February 20, 2004

D'jaknow Who?

In comments to this entry, Robbie asks a good question about my doctrine that there's no such thing as a de re belief: How do I account for a statement like (1)?

(1) There's someone in my class who Mr. Wilson thinks left a tack on his chair.

This seems to require quantifying over the object of Mr. Wilson's thought:

(2) (Ex)(Mr. Wilson thinks x left a tack on his chair)

In (2) it looks as though Mr. Wilson's thought will have to be about x. So doesn't Mr. Wilson have a de re thought here?

Well, there's a problem here, and I'm not sure how to solve it, but I don't think de re thoughts help.

Take the case where Mr. Wilson sees Scooter running away (but doesn't recognize him). He says:

(3) That boy left a tack on my chair!

But let us suppose he will also say:

(4) Scooter didn't leave a tack on my chair. In fact, no boy in this class left a tack on my chair.

In this case (1) seems false. Intuitively, there's no one in the class who Mr. Wilson thinks left a tack on his chair, because he can't identify "that boy" as someone in the class. However, (3) is taken as a paradigmatic expression of a de re belief, so on the theory of de re beliefs (1) comes out true.

My first stab at solving this is to use substitutional quantification. So instead of (2), (1) gets represented as

(5) For some ____, ____ is in my class and Mr. Wilson thinks that ____ left a tack on his chair.

Here the range of the quantifier has to be restricted to some set of acceptable descriptions--probably set by context. If Mr. Wilson knows the names of all the kids in the class, it would probably be OK to restrict the range to names. Then (1) comes out false because of the falsity of this instance:

(6) Scooter is in my class and Mr. Wilson thinks that Scooter left a tack on his chair.

Mr. Wilson's thought (de dicto, as they all are) is "That boy left a tack on my chair," and it's impermissible to ascribe that with the second half of (6); though "Scooter" and "that boy" corefer, "Scooter" isn't in the range of permissible substitutions.

I'm not completely happy with the idea of substitutional quantification, but I'm not sure how else to capture the idea that it matters how Mr. Wilson's thought picks out its object. On the other hand, the construal of Mr. Wilson's thought as de re seems to imply that it doesn't matter how the thought picks out the object, and I think (3) and (4) show that that isn't right.

One last thought: The line could be taken that in the scenario I've described, (1) is true but misleading. OK, but I can take the same tack (ha ha!). My original proposal was that "A thinks that B phis" is true iff A has the thought "X phis" for some term X that refers to B and falls within some contextually set acceptable range. I can drop the acceptable range and declare that any X that refers to B is acceptable, though we will often have statements that are true but misleading. Then (2) works as a logical form for (1)--in fact, it's true but misleading.

This proposal generates more true-but-misleading cases than the proposal that some beliefs are de re, but I think that's hardly any cost at all. The cases are misleading in pretty much exactly the same way, so why knock yourself out trying to exclude one of them?

Posted by Matt Weiner at February 20, 2004 02:23 PM

How would the substitutional proposal work with empty names? I guess we don't want to get from 'I believe Pegasus is a winged horse' to 'there's something I believe to be a winged horse' (I presume this problem will arise in more natural sounding cases).

So, I guess I'm in favour of your second suggestion, though I'm not sure I've grasped the details. It sounds pretty close to Kaplan's line in 'Quantifying In' (a very good thing, imho). In any case, here's a standard worry. We don't want to get from 'Ortcutt believes the oldest spy is a spy' to 'There's someone Ortcutt believes to be a spy'. (Replace 'the oldest spy' with the corresponding descriptive name, 'Oldie', if the mood takes you).

At this point, I'm inclined to declare 'true but misleading'. However, I've been thinking about that puzzle for a while, but have yet to find a pragmatic story that convinces me. Any suggestions?

Posted by: Robbie at February 23, 2004 03:46 AM

Shorter version of Robbie's post: "So, Mr. Weiner, tell us what you think you might be working on in five years."

As of this moment, I'm leaning toward a story on which belief-attribution always or almost always turns out to be imprecise. So Ortcutt's belief (remember that the belief itself is de dicto)

(7) Ortcutt: Oldie is a spy

turns out to be a very weak basis for saying

(8) There is someone that Ortcutt believes to be a spy.

Suppose now that Ortcutt has the belief

(9) Ortcutt: Van is a spy

where Van is the real name of someone Ortcutt has heard of, and that in fact Van = Oldie, that is, Van is the oldest spy. Then (3) is under most circumstances a strong basis for saying (2). But if Ortcutt has the belief

(10) Ortcutt: That man is a spy

that may or may not be a strong basis for (2)--it could depend on whether Ortcutt could identify the man again.

I think my problem the doctrine of de re beliefs is that it's an attempt to come up with a clear division between acceptable and unacceptable belief-ascriptions. So [in the main post] (3) makes (1) true, true, true but misleading, and in the original case "Mr. Wilson thinks he'll make your life hell" comes out false, false, false but non-misleading. And I don't see that it's useful to us to make this black-and-white division. (This paragraph needs a footnote to Elijah Millgram for inspiring some of the framework.)

Now, a problem is that I haven't really said anything about why (7), (9) and (10) justify the following ascriptions respectively:

(7*) Ortcutt thinks that Oldie is a spy
(9*) Ortcutt thinks that Van is a spy
(10*) Ortcutt thinks that the man he saw is a spy.

Maybe I can borrow from Geoffrey Nunberg's paper "Indexical Descriptions and Descriptive Indexicals", in which he argues that in some contexts the referent of an indexical/demonstrative contributes one of its properties to what is said--e.g. "This game is usually exciting" might mean that the annual Stanford-Berkeley game is usually exciting. So in (7*) Oldie/Van contributes the property of being thought about by Ortcutt as the oldest spy, while in (9*) Oldie/Van contributes the property of being thought about by Ortcutt as "Van."

Could I stretch this so far as to say that in (8) the witness to the quantifier contributes the property of being thought about by Ortcutt in some relatively definite way--"Van" counts, "Oldie" doesn't? Seems extreme.

Anyway, this needs a lot more thought!

Posted by: Matt Weiner at February 23, 2004 01:08 PM