We aimed to examine the differences between appraisal and attribution as regards their capacity to predict emotions. More specifically, we hoped to test the hypothesis, derivedfrom the model of Lazarus and Smith(1988), that appraisal's predictive potential increases with the subject's personal involvement in the events. The same is not true of attributions, which are probably determined more by implicit theories concerning such events than by the personal significance they might hold. Subjects were required to evaluate an episode relating either to them directly or to a fictitious character. The information in the episode was manipulated to induce certain attributions and appraisals. Following the evaluation, the subject was asked to report on his/her emotions (in the former case) or to infer those of the character. Anger andguilt were chosenas the emotions to be studied. The data, analysed using causal analysis techniques, revealed that appraisal was superior to attribution in terms of predictive ability. This superiority was greatest with increased personal involvement on the part of the evaluator.