G. Rosenberg’s “Ontological discussion” (pp. 59-68)
Rosenberg’s overall argument is that “if primitive identities exist, they cannot provide a basis for a posteriori necessity because primitive identities themselves would need to inherit their primitiveness from an independently based a posteriori necessity” (59). The argument is stated in outline form on pp. 59-60; I won't repeat it here.
Obviously that depends on what “basis” means, which has been unclear. (I am writing this without yet having got hold of the book and checked all the footnotes.) I assume Rosenberg is appealing to a “basing” relation, and that “inheriting” is its converse. And the argument seems to be a regress argument: An a posteriori necessity must have a basis. And some a posteriori necessity would have to bottom out by being based on a primitive identity that was not itself based on an a posteriori necessity. OK. As with any regress argument, we must ask whether the alleged regress would be vicious; but here it would be, because no one thinks that the world contains an infinite string of a posteriori necessities based on primitive identities based on a posteriori necessities. Rosenberg alludes to supervenience, so “basing” seems to be a relation that holds as between one level of nature and another; and certainly facts about macroscopic token identity must “supervene on other, more fundamental facts” (p. 59), because all macroscopic physical facts do. (I have not seen the works of Yablo’s and Johnston’s that Rosenberg cites here.)
The crucial premise is 4: "If entailment does not provide the [supervenience] connection, the primitiveness of the identities is inherited from the primitiveness of an underlying a posteriori supervenience relation [which would be an a posteriori necessity."
Three things first: (1) Rosenberg says on p. 51 that by "basis" he means the sort of principled distinction between metaphysical and merely conceptual possibility that is questioned by Jackson, Chalmers, and himself in sec. 3.3.4. But that notion of "basis" has nothing obvious to do with the basing relation aforementioned. I suppose the idea is that the Kripkean distinction must have a (so to speak) primary locus: What is ultimately responsible for the difference between a metaphysically possible world and a merely conceptually possible world? I myself have said that natures and essences are what make that difference, so I guess I understand the question despite my feeling that I don't. (2) Rosenberg talks specifically of particular facts' being "based on" underlying facts, or being "produce[d by]" them (p. 61). Reading charitably, I think this must be a noncausal but broadly explanatory relation. (3) By "supervene," Rosenberg means something stronger than supervene. It is not at all controversial that macroscopic identities supervene on lower-level facts about indiscernibility (and indexicality) (p. 60); if two worlds share an indiscernibility, they have the same corresponding identity. But Rosenberg means that identity is simply analyzed as indiscernibility, and disappears in favor of it.
He gives just one argument for the latter claim (pp. 60-61), a bad one: We test for identity by checking for sameness and difference of properties. Premise true, but it's fallacious (also silly and naughty) to infer metaphysics from epistemology. I know of no other reason to accept this analysis of identity; intuitively, I think, the indiscernibility of X and Y is explained by "their" identity, rather than the other way around.
However, suppose the analysis is correct. Then a given primitive identity is just a primitive indiscernibility. But now comes a very cryptic and puzzling argument (pp. 61-62). We have the lower-level "local" A-facts and the higher-level local B-facts supervening (in the ordinary sense) on the A-facts. The two premises ((1) and (2)) are hard to dispute. But now Rosenberg promulgates his dilemma: Either the A-facts and B-facts together (n.b.) entail that A is locally indiscernible from B or they don't.
There are already two things here that I don't understand. First, why is Rosenberg talking about the A-facts and B-facts together, rather than just the A-facts? (The primitiveness issue is that of whether macroscopic A = B is semantically entailed by the A-facts alone.) Second, what work is being done by the restriction to local properties? Clearly I am missing something.
Assuming the indiscernibility analysis of identity, consider Rosenberg's first horn: He says that if the A-facts and B-facts together entail the indiscernibility of A and B, then A = B is not primitive. Why? Again, to call A = B primitive is to say that it is not semantically entailed by the A-facts alone. And, obviously, that the A-facts and B-facts together entail it hardly shows that the A-facts alone do. I really don't get this.
Rosenberg's second horn is pretty weird in and of itself: The A-facts and the B-facts together don't entail the indiscernibility? Why wouldn't they? (OK, you'd need a "That's all" fact.) Rosenberg goes on to say that if they don't, the indiscernibility must follow from them taken together with some further facts--an indisputable logical truth. But now he starts talking about "providing a basis for" the indiscernibility. Now I'm lost. Also, why does Rosenberg bother to make a substantive argument regarding case (2) (the argument about the determination relation)? He's already assuming the indiscernibility analysis of identity, and it follows from that assumption alone that the identity itself cannot "provide a basis for" indiscernibility, whatever is meant by "provide a basis for."
In any case, in defending the Kripkean distinction I myself have appealed to the necessity of natures, so let's move on to sec. 3.3.3. Rosenberg reverts to his "canonical example," water and H2O. He says that "the conceptual content of the theoretical concept H2O entails that anything that is H2O is liquid at room temperature" (p. 65). As always, I seriously doubt that, because I don't see that there's going to be an adequate "conceptual analysis" of liquid, but let's again suspend disbelief and see what comes next. The key point seems to be in the paragraph beginning "Physicalists needs an example...." What Rosenberg seems to be asking for is an example of a microtheory that upwardly necessitates the macro-facts about its subject-matter but without semantically entailing them. He says, "It is implausible that such an example should exist, because that kind of predictive failure is always taken as compelling evidence that a theory is incomplete" (cf. the "Methodological discussion" on pp. 55-59). Well, first, we Kripkeans (already) think that nearly every natural kind is an example of supervenience without upward semantic entailment; that issue has been hashed over in responses to Jackson and Chalmers. Second, more specifically, there is in such cases no "predictive failure" worth calling that; that the way a cat's kidney functions is not semantically entailed by the quantum mechanics of the relevant subatomic particles does not prevent us understanding that function as thoroughly as we like. Nor have I heard of a single scientist who ever took a lack of upward semantic entailment as evidence, much less compelling evidence, that the lower-level theory is incomplete.
(I'm probably misunderstanding again, but I also don't see how the J-C upward entailment thesis is supposed to be an objection to grounding the Kripkean distinction in the necessity of natures. Possibly the idea is this: Although "Water = H2O" is an a posteriori identity, it is in fact semantically entailed by the microphysics of this world; so there is no even conceptually possible world that is like ours microphysically but in which water =/ H2O. But the relevant worlds are ones which are like ours microphysically except that there is XYZ instead of H2O, and in which the XYZ is water. What shows that those worlds are not even conceptually possible?)
There remains the minimal meaning postulate (MM, sec. 3.3.4). What Rosenberg says about the context-relativity of possibility statements is both true and very important. For amplification and a wealth of (if I say so myself) enchanting examples, see Ch. 8 of my Modality and Meaning. I have no quarrel with MM (= Definition 3.2). So a philosopher who distinguishes metaphysical possibility from merely conceptual possibility needs to indicate the relevant background constraints. Rosenberg is skeptical: "it seems that the relevant constraints are not being even obliquely specified, and this raises worries about whether any really exist" (p. 67). Not even obliquely? Let me articulate them once more, straight out: Natures and essences. In the preceding section, Rosenberg argued (however obscurely) that the necessity of natures does not "ground" the Kripkean distinction, but in 3.3.4 the question is only that of what the background constraints are to which "metaphysical possibility" is relativized. That question has been answered. As always, one might be skeptical about n.'s and e.'s themselves, and indeed about all of modality, but that would be a much deeper objection to the Kripkean picture and there is none of it in Rosenberg.
In closing (p. 68), Rosenberg says that the Kripkean
will have to "posit" a set of metaphysical laws and constraints along with
the physical facts and laws of nature. Well, yes, not to quibble,
the Kripkean believes in such laws and constraints, on the basis
of modal intuitions tutored and untutored. "So," Rosenberg concludes
without any supplementary premise, "physicalism would be false all the
same...." ??!--I'm baffled again.