August 18, 2004

Ascribed Quantifier Restrictions

Way down in the comments to this entry, Brian Weatherson makes an interesting claim:

I think when we’re reporting other people’s beliefs using restricted quantifiers, it is our domains that matter, not theirs. So consider the following case. A, B and C are hosting a party. A wrongly believes that there is some Glenfiddich in the kitchen cupboard. In fact there is no whiskey of any kind in the house. Now consider the following dialogue between B and C, who are working out what drinks are available for guests:

B: Is there any whiskey?
C: A thinks there is some Glenfiddich.

I think that C correctly reports A’s belief, and correctly reports a false belief. That is despite the fact that it is surely true that there is some Glenfiddich somewhere. The tacit quantifier restriction to in the house or available to drink attaches to the content of the belief, and makes it a report of a false belief.

A problem with this is that (as Angel Pinillos more or less points out in the next comment) it doesn't seem to settle whose domain is relevant. If A's domain were relevant to the ascribed belief, C would still be reported a false and relevant belief of A. But we can fix that: Suppose that there is Glenfiddich upstairs, and A knows this. B and C are discussing what's in the cupboard, because they're only allowed to drink what's in the cupboard. Then if the same dialogue ensues:

B: Is there any whiskey?
C: A thinks there is some Glenfiddich

C (it seems intuitively) has falsely reported A's belief, even if A is at this moment repeating to himself, "There is some Glenfiddich, there is some Glenfiddich." The belief C is ascribing is false; the belief A has is true; so they're not the same beliefs.

What interests me here is how differing quantificational domains, and similar things, might affect things like free indirect discourse and fulsome de re belief ascriptions.

Take the fulsome de re as described by Ken Taylor, which combines de re belief ascription combined with a description of how the believer thinks of the object of the belief. So if the amnesiac war hero Heimson thinks (reading his biography) "Heimson deserves a medal," we could report it de re by saying

Heimson thinks that he, who he thinks of as "Heimson," deserves a medal.

Now apply this to a domain restriction. Suppose that we're of a school where only Plato counts as worth talking about--so when we say "the philosopher," it refers to Plato. And we read the scholastics who refer to Aristotle when they say "the philosopher." The scholastic Eco says:

(1) No dialogues by the philosopher survive.

We can't report (1) by saying:

(2) Eco believes that no dialogues by the philosopher survive

because that would ascribe to Eco the belief that no dialogues by Plato survive. But we can go fulsomely de re:

(3) Eco believes that no dialogues by Aristotle, whom he thinks of as "the philosopher," survive.

So far so good. But in free indirect discourse, it seems to me that it's correct to let the ascribee's usage rule:

(4) Eco pored over the writings of Aristotle. What did this mean? Now he understood what the philosopher was saying.

(I've just read Spiderweb by Penelope Lively, which has a lot of FID passages in which two characters think of their mother as "she." I think that's a similar phenomenon--it's the focus supplied by the characters' context that matters, and their mother is the woman who is always in focus there, so "she" always refers to their mother in free indirect discourse from their point of view.)

I don't have a grand moral here--it just seems like domain restrictions might provide some interesting supplements to the data we get from indexicals when we're discussing belief ascriptions.

Posted by Matt Weiner at August 18, 2004 11:37 AM