The price of inscrutability

Some philosophers offer arguments contending that ordinary names such as “London” are radically indeterminate in reference. The conclusion of such arguments is that there is no fact of the matter whether “London” refers to a city in the south of England, or whether instead it refers to Sydney, Australia. Some philosophers have even suggested that we accept the conclusion of these arguments.1 Such a position seems crazy to many; but what exactly goes wrong if one adopts such a view? This paper evaluates the theoretical costs incurred by one who endorses radical inscrutability of reference (the ‘inscrutabilist’). I show that the inscrutabilist can deflect or accept many apparent costs; but that there is one particular implication of radical inscrutability which pushes the price of inscrutabilism too high. An extension of the classic ‘permutation’ arguments for radical inscrutability allow us to establish what I dub ‘radical indexical inscrutability’. This result, I argue, unacceptably undermines the epistemology of inference. The first half of the paper develops the background of permutation arguments for radical inscrutability of reference and evaluates some initial attempts to make trouble for the inscrutabilist. Sections 1 and 2 describe the setting of the original permutation arguments for radical inscrutability. Sections 3 and 4 survey four potential objections to radical inscrutability of reference. Sections 5 sketches how the permutation arguments can be generalized to establish radical indexical inscrutability; and shows how this contradicts a ‘stability principle’—that our words do not arbitrarily change their reference from one moment to the next—that I claim plays a vital role in the epistemology of inference. The second half of the paper develops in detail the case for endorsing the stability principle. In section 6 I distinguish between the deductive goodstanding of a token inference and the validity of argument types. In section 7, I use this distinction to call into question the epistemological relevance of validity of argument types; Kaplan’s treatment of indexical validity partially resolves this worry, but there is a residual problem. In section 8, I argue that the stability principle is exactly what is needed to bridge this final gap, and so secure the relevance of validity to good inferential practice. Section 9 responds to objections to this claim. An appendix to the paper provides formal backing for the results cited in this paper, including a generalization of permutation arguments to the kind of rich2 setting required for a realistic semantics of natural language. Radical indexical inscrutability results can be proved within this setting. Moreover, in the appendix I consider and reject recent criticisms of permutation arguments due to Vann McGee (2005a).