Systematicity as a selection constraint in analogical mapping

In an analogy a person's knowledge about one domain is used to understand a second domain: to highlight important similarities between domains or to predict new features of the second domain. For example, we use our knowledge about water flow to elucidate properties of electric circuitry. A central fact about analogy is that it is selective. Not all commonalities are equally important in matching or prediction. It follows that a central problem in theories of analogical mapping is to identify the selection constraints that tell us which similarities and dissimilarities count in an analogy (Kedar-Cabelli, 1985a; Hall, 1988). Within the enormous space of possible similarities between domains, what constrains the choice of information to include in an analogy? Two major classes of selection constraints have been utilized in computational models of analogy: goal-relevance constraints and structural constraints. The goal-relevance constraints direct analogical mapping toward information that is relevant to the problem at hand (Burstein, 1983; Holyoak, 1985; Kedar-Cabelli, 1985b). The structural constraints focus mapping on such properties of the matching information as structural consistency (Winston, 1980; Gentner, 1980, 1983; Falkenhainer, Forbus and Gentner, in press). For example, in her account of analog