Binding arguments and hidden variables

Abstract

Binding Arguments and Hidden Variables Jonathan Cohen and Samuel C. Rickless In recent years, several philosophers have appealed to evidence about bind- ing relations to show that various linguistic expressions are represented (at some level of linguistic representation) as having hidden variables (e.g., Stan- ley (2002), 368–369; Stanley and Szab´ (2000), 243). In particular, the idea o is that binding interactions between the relevant expressions and natural lan- guage quantifiers are best explained by the hypothesis that those expressions harbor hidden but bindable variables. Recently, however, Herman Cappelen and Ernie Lepore have rejected such binding arguments for the presence of hid- den variables on the grounds that they overgeneralize — that, if sound, such arguments would establish the presence of hidden variables in all sorts of ex- pressions where it is implausible that they exist (Cappelen and Lepore (2005), Cappelen and Lepore (2002)). 1 In what follows we respond to Cappelen’s and Lepore’s attempted reductio by bringing out crucial disanalogies between cases where the binding argument is successful and cases where it is not. But we have a deeper purpose than merely to respond to Cappelen and Lepore: we think the attempted reductio goes wrong by not taking sufficiently seriously the nature of the binding relation that holds between quantifiers and arguments/variables, and that our criticism will serve to highlight the nature and importance of this relation. From Binding to Hidden Variables As we noted, the binding argument adduces facts about variable binding as evidence for the presence of hidden variables in the representation of linguistic expressions. For example, the thought is that we can reveal the presence of a hidden variable in ‘it is sunny’ in (1) by demonstrating binding interactions between that variable and the quantifier ‘Everywhere Sally goes’ in (1b): (1) It is sunny. 2 1 Cappelen and Lepore object to the postulation of hidden indexical variables on the further grounds that these variables (i) are insufficient to demonstrate semantic context sensitivity (which is perhaps the main reason philosophers and linguists have been interested in such variables), (ii) are not bindable by anaphora in the way that ordinary indexicals are, (iii) don’t guarantee certain a priori truths of the kind that are ordinarily guaranteed for indexicals, and (iv) predict the availability of interpretations that are, they claim, unavailable. We’ll have nothing to say about these other arguments. 2 For the sake of uniformity, we’ll use ‘it is sunny’ rather than the contracted form ‘it’s