I went to an excellent paper by “Yael Sharvit”:http://www.sp.uconn.edu/~sharvit/ on free indirect discourse and _de re_ pronouns. I hadn’t seen any theoretical work on free indirect discourse before, so I spent most of the talk being flooded with lots of new information.

Sharvit argues that third person pronouns in free indirect discourse are not _de re_, which would be a really stunning result if it is true. It would really shake up a lot of the attitudes towards pronouns that have been orthodoxy ever since Kaplan’s early work. She also argues that theorising about free indirect discourse provides an argument for the existence of _de se_ pronouns in natural language.

I’d need to spend a lot of time to really evaluate the main argument, so I don’t want to endorse any of the conclusions in public yet. But I was really impressed by how sensitive the argument was to very delicate points of data. Philosophers are so often careless about the details of what they are doing when they discuss language. (I don’t mean all philosophers here, but I do mean to include me in the clumsy group.) The most obvious example is when a philosopher purports to be discussing the construction _If p then q_ and then never provides an example sentence containing the word _then_. There’s frequently an assumption that various constructions are to be treated alike, so it doesn’t matter if all of the data are drawn from one particular class within that construction. As I said, I do this too, so I’m throwing stones in glass houses here. But I’m always impressed when these assumptions get questioned, and especially when that questioning leads to results.

I could provide several examples of just this kind of questioning attitude from Sharvit’s paper, but unfortunately the paper isn’t online, so I can’t link to it, and I’d feel a little bad repeating some of the main examples before the paper goes into circulation. So I’ll just say a little bit about the background, and why we should be interested in free indirect discourse (FID). Compare first (1) and (2).

(1) John thought that the Red Sox should have won last year.
(2) The Red Sox should have won last year(, thought John).

FID is the second kind of report. The parenthetical addition is because it’s optional whether we need to say just who is doing the thinking, or even that it’s thinking, rather than say saying, that’s going on. It’s much easier in FID to leave this up to context than it is in regular indirect discourse. We see the first signs of a distinction in the way tenses are treated.

(3) John thought that it was time for a party.
(4) It’s time for a party, thought John.

Although we have to use ‘was’ in (3), we can use a present tense verb in (4). Indeed, it is better to use a present tense marker in (4). The same kind of thing applies to temporal pronouns. Assume that at 2 John thinks “I’ll will be there by three,” and the speaker is at the location denoted by John’s ‘there’.

(5) John thought that he/*I would be here/*there by now/three.
(6) He/#I will be there/*here by *now/three(, thought John).

In FID, ‘now’ is governed by the context of the thinker. Similarly the appropriate place pronoun is determined by the relation of the thinker to the place, not the speaker. But first-person pronouns seem to be different. ‘He’ is perfectly acceptable in (6), and ‘I’ is at least a little troubling. It’s not awful, I think, but ‘He’ is probably better. Still, the behaviour of ‘now’ is very strange, since we’re used to thinking that it just picks out the time of utterance in indirect reports.

There’s many many more examples where those came from, but as I said I’m a bit nervous of stealing Sharvit’s thunder. Hopefully her paper will appear shortly, and I can talk about the examples in greater detail, because the conclusions seem to be philosophically important.