Brunswikian and Thurstonian Origins of Bias in Probability Assessment: On the Interpretation of Stochastic Components of Judgment

In this paper the Brunswikian framework provided by the theory of Probabilistic Mental Models (PMM), and other theoretical stances inspired by probabilistic functionalism, is combined with the Thurstonian notion of a stochastic component of judgment. We review data from 25 tasks with representative selection of items collected in our laboratory. Over/underconfidence is close to zero in most domains, but there is a moderate hard–easy effect across task domains that is inconsistent with the original assumptions of the Brunswikian framework. The binomial model modifies PMM‐theory by allowing for sampling error in the process of learning the ecological probabilities and the response‐error model takes error in the process of overt probability assessment into account. Both models predict a moderate hard–easy effect across task environments that differ in difficulty or predictability, but it is also demonstrated that the two interpretations of random error lead to different predictions. The response error model predicts format dependence , with more overconfidence in full‐range than in half‐range assessment, and the phenomenon is illustrated with empirical data. It is proposed that a model that combines the Brunswikian framework with both sampling error and response error captures many of the important phenomena in the calibration literature. For illustrative purposes, a combined model with four parameters is fitted to empirical data suggesting good fit. © 1997 John Wiley & Sons, Ltd.