Essence, Necessity, and Counterfactuals — Epoche B2
Three Theories Often Confused: Modal Realism, Counterpart Theory, and the Semantics of Counterfactuals What is under review David Lewis is standardly charged with two related sins [1] : that his possible worlds make a mess of the difference between what a thing must be and what is merely true of it everywhere [2] , and that his account of counterfactuals is less intuitive than Robert Stalnaker's [3] . The charge is partly right and the diagnosis usually offered is wrong, because three separate doctrines get run together under the name "Lewis's modal realism" [4] . This review separates them and asks which of them the complaint is really about. The three are: modal realism, the claim that possible worlds are concrete entities as real as this one; counterpart theory [5] , the analysis of claims about what a particular thing might have been; and the closest-worlds semantics for counterfactuals [6] . I argue that the famous objection about essence is an objection to the second and not the first [7] ; that the genuine cost of Lewis's position is the inconstancy of de re modality, which he embraces deliberately rather than incurs by accident; and that the actual Lewis–Stalnaker disagreement about counterfactuals concerns conditional excluded middle and is independent of both [8] . A specialist can deny the first of these, and the argument that would do it — Lewis's own — is set out and answered below. 1. Modal realism In On the Plurality of Worlds (1986) Lewis holds that there are many worlds, that they are concrete, and that they are isolated from one another: no spatiotemporal or causal relations run between them. Ours is not distinguished by any intrinsic feature. "Actual" is indexical, like "here": each world is actual relative to its own inhabitants. The motive is reduction. If worlds are what they are on Lewis's account, then "possibly p" says that p is true at some world and "necessarily p" that p is true at every world, and these are claims about what exists — no modal notion is left unexplained. That is the pay-off Lewis is buying, and the price he is willing to pay for it is a very large ontology. 2. Counterpart theory Counterpart theory is a separate doctrine, first set out in 'Counterpart Theory and Quantified Modal Logic' (1968). It concerns claims about a particular individual: not "there might have been a winner of the 1968 election other than the one there was", but "Humphrey might have won". On Lewis's analysis, nothing that exists at another world is identical with Humphrey. What makes the claim true is that some individual at another world is a counterpart of Humphrey — sufficiently similar to him in the respects that matter in the context — and wins there. Two things follow that will matter. The counterpart relation is a similarity relation, and similarity is a matter of respects and degrees; and it is a relation, not an identity, so it need not be transitive or one-to-one. 3. The semantics of counterfactuals The third doctrine is about conditionals of the form "if A had been the case, C would have been". Stalnaker's 'A Theory of Conditionals' (1968) proposes that such a conditional is true if C holds at the world most similar to ours among those where A holds — a selection function picks that world out. Lewis's Counterfactuals (1973) gives a closely related account without the assumption that a single closest world exists. It is worth correcting a common misdescription here. Stalnaker's 1968 paper is a semantics for conditionals; it takes no position on what possible worlds are made of. Stalnaker's own view that they are abstract — ways things might have been, rather than concrete universes — is argued separately, in 'Possible Worlds' (1976) and in Inquiry (1984). Presenting the 1968 paper as the rival ontology is an error, and it makes the Lewis–Stalnaker contrast look like a dispute about concreteness when it is not. The Humphrey objection belongs to (2) The best-known complaint against Lewis is Saul Kripke's [9] , in Naming and Necessity (1980). Humphrey, Kripke objects, could not care less whether someone else, however similar to him, wins an election in another world; what he cares about is whether he might have won. Counterpart theory, on this objection, changes the subject: it substitutes a claim about a similar person for the claim we wanted to make about this one. Whatever its force, this is an objection to the analysis of de re modal claims, not to the thesis that worlds are concrete. The two come apart in both directions. An abstractionist may adopt counterpart theory — Stalnaker argues in 'Counterparts and Identity' (1986) that the counterpart apparatus can be prised away from Lewis's ontology and put to work by someone who thinks worlds are abstract — in which case the Humphrey objection applies to him too. And Lewis's own view has an alternative available in principle, since a concretist could allow that a single individual is wholly present at more than one world, an option Lewis considers and rejects for reasons of his own rather than by definition. So the review's first verdict: if the essence-and-necessity complaint is the Humphrey complaint, it is not a complaint about concrete worlds, and the standard framing of the dispute as "concreteness versus intuition" misplaces it. The strongest objection to that verdict The strongest reply is Lewis's argument from accidental intrinsics, and it is a good one. Suppose I am wholly present at two worlds, seated at one and standing at another. Shape is an intrinsic property: having it is not a matter of standing in some relation to something else. But then one and the same individual has two incompatible intrinsic properties, full stop. The available escapes are all bad. We can say that shape is really a relation between a thing and a world, which denies that it is intrinsic and makes every ordinary property relational. Or we can say the seated-me and the standing-me are different temporal-style parts, which for worl