Fintan J. Costello
Traditional approaches to semantics give a logical, settheoretic account of category conjunction, where an item is a member of a conjunction A&B only if it is a member of both single categories A and B. However, people do not always follow this logical approach when classifying items in conjunctions. For example, Hampton (1988) found that people typically classify blackboards as non-members of the single category “furniture”, but as members of the conjunction “school furniture”. This paper describes a computational model intended to explain this “overextension” of conjunctive categories, and an experiment testing this model using conjunctions of controlled, laboratory-learned categories.