Referential inconstancy in natural language

There is a lively debate between proponents of referentialist and predicativist semantics about the nature of proper names in natural language. Still, the assumption that bare singular (referential) uses of proper names in argument position of a predicate are modally constant -they have one and the same referent across modal scenarios- has become orthodoxy in the field. This is a mistake, as there are referential yet inconstant uses of proper names in argument position. Initially presented in a rather ignored passage of Lewis' (1986) seminal work on modality [Lewis, 1986:248-253] these data create a puzzle for extant theories of proper names. Referential uses in argument position are typically assumed to be constant. However, they can be turned inconstant depending on context. As a result, constancy may not be accounted for in virtue of semantics, and neither may inconstancy. Current theories of proper names must address this and account for referential constancy and inconstancy, but they cannot do so in semantic terms. The paper begins, section 1, with a brief account of the dispute between referentialist and predicativist theories of proper names, showing how both traditions are committed to referential constancy. Section 2 presents Lewis' (1986) data on referential inconstancy and why it is problematic. Section 3 shows how this result may be generalized for any ordinary proper name. Section 4 considers three possible objections and why they fail. The paper concludes, section 5, by considering Kripke's (1980) take on referential constancy as the result of stipulation and independent from semantics, suggesting a syntactic approach to referential uses of proper names.