Inferring decision processes from economic experiments

Researchers typically assume experimental subjects have rational expectations. If the object of the experiment is to learn the distibution of decision makers' types, then this amounts to assuming subjects' know the very thing the researcher wishes to learn. We propose a method of conducting controlled experiments in the absence of the rational-expectations assumption, and implement the method in simple games of proposal and response. The method does not, in general, identify the distribution of decision makers' types. But it does place informative bounds on this distribution. So the relaxing of the rational-expectations assumption is not fatal to experimental economi