CAUSAL INFERENCE, CONTINGENCY-TABLES AND THE SELECTION TASK

Determining whether or not a causal conditional, such as ''If a person has Zav's disease then they have a raised temperature'', is probably true, or probably false, brings benefits but incurs costs. We used a v ariant of the selection task in which there were four categories of pa tients (e.g., those diagnosed with Zav's disease) and participants dec ided what number of patients in each category they wished to examine. Experiment 1 showed an effect of the seriousness of the disease on the number of patients examined. Also, unlike performance on the standard selection task involving a non-causal conditional, individuals wished to search all categories of patients. Experiment 2 varied the probabi lity of the conditional being true and the budget for examining patien ts. Search became more selective under strong budgetary constraint. We discuss our results in the context of recent Bayesian analyses of the selection task and of causal inference, and propose various extension s to this line of research.