9320497 Grether Psychologists have identified a number of rules or heuristic that individuals may use in making judgements about the likelihood of uncertain events. Recent studies in experimental economics and psychology have focused on finding the best fitting model from an ex ante natural class of models to some experimental data in ways that allow for subject differences in decision rules. But these data mining exercises limit subjects to a few pre-determined rules. Proper statistical analysis of these data may be impossible, if patterns of behavior do not fit into one of the pre-specified models. The contribution of this project comes from generalizing the single model selection approach by allowing subjects to be explained by different models and generalizing the standard data mining approach by choosing rules that best explain the subjects' behavior from a large potential class of rules. The project develops a procedure for analyzing data on individual choice and market experiments that rigorously estimates the number of rules used by subjects, estimates what those rules are, and classifies subjects by decisionmaking rules used. The theoretical and practical aspects of this procedure are studied. The procedure is used to design and conduct experiments that can best discriminate between different decision rules, minimizing misclassification errors and biases in estimates. The technique is applied to data from double auction experiments.