The goal of this paper is to defend Tyler Burge’s (1973) approach to the semantics of proper names, according to which proper names are predicates in their own right. On this predicate approach, the predicative contents of proper names figure in the computation of meaning at two different levels. Depending on its linguistic environment, a proper name contributes its metalinguistic content either to the presupposition or assertive content of a sentence. The singular use of a proper name in argument position constitutes a definite description, which nondescriptively contributes an individual to the assertive content, while its predicative content is presupposed in the conversational background. Any adequate theory of the referential use of an incomplete definite description accounts for the rigidity of proper names in argument position.

I present two arguments for the predicate approach. First, I argue that the predicate approach is necessary to account for the metalinguistic implication of a use of a proper name (e.g., ‘That there is someone called by that name’). The metalinguistic implication of a proper name is a genuine presupposition (which is projectable and defeasible) rather than entailment or implicature. We have to attribute a predicative content to a proper name to derive such a presupposition. Second, I argue that the predicate approach explains the complex, modified uses of proper names better than any Millian alternative. I discuss the modified uses of proper names in English, such as 'the joking Woody Allen' and 'every Alfred', as well as the generic use of a name in Japanese. The Millian theorist of proper names cannot explain away the productive occurrences of names without introducing an ad hoc machinery.