This article looks at some phenomena that recur during expert problem solving across domains including computer science, physics, chess, and mathematics. Several of the mechanisms which have been hypothesized to underlie performance in one domain are then used to explain the results found in the others. Competing explanations, mechanisms which are not fully specified, and untested empirical issues are also discussed.
In this article I attempt to do four things. The first is to identify some phenomena that are common to several semantically-rich, problem-solving domains. The second is to cite some constructs which although proposed to account for a specific phenomenon in one domain, seem also to account both for other phenomena in that domain, and for similar phenomena in other domains. The third is to try to combine existing constructs so as to provide more complete explanations for existing data. And the fourth is to discuss issues which arise from competing classes of explanations and from incompletely specified mechanisms.
The goals motivating this attempt are two. The first is to support the position that studying semantically-rich domains can give us insights into general cognitive principles. In support of this position, the studies cited here deal with skills both developed in more widely disparate domains and studies using more highly vaned paradigms and stimuli than is usually done in experimental psychology; nonetheless, similarities appear. A word on the related issue of the transfer of skill across domains is appropriate here. Previously, researchers have asked whether principles in problem solving are general by asking whether or not problem-solving skills transfer from one task to another. Despite the fact that the results of these transfer studies have frequently been negative,' arguments will be presented supporting the position that a lack of transfer does not necessarily mean a lack of generality. It may be that processes which were formed in conjunction with particular structures cannot be induced to function in isolation from those structures; however, that does not mean that similar processes do not develop around the structures in several distinct domains.