Online research in philosophy

Recently, an orthodox Russellian tenet has come under fire from within. In particular, some Russellians now argue that definite descriptions don’t semantically encode uniqueness. Instead, Reformed Russellians, as I call them, hold that definite descriptions are truth-theoretically identical to indefinite ones. On this approach, a definite description’s uniqueness reading becomes a matter of pragmatics, not semantics. These reforms, we’re told, provide both empirical and methodological benefits over and above the prevailing orthodoxy. As I argue, however, the Russellian Reformation contains serious flaws. Indeed, it is Orthodox Russellianism that remains a superior choice, both empirically and conceptually.

A sentence containing a number of definite descriptions, each lying within the scope of its predecessor, is naturally read as asserting the uniqueness of a sequence of objects satisfying the descriptions. The project of providing a general uniform procedure for eliminating embedded definite descriptions that gets this and other logical forms right is impeded by several puzzles.

Many characterizations of definiteness in natural language have been given. However a number of them converge on a single idea involving uniqueness of applicability of a property. This paper will attempt to do two things. One is to try to unify some of these current views of definiteness, seeing them as drawing out Gricean conversational implicatures of the uniqueness concept, and the other is to try a more articulated approach to dealing with some recalcitrant counterexamples. I will focus primarily, but not entirely, on definite descriptions in English.

As is well known, Russell assigned indefinite and definite descriptions the interpretations represented schematically in (1) and (2) respectively, where “CNP” stands for “Common Noun Phrase” in the sense used by Montague (1973) – i.e. as standing for the constituent which a determiner combines with to form a noun phrase (NP). (1) a. …a/an CNP… b. ∃x[CNP(x) & …x…] (2) a. …the CNP… b. ∃x[CNP(x) & ∀y[CNP(y) → y=x] & …x…] Examples (3) and (4) are illustrations. (3) a. Mary bought a car that she liked. b. ∃x[Car(x) & Liked(m, x) & Bought(m, x)] (4) a. Mary bought the car that she liked. b. ∃x[Car(x) & Liked(m, x) & ∀y[[Car(y) & Liked(m, y)] → y=x] & Bought(m, x)] The difference, as is obvious, is the underlined clause expressing uniqueness – exhaustive possession by the entity in question of the property expressed by the CNP. Szabó (2000) and Ludlow & Segal (2002) (following Kempson (1975), Breheny (1997), and others) defend analyses on which definite descriptions are assigned the same quantificational interpretation as Russell assigned to indefinite descriptions. Thus on both accounts (3a) as well as (4a) would be given the quantificational analysis in (3b). Both proposals acknowledge that definite descriptions differ from indefinites in their implications – where “implication” is to be understood as neutral between semantic and pragmatic conveyance. One of these implications is what is commonly termed “familiarity” – an assumption that the denotation of the NP2 has already been introduced.

Philosophers and linguists alike tend to call a semantic theory ‘Russellian’ just in case it assigns to sentences in which definite descriptions occur the truth-conditions Russell did in ‘On Denoting’. This is unfortunate; not all aspects of those particular truth-conditions do explanatory work in Russell's writings. As far as the semantics of descriptions is concerned, the key insights of ‘On Denoting’ are that definite descriptions are not uniformly referring expressions, and that they are scope-bearing elements. Anyone who accepts these two claims can account for Russell's puzzle cases the way he did. Russell had no substantive argument for the claim that ‘The F is G’ entails ‘There is at most one F’; in fact, he had important misgivings about it. I outline an argument against this claim, and I argue that by holding on to uniqueness contemporary semanticists make a momentous mistake: they keep the illusion alive that there is a way to account for linguistic meaning without addressing what linguistic expressions are for.

Both proposals acknowledge that definite descriptions differ from indefinites in their implications. (Two parenthetical clarifications: (i) "implication" is to be understood here and below as neutral between semantic and pragmatic conveyance; (ii) "semantic" is to be understood to mean "conventional", that is including, in addition to truth conditional impact, anything else that is linguistically encoded.) One of these implications is what is commonly termed "familiarity" ? an assumption that the denotation of the NP has already been introduced, as such, to the addressee of the utterance. The other is uniqueness, or more properly exhaustive application, within the salient discourse context, of the descriptive content of the NP to the intended denotation. However both analyses attempt to derive one or both of these implications pragmatically. Ludlow & Segal propose that familiarity is a conventional implicature and uniqueness a conversational implicature. Szabó concurs with Ludlow & Segal that familiarity is more essential to definite descriptions, but attempts to derive both implications pragmatically.