I am still thinking about belief reports. As mentioned earlier, I like the description theory defended by Kent Bach. On the description theory, 'that' clauses are not semantically singular terms; they do not refer to exact belief contents; they merely describe them. The description theory is committed to (i) the truth-value of 'S believes p' varies with context; or more precisely: it varies with the standards for belief description salient in the context, and (ii) 'S believes p' ascribes a dyadic relation. But (i) and (ii) are compatible with at least two different theories. On one theory, 'S believes that p' has different contents in different contexts. On another, 'S believes that p' has the same content in all contexts (unless 'p' contains indexical terms like 'I') but its extension varies across contexts. Kent says he holds the second view. But on the second view, I think that we need standards for belief description in the index. To see this, consider:
(1) Peter believes that Paderewski has musical talent
(2) Peter disbelieves that Paderewski has musical talent
Assume the second view. And assume for reductio that the index is a pair of a time and a world. On the second view, (1) and (2) can both be true in different contexts where different standards for belief description obtain (in one context we might be intending to describe the exact content Peter believes that Paderewski, the musician, has musical talent, in another context we might be intending to describe the exact content Peter disbelieves that Paderewski, the statesman, has musical talent). Moreover, since (1) and (2) do not contain any indexicals, (1) and (2) have the same contents in all contexts. So, if the index is a pair of a world and a time, then the semantic contents of (1) and (2) may both be true at one and the same index. So, the index cannot be a pair of a time and a world. Instead, it must be a triple of a world, a time, and a standard for belief description. Now, this is interesting, because if I am right, then Kent's proposal is in fact a form of MacFarlane-style non-indexical contextualism.
Sunday, August 06, 2006
Belief Reports and Enriched Indices
Subscribe to:
Post Comments (Atom)
5 comments:
The view proposed in my paper is not contextualist. According to the paper, belief sentences (of the form 'S believes that p') are semantically incomplete, i.e., do not fully express propositions. Contextualism, whether indexical or non-indexical, says something different. Contextualism about a given class of sentences says that they express propositions but possibly different ones relative to different contexts.
But I think I understand why you interpret me as a (non-indexical) contextualist. Reading over sec. 4 of my paper, I can see that if you equate belief reports with the sentences used to make those reports, what I say about what varies with context (specifically, the difference that substitution sometimes makes) would make my view sound contextualist. But you need to distinguish belief reports from the sentences used to make them. It's only the contents of the reports that I say vary with context. (Today, in light of things I say in "Context ex Machina," I'd put that rather differently. )
One other thing. At one point in your post you make the following contrast:
"in one context we might be intending to describe the exact content Peter believes that Paderewski, the musician, has musical talent, in another context we might be intending to describe the exact content Peter disbelieves that Paderewski, the statesman, has musical talent)."
For what it's worth, in section 6 of my paper, "Every case a Paderewski case," I suggest that inserting additional material, like 'the pianist' (or 'the musician' or 'the statesman'), leaves open the possibility of a further Paderewski case, this time involving 'Paderewski the pianist' (classical pianist vs. jazz pianist). And it can keep going with 'Paderewski the classical pianist' (playing Rachmaninov vs. playing Mozart), and going and going. So, for example, in a given context belief reports made using (i) and (ii),
(i) Peter believes that Paderewski the classical pianist has musical talent
(ii) Peter disbelieves that Paderewski the classical pianist has musical talent
could both be true because Peter believes that Paderewski the classical pianist playing Rachmaninov has musical talent and disbelieves that Paderewski the classical pianist playing Mozart has musical talent, where Peter does not believe that Paderewski the classical pianist playing Rachmaninov is Paderewski the classical pianist playing Mozart.
This illustrates why I say that 'that'-clauses are not inherently capable of fully specifying the contents of beliefs.
Hi Kent. Thanks for your comment! It seems to me that you're committed to the following claims.
(i) The semantic content of the sentence 'Peter believes that Paderewski has musical talent' does not vary with context.
(ii) The truth-value of the semantic content of the sentence 'Peter believes that Paderewski has musical talent' may vary with context (i.e., at one context the semantic content of the sentence 'Peter believes that Paderewski has musical talent' may be true, and at another context the semantic content of that sentence may be false).
But if the semantic content of the sentence 'Peter believes that Paderewski has musical talent' is invariant, but the truth-value of the semantic content of the sentence 'Peter believes that Paderewski has musical talent' may vary with context, then why would that not be a kind of non-indexical contextualism?
Or are you saying that the semantic content of 'Peter believes Paderewski has musical talent' relative to a context is not truth-evaluable?
Yes. If such sentences are semantically incomplete (do not fully express propositions), their semantic contents are merely propositional radicals and they are not truth-evaluable.
In your paper you say:
"Semantic content of a belief report: An utterance of 'A believes that S' is true iff A believes a certain thing which requires the truth of the proposition that S."
and:
"For any 'that' clause 'that S', there could be circumstances in which it is true that A believes that S and true that A disbelieves that S but without A's being illogical".
So is the truth-value of the content of a belief report, viz. A believes that S or A disbelieves that S, relative to a way of describing the exact belief content?
No, I don't say that the truth-value of the content is relative to anything (other than circumstances). Notice that the first sentence you quote contains 'a certain thing', not 'something', and 'a certain' is tricky is its own right.
That sentence is one (#5) of eight descriptivist theses. I 've pasted them in below, with several corrections in wording in brackets. Their rationales won't be evident to anyone who hasn't read the paper.
Descriptivist Theses
1. ‘That’-clauses don’t refer, they describe (‘that’ is not a term-forming operator on sentences, and ‘that’-clauses are not noun phrases).
2. ‘That’-clauses generally do not specify complete contents of beliefs (belief-predicates do not individuate belief contents)—a belief report can be true even if the person does not believe the proposition expressed by its ‘that’-clause.
3. Coherent belief/disbelief pairs can be described with the same ‘that’-clause, or with semantically equivalent ones.
4. Belief reports do not distinguish the “how” from the “what” of belief. They make no reference to ways of taking propositions (or to representations or modes of presentation of their constituents).
5. Semantic content of a belief report: An utterance of ‘A believes that S’ is true iff A believes a certain thing which requires the truth of the proposition that S .
6. Belief reports [I should have said 'belief sentences'] are semantically incomplete (not true or false independently of context)—they [belief reports] are sensitive to contextually variable conditions of difference of thing believed.
7. Substitution of coreferring terms in the ‘that’-clause of [the sentence used to make] a belief report can affect its content and change its truth value, but not because of anything encoded by the different terms.
8. Presumed identities license substitution; contextually relevant differences in semantically identical ‘that’-clauses are due to the absence or suspension of a presumed identity.
Post a Comment