Actor–observer asymmetry
Also known as: Actor–observer bias
We blame our own behavior on the situation but others' on their character.
What it means
The tendency to attribute one's own actions to external, situational causes while attributing the identical actions of others to stable internal dispositions. The classic explanation is perceptual: when we act, the situation fills our visual field, whereas when we observe others, the person is the salient figure against a background we barely notice. We also simply have more information about the variability of our own behavior across contexts. A large meta-analysis (Malle, 2006) found the effect is far weaker and more conditional than once assumed, surviving mainly for negative and intentional events. It matters because it systematically distorts conflict, performance reviews, and how blame is assigned.
What the evidence shows
The original demonstrations were vivid. Nisbett and colleagues (1973) asked students to explain their own choice of major and their best friend's; self-explanations leaned on properties of the field, friend-explanations on properties of the friend. Storms (1973) then made the perceptual account testable, filming two-person conversations and replaying the conversation from the opposite camera: actors saw themselves, observers saw what the actor had been looking at. The attributions moved with the camera. That looked like proof of the mechanism, and it did not survive. Storms found the standard asymmetry in his control condition (d = 0.482) and the predicted reversal when perspectives were switched (d = -0.636), but five subsequent studies between 1975 and 1984 averaged a reverse asymmetry in the control condition; across all six studies the control condition sat at d = -0.077 (ns) and the switched condition at d = -0.280. As Malle (2006) puts it, given that no asymmetry occurred in the control condition, it remains unclear what the switched perspective achieved. One candidate account, reported in the same review, is that perspective switching is inert and merely imposes a delay: a switched-perspective video moved explanations no more than an irrelevant music video did. Malle's meta-analysis of 173 studies then found average effect sizes between d = -0.016 and d = 0.095. Correcting for the pool's tilt toward negative-valence stimuli brought the average to d = 0.042; a separate trim-and-fill correction for publication bias, imputing 25 omitted studies, brought it to d = -0.011 — indistinguishable from zero. The asymmetry textbooks called pervasive is, in aggregate, absent.
Related but distinct
Three attribution effects get collapsed into one. The fundamental attribution error concerns observers alone: we over-weight disposition when explaining behavior, whoever performed it. Self-serving bias concerns valence: we claim credit for successes and blame circumstance for failures, regardless of perspective. Actor-observer asymmetry makes a narrower claim, that perspective by itself shifts attributions with valence held constant. That is the claim the meta-analysis dissolves. Malle found the predicted pattern for negative events and a reversed pattern for positive ones, which is the signature of self-serving bias wearing the actor-observer costume. Much of what practitioners label actor-observer bias in the field is motivated self-protection, or simply the actor knowing more, rather than a quirk of where the eyes happen to point.
What survives
Malle's moderator analysis left the effect alive but narrow: it holds when the actor is portrayed as idiosyncratic, when events are hypothetical rather than lived, when actor and observer are intimates, and when explanations are free-response rather than scaled. Malle, Knobe and Nelson (2007) then argued the original framing was wrong at the root. People do not sort causes into person and situation; they separate intentional actions, explained by the agent's reasons, from unintentional ones, explained by causal history. No reliable person–situation asymmetry survived aggregation (pooled ds -0.08 to 0.10). It surfaced once, in Study 2 (d = 0.49), and dissolved on inspection: the gap was carried entirely by causal-history explanations (d = 0.80), with actors and observers indistinguishable on cause explanations (d = -0.05). Three other asymmetries did hold: actors give more reasons and fewer causal histories (never below d = 0.55), more belief than desire reasons, and they leave their beliefs unmarked. The gap is in the grammar of explanation, not the person-situation ledger.
Using it in practice
The standard countermeasure — hand the reviewer the actor's constraints, the queue, the deadline, the guidance available at the time — is the one Malle, Knobe and Nelson tested and found wanting. Their knowledge hypothesis failed repeatedly: close observers who knew the actor well explained no differently from distant strangers (ds = 0.08, -0.02, -0.04). For the belief asymmetry they conclude that observers must both gain access to more information and be motivated to use it, and that neither process is sufficient on its own; the reason asymmetry they attribute primarily to motivational factors. What moved observers in their Study 6 was impression-management instructions, not facts. So the lever is to assign the reviewer the job of stating the actor's case, not merely to give them the context file. Yelderman and colleagues (2021) had 1,317 participants act as a tie-breaking parole board member or merely read a news report of the parolee's later murder of a child; the reader-observers blamed the board more than the mock members blamed themselves — reliably, though the effects were modest (η² = .03–.06 on blame). Those mock members had both a self-protective motive and private access to their own reasoning, and in applied settings the two are rarely separable — which is tolerable, because the intervention is the same either way. Treat the asymmetry as a hypothesis about a particular case rather than a law.
Examples
When I snap at a colleague it's because I'm under deadline pressure; when they snap at me, they're just hot-tempered.
The driver who cuts across me is a selfish idiot. When I do it, it is because the lane was ending and nobody would let me in.
A mother explains her toddler's meltdown by the missed nap and the heat, while reading the child screaming two tables away as simply badly behaved.
At an outage postmortem the on-call engineer explains in reasons — the runbook said the queue was safe to drain, so he drained it — while the review writes causal history: fatigued, under-trained. One incident, two grammars of explanation.
A patient who skips doses explains it by the cost and a confusing schedule; the clinician charts non-adherence as a trait of the patient and prescribes the identical regimen again.
First described in Jones & Nisbett (1971).
Key references
- Yelderman, L. A., Lawrence, T. I., Lyons, C. E., & DeVault, A. (2021). Actor–observer asymmetry in perceptions of parole board release decisions. Psychiatry, Psychology and Law, 28(5), 623–644. doi.org/10.1080/13218719.2020.1821826
- Malle, B. F., Knobe, J. M., & Nelson, S. E. (2007). Actor-observer asymmetries in explanations of behavior: New answers to an old question. Journal of Personality and Social Psychology, 93(4), 491–514. doi.org/10.1037/0022-3514.93.4.491
- Malle, B. F. (2006). The actor-observer asymmetry in attribution: A (surprising) meta-analysis. Psychological Bulletin, 132(6), 895–919. doi.org/10.1037/0033-2909.132.6.895
- Storms, M. D. (1973). Videotape and the attribution process: Reversing actors' and observers' points of view. Journal of Personality and Social Psychology, 27(2), 165–175. doi.org/10.1037/h0034782
- Nisbett, R. E., Caputo, C., Legant, P., & Marecek, J. (1973). Behavior as seen by the actor and as seen by the observer. Journal of Personality and Social Psychology, 27(2), 154–164. doi.org/10.1037/h0034779