The reading group on Scott Soames' Reference and Description: The Case Against Two-Dimensionalism (see this entry) has met twice so far. The first meeting covered chapters 1-3, which are mostly just background. In chapter 3, Soames does consider one of Frank Jackson's arguments for descriptivism, remarking somewhat incredulously that if it works, it is a priori and irrefutable -- prompting Frank's response, "That's an objection?". At the end of the chapter he also gives initial characterizations of "strong" and "weak" two-dimensionalism, both of which seemed fairly unrecognizable to the two-dimensionalists present. But all that is discussed in more detail later on in the book.
Things get going in chapter 4 on "Roots of Two-Dimensionalism in Kaplan and Kripke". The main theme here is that while Kaplan and Kripke are the main heroes of the anti-descriptivist revolution, their work contains "errors, slips, and misleading suggestions" that gave too much encouragement to later two-dimensionalists. Soames points to various errors, but the main "error" for both of Kripke and Kaplan is the suggestion that when a name or a 'dthat'-expression is introduced using a reference-fixing description, this can give rise to contingent a priori knowledge.
The relevant sorts of sentences are the following:
(1) Dthat[the youngest Chinese spy] is the youngest Chinese spy (if anyone is) [p. 52]
(2) Lee is the youngest Chinese Spy (if anyone is). [p. 55]
Here 'dthat' is Kaplan's familiar expression that turns a description into a rigid singular term, and 'Lee' is a name introduced with the stipulation that it refers rigidly to the youngest Chinese spy (if there is such a spy). (Soames also discusses analogous sentences involving natural kind terms, but I'll leave those aside, as the issues here are very close to the issues for names.) Kaplan holds that sentences such as (1) are examples of the contingent a priori, while Kripke holds that sentences such as (2) are examples of the contingent a priori. Soames holds that these claims are errors. Here I'll look at his argument for this claim, concentrating mostly on the Kripke-style cases.
Soames' argument can be summarized as follows. He says that if (1) and (2) are examples of the contingent a priori, then it is possible to know a priori, of a certain individual, that he or she is the youngest Chinese spy (if anyone is). And he says that this sort of a priori knowledge is impossible. He concludes that (1) and (2) are not examples of the contingent a priori. (This argument is fairly closely related to things that Donnellan and Salmon say about the contingent a priori, and my reaction to it in the following is related to my reaction to Donnellan and Salmon. But I don't have my copies of those papers handy, so I'll stick to Soames.)
What's odd is that Soames gives hardly any argument in the chapter for the claim that if (1) and (2) are examples of the contingent a priori, then it is possible to know a priori of a certain individual that he or she is the youngest Chinese spy (if anyone is). He seems to think that the claim is obvious. But pretheoretically, the claim doesn't seem especially plausible, especially in the case of (2). So there seems to be a big hole in the chapter's argument. Here I'll discuss options for how the argument might go.
One might put the overall argument in a little more detail as follows, with reference to the following sentences.
(3) Someone can know a priori that Lee is the youngest Chinese spy (if anyone is).
(4) Someone can know a priori, of the individual who is the youngest Chinese spy, that he or she is the youngest Chinese spy (if anyone is).
Soames argues, in effect:
(S1) If (2) is an example of the contingent a priori, then (3) is true.
(S2) If (3) is true, then (4) is true.
(S3) (4) is false; therefore
(S4) (2) is not an example of the contingent a priori.
I think that (S1) and (S3) are nontrivial (I recall Robin Jeshion arguing for de re a priori knowledge in such cases), but I'll grant them here. For present purposes the action is with (S2). Presumably, most of those who hold that (2) is contingent a priori (including the Kripke of Naming and Necessity) will be inclined to deny that this claim entails anything about de re a priori knowledge. And prima facie, it is pretty natural to do so by denying (S2). It's a commonplace, after all, that one must be cautious about moving from de dicto attitude ascriptions, such as (3), to de re attitude ascriptions, such as (4). But Soames seems to take (S2) for granted.
How might one argue for (S2)? One way would be to appeal to a general exportation principle (E) that Soames discusses later in the book (simplifying slightly, and ignoring quotational niceties):
(E) For any name n and predicate F, if 'A knows/believes that n is F' is true, then 'A knows/believes of x that it is F' is true, where the referent of n is assigned to 'x'.
When he introduces (E). Soames says that it is "intuitively compelling" (p. 261). Elsewhere he says that a variant of (E) with 'knows a priori' instead of 'knows' or 'believes' is also obviously correct. If we accept this variant (and we accept that 'Lee' above is a name), then (S2) follows.
Now, I think that (E) has some intuitive plausibility where ordinary proper names are concerned. But it's far from obvious that it's correct where descriptive names such as 'Lee' are concerned. Note that nothing substantive here turns on the issue of whether these expressions are really names. If they're not, we just need to consider the plausibility of a modified principle (E') that applies to these expressions.
Consider a scenario where Jones has just introduced the term 'Lee' (to himself and Smith) with the stipulation that it refers to the youngest Chinese spy, if there is one, and consider the following knowledge ascriptions (N.B. not a priori knowledge ascriptions) made by Smith:
(5) Jones knows that Lee is the youngest Chinese spy (if anyone is).
(6) Concerning the individual who is the youngest Chinese spy: Jones knows that he or she is the youngest Chinese spy (if anyone is).
Principle (E) entails that if (5) is true, then (6) is true. But to my ears, (5) is true while (6) is false. And I suspect that this will be the most common pretheoretical judgment about these knowledge ascriptions. If we take students who have learned just enough philosophy to get a sense of the difference between de dicto and de re knowledge ascriptions, but haven't yet taken on board all the theoretical apparatus of direct reference, etc, my expectation is that a majority will judge that (5) is true while (6) is false. (I'm interested to hear about others' intuitions here -- though note that post-theoretical intuitions aren't really to the point here.) If so, then one certainly can't claim that principle (E), in the version that applies to descriptive names, is intuitively compelling. But it looks like this is the principle that Soames is presupposing.
One way to argue for principle (E) would be as follows. Let's assume that 'n is F' expresses a singular proposition P. Then if 'A knows that n is F' is true, then A knows the proposition P. And if A knows P (a singular proposition), then 'A knows/believes of x that it is F' is true, where 'x' is assigned the referent of n. So (E) is true.
Obviously, this argument is only as good as its assumption. In particular, the second step requires that the claim 'n is F' expresses a singular proposition. But clearly, to argue that 'Lee is the youngest Chinese spy (if anyone is)' is not contingent a priori, one can't simply assume that the sentence expresses a singular proposition. Plenty of people (including Soames himself, I seem to recall) will deny that the sentence expresses such a proposition. And certainly, many people who think that (2) is contingent a priori will deny that it expresses such a proposition. So one can't simply assume that the proposition expressed, if any, is a singular proposition.
All this makes Soames' objection to Kripke a bit baffling. If we could assume principle (E) for descriptive names, or assume that sentences such as (2) express singular propositions, then one could plausibly infer (S2), and with it the claim that (2) isn't contingent a priori. But principle (E) for descriptive names is prima facie counterintuitive, and the claim that (2) expresses a singular proposition is a controversial theoretical claim. So simply assuming these claims can't do any work in this context.
(Question: can anyone give any pointers to literature on the exportation principle for descriptive names? For example, I wouldn't be surprised if the objection I'm making to Soames is related to an objection that some people have made to Donnellan on the contingent a priori.)
At this point I'll quote the only paragraph from Soames that looks like it might be giving some further relevant argument (pp. 55-56):
In Kripke's case we may put the point by contrasting two different scenarios in which one might attempt to use a description to introduce and fix the referent of a name. In the first scenario, one does not know, or believe, of any object that it is denoted by the description introducing the name -- e.g. one doesn't know, or believe, of any individual i that i is the youngest Chinese spy. Clearly the mere performance of a linguistic ceremony of using the description to introduce a name can't change this. If, despite this, one's introduction of the name 'Lee' for i is successful, then [(2)] will come to express a proposition about i that one doesn't know to be true, and can't come to know except by gaining de re knowledge of i through further empirical investigation. Regarding this scenario, one must say either (i) that the name hasn't been successfully introduced after all, (ii) that the speaker doesn't understand the name he has introduced, or (iii) that understanding and justifiably accepting a true sentence containing the name is not sufficient for knowing the proposition p which it expresses to be true. Either way, a priori knowledge of a contingent truth has not been achieved.
This "first scenario" is the relevant one for our purposes. Sentence 3 amounts to (S3) above. The following sentence, sentence 4, seems to be the crucial one. Prima facie it's very unclear why anyone who believes in the contingent a priori should accept it. Again, if one assumes that if (2) expresses a proposition, it expresses a singular proposition, then one might accept the claim. But one can't simply assume that here. As for the following sentence, it's far from clear that (i)-(iii) exhaust the options. If we assume that the relevant proposition has to be singular, these might be the main options. But one who holds that (2) expresses a non-singular proposition can happily deny each of (i) through (iii).
[It may that the notion of a "name" is doing some work in the above passage: e.g. Soames might be assuming that a name must have singular content by definition, or that names require some sort of strong relation to their referent. But this doesn't change much. An opponent can simply deny that 'Lee' is a name in this sense, while continuing to maintain that (2) is an example of the contingent a priori.]
Of course there's something odd about using all this "proposition" talk to argue against Kripke, given that he resolutely refuses to say anything about propositions in Naming and Necessity (here I set aside the later Kripke of "A Puzzle about Belief", as his views clearly evolve in the meantime). He might respond simply by denying that the contingent apriori status of (2) is determined by the contingent a priori status of an associated proposition. But if he accepts that it is so determined, then he'll presumably deny that (2) expresses a singular proposition. Just what sort of proposition it expresses is up for grabs, but there are all sorts of theoretical views that are quite consistent with the contingent apriori status of the relevant proposition. It looks like Soames' argument just begs the question against these views.
So, I'm a little perplexed about what's going on in this chapter. If Soames is just indicating his disagreement with Kripke, that's fine -- it's clear how certain theoretical views that Soames holds entail that Kripke's claim is false. But it seems pretty clear that the chapter is more ambitious than that: it's trying to demonstrate an error in Kripke, presumably based on premises that don't presuppose controversial theoretical views (otherwise, none of this can do much work in arguing against alternative views, as it is intended to do). I don't see how this demonstration is supposed to go. Can anyone help out here?
Soames may be on somewhat firmer ground in the case of Kaplan. That depends on just how the 'dthat' operator is understood. If one understands it so that the content of 'dthat[the F]' is singular by definition (if it has content at all), then the argument gains purchase. To resist it, Kaplan has to either accept that we can have the relevant de re knowledge a priori, or deny that the apriority of 'dthat[the F] is the F (if anything is)' is determined by the apriority of the relevant content (e.g., maybe it's determined by the apriority of the relevant character). Maybe Kaplan would do one of these things (elements of his work suggest both responses), but one can at least make a case that there are costs here. My own view is that if 'dthat' is defined as above, it's far from clear that there can be an expression that satisfies the definition. If instead one simply defines 'dthat' as a rigidifying operator, then the definition is perfectly coherent, but Soames' argument now doesn't gain any more purchase than it does in the previous case of descriptive names.
Two final notes. First, Soames himself grants that his argument against the contingent a priori gains no purchase where 'actually'-involving rigid descriptions are concerned. E.g. he allows that 'The actual youngest spy is the youngest spy (if anyone is)' is contingent a priori. This means that the arguments in this chapter won't be of great concern to two-dimensionalists such as Davies, Humberstone, and Jackson, who base their framework on 'actually'-involving descriptions, and who hold that expressions such as 'Lee' are equivalent to expressions such as 'The actual youngest spy'.
Second: Throughout this chapter, Soames suggests that advocates of the contingent a priori confuse the claim that the relevant subjects know the proposition expressed by sentences such as (1) and (2) with the claim that subjects know that sentences (1) and (2) express a truth. As Daniel Stoljar pointed out, this claim itself has an interestingly two-dimensional flavor. In particular, it's very close to the strategy that Pavel Tichy, an early two-dimensionalist, used to argue against Kripke in his paper "Kripke on necessary a posteriori" (Philosophical Studies 1984). Tichy said that sentences such as 'Hesperus is Phosphorus (if it exists)' can be associated with two propositions: (i) the proposition expressed by the sentence, which is just the proposition that Venus is Venus (if it exists), and which is necessary and a priori; (ii) the proposition "associated" with the sentence, which is the proposition that 'Hesperus is Phosphorus (if it exists)' expresses a truth, and so is contingent and a posteriori. Tichy thinks that Kripke confuses the two propositions, and that once we've clarified the distinction, it's clear that there are no necessary a posteriori propositions in the vicinity. It's interesting to note that Soames' strategy, in the case of the contingent a priori, is so similar to Tichy's.