Numerous difficulties arising in connection with developing an ontology for linguistic entities can be thought of as manifestations of a more general problem, aptly characterized by David Lewis (1975) as a tension between two conflicting conceptions of language. On the one hand, our best theories model languages as abstract semantic systems—roughly, functions assigning meanings to expressions. On the other hand, we think of languages as contingent and changing social constructs—both grounded in, and grounding, various social relations and institutions of human (...) beings. There are various ways in which these conceptions appear to be in conflict. For instance, if languages are set-theoretical entities—as our best logical and linguistic theories would have it—how do we account for the fact that they change? Or that they could have been different in various respects? And how do we provide principled and noncircular conditions for set-membership, when there appears to be nothing in common to all tokens of the same expression other than belonging to the same type? This paper aims to develop an ontology of linguistic entities—specifically, of languages and linguistic communities—that can resolve these apparent tensions. (shrink)

A widespread intuition is that words, musical works, and flags are intentionally produced and that they’re abstract types that can have incorrect tokens. But some philosophers, notably Julian Dodd and Nicholas Wolterstorff, think intention-dependence isn’t necessary; tokens just need to have certain relevant intrinsic features to be tokens of a given type. I show how there’s an unappreciated puzzle that arises from these two views: if tokens aren’t intention-dependent and types can admit of correct and incorrect tokens, then some driftwood (...) that washes up and forms what seems like the word ‘cat’ may simultaneously be a misspelling of ‘act’ and innumerable other misspelt words. I consider various ways Dodd and Wolterstorff can respond to this counterintuitive result and argue that biting the bullet, as well as nearby views, are implausible. Thus, they need to give up one of the two commitments, and I argue that requiring the intention-dependence of tokens, rather than the mere possession of certain intrinsic properties, is the best option. (shrink)

Existing metasemantic projects presuppose that word- (or sentence-) types are part of the non-semantic base. We propose a new strategy: an endogenous account of word types, that is, one where word types are fixed as part of the metasemantics. On this view, it is the conventions of truthfulness and trust that ground not only the meaning of the words (meaning by convention) but also what the word type is of each particular token utterance (words by convention). The same treatment extends (...) to identifying the populations through which the conventions prevail. We consider whether this proposal leads to new underdetermination challenges for metasemantics, and make a case that it does not. (shrink)

It is plausible to think that our knowledge of linguistic types can bejustified by what we know about the tokens of these types. But one then hasto explain what it is about the relation a type bears to its tokens that makespossible the move from knowledge of the concrete to knowledge of theabstract. I argue that the standard solution to this difficulty, that the relevant relation is instantiation and that the transition is inductive generalization, is inadequate. I propose an alternative, (...) according to which tokens are representations of the type they belong to. I also defend this view against the charge that it cannot account for the systematic ambiguity of expressions like ‘word’ or ‘sentence’, and the objection that it leads to an implausible form of Platonism. (shrink)