Structural alignment in induction and similarity.

Abstract

According to a structural alignment view, representational commonalities contribute to the perception of similarity, whereas nonshared attributes related to commonalities are candidate inferences in induction. This view was tested in 5 experiments. Novel animal pairs varying in the number of attributes and relations they shared were used to assess the relationship between induction and similarity. In Experiments 1 and 2, the number of shared attributes and the presence of 2 kinds of causal relations between attributes varied. Shared attributes increased both similarity and inductive strength judgments. Shared causal relations, possessed by both animals, influenced perceived similarity, but binding causal relations, which connected a shared attribute to a candidate inference in the induction task, were important for inductive strength. In Experiments 3-5, these results were extended through use of a noncausal relation and familiar animal categories.