The informal attempt in Chapter 1 and subsequently to reach a general definition of definiteness ran into a puzzle. Definiteness seems empirically to be a unified phenomenon, on the evidence of the way languages represent it, but it is not straightforwardly characterized. Two characteristics are prominent, but neither is apparently fully adequate as the defining feature. Identifiability is particularly attractive for referential uses, especially where the referent is a physical entity locatable in a physical context, and inclusiveness is particularly attractive for nonreferential uses. Indeed many uses are readily handled by either one of these concepts. But neither works for all uses. In this chapter we survey attempts to analyse definiteness within various theoretical frameworks. As will be seen, writers have variously argued for versions of identifiability or of inclusiveness, or have simply assumed one or the other, as the basic descriptive insight. The general tendency is for logicians and semanticists to prefer inclusiveness (or, very often, uniqueness – thus limiting themselves to accounting for singular definites), and pragmatists to prefer identifiability. But this is by no means a general rule; some have indeed sought to combine the two.

After outlining some major approaches I will argue (following up hints dropped in preceding chapters) that the attempt to find a fully unified characterization of definiteness in semantic or pragmatic terms is misguided. I will propose an account of definiteness as a grammatical category which, like other such categories, cannot be completely defined in semantic or pragmatic terms, though it represents the grammaticalization of some category of meaning.