Chapter 7 of Soames' Reference and Description begins the third and by far the longest section of the book (comprising about 200 pages), devoted to a critique of the "ambitious two-dimensionalism" attributed to Frank Jackson, David Lewis, and me. (Previous entries: introduction, chapter 4 on Kripke and Kaplan, chapter 5 on Stalnaker. Chapter 6 on Davies and Humberstone was postponed until Martin Davies returns from UCLA in April.) In this chapter, Soames lays out the main theses of what he takes to be the two main versions of ambitious two-dimensionalism: "strong" and "weak" two-dimensionalism.
These theses are almost identical to the theses quoted in my Arizona State commentary "Soames on Two-Dimensionalism" (which can be used as a reference by people who don't have the book, though most of the material in this post should be comprehensible without reading either). My reactions are similar to the reactions in the commentary. I reject many of the theses of both "strong" and "weak" two-dimensionalism. These include the first thesis of both, which characterizes primary intensions in terms of Kaplanian characters; the fourth thesis of both, which says that names and natural kind terms are synonymous with rigidified descriptions; the fifth theses of both, which give (different) accounts of the truth-conditions of attitude ascriptions; and the seventh/eighth thesis of strong two-dimensionalism, which denies that there are necessary a posteriori propositions. I'm also somewhat doubtful about the second thesis of both, which makes claims about understanding and about our knowledge of the relevant intensions. It turns out that Frank Jackson is happier to go along with some of these theses: he's inclined to endorse the second, fourth, and the seventh (perhaps in slightly modified forms), and doesn't have really strong objections to the first. But like me, he strongly rejects both versions of the fifth thesis concerning attitude ascriptions.
It turns out that later in the book, most of Soames' objections to strong and weak two-dimensionalism consist in objections to these theses concerning attitude ascriptions. So here I'll concentrate on these theses (focusing on the case of belief ascriptions), and on why I take them to be pretty obviously false. In the second part of this entry I'll talk about some closely related issues about ascriptions of a priori knowledge.
The two crucial theses of "strong" and "weak" two-dimensionalism concern the truth-conditions of attitude ascriptions such as 'A believes that S'. (Also 'A knows that S', 'A knows a priori that S', and other attitude ascriptions, where S is a sentence. For now I'll concentrate on belief ascriptions.) I'll call these theses S5 and W5, though in the book they're called T5b and WT5b. To a first approximation, S5 says that such an ascription ascribes a relation between the subject and the primary intension of S, while W5 says that such an ascription ascribes a relation between the subject and the secondary intension of S. We might put the claims as follows. The theses as Soames states them are somewhat more complex (see the Arizona State piece) but this gives the basic idea.
S5: 'A believes that S' is true iff the referent of A has a belief whose primary intension is the primary intension of S.
W5: 'A believes that S' is true iff the referent of A has a belief whose secondary intension is the secondary intension of S.
Thesis W5 has some pretty familiar problems. W5 entails that if (1) is true, (2) is true:
(1) Lois believes that Superman is Superman.
(2) Lois believes that Superman is Clark Kent.
But intuitively it seems clear that (1) can be true while (2) is false. Some Millians (such as Soames himself) try heroically to deny the intuitions, but this is obviously biting a huge bullet, one that two-dimensionalists needn't bite. To a two-dimensionalist, the obvious upshot is that the truth of (1) and (2) depends not only on the secondary intension of Lois's beliefs, but also on the primary intension of her belief (so that a belief that satisfies (1) won't have the right sort of primary intension to satisfy (2)). Maybe in principle there could be a "two-dimensionalist Millian" who holds that the truth-value of the ascription depends only on the secondary intension of the belief, as W5 says, while allowing that the information conveyed by the ascriptions is sensitive also to the primary intension of P (this might be a 2D version of Nathan Salmon's view). But there's not much motivation for the view, given that two-dimensionalists have non-Millian semantic values available to ground the intuitive non-Millian truth-conditions.
Thesis S5 suffers from multiple problems. The first is the converse of the problem for W5: S5 entails that the truth of an ascription depends only on the primary intensions of the subject's beliefs, but it seems clear that primary intensions aren't all that matter. Consider e.g.
(3a) Oscar believes that water is wet.
(3b) Twin Oscar believes that water is wet.
(4a) Fred believes that George Bush is a Republican.
(4b) Twin Fred believes that George Bush is a Republican.
Here Oscar and Fred are typical Earth inhabitants, while Twin Oscar is a duplicate of Oscar on Twin Earth (where the watery stuff is XYZ), and Twin Fred is a duplicate of Fred millions of years ago, or in a distant part of the galaxy. In these cases, the standard intuitions are that (3a) and (4a) are true, while (3b) and (4b) are false. But in these cases, Oscar and Twin Oscar may have beliefs with the same primary intensions, as may Fred and Twin Fred. The moral is that the truth of these ascriptions requires more than a belief with an appropriate primary intension: it also requires an appropriate environment, and in particular an appropriate relation to water or to George Bush respectively. For a two-dimensionalist, the obvious diagnosis in both cases is that the truth of these ascriptions depends not only on the primary intensions of the subject's beliefs, but also on their secondary intensions.
This suggests that the right-to-left component of S5 is false. There are also reasons for thinking that the left-to-right component of S5 is false. (This contrasts with W5, where the right-to-left component looks false but the the left-to-right component looks OK.) Relevant examples here include
(5) Fred believes that I am hungry.
(6) Pierre believes that London is pretty
To satisfy the first ascription, Fred needn't have a belief with the same primary intension as 'I am hungry': if he did, he would believe that he is hungry. Rather, Fred can satisfy the ascription with a belief that picks the ascriber out via a quite different primary intension. Likewise, the second ascription may be true even though Pierre use of 'Londres' and the ascriber's use of 'London' have somewhat different primary intensions. So here again, Pierre needn't have a belief with the primary intension of 'London is pretty' in the mouth of the ascriber.
So it looks like S5 and W5 are pretty clearly false, at least if one accepts standard intuitive judgments about belief ascriptions. The natural way for the two-dimensionalist to embrace all the data above is to hold that belief ascriptions are sensitive to both primary and secondary intensions of the subject's beliefs, in slightly different ways (reflecting the fact that the left-to-right component of W5 appears to be correct, while that of S5 appears to be incorrect). The obvious resulting account is something like the following:
(A) 'A believes that S' is true iff the referent of A has a belief with the secondary intension of S (in the mouth of the ascriber) and with an appropriate primary intension.
"Appropriate" is used to accommodate the fact that the truth of the ascription is sensitive to the primary intensions of the subject's beliefs (as example (2) suggests) but does not require a belief with a specific primary intension (as examples (6) and (7) suggest). Rather, it seems to require a belief with a primary intension that falls into a certain class. (For a subject to believe that Clark Kent is handsome, the relevant belief can pick out Clark under various different primary intensions, but not under just any primary intension.) Of course (A) leaves open just what it is for a primary intension to be "appropriate"; the relevant class may well be determined in a complex and context-sensitive way. Strictly speaking, (A) should invoke structured primary intensions and structured secondary intensions, but the approximation above will do for present purposes.
I take it that all of the problematic ascriptions discussed above are familiar, and that all have been used to raise problems for various Millian and Fregean theories of belief ascription. For example, the point about variation in primary intension between ascriber and ascribee is an analog of a familiar point about variation in Fregean senses (or of modes of presentation) between ascriber and ascribee, used to make trouble for Frege's original semantics of attitude ascription. And I take it that the shape of the resulting account is also pretty familiar: it's very close to the shape of "hidden-indexical" or "notion constraint" accounts of belief ascriptions (though it's not committed to any particular account of the logical form of such ascriptions).
I'm not sure why Soames spends so much time discussing and attacking such implausible theses (interestingly, his criticisms of these theses are largely different from those above). I don't know of any writings by two-dimensionalists that endorse either thesis. As far as I know, the only published analysis of attitude ascriptions by an "ambitious two-dimensionalist" is my analysis in "The Components of Content" (section 8), where I give something like the analysis above. Soames does get around to discussing this view later in the book, but it's clear that he views it as some sort of baroque defensive maneuver. He says that it's a change of mind on my part, dating from six years after The Conscious Mind (where he thinks I endorse strong two-dimensionalism and S5). In fact the view is explicit in the 1995 version of "The Components of Content", which is cited in the 1996 book (which otherwise says hardly anything about attitude ascriptions). Since I started writing on two-dimensionalism, it's always seemed pretty clear to me that this view is the natural two-dimensionalist approach to attitude ascriptions.
A Priori Knowledge Ascriptions
On a closer examination, it looks like the main reason for Soames' attributing thesis S5 to Jackson and to me involves ascriptions of a priori knowledge. Some of the relevant discussion is in Chapter 9 of Soames' book (especially pp. 236-239, and pp. 250-265), but it's useful to spend time on it now, as the issues are closely connected to the issues above, and issues about a priori knowledge ascription arise repeatedly in the remainder of the book.
The reasoning is roughly as follows. Soames ascribes the following two theses to me (p. 250):
(8) A sentence S is a priori iff S has a necessary primary intension.
(9) A sentence S is a priori iff 'It is knowable a priori that S' is true.
He infers that I am committed to:
(10) 'It is knowable a priori that S' is true iff S has a necessary primary intension.
From here, he infers that I am committed to the claim that 'It is knowable a priori that' operates on the primary intension of the embedded sentence S (or at least that only the primary intension of S matters to the truth-value of the apriority claim). From there, he infers that I am committed to corresponding claims involving 'A knows a priori that S', 'A knows that S', 'A believes that S', and 'A asserts that S', because of the "evident semantic connection" between these locutions. That is, he infers that I am committed to his thesis S5.
I think that a number of the steps in this reasoning are problematic. The last step is especially problematic, for reasons I'll discuss shortly. Further, while I'm committed to (8) and I think that (9) is not unreasonable, it's not entirely obvious that (9) is correct, and I'm not committed to it. One might think that (9) is true by definition, but I don't define the apriority of a sentence this way. As I define apriority, it's a property of sentence tokens (or utterances) rather than sentence types. This allows that apriority can vary between speakers -- cf. Kripke's locution "a sentence is a priori for a speaker" -- and also that it can vary between utterances by the same speaker, because of contextual variation. A sentence token S is a priori iff it expresses actual or potential a priori knowledge: that is, if it expresses a thought that can be justified a priori, yielding a priori knowledge. Here, the apriority of sentence tokens is defined in terms of the apriority of certain mental states (this is done partly in order not to presuppose any apparatus involving propositions). This usage seems to correspond reasonably well to the intuitive classification of sentences as a priori and a posteriori, and also seems fairly close to Kripke's usage.
Note that on this approach, apriority isn't defined in terms of attitude ascriptions in which the relevant sentence is embedded. In fact, it seems pretty clear that the two can come apart, at least to some extent. For example, an utterance by Huey of (11a) below is plausibly a priori, but an utterance by Dewey of the attitude ascription (11b) is plausibly false.
(11a) I am here now (if I exist and am spatiotemporally located)
(11b) Huey knows a priori that I am here now (if I exist and am spatiotemporally located).
One reason is that the truth of Dewey's utterance of (11b) requires an appropriate relation between Huey's a priori knowledge and Dewey, but the a priori knowledge that Huey would express by uttering (11a) bears the relevant relation to Huey. This is what one would expect if we generalize account (A) of belief ascriptions to a priori knowledge ascriptions, as follows:
(A*) 'A knows a priori that S' is true iff the referent of A has a priori knowledge with the secondary intension of S (in the mouth of the ascriber) and with an appropriate primary intension.
Given account (A*), a priori knowledge that Huey would express by uttering (11a) clearly can't make Dewey's utterance of (11b) true, as this knowledge doesn't have the secondary intension of 'I am here now...' in Dewey's mouth.
Now, in the special case of a first person a priori knowledge ascription, such as (11c), this sort of difference doesn't arise.
(11c) I know a priori that I am here now (if I exist and am spatiotemporally located)
It's plausible that whenever Huey's utterance of (11a) expresses a priori knowledge, his utterance of (11c) will be true (assuming that the occurrence of (11a) embedded in (11c) has the same primary and secondary intensions as the original utterance of (11a)). To see this, first note that whenever an utterance expresses a belief, the utterance and the belief have the same primary and secondary intensions. It follows that the a priori knowledge expressed by Huey's utterance of (11a) will have the same primary and secondary intensions as the utterance itself, which are the same as the primary and secondary intensions of the occurrence of the embedded sentence in (11c). It follows that this a priori knowledge will automatically be enough to make the utterance of (11c) true, as long as we assume (as is plausible) that in a first-person attitude ascription, the primary intension of the embedded sentence itself always counts as "appropriate".
For similar reasons, it is plausible that whenever Huey's utterance of (11a) is a priori (even if it merely expresses potential a priori knowledge), an a priori knowability claim such as (11d) is true (at least if the 'can' here is sufficiently idealized).
(11d) I can know a priori that I am here now (if I exist and am spatiotemporally located).
Of course (11d) entails (11e)
(11e) It is knowable a priori that I am here now (if I exist and am spatiotemporally located).
So all this suggests that the apriority of (11a) entails the truth of (11e). More generally, the apriority of an English utterance S plausibly entails the truth of a corresponding utterance of 'It is knowable a priori that S'. This is more or less the left-to-right direction of Soames' principle (9) above. But importantly, note that nothing here suggests that the truth-conditions of 'A knows a priori that S' involve only the primary intension of S. In fact, the reasoning above that yielded the entailment assumed account (A*), on which both primary and secondary intensions are relevant in a priori knowledge ascriptions.
What about the right-to-left direction of (9)? Given what's gone before, one might have doubts about this. One might worry that there could be cases where 'A knows a priori that S' is true, even though S (in the mouth of the ascriber) is not a priori, due to the fact that the ascription allows some variation in primary intensions between ascriber and ascribee. Here's a candidate for such a case. Consider Huey's utterances of (12a) and (12b), where 'he' and 'here' refer to Dewey and his location.
(12a) He is there now (if he exists and is spatiotemporally located).
(12b) Dewey knows a priori that he is there now (if he exists and is spatiotemporally located).
Here, (12a) seems pretty clearly not to be a priori for Huey, but it's not wholly unreasonable to take (12b) to be true, in virtue of the fact that Dewey has the a priori knowledge that he expresses by uttering 'I am here now, if I exist and and am spatiotemporally located'.
Here's another candidate. Huey uses 'Lee' as a descriptive name for the shortest Chinese spy (if there is one). By contrast, Dewey uses 'Lee' as a descriptive name for the shortest male Chinese spy, if there is one. Perhaps the two uses have a common origin, with small variations in the community. We can stipulate that the youngest Chinese spy is in fact male. Then consider Huey's utterances of (13a) and (13b).
(13a) Lee is the shortest male Chinese spy (if there is one).
(13b) Dewey knows a priori that Lee is the shortest male Chinese spy (if there is one).
Again, it seems clear that (13a) isn't a priori for Huey (it doesn't express a priori knowledge or an a priori justifiable belief). But there's at least a case for saying that Huey's utterance of (13b) is true. My intuitions about (13b) waver, but I think it's not entirely unreasonable to hold that Huey's utterance of could be true. (I'm interested to hear about others' intuitions concerning (12b) and (13b). Maybe I'll take another poll, but if you've read this far, feel free to comment here.)
Furthermore, it's not unreasonable to hold that if Huey's utterances of (12b) and (13b) are true, so are his utterances of (12c) and (13c).
(12c) It is knowable a priori that he is there now (if he exists and is spatiotemporally located).
(13c) It is knowable a priori that that Lee is the shortest male Chinese spy (if there is one).
Interestingly, although (12c) and (13c) seem to follow logically from (12b) and (13b), my intuitions about them (taken alone) are somewhat more negative than my intuitions about (12b) and (13b). One might diagnose this by suggesting that utterances of 'It is knowable a priori that...' tend to introduce somewhat different "appropriateness" constraints on primary intensions than do utterances of 'Dewey knows a priori that...' and the like (e.g., maybe the former but not the latter tend to require that the relevant a priori knowledge has the same primary intension as the relevant sentence in the ascriber's mouth). But I can also hear (12c) and (13c) as true, especially in contexts where they are uttered immediately after (12b) and (13b). If (12c) and (13c) can be true, then it looks like while the apriority of S entails the truth of 'It is knowable a priori that S', the reverse entailment does not always hold. If this is so, then the right-to-left direction of (9) is false.
The upshot of all this for Soames' discussion is the following. First, I'm not strongly committed to (9) or to (10). Neither is true by definition, and the cases above cast at least some doubt on whether they are true at all. Soames' evidence that I'm committed to these claims is a single sentence (!) on p. 64 of my book: "The primary intensions of 'water' and 'H2O' differ, so that we cannot know a priori that water is H2O." Soames concludes from this sentence that I use "it is knowable a priori that S" as if it is interchangable with the claim that S is a priori. I don't think one can draw this conclusion from the sentence, but I think he's probably correct that I was using the phrase on this occasion as if the two were interchangable. This reflects the fact that there's plausibly a use of 'It is knowable a priori that S' (i.e. the use with the stricter "appropriateness" constraint above) such that two really are interchangeable, and in context it's pretty clear that this sort of use is intended. Still, when one is being careful, it's probably best to use a different locution such as 'It is a priori that S' (which I use much more often) for the stricter claim. E.g. we can stipulate that a token of 'It is a priori that S' is true iff the embedded token of S is a priori. Of course one then needs to be careful about moving between 'It is a priori that S' and the quantified modal a priori knowledge ascription 'Someone can know a priori that S'.
Second, and more importantly, even if (9) and (10) are correct, one can't infer that the truth-conditions of 'A knows a priori that S', 'A knows that S', 'A believes that S', and so on all involve only the primary intension of S. We've seen above that the truth of (9) and (10) is quite compatible with the claim that the truth-conditions of attitude ascriptions such as 'A knows a priori that S' involve the secondary intension of S. So even if Soames were correct in ascribing (9) and (10) to me, the more general claim about attitude ascriptions doesn't follow.
On p. 250 of the book, noting the tension in his interpretation, Soames says that the account of attitude ascriptions in "The Components of Content" "is not easy to square with the account of 'it is knowable a priori that' which emerges from The Conscious Mind. I think we've seen here that when we work through the details, it is easy to square the two accounts with each other.