Casual judgements from contingency information : competition between multiple causes of a single outcome

In causal reasoning the presence of a strong predictor of an outcome interferes with causal judgments of a moderate co-occurring predictor. Causal competition effects have generally been demonstrated with a strong competing predictor that is followed by an outcome with a higher probability than a moderate target predictor, and that also signals as many or more of the total outcome occurrences than the moderate target predictor. Confounding these two distinct aspects of predictiveness has constrained the ability to examine their respective importance for the relative validity of predictors in causal competition. By examining the effects of one and two strong competing causes on judgments of a moderate cause, varying the proportion of total outcomes that the competing predictors are paired with while holding overall outcome frequency constant, this series of experiments begins to disentangle these aspects of predictiveness. It demonstrates competition effects with a strong predictor that predicts fewer outcomes than the moderate target predictor. In addition, causal competition was examined between positive predictors (those signaling the occurrence of the outcome), between negative predictors (those signaling the absence of the outcome) and between predictors of opposite polarity (positive and negative). Causal candidates of opposite polarity were found to enhance rather than reduce causal judgments of moderate positive and negative predictors, posing a challenge for some of the most influential theories of causal learning that explain competition effects as the discounting of the moderate predictor or a failure to learn its association with the outcome. Rather, these results are consistent with a contrast mechanism whereby causal judgments of moderate predictors are not necessarily reduced toward zero in the presence of stronger predictors, but are adjusted along the causal judgment scale in opposite direction from the strong predictors. When the competing predictors are of the same polarity causal judgments of moderate predictors appear to be reduced, but when they are of opposite polarity judgments are enhanced. The implications for various associative and statistical models of causal learning are discussed