possible processes. Situational information may be less salient and require extra cognitive work. Observers may underestimate the strength of a constraint. Or they may begin with a dispositional inference and fail to correct it sufficiently.
The task remains instructive and has historical and methodological limits. Writing an assigned essay is not equivalent to every complex act in daily life. Participants could know that the position was assigned while believing that authors retained choices about how enthusiastically to argue it. Correspondence bias describes an average tendency under a family of conditions, not a single mechanism governing every trait inference.
How a textbook law lost its universality
The more famous proposition says that when explaining our own behaviour, we emphasise the situation; when explaining somebody else, we emphasise personality. This actor–observer asymmetry feels immediately plausible and was long presented as a firmly established fact.
Bertram Malle’s 2006 meta-analysis of 173 published studies produced a surprising result. Average effects were close to zero. The classic asymmetry appeared under some conditions—for negative events, hypothetical scenarios, intimate relationships or particular coding of free responses. Positive events could produce the reverse direction. Across valence, there was no general actor–observer asymmetry.
This does not show that people never neglect circumstances. It shows that self versus other is not sufficient to determine explanation. Whether an event is good or bad, whether behaviour was intentional, the relationship between explainer and actor, and the question used by the researcher all alter the pattern. Self-serving explanation—claiming credit for good outcomes and assigning bad outcomes to the environment—also differs from a simple effect of observational position.
The history is itself a lesson in psychological knowledge. A concept can be memorable without surviving cumulative evidence in its original neat form. The science has to revise the sentence; a good sentence cannot substitute for the meta-analysis.
Why circumstances are harder to see than actions
An action is concrete: a sentence, a delay, a gesture. Circumstances are dispersed. We may need to know a schedule, a language background, an organisational rule, the preceding event or a temporary bodily state. A trait word compresses those unknowns into a convenient label.
The environment also works through behaviour. That makes it easy to treat the constrained result as a free expression of the individual. An organisation gives a newcomer no room to speak and records the resulting silence as limited capacity. Insufficient resources create delay, and a system then uses delay to score conscientiousness. Attribution here can become more than a private error; it can be stabilised in procedure.
The opposite mistake is possible. Persistent deception cannot be explained away indefinitely by pressure, and recurrent harmful conduct requires judgments of responsibility. A situational account answers what influenced an act. It does not automatically decide who should bear consequences. Causal understanding and responsibility are connected but distinct tasks.
Keeping both classes of explanation available
After one troubling event, begin with a behavioural description that avoids trait language: “After I asked a question, he did not answer and left ten minutes later.” Then generate at least one situational and one dispositional possibility. The purpose is not to give each exactly half the probability. It is to prevent the first available cause from monopolising the explanation.
Next look for evidence across settings. Does the behaviour recur? Does it persist after the constraint changes? How does the person explain it, and is that account consistent with checkable facts? Do other people in the same environment behave similarly? These questions can progressively alter the weight of each explanation.
Institutions should also examine whether evaluation metrics individualise environmental products. Sales, speaking frequency, attendance records and customer ratings may reflect resources, role design and other people’s biases as well as individual conduct. Fair assessment does not exclude personality or ability. It asks whether the measure has disguised controlled conditions as personal essence.
An apology or repair offers further evidence. A one-off failure followed by recognition and change has a different structure from a repeated pattern accompanied by denial. Behaviour after feedback often tells us more than the initial event alone.
Repair does not automatically erase a consequence already produced. It is new evidence for the later judgment, not a cancellation of the original event.
A person is not one hidden cause behind behaviour
We often imagine understanding a person as passing through the surface act to find the real individual underneath. Attribution research suggests that behaviour is not emitted by personality alone. Person, task, relationship, institution and moment meet in an action. Stable tendencies exist, but are realised through conditions.
This does not dissolve personality. It gives trait judgment a stricter meaning. When a relevant pattern persists across settings, continues as conditions change and is supported by multiple sources, a relatively stable disposition becomes a better inference. One behaviour can be evidence; it cannot supply the definition by itself.
The version 0.1 judgment is that correspondence bias under known constraint has credible evidence. The slogan “we blame circumstances for ourselves and character for others” should not continue as a universal law. The practical principle is not to excuse every person forever. It is to let circumstances enter the explanation before behaviour is promoted into character. Responsibility can still be assigned, but it should rest on a structure broader than one salient act.
Primary research sources
- Jones, E. E., & Harris, V. A. (1967), “The Attribution of Attitudes”.
- Gilbert, D. T., & Malone, P. S. (1995), “The Correspondence Bias”.
- Malle, B. F. (2006), “The Actor–Observer Asymmetry in Attribution: A (Surprising) Meta-Analysis”.
- Choi, I., Nisbett, R. E., & Norenzayan, A. (1999), “Causal Attribution Across Cultures”.
Series navigation: How Everyday Psychology Takes Shape — Season Two overview
Discover more from Geoffrey Chen
Subscribe to get the latest posts sent to your email.