Kai von Fintel (MIT): Variable costs

(joint work with David Beaver)




Unbound pronouns as in He's gone on vacation are standardly treated as free variables semantically. It is assumed that they receive a value based on contextual information. Using such a free variable pronoun incurs a debt on the part of the speaker: the context must provide a determinate value (Tonhauser et al. 2013 calls this constraint "Strong Contextual Felicity"). The device of free variables is also used quite extensively to capture other types of contextually filling in of information. In many cases, this is a way of providing contextual glue to semantically underspecified combinations, consider for example, compounds (apple juice seat, Downing 1977), genitives (John's team, Partee 1984), free adjuncts (Having grown up in Sweden, Lars can touch the ceiling, Stump 1985), elided verbal arguments (We ate, which is understood existentially vs. We pushed, which requires a salient object needing a push, Condoravdi and Gawron 1996), and filling in of temporal reference intervals required by tenses (e.g. Reichenbach 1947). These glue uses of free variables, like presuppositions more generally, are not uniformly subject to the same strong contextual felicity constraint. One might use John's team in a context that does not uniquely determine a salient relation between John and the team; in this case, and depending on context, hearers might assume some kind of ownership or membership relation as a default. A speaker might use apple juice seat and hearers might simply assume that the speaker has some relation or other in mind between the apple juice and the seat. It appears then, that there is more than one way in which free variables can interact with the context in conversations. We attempt to motivate a taxonomy of the contextual effects of free variables and suggest some ways in which theories of semantics and pragmatics might capture the data. (There is some precedent in work by Schwarzschild 1999, 2002, Kratzer 2003. There are many other connections to recent literature, e.g. to von Fintel & Gillies' 2011 account of contextual indeterminacy with epistemic modals, to issues in the interpretation of contextually incomplete definite descriptions, cf. e.g. Ostertag 1999, Buchanan & Ostertag 2005, and to differences that have been reported in cross-linguistic studies of the strength of presuppositional requirements, c.f. Tonhauser et al. 2013 and Matthewson 2006.)