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...