Blog powered by Typepad

« Matrix philosophy | Main | Upcoming conferences »

February 04, 2005


Istvan Aranyosi

If in the case of a posteriori necessities we have a necessary secondary intension and a contingent primary intension, then in the cases of the contingent a priori we will have a necessary primary intension and a contingent secondary intension. Since the two-dimensionalists' main point is to talk in terms of these two dimensions, they will not accept Soames' assertion that if (3) is true, then (4) is true, because it is a move from one dimension (the primary intension) to another (the secondary intension). The 2D-ist holds that de re ascriptions depend on secondary intensions, and de dicto ascriptions depend on primary intensions. In the case of a potential contingent a priori sentence, the apriority comes from the primary intension, while the contingency from the secondary intension. In the case of "Lee is the youngest Chinese spy", we will say that it is knowable a priori because the primary intension (the a priori aspect) resulting compositionally from the primary intensions of the component terms "Lee" and "the youngest Chinese spy", is such that it is knowable a priori: "The youngest Chinese spy is the youngest Chinese spy".

When we move to considering de re knowledge, things change in respect of which intension we consider. In our case, ascriptions involving the actual referent of 'Lee' will involve the secondary intension, which is the a posteriori aspect of the meaning of the sentence under consideration. Therefore, the 2D people will not accept the move as regards a priori knowledge, in the case of contingent a priori sentences, from de dicto to de re.

I think Soames' premise would be acceptable for terms that have coinciding primary and secondary intensions. But I think cases of contingent a priori knowledge are those involving terms with differing such intensions. Am I right here?


Well, I didn't want to presuppose two-dimensionalism in an analysis of this chapter, as Soames is suggesting that 2Dism itself is grounded partly in these "errors" by Kripke and Kaplan. But to think about what a 2Dist should say about these cases, I think you're right that they can reject the move from (3) to (4). It's not quite as simple as saying that de dicto attitude ascriptions operate on primary intensions and that de re attitude ascriptions operate on secondary intensions, though. If that were so, then the exportation principle wouldn't work even for ordinary proper names, but it looks fairly plausible there. Instead, I think that both sorts of ascriptions involve both sorts of intensions, but in different ways, roughly as follows (see section 8 of "The Components of Content" for more details):

The de dicto ascription 'S believes that n is F' is true iff the subject has a belief with the secondary intension of 'n is F' and with an 'n is F'-appropriate primary intension (this needn't always be the primary intension of 'n is F' in the ascriber's mouth -- but this point doesn't matter much for present purposes).

The de re ascription 'S believes of n that it is F' is true iff the subject has a belief with the secondary intension of 'n is F', where this belief picks out the referent of n under a de re-appropriate primary intension (roughly, a primary intension that involves acquaintance; cf. Kaplan's "vivid names").

As for the move from the de dicto ascription to the de re ascription: On this account, the two ascriptions constrain the secondary intension in the same way, so this part will be fine. The crucial question concerns the primary intension. Is it the case that when a belief (with the secondary intension of 'n is F') has an 'n is F'-appropriate primary intension, it will pick out the referent of n under a de re appropriate primary intension? I think the answer here is: plausibly yes when n is an ordinary proper name, but no when n is a descriptive name. When n is a descriptive name such as 'Lee', a 'Lee is F'-appropriate primary intension may pick out the referent under a non-acquaintance-involving descriptive intension, so a belief with this sort of primary intension may satisfy the de dicto ascription, but it won't satisfy the de re ascription.

To address your last point: If the above is right, then exportation doesn't require that the relevant singular term has the same primary and secondary intensions (cf. the case of ordinary names). It's an interesting question whether this feature would suffice for exportation to work, though. I think exportation pretty clearly fails for nonrigid descriptions with the same primary and secondary intensions ('the largest object in the universe', say). But I presume we're restricting ourselves to rigid singular terms. If so, then sameness of primary and secondary intension requires a rigid primary intension, as happens with what Martine Nida-Rumelin calls "super-rigid" expressions (those that pick out the same referent throughout a 2D matrix). Super-rigidity is pretty rare, but it's plausibly that some singular terms, such as mathematical expressions for numbers and maybe some terms for abstract objects, are super-rigid. It's plausible that in at least some of these cases, exportation will work. And it's arguable that these cases are the only cases where one can have full-scale de re a priori knowledge.

Jason Stanley


I heartily endorse your comments (though I'm relying on your interpretation of his arguments, since I must confess to not yet having read Soames's new book -- for gosh sake, I'm still in the middle of Volume 1 of the history of analytic books).

In particular, isn't the central purpose of Evans's "Reference and Contingency" (one of the founding 2-dimensionalist papers) to challenge (S2)? In that paper, Evans advances descriptive names such as "Julius" as counterexamples to (S2). That motivates a distinction between a name licensing epistemic de re exportation and modal de re exportation. According to Evans, descriptive names allow the latter, but not the former. Certainly, this is how I use descriptive names in my papers on this topic -- as counterexamples to (S2), counterexamples that motivate making a distinction between epistemic content and modal content.

So, one of the original 2-D moves was to introduce descriptive names as cases in which you can have a modal content that is singular, but an epistemic content that is non-singular (hence requiring two notions of content -- thought vs. modal profile, in roughly Evans's terminology, or assertoric content and ingredient sense, in Dummett's terminology). Given this dialectical situation, it is illegitimate to presuppose that descriptive names express epistemically singular propositions (that is, express singular propositions when embedded under epistemic operators such as attitude verbs). As you say, this might not be a problem for Soames, since he is here not here arguing against Evans or Dummett's positions, or subsequent 2-Ders; he's making clear his disagreements with Kripke and Kaplan. But some of the principle original arguments in favor of 2-Dism involve using descriptive names to reject (S2).


Hi Jason, thanks for this. I finally dug up my copy of Evans. You're right that his theoretical account in that paper is well-suited to resist exportation. I almost cited him in the entry, but didn't on the grounds that Evans' account takes for granted the claim that the relevant sentences are contingent a priori, and Soames seems to think that this claim already involves some relatively pre-theoretical "error".

Also, on looking through the Evans paper, there's not really any discussion of sentences embedded in epistemic contexts. So there's also not really any discussion of principles such as (S2), involving exportation in these contexts. There's a lot of discussion of sentences embedded in modal contexts, but there's his discussion of epistemic issues is largely cast in terms of metalinguistic locutions such as "the truth of (S) can be known a priori". Of course there must be some connection between these metalinguistic epistemic claims and ordinary epistemic contexts, but the connection isn't completely straightforward (see e.g. this entry). So I'm somewhat hesitant to ascribe views about ordinary epistemic contexts (or about exportation in these contexts) to Evans on the basis of this paper. It's certainly true that the view in the paper lends itself naturally to claims about exportation in epistemic contexts, though.


There are just a few comments i want to make about your reconstruction of Soames' arguement against the contingent a priori, in particular S2. you complain that Soames does not offer any argument for S2 in the chapter, but rather assumes it. That is because he has already presented Kripke's decisive(epistemic and semantic)arguments that ordinary proper names cannot be equivalent to descriptions in chapter 2. If the description "the youngest chinese spy" gave the meaning (semantic content) of "Lee", then if we know the proposition expressed by "The youngest chinese spy is the youngest chinese spy" a priori, then we should know the proposition expressed by "Lee is the youngest Chinese Spy" a priori. Chalmers and others might respond, "well, what's the problem? We do know the proposition expressed by 'Lee is the youngest chinese spy' a priori". This is not in general correct. It is an open question whether Jones, (the intoducer of the name "Lee" via the description "the youngest chinese spy") knows the proposition a priori. But, it is not an open question whether everyone who is a competent user of "Lee" knows it a priori. The answer is no, not everyone knows it a priori. We can imagine Jones making all kinds of statments regarding Lee, e.g, "Lee is F, and Lee is G, and Lee is not H, etc. Smith might hear some of the things Jones is saying and pick up the name "Lee". So,then Smith can use the name. Among other things, Smith believes that Lee is F and G and not H. Yet, Smith may not know that she is the younest chinese spy. In order to know that, Smith would need the (empirical) information from Jones that Lee is the youngest Chinese spy. Well, a proposition is either knowable a priori or it is not. If some competent speaker grasps a proposition P but cannot determine whether it is true simply on reflection, then P is not knowable a priori. So, then Jones probably does not know it a priori either.

Whether this arguement is convincing or not is a separate matter. the point is that Soames has given (Kripke's) arguement that the meaning of "Lee" cannot be "the youngest chines spy". At most, the description fixes the reference.

Regarding Kaplan's Dthat, Soames is dead on. "Dthat [the youngest chinese spy]is the youngest chinese spy" expresses a singular proposition. How do we know? "Dthat" is an artifical term and Kaplan has decreed that any supplemented "Dthat" term, e.g., ^Dthat [the youngest chinese spy]^ (^ is a corner quote) is directly referential. So, the content of ^Dthat [the youngest chinese spy] is the youngest chinese spy^ is the singular proposition about Lee, that she is the youngest chinese spy. In order to grasp this content, one needs de re knowledge of Lee. In order to have the requisite de re knowledge regarding Lee, One needs empirical information that she is the younest chinese spy. And, this empirical information, which is necessary to grasp the content, is also necessary to justify our belief in the truth of the content expressed by the sentence. Thus, we do not know the content of the sentence to be true a priori.

Of course Kripke never comes out and says that proper names are directly referential. the only conclusion he arrives at is that they are rigid. Well, what does their rigidity consists in.

Option 1: The semantic contents of names are descriptive, but the descriptions must be rigidified in some way. However, Kripke's semantic and epistemological arguments show that they cannot be descriptive.

Option 2: Names are directly referential.

Given that there are only two viable optins, and that Kripke has refuted option 1, it follows that option 2 is the only option.

For these and other reasons, Soames believes it is safe to accept S2.


Thanks for this. Your argument seems to be: (1) Names such as 'Lee' are not descriptive; (2) If these names aren't descriptive they are directly referential; (3) If these names are directly referential, sentences involving them can't be contingent a priori. My initial reaction is that none of the three premises is obviously correct, and I don't think any of them would clearly be accepted by Kripke.

Re premise (1): It's notable that Kripke never runs his epistemic or semantic arguments for names (such as 'Lee') whose reference is fixed by a description, and he seems to exempt such names from these arguments. Certainly there will be later users who pick up the name without the associated descriptive information, and for whom the sentence won't be contingent a priori. But unless one makes a host of additional theoretical assumptions, that doesn't show that the sentence isn't contingent a priori for the original speaker. Recall that for Kripke the apriority of a sentence is always relativized to a speaker.

Re premise (2): One can't presuppose that the only theoretical options are descriptions and direct reference. Certainly Kripke doesn't want to presuppose that these are the only options. And later work suggests all sorts of intermediate options: cf. two-dimensionalism and the views of Evans and Burge, to pick three. No doubt there are others.

Re premise (3): One can hold that names are directly referential and that the sentences involving them are contingent a priori, as long as one holds that the apriority of a sentence is not determined by the apriority of the (singular) proposition it expresses. A number of people have done just that.

I don't know which, if any, of these options Kripke was inclined to take at the time of Naming and Necessity, but all of them seem compatible with the views therein, and none of them can be ruled out without a lot of additional argument, or without heavy-duty theoretical assumptions. So if this is really Soames's tacit argument that Kripke is mistaken, then there's a lot more work to be done.

The comments to this entry are closed.