Ortega-Castro, N. 1 , Barberia, I. 2 , Vadillo, M. A. 1 & Baker, A. G. 2
1 Deusto University, Bilbao, Spain
2 McGill University, Montréal, Canada
Many associative and rule-based theories assume that the probability of a binary outcome given a set of potential binary causes should equal the sum of the independent causal tendencies of each cue or cause. By contrast, some other models such as the Power PC theory of causal learning assume that the causal power of a compound cue or combination of causes should be computed in a more rational way. The causal power of each cue is added but the sum should be corrected by subtracting the overlap between them. We conducted a series of experiments testing these predictions, using different sets of probabilities, several cover stories and different formats of presenting information. We found that participants usually chose the simplest, though not normative, strategy to combine the influence from several causes. Most of the time they simply added the probabilities without considering the potential overlap; other times they averaged both conditional probabilities. Finally, we argue that, based on our experiments, it seems reasonable to conclude that there are a number of factors that might promote one strategy or another.