Knowledge-Based Assumptions in Causal Attribution

Deviations from covariation-based predictions are common in the causal attribution literature. Cheng and Novick (1990a) proposed that these deviations are due to previous researchers' failure to accurately identify the input to the causal inference process. More generally, they proposed that when the set of events considered relevant by an attributor (termed the focal set) is identified, the inference process may be found to reflect an unbiased assessment of covariation over this set of events, as specified by their probabilistic contrast model. We report an experiment using college students that measured both (a) the hypothesized focal sets resulting from subjects' knowledge-based assumptions and (b) subjects' causal attributions. Our results provide strong support for this hypothesis.