Abstract

Similarity is a focal idea in cognitive theory. It is used to help explain conceptual structure, transfer of learning, generalization, and other basic phenomena. Despite its importance no satisfactory process model for similarity has yet been proposed. Previous work has generally assumed that similarity requires a simple measurement of mental distance or a straightforward count of matching and mismatching features . In this chapter we propose that the similarity process involves the alignment of relational structures . This proposal has its roots in research on analogical reasoning, which has long been viewed as a mapping of complex representations (Hesse 1966) . We begin by reviewing two current approaches to similarity . Then we describe a framework for characterizing analogical processing and show how this framework can be extended to include ordinary literal similarity. We suggest some processing principles for human similarity and discuss some supporting empirical findings . Finally, we describe a computer model that makes the processing principles more explicit and demonstrates that analogical mapping is a computationally viable process.