First a couple of examples, and then some rants about why I think they're relevant.
(1) Scooter has put a tack on Mr. Wilson's chair. Karl knows that Scooter did this, but Mr. Wilson doesn't know. Mr. Wilson says to Karl, "I will make life hell for the boy who put the tack on my chair." Karl says to Scooter, "Mr. Wilson thinks he'll make your life hell."
(2) At the baseball stadium, they show pictures of fans on the Jumbotron; whoever gets the most applause wins a free ticket. They show Brian's picture on the screen; Brian does not recognize himself, but Luz recognizes him. Brian points to the screen and says, "He won't win the free ticket." Luz says, "Brian thinks that he won't win the free ticket."
I think (1) is unexceptionable. On the other hand, (2) sounds pretty weird; if Luz's audience doesn't know that Brian has failed to recognize himself, they will be seriously misled. But on most theories of de re attitudes, (2) is fine and (1) isn't.
The idea behind theories of de re belief is that certain beliefs are based on direct access to the object of the belief, such as direct perceptual links (often expressed with proper names). To say the beliefs themselves are de re is to say that the object of the belief is a constituent of the belief--as opposed to the way that the object is thought about. So the belief Brian expresses in (3) would be [Brian, "won't win the free ticket"] as opposed to something like ["That guy won't win the free ticket"] or ["dthat (the guy on the screen) won't win the free ticket"]; where the stuff in quotes is in mentalese. If the belief is de re, then Brian himself is part of the belief.
Since de re beliefs incorporate their objects, it shouldn't matter how you refer to the object when you report the belief. "Brian" and "that guy" and "the guy in the seat next to me" and "the proprietor of Thoughts Arguments and Rants" all refer to the same object, and that object is a component of Brian's belief. So the report in (2) comes out OK on this theory, because Luz's "he" refers to Brian.
De re beliefs are traditionally contrasted with de dicto beliefs, which incorporate a definite description. This description does not reflect direct acquaintance with the object of the belief; you think about the object by means of the description. It's generally taken that a description of a de dicto belief is loose, unless it exactly reproduces the description involved in the original belief.
So (1) should be unacceptable. Mr. Wilson isn't directly acquainted with Scooter as "the boy who put the tack on my chair." He might not be acquainted with Scooter at all. So his belief doesn't reach out and incorporate Scooter, and it should be illegitimate for Karl to substitute the coreferential term "you" in describing Mr. Wilson's belief.
So what do I think is going on?
Well, as I ranted before, I don't think that it makes sense to split beliefs themselves into de re and de dicto. In (2), and in every other situation, Brian's belief doesn't just swallow him. Brian has to think about himself under some guise; this time it's the guise of "The guy on the screen" or just "That guy" or even a mental picture. The whole belief can be put in mentalese in that way. There's an important difference between Brian thinking "I won't win the ticket" and "That guy won't win the ticket"; but if we construe them as de re beliefs, they both come out as [Brian, "won't win the ticket."] That, to me, indicates that construing them as de re destroys important information.
Suppose, then, we abandon the division of beliefs into de re and de dicto. All beliefs are to be represented as mentalese sentences. Then an ascription of a belief has the potential to be correct if every term that the ascriber uses corefers with the corresponding term of mentalese as thought by the original believer. Karl's ascription in (1) has the potential to be correct because his "You" and Mr. Wilson's "the boy who put the tack on my chair" both refer to Scooter; Luz's ascription in (2) has the potential to be correct because her "he" and Brian's "He" both refer to Brian.
Why is there a difference between (1) and (2)? If someone wants to get the clearest possible picture of the original belief from a report, they have to figure out the original mentalese phrasing of the belief. In (1), it's obvious that "you" can't be exactly what Mr. Wilson thought. So Scooter can ask himself, when Mr. Wilson thought, "I'm going to make [blank's] life hell," what term referring to me would have gone in the blank? "The boy who put the tack on my chair" is a natural answer--even if Mr. Wilson knows Scooter's name.
In (2), Luz's audience can ask themself, when Brian thought "[Blank] won't win the ticket," what term referring to Brian would have gone in the blank? Well, unless you know most of the backstory, the first term you'll think of is "me." Most people frequently think of themselves in the first-person (or in terms that they know to be synonymous with the first person, like "Bob Dole"). In (2), if you discover that Brian is going around saying, "I'll win the ticket," you'll probably take Luz to be proved wrong, because the most natural interpretation of her utterance has been proved wrong.
In general, I think there's going to be a range of permissible substitutions for the original terms of the belief. The range may depend on the context--what you're trying to convey with the report, how important it is to give a precise idea of the original belief, which assumptions the hearer will most naturally make. I want to suspend judgment about whether an impermissible substitution falsifies a belief report, but some substitutions will certainly mislead.
The problem with the division of beliefs into de re and de dicto is that it creates a black-and-white picture of permissible substitutions. If you're reporting a de re belief, you can use any term that refers to the res; if you're reporting a de dicto belief, you have to use the exact description that appears in the original mentalese. But it'll be rare that the range of non-misleading substitutions hits either extreme. And there's no particular reason to think that the range will always be narrower for thoughts containing definite descriptions as opposed to thoughts containing indexicals or demonstratives.
(This rant is occasioned because I've been reading up on the linguistics literature that Kai von Fintel cites here concerning the de re and the de se, and the division of beliefs into de re and de dicto seems to go pretty much unquestioned throughout. I think this gets the debate off on the wrong foot. For instance, Higginbotham thinks de se thoughts are a will-o'-the wisp; but surely we understand what it is for me to think "My pants are on fire" better than we understand what it is for me to think [x's pants are on fire] with Matt assigned to x? There's a lot in this literature I need to think through, but taking de re beliefs as fundamental is a philosophically fraught move.)Posted by Matt Weiner at February 13, 2004 12:21 PM