Contextual Indefiniteness and Semantic Theory

Call for abstracts for a two-day workshop organized by John MacFarlane, Paris IAS fellow

The Institut d’Études Avancées de Paris will host a two-day workshop on “Contextual Indefiniteness and Semantic Theory.” The workshop will take place April 19-20, 2017 at the historic Hôtel de Lauzun in Paris.

If you are interested in presenting a paper at the workshop, please submit an abstract of no more than 2 pages, in PDF format. We welcome submissions from anyone who works with formal semantics, whether in linguistics, philosophy, or another discipline. Abstracts should be sent to contextual.indefiniteness@gmail.com by February 15, 2017. Notifications of accepted abstracts will be sent out by February 28.

The workshop will be jointly funded by the IEA de Paris, the Institut Jean Nicod, and UC Berkeley (via John MacFarlane).

Prospectus

The dominant framework for work in the semantics of natural language is truthconditional semantics. This approach has its roots in the work of Gottlob Frege and Alfred Tarski, who showed how to give recursive truth definitions for subsets of the language of mathematics. For this language, devoid of vagueness and contextual sensitivity, it made sense to talk of the condition for a sentence to be true, simpliciter. But natural language sentences like “I have been there” can only be classed as true in relation to a context. (Who is speaking? At what time? What place is meant by “there”?) So, natural-language semanticists replaced the project of giving conditions for sentences in a language to be true with the project of giving conditions for sentences to be true at a context. It has now become standard to talk of a context as determining a speaker, addressees, a reference time, referents of demonstratives and indexicals, a “modal base” of possibilities, quantifier domains, similarity metrics or selection functions for conditionals, reference classes and cutoff points for gradeable adjectives, and much else. (See any textbook in semantics.)

There is reason to wonder whether semanticists have become too comfortable with this formalism. In the cases that initially motivated the introduction of contexts (speaker, place, time, demonstrated objects), speakers and hearers will normally have mutual knowledge of the values of the relevant contextual parameters. When they don’t—for example, when the hearer is not sure which of several dogs the speaker meant to be referring to in using “that dog”—it is a defect in communication (which is not to say that the hearer cannot gain information from the speaker in such cases). However, semanticists routinely posit contextual parameters whose settings cannot possibly be mutually known between speakers and hearers. Three examples:

1. Kennedy (2007) posits a contextually determined function sc, which maps degree functions for gradeable adjectives on to cutoff points for the positive degree of the adjective. But it is not plausible that every time I say that something is “large,” I must have a definite cutoff point function in mind, and even less plausible that speakers and hearers must converge on one in successful communication.
   
   
2. Standard semantics for counterfactual conditionals posit a contextually determined similarity metric. Allowing contextual variation is well motivated: in some contexts, “If Caesar had been in command in Vietnam, he would have used catapults” expresses a truth; in others, a falsehood (Caesar would have used the A-bomb!). But a similarity metric settles not just how to evaluate this counterfactual, but how to evaluate indefinitely many others. (Would Caesar still have shaved? Would he have spoken English with a Latin accent?) It hardly seems plausible that a speaker must have the kind of determinate intentions that would be required to pick out one of the indefinitely many similarity metrics that would be compatible with the truth of the sentence asserted, or that hearers must converge on such a similarity metric in order to have understood.
   
   
3. As François Recanati pointed out long ago (Recanati 2003), even a simple indexical like “here” can be used to denote differently sized regions, depending on the speaker’s intention. But it seems implausible that speakers must, in general, have in mind a particular determinate region. If I say “it’s warm here” and you ask, “in this room? in this suite? in this building?”, you are very likely going beyond what my intentions determine. Yet my intentions do rule out some possibilities: I didn’t mean to include the whole city, for example.

Although examples like this abound, it is rarely noted explicitly that the settings of the posited contextual parameters are not mutually known. Even when it is noted, there has been little explicit recognition that it might be problematic.

But, arguably, it is problematic. If I know that the sentence you’re using is true in this context just in case p, and I know that you know that I know this, then I can infer from your utterance that you’re trying, by means of your utterance, to communicate that p. I can figure out what you, the speaker, mean by leveraging our mutual knowledge of the truth conditions of the sentence you use. Indeed, it is precisely because mutual knowledge of truth conditions can play this role that a theory of truth conditions can play a central role in a theory of linguistic meaning. But when there is no mutual knowledge—when the hearer and hearer lack a shared understanding of what is required for the speaker’s sentence to be true in the context—this explanatory schema breaks down. We may assign truth conditions, but it is hard to see how this assignment can be part of a broader explanation of meaning (for this general argument, see MacFarlane 2016).

The purpose of this workshop is to assess whether it is problematic to posit contextual parameters whose settings are not mutually known (even in normal cases of successful communication), and to consider the upshots. Is it plausible to think of definitions of truth at a context as theories of meaning when the relevant contextual parameters are not known in common between speakers and hearers? How do such definitions figure in explanations of communication? If it is problematic to posit inscrutable settings for contextual parameters, what are the upshots for the abundant work in semantic theory that seems to posit them? Are there ways of reconceiving this work without throwing away its insights? Is it time to rethink the way formal semantic theory relates to the use of language?

Papers may address these questions directly or indirectly, through their connections with recent discussions of metalinguistic negotiation, semantic minimalism, loose talk, vagueness, expressivism, relativism, dynamic semantics, presupposition and the common ground, knowledge which, or the semantics of particular expressions such as quantifiers, conditionals, gradeable adjectives and modals.

References

Kennedy, Christopher. 2007. “Vagueness and Grammar: The Semantics of Relative and Absolute Gradable Adjectives.” Linguistics and Philosophy 30: 1–45.

MacFarlane, John. 2016. “Vagueness as Indecision.” Proceedings of the Aristotelin Society s.v. 90: 255–83.

Recanati, François. 2003. Literal Meaning. Cambridge: Cambridge University Press.