Insensitive Semantics, Chapters 3 and 4

Marie Guillot (who’s visiting MIT this year) led our discussion today. Here is the handout she made. In many ways, chapter 3 is a crucial chapter in the book, because it’s supposed to establish the following conditional. In the post on the first couple of chapters, I talked about the two kinds of philosophers Cappelen and Lepore officialy oppose: moderate and radical contextualists. Chapter 3 is aimed at establishing that the grounds that philosophers usually cite in favor of moderate contextualism rationally force them to endorse not just moderate contextualism, but radical contextualism. Many theorists find the latter position untenable, and C&L will attempt to show that radical contextualism is as bad as many people think it is: empirically inadequate and even internally incoherent. And it indeed seems to me as if of the two steps (if moderate contextualism, then radical contextualism, and radical contextualism is false), many people will accept the latter. So most of the action is going to be here.

The basic argument is this: context-shifting arguments are arguments that moderate contextualists accept as sound. Such an argument focuses on a particular expression and is establishes, according to the moderate contextualist, that that expression is context-sensitive. C&L then argue that the very same argument can be constructed for just about any sentence in the language. So to the extent that radical contextualism is the claim that (just about all) expressions in the language are context-sensitive, radical contextualism follows by the lights of the moderate contextualist. Hence, moderate contextualism collapses into the radical variety.

I think that C&L succeed in this much: they can show for just about any sentence that there is a situation in which that sentence is appropriate, and there is another context in which it is not. For many of the examples, in fact, they even succeed in showing that there are contexts in which we want to respond to an utterance by saying “that’s true,” while in another context we want to say “that’s false.”

It’s also pretty clear that if you’re a moderate contextualist, you’re not going to deal with the all of the examples in the same way. Maybe in some cases, you’ve got a situation in which appropriateness doesn’t track semantic content; maybe in other cases, you’ve got a hidden indexical; in other cases, still, you’ve got a different mode of syntactic combination.

But what’s less than clear is whether the fact that, as a moderate contextualist, you have to go case-by-case, maybe dealing with classes of examples at a time, has any bearing on the debate. Certainly, C&L seem to demand that their opponent offers a unified response to all (or almost all) cases. It seems to me that there could be two reasons for imposing this condition.

1. Evaluating a degenerating research program. Moderate contextualists really have a research program. C&L somewhat snarkily describe that fact by saying that a moderate contextualist will view their examples as a challenge and a chance to generate some more publications. But the fact remains that for someone who’s working in a research program, the response to apparently recalcitrant data is to try and fit those data into the program. That looks like exactly the right response. However, at some point the fitting of data no longer seems like it’s offering enlightenment, and the research program degenerates. So perhaps they want to impose a uniformity condition in order to guard against dealing with a degenerating research program.
2. The nature of semantic knowledge. The previous point is about how to theorize, and it’s independent of considerations peculiar to the domain. Alternatively, the demand for a (relatively) uniform account of context-sensitivity might derive from some feature specific to semantic knowledge. For example, one might think that semantic knowledge must be very simple by some measure of simplicity, because semantic knowledge is acquired by very young people, and it’s acquired uniformly across a wide array of learning environments.

Chapter 4

This chapter is mostly concerned with trying to say where things go wrong. C&L suggest this: given what they’ve argued in the previous chapter, you cannot think that context-shifting arguments tell you much, or indeed anything, about the semantic content of the sentence the argument is about. Thinking that they do is what they call the mistaken assumption, and here’s how they formulate it.

A theory of semantic content is adequate just in case it accounts for all or most of the intuitions speakers have about speech act content, i.e., intuitions about what speakers say, assert, claim, and state by uttering sentences.

As Marie pointed out today, this sounds pretty clear, but once you think about it, it’s not really clear what it says. A lot of the issue focuses on what accounts for means in that formulation.

• It might mean is part of the explanation. In that case, C&L accept this principle, since obviously, the semantic content is going to be part of the explanation of our intuitions about speech act content, at least if you think that we’re pretty good at detecting speech act content, even if we suck at detecting semantic content. For after all, C&L agree that semantic content plays some role in explaining why a particular speech act has the content it does. It’s just kind of circuitous on their view.
• It might mean is identical with, i.e., if we intuit that the asserted content is that p, then the semantics has to assign that proposition as the semantic value of the sentence in context. The problem here is that nobody their attacking can possibly hold this position, since none of the people their attacking think that the semantic value of a sentence, in context, is a proposition. (This by their own characterization of moderate and radical contextualism.) To the extent that there’s anybody aside from them who agrees with this, it’s people who do semantics in the tradition of Kaplan—people like Jason Stanley or Zoltan Szabo. They think that the semantic value of a sentence in context is a proposition, and that at least many intuitions about the truth-value of a sentence should be taken as good evidence for the truth-value for the sentence in context.

So it’s got to mean something else. But what?