Actor–observer asymmetry

The missed deadline

A project lead misses a deadline and tells the team, “The client changed the brief twice this week.” Two weeks later a colleague misses one, and the lead says afterward, “He’s just not very organized.”

This is the claimed pattern in its clearest form: a situational explanation for your own lapse and a character explanation for the same lapse in someone else. It is also the kind of case where studies come closest to finding it, because it’s a bad outcome (see Evidence). The move would fail if the lead had no more reason to think the colleague’s week was calm than their own was hectic.

Reasons that sound like circumstances

Asked why she took her new job, Maya says, “The commute is twenty minutes and the team was great.” Asked why her friend took his, she says, “He wants to be a manager someday.”

On a person-versus-situation scale this looks like the asymmetry: a commute and a team for herself, ambition for him. But “the commute is twenty minutes” is her reason, something she believed and cared about, stated without the “I thought” or “I wanted” in front. Research on how people word explanations finds that people explaining their own actions use those markers less often, and a person-versus-situation coding scheme can then count the reason as “the situation.” Part of what looked like an actor–observer difference in early studies may have been this difference in wording.

The same people, a good outcome

The same project lead finishes a proposal early and says, “I planned it properly this time.” When the colleague finishes his early, the lead says, “The client gave him an easy one.”

The asymmetry predicts the opposite: circumstances for your own behavior, whatever the outcome. Here the lead credits their own effort and the colleague’s circumstances. In studies of good outcomes, the average pattern was this reversed one, which is why the evidence points to people favoring themselves rather than to a general difference between explaining yourself and explaining others.

Explaining your own behavior differently from someone else’s is often justified:

• When you know things about your situation that the other person doesn’t. You may know you were sick, or that the instructions changed. Using information you really have isn’t a double standard.
• When the behavior is unusual for you but typical of them. If you usually do well, one bad result is reasonably treated as a fluke. If you’ve only ever seen someone do something once, you have less to go on, and if you’ve seen them do it many times, a lasting explanation is fair.
• When the situations really were different. Two missed deadlines under different workloads don’t call for the same explanation.
• When the “situational” explanation is really a reason. Saying you chose something “because it was close” doesn’t deny that you chose it.

The test: if you had the same information about them that you have about yourself, would you explain it the same way?

A fluke for you, a pattern for him

A student who has passed every chemistry quiz this term fails one after a night with almost no sleep, and puts it down to the bad night. A classmate who has failed every quiz this term fails this one too, and the student guesses he isn’t keeping up with the reading.

The explanations differ, but so does the evidence. For herself she has a long record and a known disruption; for the classmate she has a consistent record and no known disruption. Treating a one-off as circumstance and a steady pattern as something about the person is sound. This is the unusual for you, typical of them condition above.

Knowing each situation

Frank takes the bus to work for two weeks and tells a neighbor his car is in the shop. He notices that another neighbor, who has a car in her driveway, takes the same bus every day, and he figures she prefers not to drive.

Frank explains his own behavior by circumstances and his neighbor’s by her preference, but he knows his car is out of action and can see hers isn’t. She has an easy alternative every day and keeps choosing the bus, so inferring a preference is reasonable. This is the situations really were different condition above.

The hypothesis is intuitive, and Jones and Nisbett proposed reasons it should be true:

• Information. You know your own history, feelings and intentions; an observer usually doesn’t.
• Attention. When you act, you’re looking at the situation; an observer is looking at you, so you stand out against the background.
• Motivation. Researchers suggested people prefer not to see their own behavior as fixed by their traits.

Each of these is plausible, but none has produced a reliable person-versus-situation difference. In the meta-analysis, observers who knew the actor well showed a larger gap, not the smaller one the information account predicts, and showing actors a video of themselves from the observer’s viewpoint reversed the pattern in the first study but not reliably in later ones.

Why the idea has seemed obviously true is a separate question. Bertram Malle, who ran the meta-analysis, pointed to its simplicity and to the self-favoring pattern for bad outcomes, which looks just like it in any single case. Differences in how actors and observers word their explanations may also have made it look supported in studies that coded what people said. Actors and observers do explain behavior differently, in which reasons they give and how they phrase them, just not along the person-versus-situation line this hypothesis draws.

Status: failed. This site’s criteria for “failed” name preregistered or multi-lab replications, and none was found for this hypothesis. The label rests instead on a meta-analysis of 173 studies whose average effect was close to zero and, after correction for likely publication bias, indistinguishable from zero; the criteria for “robust” accept a bias-corrected meta-analysis as strong evidence, and the same evidence counts here in the other direction. A follow-up set of six studies found no support either. The weaker point is the recent reviews: the case against the hypothesis has been made mainly by the meta-analysis’s author, and no later large test that revives it was found.

• Jones and Nisbett (1971) stated the hypothesis that “actors tend to attribute the causes of their behavior to stimuli inherent in the situation, while observers tend to attribute behavior to stable dispositions of the actor.” It became one of the most widely cited ideas in social psychology and a textbook staple.
• Storms (1973) had participants rate the causes of behavior in a getting-acquainted conversation, and showed some of them a videotape of it from the opposite point of view. The standard asymmetry appeared, and it reversed after the switch in viewpoint, supporting the attention account.
• Watson (1982) reviewed the studies then available and called the effect firmly established, concluding that actors and observers differed mainly in how much they cited the situation, not in how much they cited the person.
• Malle (2006) meta-analyzed 173 published studies (113 articles, 14,686 participants, 1971–2004). The average asymmetry ranged from d = −0.016 to d = 0.095 depending on the measure, where 0.2 is conventionally a small effect. In studies run under standard conditions it was essentially zero, and a correction for publication bias brought the overall estimate to −0.011. Malle concluded that the hypothesis “appears to be a widely held yet false belief.”
• The exceptions. A gap in the predicted direction appeared for bad outcomes such as failures (about d = 0.24), but it reversed for good outcomes (about −0.15). Because the hypothesis claims a difference whatever the outcome, Malle read this as a self-serving pattern, people crediting themselves for successes and not blaming themselves for failures, rather than an actor–observer one. The predicted gap also appeared when people explained hypothetical rather than real events, when researchers coded free-form explanations rather than using rating scales, when actor and observer were close to each other (the opposite of what the theory predicted), and when the actor was portrayed as behaving very differently from other people (a result driven largely by one article). Visual perspective switching, time delay and observer involvement did not reliably change it.
• Malle, Knobe and Nelson (2007) ran six studies with different settings and methods. They found no evidence for the traditional hypothesis, but consistent differences of another kind: for example, actors explaining their intentional actions more often cite the reasons they acted for, and state them differently from observers.

The relationship to the fundamental attribution error. Many textbooks present the Fundamental attribution error as the observer’s half of this asymmetry. Malle argues that pairing is misleading. The fundamental attribution error is about inferring what someone is like from behavior the situation largely explains (an essay writer assigned a side is judged to believe it). The actor–observer hypothesis is about how people explain why a behavior happened, and about a difference between explaining yourself and explaining others. Finding no such difference, he writes, does not show that the fundamental attribution error doesn’t exist. It has since replicated strongly, including in a preregistered test across dozens of labs (Many Labs 2, 2018), with a large effect (d = 1.82). So the common impression that this asymmetry “contains” the fundamental attribution error is half right: the popular account nests the two, but the research measures different things, and the observer’s tendency held up while the claimed difference between actors and observers did not.

Popular version. The asymmetry is often summarized as “we excuse ourselves and blame others.” The part of that which has some support is narrower: when things go badly, people are less inclined to blame their own character than an observer is, and when things go well the pattern tends to flip. That is a matter of favoring yourself, not of being the actor.

What remains uncertain. The meta-analysis covered published studies, and the variation between studies was much larger than chance would produce, so conditions under which a real gap appears may still be found. The self-serving pattern for bad outcomes may itself be partly reasonable: people often have good grounds to see their own failure as out of character, while an observer has only the single failure to go on.