The difficulty for using a “conceptually prior” class of possible worlds — more about what is meant by the phrase in scare quotes in a moment — to the purpose of giving an account of the intension of predicate terms (at least one which is epistemically responsible) is more apparent. Say we assume, as Lewis does, that for the purposes of explaining the truth of modal claims, there are possible worlds as he understands them and there is a class of all such worlds. If this account of modality is to be a reductive one, then to avoid circularity in it, we must assume that the worlds exist independently of our conceiving of them, else we would be attempting to reduce the modal to a class that was delimited with the aid of some sort of modal ability (our conceiving of the these worlds — this is why I say that the worlds must be “conceptually prior”). But, if we have a mind-independent class of possible worlds to explain the truth of modal claims, then we’re unable to use this class of worlds to explicate the intension of predicate terms given that we understand those intensions.
Why? Well, say for example, we want to use possible worlds to understand the intension of the predicate ‘is a mountain’. What do we do? Simple: the intension of ‘is a mountain’ is the set of all individuals picked out by ‘is a mountain’ in each possible world, or, rather, the set of mountains in all possible worlds. The trouble is that we do, prephilosophically, understand the intension of ‘is a mountain’ and so can use the predicate appropriately in conversations about counterfactual situations, but on the possible world account of the intension of ‘is a mountain’, it remains mysterious how we can have knowledge of the intension of ‘is a mountain’ given that the class of possible world exists mind-independently. Any claim that we have some sort of special access to possible worlds runs counter to the idea that a class of possible worlds serves as the reductive ground for modality (and so counter to the claim that the worlds are “conceptually prior”).
From part II, we’ve seen that Ray’s adaptation of Menzel serves only to place the formal machinery for providing the semantics of a language with modal operator atop notions which are fundamentally modal (meaning notions), and so can’t shed light on more fundamental modal notions. If I’m right about the foregoing in this post, then any method which uses Lewisian possible worlds to account for the intensions of predicate like ‘is a mountain’ by identifying those intensions with the set of all individuals in the extension of ‘is a mountain’ in each possible world can’t also be part of the theory which is reductive account of alethic modality.
One might say that the use of Lewisian possible worlds to explicate modal semantics could be part of a non-reductive account, but this leaves one with the question of why use complete, non-temporally related, physical universes as the basis for explaining something like fundamental modal properties or dispositions. Perhaps objects this “large” are the sort of thing to explain these sorts of properties or dispositions, but I suggest in the final post that we need something different (yet very similar) to explain de dicto modal claims and simultaneously explain intensions of predicate terms.
Okay, so this seems wrong…
“if we have a mind-independent class of possible worlds to explain the truth of modal claims, then we’re unable to use this class of worlds to explicate the intension of predicate terms given that we understand those intensions… on the possible world account of the intension of ‘is a mountain’, it remains mysterious how we can have knowledge of the intension of ‘is a mountain’ given that the class of possible world exists mind-independently.”
The question is: what makes this especially hard to understand? We think that the actual world exists mind independently, and we think that we can have knowledge of what goes on in the actual world, so bare mind independence cannot be a barrier to the sort of knowledge you are asking after. Now, one might think that there *is* something especially hard to understand about the epistemology of Lewisian worlds because they are causally unrelated to the actual world, but that depends on how liberal one’s conception of routes to knowledge is, I suppose. Lewis was well aware of the epistemological burden faced by his theory, but he offered some tips. We can use a principle of recombination, tempered by the fact that this is at best imperfect, as a guide to the facts about other possible worlds. If we can take that as a serious and plausible response to the philosophical worry about how it is that we do modal epistemology, then we can simply extend the idea to give a treatment of your above worry. How is it that, even pre-theoretically, we understand the intensions of concepts that require modal knowledge? By some sort of pre-theoretic (perhaps even largely subconscious) deployment of a principle of recombination. What really concerns me is that you suggest that *if* we have knowledge of other possible worlds on the Lewisian realist construal of the nature of said worlds, it would undermine the reductivity of the account. I just don’t see this. Could you expand on that idea?
Colin — thanks for this comment. There are a few different points let me try to respond to each one.
By way of preface, one of my concerns in this for post is, of course, to sketch a way to understand an epistemology of intensions of predicate terms. We shouldn’t be satisfied with a claim that a necessary and sufficient condition on knowing the meaning of ‘is a mountain’ is knowledge of when to apply the predicate to merely actual objects. This condition is necessary, but not sufficient. One who knows the meaning of ‘is a mountain’ must also know when it’s appropriate to use the predicate when speaking of counterfactual situations or about “what could have been the case”.
I’d argue that mind independence is not a barrier to a bit of (non-mathematical) knowledge K if there is some sort of causal connection between the fact that K and my knowing that K. In the actual world, this causal connection is had by information one gets via the senses. I know my house is blue because I can see that it’s blue.
I’m not exactly sure what a principle of recombination is, but I’ll take a guess. From things I know about the actual world, I might be able to imagine or conceive a (recombined) world in a compositional way (meaning I mix-and-match objects and features that are found in the actual world) in which there were a 150m tall tree. By this sort of method of recombination, I could come to know that it’s possible for a tree to be 150m tall. What I find troubling about this route to knowledge is that the sentence ‘possibly, there is a tree that is 150m tall’ is made true by the fact that in some nearby possible world, there is a tree that is 150m tall but I come to have knowledge of the truth of this sentence by some other means. I can’t have the sort of knowledge about that nearby possible world that I have about the actual world — I can’t perceive the tree in the nearby possible world, but I could perceive a tree in the actual world. The gap between (for lack of a better term) truth-making and epistemology bothers me. In these four posts, I’m trying to sketch out a view according to which a modal claim is made true by the same way in which we know it. I hope this comes out in the final post — hopefully up later today.
Another way to put the point is with the following. In “New Work Towards a Theory of Universals” (I think), Lewis gives a set-theoretic treatment of properties: a property is a set of individuals from various possible worlds. Properties carve the world as the joints (the set of all actual and possible mountains) and everywhere else (a random set of individuals from the collection of possible worlds). I want to hold the traditional view that predicates express concepts and pick out properties. So a predicate like ‘is a mountain’ would pick out that set whose membership carves the world at the joints and expresses a concept (call it ‘MOUNTAIN’). What troubles me is that I can know the meaning of ‘is a mountain’ and so have the concept MOUNTAIN, yet I can know only indirectly about what I make a modal claim about with a sentence which includes ‘is a mountain’. I certainly can’t have direct epistemic access to the (Lewisian) property; I must have access to those things picked out by my use (in a modal context) of ‘is a mountain’ by the method of recombination. My epistemology of intensions of predicate terms in modal claim contexts is not of a piece with my epistemology of intensions of predicate terms in the context of non-modal claim contexts.
The argument that possible worlds cannot be used to give an epistemically responsible, reductive account of modality is advanced by Scott Shalkowski in The Ontological Grounds of the Alethic Modality. I think Ted Sider has a response (that may miss Shalkowski’s call for epistemic responsibility) in Reductive Theories of Modality (pages 19-20).
“My epistemology of intensions of predicate terms in modal claim contexts is not of a piece with my epistemology of intensions of predicate terms in the context of non-modal claim contexts.”
Absolutely, I’ll grant as much, but one wonders just why this should be so bothersome. When you say that you can’t have direct epistemic access to the Lewisian modal facts, but rather have only indirect access, I take it this is simply a way of alluding to the non-causal nature of Lewisian modal epistemology. Again, I grant this much, but I am not yet clear on why we should see this as a problem for treating our understanding of intensions in modal terms. I guess what I really wonder is: if you find it even the least bit plausible that some recombination tactic (you roughly describe the idea accurately in your comment) can deliver knowledge of the modal facts, then why should we need more? Why should we care about the channel by which we got the knowledge?
There’s a principle of parsimony in my background assumptions. Say, for the sake of argument, that I can gain the modal knowledge expressed by the sentence ‘necessarily, S’ by means of a recombination tactic, and that the sentence ‘necessarily, S’ is made true because in every possible world, S is true. Now, just from what we’ve learned by use of the recombinant tactic, it is the case that ‘necessarily, S’ is true because it is knowledge after all. I’m suggesting that we don’t need to invoke possible worlds to be the truth-makers for sentences like ‘necessarily, S’ (and so we shouldn’t) — it’s enough to give an account of modality that gives us the right semantics for sentences whose primary operators are ‘necessarily’ or ‘possibly’, given that we can recognize the truth of those sentences.
This post is part of a larger project in which I argue for the view that a sentence of the form ‘necessarily, S’ (where S is either de re or de dicto) is true just in case S is (roughly) analytic. If we can make a convincing case that sentences like ‘nec. S’ for de dicto S, are true iff S is analytic, then if we have knowledge of the predicates of S we’d be able to recognize the truth of sentences of this sort. Dealing with de re embedded sentences is harder.
That seems reasonable, and I think parsimony is exactly the right motive. I’m not clear that someone who has independent commitment to Lewisian possible worlds need be bothered by the way you’ve sketched this issue above. They might be happy to account for knowledge of intensions in the way I’ve mentioned. But many philosophers seem to share the methodological stance that if we can do semantics without possible worlds, then we don’t need possible worlds at all, and that would be a good thing. So, I can see why you would pursue this line.