I was thinking about Wo’s comments in the post below about Laurie Paul’s paper on contextualism, and I realised I’d overlooked some options for Lewis. Rather than the five options I present, there are at least seven. One more is saying that the predicate is context-sensitive. This won’t really do. Lewis says that the semantic values of (ordinary) predicates are sets of world-bound individuals, and which set is picked out by, say, ‘was brought by a stork’ won’t be sensitive to modal context in the right way.
The seventh option is more interesting. Lewis could say that changes in context change the composition rules. At the time I wrote the post I thought this would violate compositionality. What I realised on reading Wo’s comments is that this need not be so. We can say that which proposition is denoted by ‘a is F’ is sensitive to the contextually determined counterpart relation without violating compositionality, because within a given context the proposition picked out by ‘a is F’ just is a function of the semantic value of ‘a’ and the semantic value of ‘is F’. It almost goes without saying that this option leads to an utterly bizarre semantic theory. If I had a choice between Laurie’s theory that context changes the semantic value of names and the theory that context changes the semantics of predication generally I’d pick Laurie’s theory every day of the week and twice on Sundays. But I think Lewis really is committed to the latter theory.
Consider two contexts such that in the first a has no counterparts that are F, and in the second it does have some. In the first, ‘a is F’ expresses the necessarily false proposition. In the second, it expresses a contingent proposition. By hypothesis, ‘a’ does not change its content between the contexts, and neither does ‘is F’. But the proposition expressed changes. Unless we just abandon compositionality altogether, we have to say that the function that takes a subject and a predicate as input and delivers a proposition as output changes between contexts. I don’t think this is very plausible, but it looks at least consistent.