I’m always reluctant to start thinking about Frege’s puzzle, much less talking about it. It’s not just a hard problem but also one of the most discussed in recent philosophy of language. It’s far too difficult to get clear on how the debate has gone to be confident that my point hasn’t been refuted in a paper written before I was born. I feel the same about the internalism/externalism debate, and probably some others. In any case I want to blog a bit more about it and hopefully find out what some of those anticipatory papers are.
I was thinking today about how to set up the problem. It struck me that there might be two versions that it would be useful to highlight given my interest in names. Both rely on the classic belief report pairs, i.e.:
(1) Lois believes that Superman flies.
(2) Lois believes that Clark Kent flies.
Puzzle A: Whatever we think of the of the syntax and semantics of (1) and (2) let’s stipulate that they’re the same except for the proper names ‘Superman’ and ‘Clark Kent’. Suppose that we stipulate that [['Superman']] = [['Clark Kent']], as the Millian would have us believe. Then, if our semantics is compositional, (1) and (2) can’t differ in truth-value. They both have the same structure and every element has the same semantic value as its counterpart in the other sentence.
Puzzle B: Suppose that we say that when we have attitude verbs like ‘believe’ and a complement clause that has as its semantic value a proposition p, the complex sentence is true iff the subject of the sentence stands in an appropriate relationship (i.e. belief) to that proposition. Suppose that names contribute their semantic values to the propositions that the sentences they appear in express, and make the Millian assumption relied on in A. It looks like the proposition for both (1) and (2) will be the same. In that case, given that they are otherwise identical, they can’t differ in truth-value.
Here is why I was thinking that separating puzzle A from puzzle B might be a good idea. On the view where the semantic value of ‘Superman’ is not Superman but something like [λx. x is called 'Superman'] we can distinguish (1) and (2) on the basis that they have constituents with different semantic values. So, puzzle A as I have worded it doesn’t arise. But if this view is combined with the view that occurrences of names contribute objects to propositions, then we nonetheless get puzzle B.
I’ve been thinking about arguments for semantic underdeterminacy that use the non-existence of eternal sentences as a premise. For instance, Robyn Carston gives an argument like that in the first chapter of her book Thoughts and Utterances. I’d like to say something about the whole issue, but I don’t really know what the literature on it is like. Carston cites papers by Wettstein and Sayward. They’re responding to Quine’s argument in Word and Object that we can replace talk of propositions once we recognise that eternal sentences do the job we want propositions for just as well. Is there anything else that I really have to read before I get started?
I’m visiting CSMN for two months. So far I’ve found Oslo in January to be cold and slippery, but I think things can only get better.
I have a tentative plan to get clear on what, if anything, the semantics of propositional attitude reports has to be with my thesis and the contextualism debate in general.
I like direct reference theories of proper names, i.e. I think that the proposition expressed by a simple sentence (one with no intensional operators) including a proper name will have as a constituent the object named. I also like theories that treat names syntactically and semantically as combinations of a determiner and a predicate. To be compatible with the direct reference picture the determiner will have a semantics rather like David Kaplan’s ‘dthat’ (or ‘dthe’). This seems fine to me as a consequence. I think there are three reasons for holding the general picture I have just sketched. (i) It unifies names and other NPs (as Richard Montague’s treatment of names and quantifiers is rightly praised for doing). (ii) There is linguistic evidence for this (as has been argued by Tyler Burge and others). (iii) There is no need for the claim that names are ambiguous (I take this to be an improvement over Nathan Salmon’s account).
The obvious question to ask is what sort of predicate is to be combined with the determiner. I think that there are good reasons, famously given by Saul Kripke, to think that ordinary properties or bundles of properties won’t do. Therefore the property I favour is the property of being called ‘N’. This has been argued by several people, probably first by Kent Bach.
I take it that the major sort of objection to this kind of view will be an alleged loss of rigidity, where this is a way to make precise in possible world semantics a problem with the intuitive truth conditions of expressions involving names in modal contexts. One thing to note is that there is reason to think that straightforward definite descriptions exhibit a kind of ambiguity. For instance,
(1) Necessarily, the Prime Minister is male.
I get a reading where this is true, the Prime Minister is Gordon Brown and necessarily he is male. (I’m assuming that a different sperm and egg would have meant a different person. Use a different example if you don’t like that.) I can also get a reading where this is false, I think it’s perfectly possible for there to be a female Prime Minister, in fact for us to now have one. If names are really just like definite descriptions then the issue is how to avoid the possibility of this latter reading. Stipulating the directness of a ‘dthat’-style operator would presumably do it, but would look ad hoc.
An idea that I have toyed with, but that I am still rather unsure about, is to put this down to the use to which names are put. I was inspired to do this after reading a paper of Robin Jeshion’s, who does not propose this but who does present some interesting data on the way people go about naming or not naming things in their environment. The idea would be that we only give names to things that we are interested in as individuals, rather than just as tokens of a type. This makes the reading where we are interested in types of the token ‘called N’ very hard, if not impossible, to get. Is this at all plausible? Or any less ad hoc?
I’ve been reading Jeff King’s book The Nature and Structure of Content. It’s an interesting read. I’ve been thinking about an objection that seems so simple it must be wrong-headed. My assumption is that one of the things a theory of propositions must do is explain, or explain away the intuition that
(1) Tom believes that Hesperus is bright.
and
(2) Tom believes that Phosphorus is bright.
can differ in truth-value. Now I wonder whether King’s can do it. His view about what propositions are is somewhat tricky to summarise. Here is something from p. 40 of the book. A proposition is a fact, namely the fact that:
[T]here is some context c and some words a and b of some language L such that a has x as its semantic value relative to c and occurs at the left terminal node of the sentential relation R that in L encodes the instantiation function and b occurs at the right terminal node and has as its semantic value in c y.
Take the embedded clause in (1) as an example:
(3) Hesperus is bright.
(3) is an expression of a language (English) used to say that some object, the referent of ‘Hesperus’ i.e. Venus, falls in the extension of some property, the referent of ‘bright’. This meets the conditions: Two expressions of English stand in the relation described and have certain objects as their referents. The fact that this is so is the proposition, on King’s view. The embedded clause in (2) gives rise to just the same fact because (3) and the embedded clause in (2) have identical structures and [['Hesperus]] = [['Phosphorus']]. As far as I can see, nothing about propositions so construed can explain the intuitive possibility of (1) being true and (2) being false. Certainly it would be hard to say that it is because the relata in the belief reports are different, as surely in both cases they are Tom and one and the same proposition. Am I missing something?
Rafal Urbaniak was at our conference and he’s posted his thoughts on one of the papers here.
My friend Alex is writing a book about hope based on interviews with people who have interesting stories to tell. As I understand it he’s planning a sort of contemporary social history. He has a blog about the project here.
I’ve found two seminars to go to at the University of London. Barry Smith is convening one called ‘Philosophical Linguistics and Philosophy of Language’. At King’s College there’s a seminar called ‘Formal Semantics and Pragmatics’ convened by Eleni Gregoromichelaki, Ruth Kempson, and Wilfried Meyer-Viol which looks like a lot of fun as well. I’m finding it hard to get my hands on a copy of Bob Carpenter’s book Type-logical Semantics for the latter, but I’m not discouraged.
