Subjective Sampling Distributions and Conservatism.

When people revise subjective probabilities in light of data, revisions are less than the amount prescribed by the normative model, Bayes's theorem. Previous research suggests that this results from the subjects' lack of understanding of the implications of the data; i.e., from inaccurate subjective sampling distributions. This experiment examined the effects on conservative revisions of training subjects about the implications of data. The subjects estimated sampling distributions for two binomial populations, were shown samples from the populations in order to teach them veridical distributions, and again estimated sampling distributions. Estimated sampling distributions were good predictors of revisions and, as a result of training, both the sampling distributions and the revisions became more veridical. [ABSTRACT FROM AUTHOR]