In this article, we call for a formal approach to modeling the development of reasoning. We discuss the need to operationalize the nature of encoding and processing. These problems are reviewed in the context of fuzzy-trace theory with particular emphasis on class-inclusion reasoning experiments. We argue that salience manipulations (perceptual or linguistic) give the illusion of inclusion by promoting subclass-subclass comparisons. We present an alternate, mathematical model of class-inclusion reasoning in which memory and reasoning parameters are estimated and integrated in a skills analysis of performance. We conclude that without similar formal modeling efforts, the nature of children's reasoning will remain elusive. © 1991.