Page 379 - Encyclopedia of Philosophy of Language

worlds' are, the theses (i), (iii), and (iv) are rather harmless. It is only in combination with thesis (ii), which postulates a super universe where our world is found among other worlds, that they yield a strong form of realism. Note that on this view the actual world is not identified with the totality of all there is, as is commonly done. The actual world is one out of many worlds which differ in what goes on in them but are all of the same kind.
Robert Stalnaker (1984: 44-50) argues for what he calls a 'moderate realism.' He holds against Lewis, not that there may be modal facts which might never be known of, but rather that on Lewis's view all modal facts are in principle unknowable:
Presumably, any part of reality that is spatially or caus- ally connected with something in the actual world is itself part of the actual world— But if other possible worlds are causally disconnected from us, how do we know any- thing about them?
(Stalnaker 1984:49)
Most people find it reasonable to hold that there are countless ways in which things might have been. How- ever, the reasonableness of this view is hard to come by if it depends on the existence of worlds which are farther removed from us than the remotest corners of our universe. The problem is how talk about such alien entities could play a role in the semantics of ordinary everyday speech.
This argument against Lewis strongly depends on the assumption that the worlds in the super universe are spatially and causally disconnected, and hence inaccessible. But a modal realist does not seem to be committed to such a universe. Perhaps, it is more like the universe used in the many-worlds interpretation of quantum mechanics, developed by Everett and Wheeler in 1957, where the worlds are paths through a many-dimensional branching structure (a popular account is in Gribbin 1985). Branching occurs when there are different possibilities; a point which is nicely illustrated by means of Schrddinger's cat paradox. (Lewis 1986: 124-25, discusses an ethical variant of this example invented by Larry Niven.)
Imagine a nontransparent box with a live cat in it beside a fully random device which might break a bottle of poison. Due to the randomness of the device one is totally ignorant of whether or not the cat is still alive, unless one opens the box. Now, before observing the cat, is it alive or dead? On one interpretation the states of the cat are undetermined, for equally probable. The poison is definitely emitted and the cat definitely dead only the instant one looks in the box and finds the cat in that state. Before that moment the living cat and the dead cat are equally probable and therefore equally 'unreal.' On Everett's interpretation both cats are real but located in different worlds. There are no nonactual possibilities which become actual through observation. All possibilities exist, though in different mutually incompatible worlds.
In Everett's interpretation there is a universe in which the worlds may be spatially or causally related to each other. Yet from within a world all other worlds are inaccessible. What is important here is that this inaccessibility, so crucial for Stalnaker's argument, does not prevent one from reasoning about the non- actual possibilities. In particular, the reasonableness of holding that there are many ways in which things that might have been can be sustained even in the absence of immediate access to the alternatives them- selves. What is required is, rather, some insight into how the alternatives come about.
It may also be concluded that one should resist taking the entire structure to be the actual world; a world that happens to have many world-like parts. It might appear that such a move is mainly termino- logical, but this is not so. For on the new interpret- ation, all world-like parts are actual in an absolute sense. As a consequence, the desired contingency is absent because there are no alternatives and hence nothing to vary. Contingency just occurs if the bran- ches themselves are the worlds, each one actual for its inhabitants.
4. Abstract Worlds
Only a few philosophers feel comfortable with the modal realism defended by Lewis. They say that his position, which takes possible worlds as concrete par- ticulars, is induced by an unfortunate terminology. The use of 'possible world' suggests the possibilities to be the wrong kind of entity. Nobody mistakes the euphoric state of John for John himself. Similarly, a state the actual world might be in should not be con- flated with that world itself. The world is a particular, the largest one possible, but its states are abstract 'ways things might have been.' Therefore it would have been better, if they had been called 'possible states,' 'histories,' or 'counterfactual situations' (Kripke 1980: 20).
On this view, perhaps more needs to be said about the states and how they relate to the unique concrete world. There are some proposals that try to do so. Most of them are variants of Carnap's 'state- descriptions' (Carnap 1956: 9), where possible worlds are maximal consistent sets of basic sentences (a basic sentence is either an atomic sentence or the negation thereof). In this case a possible world is a structured object built from basic sentences by set-theoretical means. A possible world is related to the real world via the basic sentences. It is actual, yet abstract, if and only if all its basic sentences are true (they correspond to the atomic facts in the real world).
As Lewis (1986: §3.1) points out, there are some objections to this approach, only some of which can be countered. For one, it must be assumed that every- thing has a name. Otherwise some individuals cannot be represented within a possible world, so that not all
Possible Worlds
357