Monday, April 05, 2010

Are King's propositions too fine-grained?

Jeffrey King is well-known for his account of propositions as worldly entities, as facts consisting of objects, properties and relations. The fact that King claims is a propositions is of the following sort: there is a language containing some expressions that stand in certain sentential relations (basically, the way they got syntactically combined) , with each expression having as semantic value an object, a property or a relation. (This is the basic set-up, he adds more bells and whistles on top of that.) One main advantage of King's view is the ability to solve a major problem for unstructured views of propositions (especially for the propositions-as-possible-worlds view): namely, accounting for necessary truths (or falsehoods) in a way that doesn't make them all equivalent. Since King's propositions inherit their structures from the sentential relations that bind together the words in the sentences expressing those propositions, each proposition (including necessary ones) will have a different structure, closely related to the sentence used to express it. A related problem that is nicely solved in King's framework are the different puzzles arising from embeddings under propositional attitude verbs: for example, the propositions expressed by the sentences "Annie ran 20 kilometres" and "Annie run 12.43 miles" are different, and that accounts for Bill's (who's ignorant about lenght measures) believing one and not the other. It thus seems that on King's account propositions have enought structure as to count as different when we want them to count as such.
This is all nice and good. The question now is: doesn't this positive feature of King's view turn on closer inspection into a negative one? For, as it has been pointed out, it could be that now propositions are too fine-grained. For King claims that not only the propositions expressed by "Annie ran 20 kilometres" and "Annie ran 12.43 miles" are different (which might be easier to accept), but also, for example, that the propositions expressed by the sentences "1=2" and "2=1" are different. This might very well strike some as being utterly counterintuitive.
King is aware of the counterintuitiveness of his claim, and therefore tries to alleviate the worry. To this effect, he asks the reader to compare the propositions expressed by the sentences above with those expressed by the following ones: "1<2" and "2<1". Do these sentences express different propositions? They clearly do. But notice now that what makes the latter sentences express different propositions is just the different order of their constituents. But if that's the case, why shouldn't we accept that "1=2" and "2=1" also express different propositions? Our reluctance to do so is traced down by King to one peculiar feature of the relation that the equality sign stands for: namely, its transitivity. It's true that equality is transitive, King says, but that is just a feature of that particular relation, and it shouldn't bear on the issue whether the same propositions is expressed or not by sentences that differ only in the order of their constituents. I find the explanation involving the peculiaity of equality convincing, but I also understand that one could still feel that one's intuitions about the identity of propositions have not being attended to. To be sure, King has a shot at dispensing with those intuitions; usual motives are invoked - their unreliability, them tracing other kinds of content than propositional content, etc. But this seems to me problematic, at least for the following reason: if King has to give up intuitions at some point, the defendant of the unstructured propositions view could do the same. The set of intuitions given up by each camp will be different, of course, but what becomes unclear is whether King can still claim the advantage he thinks his view has over the unstructured propositions view. So, the questions to be answered are: Do you believe that King has a problem here - are his propositions too fine-grained? Do you find his explanation in the case of the propositions expressed by "1=2" and "2=1" correct? What do you think about his dismissal of (some) intuitions? [This post is a late semi-transcription of a discussion that took place at a reading group in Paris, and the points raised here were originally raised by Michael Murez and Adrian Briciu.]

13 comments:

Dan López de Sa said...

Pardon my ignorance, but which role characterizes the notion of a "proposition" in King's? Being the semantic value of sentences in context? Being the object of psychological attitudes such as believe? Being "what is said" in certain linguistic exchanges? Or other?

Clas said...

Not quite sure what he says in the book... However, in his "Tense, Modality, and Semantic Values" he characterizes propositions as entities that play all of the roles you mention; plus a couple of others.

Dan López de Sa said...

That there is something playing those various apparently independent roles is part of what Cappelen and Hawthorne call Simplicity. Does King argue for this or just assumes it?

Clas said...

In the paper, he takes it for granted that the view is "attractive"- because it is simple, I assume. However, he tries to defend it against the Operator Argument, since the OA shows that we will need temporal propositions as comp. semantic values, but, according to King, we need eternal propositions as objects of attitudes. I have just tried to show in a paper that his defense of Simplicity is unsuccessful, since on his schmentencite alternative, we will also have two different entities playing the respective roles. We will have sets of assignments (schmopositions) as arguments for temporal operators (quantifiers), i.e. compositional semantic values, but eternal propositions as objects of attitudes.

Clas said...

And once we give up the implausible claim that only eternal propositions can be the objects of attitudes, it turns out that, ironically, King's opponent - the modalist - is in a much better position to hold onto Simplicity than King himself. Since then, we will have temporal propositions as comp. semantic values and as objects of attitudes.

Dan López de Sa said...

Even if it turns out that the features shiftable by operators happen to be those with respect to which the objects of attitudes' values are relative to (say: worlds and times), lacking a principled reason why this need be so makes the Simplicist identification of semantic values and objects of attitudes ungrounded, do you agree?

Clas said...

I agree. However, one might try to use propositional attitude ascriptions as a kind of bridge from semantic values of sentences to the contents of propositional attitudes. This seems to require assuming compositionality + the claim that the content of that-clauses and that of the reported attitudes are the same (which is not trivial). And even if there is this link, it is not obvious that the relation has to be as strong as identity. It might e.g. be enough that semantic values of sentences somehow encode the relevant content of propositional attitudes.

Another link might be that between assertion and belief: we utter sentences to express our beliefs. However, this also does not commit us to identify the two contents. Furthermore, there might be another gap between semantic values of sentence types in contexts and the content of assertions; e.g. as in Lewis' framework.

Dan Zeman said...

OK, I don't mean to be disrespectful to your Lewisian proclivities :-), but I think that whether King accepts or not Simplicity is irrelevant to the problem at issue. Indeed, as Clas said in the first post, he takes propositions to play all these roles (and more). But even if we need more than one thing to play those different roles, under the assumption that we need propositions to play at least one of them, the issue of what proposituions ARE still arises. So, King's book connects with such issues as the unity of the proposition, how can propositions bear truth-values, what exactly are they from a metaphysical point of view, etc. Actually, I take it that King's pride and joy is the thought that his view has solved the problem of the unity of the proposition (no more nor less). He does that by conceiving propositions as facts of a certain kind, facts that have as one of their constituents sentential relations, which is what binds words together in a sentence and from which propositions themselves inherit "the glue" that keeps their constituents together. Also, the book has a chapter dedicated to the defense of the structured view of propositions, which makes it clear that the debate King is entering is that between structured and unstructured propositions (hence my mentioning in the post of the unstructured view and ways out for it). My impression was that his dismissal of some intuitions is something that the unstructured view guys could appeal as well.

Clas said...

Hi Dan,

sorry for hijacking your post to discuss Simplicity. However, I think that Dan's question is quite relevant for your original worry. If we're not really clear about the theoretical role that propositions are supposed to play, then we're stuck in the unfortunate situation where King ends up in. I.e. the mere trading of intuitions about when two sentences express the same propositions and when don't. This is bad methodology, I think.

Furthermore, if for instance we think of propositions as compositional semantic values, then King's rejection of principle P in chapter 3 seems rather unmotivated [roughly: two sentences S, T express different propositions iff there is a sentential connective O such OS and OT differ in truth value]. Why assign two sentences different semantic values, when there is no linguistic context in which that difference would ever show up?

Something more to the point of your initial question: David Ripley has given an argument in an unpublished paper that involves sentences of different languages. Let's assume we have several sentences from different languages that all clearly mean that snow is white. However, the sentences fall into two groups. One has three terminal nodes at LF, the other only two. We might want to individuate proposition so finely that sentence from different languages never express the same proposition. However, King allows that they sometimes do. This forces him into a rather unhappy situation. He is committed to claiming that the sentences in the first group express the same proposition, and that the sentence in the second group express the same proposition, but that the two groups express different propositions, even though all sentence mean that snow is white. This seems to be a rather bad result.

Dan Zeman said...

Yeah, I agree that is could be a worry for King. One question is, is it a worry beacause people have different intuitions about that kind of case or because there are other substantial reasons? I have to confess that I don't find David's case that counterintuitive. First of all, I guess King's opponent should present us with an actual case, and not just a possible one. All languages I know of (very few, indeed) have two nodes for "snow is white". That, I guess, is because translating "snow" in these languages gets you one word, not many. But imagine now that the word for snow in a weird language is something that would more literally be translated as "the white thing falling from the sky" (say, the language is spoken by people who've only heard of snow, somewhere on the equator :-)) Don't you have the intuition that there is another proposition expressed? King talks about these issues in connection with translation, and I tend to find what he says quite compelling. If I remember Dave's point from our discussions, I guess his way to resist to King is to insist that principle P is true. I myself am not sure what to think of principle P :-)

Clas said...

I think the situation would be clearer, if King had actually considered the question as to what kinds of entities propositions are in relation to the question of what theoretical roles they should perform (tell me what they do, and I tell you what they are). In contrast, what he does is to state something like Simplicity as a reason to believe that propositions exist, and then develop his account of what they are mostly independent of the roles they play. The main constraints he imposes on his account seem to be i) that the account should clearly state what they are and ii) how they manage to have truth conditions.

For instance, he seems to leave it somewhat mysterious why we should expect them to be as fine-grained as he thinks they are. Except that this follows from his transparent account of what they are. On the other hand, it we follow proper methodology and answer the question as to what they are in relation to what theoretical roles they perform, I think their role as compositional semantic values gives us reason to uphold principle P. (One of King's motivation seems to be to account for the semantics of that-clauses; but this motivation is in accord with principle P, since that's just the question of how they embed within "believes-that" contexts).

Moreover, it seems that King's level of fine-grainedness makes them inapt to play some of the roles propositions traditionally were supposed to play: what is preserved under (strict) translation, what is common to beliefs/sentences with the same representational content... (Maybe King can give us some kind of ersatz-objects to play these roles; e.g. equivalence classes of distinct fine-grained propositions.)

I thought that David's example was a real life case. Also, as far as I know there are languages that differ syntactially just in whether they put they verb in initial or final position. Surely, we can have sentences there that undeniably have the same meaning.

Bottom line: I can't really see what motivates King's account of propositions. And he seems forced to give up on some of the theoretical roles that they are traditionally assumed to play.

thealcoholics said...

If, as you say, King's position is that a proposition IS a fact that consists of an object standing in certains relations or having certain properties, then how does he account for the difference between the propositions expressed by the sentences, "Annie ran 20 kilometres" and "Annie ran 12.43 miles"? In other words, what is the difference between the fact consisting of Annie running 20 kilometeres, and the fact of Annie running 12.43 miles? They both appear to be the same fact: Annie is identical to herself; the running relation that she bears to the length that she ran is identical to itself; and the length that she ran is identical to itself. Therefore, the difference being expressed by the sentences "Annie ran 20 kilometres" and "Annie ran 12.43 miles" is not dependent on the fact that makes the sentences true. The difference is dependent on the mode of presentation (as Frege would say) of the length that she ran: in the first sentence, the length she ran presents itself as 20 kilometres; in the second sentence, the length she ran presents itself as 12.43 miles. The mode of presentation of an object is supposed to be a constituent of a proposition. But if a proposition is just a fact, then it cannot be: because the length of 20 kilometres is the same thing as the length of 12.43 miles. So, a proposition must be something other than a fact.

thealcoholics said...

Couldn't King reply to what I said by saying: the length of 20 kilometeres is not the same as the length of 12.43 miles, because the first length has the property of being 20 kilometers in length, and the second length has the property of being 12.43 miles in length, and the property of being 20 kilometers in length is not the same as the property of being 12.43 miles in length. The problem with that position is this: any single thing has more than one property, and many of the properties it has are different from one another; so, if the fact that properties a thing has are different makes the thing more than one thing, then any single thing would be many things. Let me illustrate this point by example: Annie has the property of being female, being warm blooded, being human, being bipedal, being a mammal, among others. These properties are all different. If the fact that the properties a thing has are different makes the thing more than one thing, then Annie is, at least, five different things. But, of course, Annie is just one thing. Therefore, the fact that the properties a thing has are different does not make the thing more than one thing.