I’m teaching a seminar on David Lewis in the spring, using Daniel Nolan’s book as one of the texts, but primarily using primary sources. (Note the link is now to Amazon US, not to Britain.) There’s one primary historical claim I want to make through the course, and I might post stuff here as I write it to try and back up this claim.

The claim is that concrete possible worlds aren’t that important to Lewis’s overall philosophy.

Now this might seem like a rather odd claim to make, since the existence of concrete possible worlds is the core claim of Lewis’s longest and most important book, and most summaries of Lewis start with this belief as the lead. So I’m swimming a little against the interpretative tide, such as it is at this early stage. But I do have something like an argument for my preferred interpretation, or perhaps better, preferred sense of what is central and what is relatively peripheral.

Although Lewis held fairly consistently to most of his views throughout his career, he was amenable to changing his mind in the face of good argument. He was constantly tinkering with the analysis of causation, he accepted some tinkering to the analysis of intrinsicness, and most importantly, he abandoned egalitarianism about properties. So we can coherently, even easily, imagine him changing his mind on some of his conclusions.

Now for any paper or work of Lewis’s, do the following thought experiment. Imagine someone giving him a convincing argument that the core conclusion of the paper failed, shortly after the paper appeared in print. Try to follow through then how many changes would have been needed to later work if he’d accepted this criticism, and hence not been able to presuppose the truth of his earlier claims.

In some cases the changes would be immense. It is almost impossible to imagine what Lewis’s work would look like if he didn’t have the Ramsey-sentence definition of theoretical terms, and his convention-based account of language to work with throughout. Similarly we would need massive revisions if he’d been convinced that Humean supervience was fundamentally flawed, or that egalitarianism about properties was true after all, or that one of the paradoxes from chapter 2 of Plurality was fatal to the possible worlds apparatus. These I think are the five core claims that Lewis’s philosophy is built around, and if he’d given any of them up, it would have forced massive revision to the rest of his work.

Not so for concrete possible worlds. If someone had written a version of ersatzism in 1987 that Lewis thought gave us all we needed from a theory of possible worlds, and wasn’t vulnerable to the chapter 3 objections, then he could have accepted it, and, more importantly, he would have had to make next to no substantive changes to anything else he wrote. What is of broad philosophical importance for Lewis (and others such as Stalnaker and Jackson are in this camp too) is that the possible worlds framework is coherent and applications of it can be philosophically enlightening. Lewis had opinions on the metaphysical foundations of that framework, but they weren’t central to the philosophical picture he gave us.

The test I’m using for importance isn’t perfect, because it isn’t clear always how to apply it. I’ve been deliberately silent in what I’ve said above about the importance of the analysis of counterfactuals and of causation to the picture, as it is being sketched here. I tend to think that in both cases the details (in some sense of detail) are not important, any more than the details of the account of intrinsicness or colour or dispositions or chance or laws are important. (Which is not to say they don’t at all matter – Lewis worked hard on finding reasons to support particular analyses of all of these.) But both causation and counterfactuals seem more central to the project than these other cases, especially counterfactuals. Saying just why I think that needs more work, and hopefully I’ll be able to have something more coherent by the time next spring rolls around.

I have a vested interest in this interpretation being taken seriously, because broadly speaking I’m a Canberra Credo guy, and “We [Credoers] believe in the substantial correctness of the doctrines of David Lewis about most things (except the nature of possible worlds).” It would be nice if (a) we could argue that this is a coherent position, which it isn’t if concrete possible worlds are central to Lewis’s picture, and (b) one which involves only a small deviation from Lewis’s theories, since it’s odd to look up to someone on so many matters while disagreeing with them about what is fundamental to their work.

10 Replies to “Lewisiana”

  1. I’m interested that you think that the convention-based account of linguistic content is essential. It seems to me that the “head-first” approach is vital, so that you’ve got beliefs and desires to play with when fixing linguistic content; and perhaps the general functionalism about mental content too.

    But is the appeal to conventions itself that important? I would have thought that Lewis could etreat to a different broadly interpretationist treatment of linguistic content, if someone convinced him that his game-theoretic analysis of convention was screwy.

    For example, consider global descriptivism, in the form where we take someone’s “folk theory” of the world and ask for an interpretation that makes it (or enough of it) true. (To characterize the folk theory, we can appeal to dispositions to assert sentences, and also strengths of beliefs that sentences are true, etc—-so it’s still a head-first account).

    Basically, the convention-account will give you pairings of sentences with propositions, whereas global descriptivism (as I understand it) just gives you a pairing of sentences with truth values. Why does Lewis need the richer version, given what he says in defence of g.d. in “Putnam’s Paradox”?

    None of this is to say that Lewis didn’t maintain the convention-account throughout—-though it’s appealed to surprisingly little…

  2. The reason I think the convention account, in some amount of detail, is relevant because Lewis constantly appeals to it when working out how much theory he is supposed to provide. To take one instance more or less at random, I was just teaching “Mad Pain and Martian Pain”. Lewis says that it is sometimes unclear what ‘pain’ denotes on his theory. But he says that this is OK, because there’s no reason our conventions should have settled what to say in just those situations. This kind of thing happens all the time. I think without the convention account he’d have to say very different things about what questions he has to ask and answer.

  3. The motivation for concrete possibilia is not just that they can ‘do the work’ needed while ersatz possibilia cannot—-by providing enough merely possible individuals, say. It’s also that by believing in them one avoids having to believe in any primitive modal facts. Aren’t the sorts of thoughts that lead you to abhor primitive modality the same as the sorts of thoughts that motivate humean supervenience? If so, belief in concrete possiblia does look like an important part of Lewis’s philosophy.

  4. Brad,

    That’s an interesting link between the concrete worlds stuff and the rest, but I’m not totally convinced for a couple of reasons.

    First, even if there is some common motivation to have Humean supervenience go with no primitive modality, the two aren’t logically connected. One could accept primitive modality and say that it is part of the (primitive) concept that there are no necessary connections between distinct existences. And it’s the no necessary connections that is the foundation for Humean supervenience, at least as I read it.

    Having said that, I’m not sure exactly what motivates the ‘no primitive modality’ line. Ted Sider says it is a version of the ‘everything fundamental is categorical’ principle. Maybe that’s right, and maybe that could motivate Humean supervenience. But it’s a questionable application of the ‘everything fundamental is categorical’ principle, since primitive modality wouldn’t violate any supervenience or non-supervenience claims we want to make, and I certainly think Lewis could have come to accept that there’s no argument against primitive modality without thinking that anything else had to change.

    Second, it isn’t altogether obvious that giving up concrete possible worlds requires accepting primitive modality. Ted has an ersatz theory and a reductive account of modality. Now Lewis wouldn’t accept either part of Ted’s theory as it stands, but it is possible to have those two come apart.

  5. In one sense, it shouldn’t be controversial to claim that modal realism (as opposed to sympathy towards possible worlds in some more general sense) is not that important to the Lewis program: after all, many people seem to buy the Lewis program minus the possible worlds. I don’t see any super-special need for modal realism in counterfactual analyses of causation either, as long as we are clear about what we are analyzing. I think that books tend to focus on Lewis’s modal realism in part because it is such a distinctive view, and in part because it is the basis Lewis uses (instead of some sort of ersatzist view) to develop other interesting aspects of his position.

    In another sense, modal realism, as part of a take-no-prisoners approach to reductionism about everything except qualities scattered across the manifold, is quite an important element of the overall Lewisian perspective (I think Brad Skow was suggesting something like this in his earlier comment). But this doesn’t mean that Lewis could not have dispensed with it if an appropropriate ersatzist program were on offer (I think he even says this in Plurality). One last thought about this: one way in which modal realism might be important to the Lewisian approach involves the need to explain the motivation for taking on some of the various details of the ersatzist program. E.g., counterpart theory minus modal realism seems to require some independent justification.

  6. I suppose that realism about possible worlds became much less important once Lewis had given up egalitarianism about properties. Egalitarianism is nominalistic in spirit, and motivated by a suspicion against everything non-concrete. Nominalists take predication as primitive, and may reconstruct talk of (relatively sparse) properties as meta-linguistic talk of semantic values of predicates. Without other worlds, these values can only be extensions, which do not have the fineness of grain that Lewis needed. For an egalitarian or quasi-nominalist, one obvious way to get intensions is to postulate concrete other worlds. (Another way, taking analyticity as primitive in the Linguistic Ersatzist fashion, was not open to Lewis on pain of circularity. In “Convention”, he had already analysed analyticity in terms of possible worlds.) But when Lewis started to follow Armstrong rather than Goodman in accepting inegalitarianism, very natural ways to take possible worlds as constructions of some sort out of properties became available. Of course he had objections to all the proposed ways, but I agree that they are now largely independent of his main philosophical views.

  7. Brian, I completely agree that Lewis’ account of possible worlds is independent of must of his other important views. Specifically, his accounts of natural properties, laws, chance, counterfactuals, and causation (a formidable bunch) are all compatible with pretty much any view about possible worlds (this is sometimes not appreciated by those who worry about what a match lighting in another world could have to do with what would happen if this match were struck)though these accounts are closely connected with one another (I agree that they all need adjustments). Same for Humean Supervenience and Physicalism which are contingent claims about our world (the first wrong the latter arguably correct). Same for his idea that fundamental properties are categorical. Same for his account of reference. And while counterpart theory (denial of individual essences), and indexical account of actuality are closely onnected to his view of worlds they can be held without it.

  8. Brian’s correct here. Lewis was always careful about distinguishing the essential from the inessential presuppositions of his papers. The concreteness of the worlds was always mentioned as falling on the inessential side.

  9. I also agree that the concreteness of other worlds doesn’t matter much for Lewis’s overall philosophy. But their existence matters. That is, I think an ersatzist can follow Lewis much further than a fictionalist. For instance, if possibilia only exist according to some fiction, properties and propositions can’t be sets of possibilia (unless they, too, only exist according to a fiction).

  10. I’m not sure if Lewis’s account of properties as sets will work for an ersatz theory. At least, it seems to me that things would be messy. The set corresponding to the property “donkey” would include all actual (concrete) donkeys, and it would include whatever we’re using as ersatz-donkeys. In Lewis’ original theory, the set would consist only of concrete donkeys. This might not be a problem (and I might not know what I’m talking about), but it’s ugly. To put it another way, it seems to me that Lewis’ original account treats properties as sets of concrete objects; without concrete possible worlds, properties would have to be sets of concrete objects and abstract representations of possibilia.

Comments are closed.