I wanted to write something systematic about Laurie Paul’s paper on contextualism and essentialism, but doing the metaphysics full justice would require that I either be better informed, or smarter, or less lazy, so I can’t really say anything too comprehensive. But the points I wanted to make didn’t have much to do with the metaphysics anyway, so let’s try the following plan. I’ll start with a relatively crude caricature of Laurie’s position, make a few criticisms of the caricature, silently bet that filling in the details of the metaphysics won’t make much difference to the objections, then go on to a puzzle for Lewis that Laurie’s paper raises. Then I’ll turn over to watching television.
Here’s the puzzle. In some contexts we feel happy assenting to essentialist claims like "Queen Elizabeth II (QE2 for short) could not have been brought by a stork". Call these the Kripkean contexts. For definiteness, imagine the Kripkean context to be a seminar on essentialism run by Kripke, in particular a time in that seminar when Kripke has just said "QE2 could not have been brought by a stork" in a particularly authoritative fashion. But in other contexts, the contrary assertion "QE2 could have been brought by a stork" sounds fine. Call these the Lewisian contexts. For definiteness, imagine the Lewisian context to be a seminar on counterpart theory run by Lewis immediately after that Kripkean seminar, in particular a time in Lewis’s seminar when he has just said "QE2 could have been brought by a stork" in a particularly convivial fashion. How can we explain what is going on here?
Laurie’s solution is that in the two contexts, ‘QE2’ names different objects. For definiteness (and here I’m really caricaturing) we can imagine that in each context QE2 names a trans-world fusion, a modal continuant as Lewis calls it ("Postscript to Counterpart Theory", pg. 40-42), but it names different modal continuants in different contexts. In the Kripkean context it names a fusion every one of whose modal parts is born of human parents. (Germanic royals as it turns out). In the Lewisian context it names a much larger fusion, some of whose modal parts were brought into the world by a stork. Now we have it that both Kripke and Lewis speak truly in their respective contexts.
This isn’t a million miles from Lewis’s own solution, as Laurie acknowledges. Lewis also favoured a contextualist solution to the puzzle. But for Lewis the contextualism seems to reside in the modal terms, not in the names. (It’s a slightly tricky matter of interpretation determining whether Lewis is committed to this. But I think Laurie’s right that he is, for reasons that should become clear in what follows. Laurie argues that it is an improvement to have the contextual variability in the name rather than the modal terms. I think she’s right – Lewis’s position here is I think inconsistent. But that doesn’t mean contextualism here of any kind is a good idea.
I think some fairly simple considerations about speech reports pose pretty desperate problems for the contextualist theory. A few years ago I noticed that contextualist theories had problems with indirect speech reports. (I wasn’t the only one to notice this, and I probably wasn’t the first. But it was an original thought on my part at the time.) We can illustrate this fairly well with the seminars above. Imagine a student responds to Lewis’s convivial statement with (1).
(1) Professor Kripke said that QE2 could not have been brought by a stork.
Our intuition is that what the student said is true. Kripke did say just that. But if the student is still in the Lewisian context – and since she’s still in the Lewisian classroom she probably is – then ‘QE2’ in her mouth does not denote the modal continuant that Kripke was talking about. And we can hardly report Kripke’s speech by talking about a thing that he did not even denote. Even worse, it’s hard to see how the ‘that’-clause in (1) fails to name a false proposition in the context, so (1) reports Kripke as having expressed a false proposition. Yet the contextualist thinks Kripke spoke truly.
(It should be clear enough how to extend this kind of criticism to contextualism about knowledge, or truth, or ethics, or victory. The last would be amusing – it would be nice to know that after the game we could always find a context in which we can truly say "We won.")
Ernie Lepore and Herman Cappelen (in thus far unpublished work) have noted that contextualism also makes a mess of direct speech reports. This is actually a much stronger argument against contextualism than the argument from indirect speech reports just mentioned, since it relies on fewer theoretical overheads. On the contextualist view, the student instead of saying (1) could have truly said (2).
(2) When Professor Kripke said, "QE2 could not have been brought by a stork" he spoke truly, even though QE2 could have been brought by a stork.
But this sentence could not possibly be true. Note that this is not because we cannot contradict what appears inside quote marks in the rest of the sentence. When there really are contextually sensitive terms in what is quoted, this is perfectly possible, as in (3).
(3) When Professor Kripke said, "I wrote Naming and Necessity" he spoke truly, even though I did not write Naming and Necessity.
If ‘QE2’ is context sensitive, just like ‘I’, then (2) should sound just as plausible as (3). But it doesn’t – it sounds awful.
We can put these two complaints together. On the contextualist theory the student could also say (4), which also sounds very bad.
(4) When Professor Kripke said, "QE2 could not have been brought by a stork" he did not say that QE2 could not have been brought by a stork.
Contextualists have a response to these arguments, though I don’t think it’s a very persuasive one. They can just say that we should attribute mass error to people about direct and indirect speech reports. People they just ain’t no good – at producing reports containing terms with hidden contextual sensitivity. I think this is more plausible for the indirect speech reports than for the direct speech reports, which is why I think Lepore and Cappelen’s argument is better than the one I first came up with. (The Lepore and Cappelen paper was presented at the Central APA this year, so hopefully it shouldn’t be too far from publication.)
This isn’t meant to be a particular criticism of Laurie. I think her contextualist theory is much more plausible than the majority of contextualist theories floating around these days – as far as I can tell it isn’t vulnerable to any particular criticisms, just these general criticisms of contextualist theories. And everyone except Kent Bach is probably vulnerable to Lepore and Cappelen’s criticism somewhere. (To pick a non-random example, my views on conditionals don’t look very good in light of this argument.) The main reason I wanted to work through this here was because it provides a nice way of illustrating the problems contextualisms have with reports.
And I do think that if we must go contextualist, Laurie’s version of contextualism is better than Lewis’s, which looks inconsistent to me. The problem is that I think Lewis is committed to the following five theses, which look inconsistent.
- Meaning is compositional
- Contextualism about de re modal statements is true
- Modal locutions in ordinary language are operators not quantifiers
- Propositions are unstructured
- Names are not contextually variable (or at least they do not vary between the Kripkean and Lewisian contexts)
Here’s the argument that these are inconsistent. Assume that (5) is uttered in a context such that it is true.
(5) Possibly, a is F.
Assume it is later uttered in a context where it is false. That is (5′) we will assume is false.
(5′) Possibly, a is F.
It will be helpful to have the following two sentences to compare with (5) and (5′), the first uttered in the context of (5), the second uttered in the context of (5′).
(6) Possibly, 2+2=5
(6′) Possibly, 2+2=5
By contextualism, it is possible that (5) and (5′) exist. Note that (6) and (6′) are false no matter what we know about the context. If ‘possibly’ is an operator, then the logical form of each of these sentences is OS, where O is a sentence-sentence operator and S a sentence. That sentence denotes a proposition, or something of the sort. Note that this is an unstructured entity, a function from something (possible worlds or possibilia probably) to truth values. By the assumption that the name is not contextually variable (and I assume we pick an F that is also not contextually variable) we get that the content of S is the same in both (5) and (5′). So hence the content of ‘possibly’ must be different in the two cases. But now note that in the (5′) context, the content of a is F is the constant function that maps everything to false, the same as the content of ‘2+2=5′ in (6) and (6’). Since the content of the name and the predicate don’t change between contexts, that is still the content of S in (5). So in (5) ‘possibly’ denotes an operator that maps the constant false function into a true proposition. Hence (6), which consists of the very same operator followed by a sentence denoting the constant false function is true, which is absurd.
There’s lots of ways out of this. Possibly being more careful than I’ve been about what kinds of functions are the contents of sentences will help, though I can’t really see how it could. The argument appeals to compositionality several times and one could deny that, though Lewis does not. See, I think, General Semantics. One could give up the contextualism, which I think one should, though again Lewis does not. See chapter 4 of Plurality. One could deny that ‘possibly’ here is an operator. I might write a bit more on this next week, because it’s a more plausible position than I realised. But again Lewis does not. See Index, Context and Content. (Note that the point here is not whether the truth conditions for modal sentences should be stated using operators or objectual quantification over worlds. The point rather is whether English contains modal operators or its apparent modal operators really are objectual quantifiers.) One could accept structured propositions, as Jeff King and others have argued, but Lewis does not. I’m not sure where he explicitly says this, but it’s implicit in almost every paper on philosophy of language that he wrote. I guess there’s an explicit acceptance of unstructured propositions somewhere in Plurality. This move would fit quite nicely with the counterpart theory I think – it’s good to give the modal operators, if there are any, something in the propositions they operate on to bite into. And finally one could adopt Laurie’s position and say that the variation is in the content of the name. This position has the nice virtue of being consistent, unlike what I take Lewis’s position to be.