The Actor-Observer Asymmetry in Attribution: A (Surprising) Meta Analysis
From PsychWiki - A Collaborative Psychology Wiki
Bertram F. Malle University of Oregon
Reference: Malle, B. F. (2006). The actor-observer asymmetry in causal attribution: A (surprising) meta-analysis. Psychological Bulletin, 132, 895-919.
Summary A meta-analysis of 113 articles and 173 studies showed scarce support for an actor-observer asymmetry in attribution, with overall averages ranging from = –0.015 to = 0.095, depending on statistical models and specific attribution scores. (See Fig. 1 below.) Corrections for the preponderance of negative valence studies as well as for possible publication bias made the average effect size converge to 0. Under particular conditions, this value can increase to a medium effect size, primarily when (a) negative events are explained in between-subjects designs; (b) the experimenter provides base-rate information that portrays the actor as idiosyncratic; (c) hypothetical events are explained; or (d) explanations are verbally expressed and later coded. However, even these conditions do not provide compelling support for the classic actor-observer hypothesis. (a) A sole focus on negative events is not compatible with the hypothesis because the actor-observer asymmetry requires a main effect across valence. (b) Base-rate studies do not appear to speak to an actor-observer asymmetry per se but rather to the effects of creating specific information differences between actors and observers — differences that can easily be reversed and are not representative of the information actors and observers have outside the laboratory. (c) Similarly, an asymmetry for hypothetical events is of limited force given that outside the laboratory people predominantly explain real behaviors and outcomes. (d) Finally, codings of verbal explanations into person and situation categories appear to be oversensitive to linguistic surface patterns and therefore support an actor-observer asymmetry only in formulating explanations, not one in ascribing causes. The disconfirmation of an actor-observer asymmetry along person and situation attributions should not be mistaken for the disconfirmation of actor-observer asymmetries in general. Recent evidence suggests that actors and observers differ reliably in multiple (and psychologically significant) features of explanation, but these features are not captured by the classic person-situation distinction (Malle, Knobe, & Nelson, in preparation. Actor-observer asymmetries in behavior explanations: New answers to an old question.). Thus, the present meta-analysis both resolves a long-standing empirical question and sets the stage for a new approach to studying how people explain human behavior.