Conceptual combination: does similarity predict emergence?

Abstract

Conceptual combination is used as a paradigm for investigating the influence of similarity on emergence. Subjects were 180 undergraduates recruited from the psychology subject pool. Pairs of parent concepts were selected for study using a similarity rating task, in which the similarity of various word pairs was rated. The eight most similar and least similar pairs were used in a conceptual combination task, where subjects wrote two definitions for each pair and listed important features of each definition. The features taken from the conceptual combination task were compared to a database of features typically associated with the parent concepts when presented individually. Any features that were found in the combinations but not in the parent concepts separately were considered emergent. Results indicated that there were effects of both similarity and order of attempt at defining the combination such that there was more emergence in dissimilar pairs, and in second attempts at defining the combinations. There was also a significant interaction between similarity and order of attempt at defining. These results are explained by the influence of structural alignment on conceptual combination. Similar pairs which could be easily aligned resulted in few emergent features, but when structural alignment was more difficult, as in the dissimilar pairs, the additional work resulted in increased emergence. This effect supports the quick fix hypothesis, that easy alignments do not require much work, but that more difficult alignments result in more activation of world knowledge and thus lead to increased noncompositionality.