Are There Possible Worlds?
Possible world realists (Lewis being the one that usually most readily springs to mind) claim that there are any number of worlds just like ours, and that in these parallel universes anything that could have happened has happened. They tend to back this up with the stronger claim that this occurs because this is just what we mean by 'could' - to say that I could win the lottery means that in some (or many) possible worlds I have won the national lottery - and similarly with causal notions and those expressed by 'might' or 'would have'. Conversely, it is argued, to invoke ideas of necessity, that I must buy a lottery ticket if I am to win, means that there is no possible world in which the opposite is true - that no other version of me wins the lottery without first buying a ticket. There are occasions when we use such words to express notion which do not appear to be modal, but this should not be a problem for the claim that very frequently they are used in this way, and could be usefully expressed in new terms with notions of possible worlds.
Lewis argues further that we should be realists about possible worlds - that we should believe they literally exist - because it is in his mind a very useful way of explaining much of our usage of language, our notions of possibility and necessity, as well as those of causation, and he argues that usefulness should be viewed as a viable reason for believing that something is true. In this essay I shall argue against Lewis' (and other possible world realists) position on three different counts, the first of which seems the most important, and which I will spend the majority of the essay on, and two further simpler points: firstly I shall argue that possible worlds realism doesn't always seem to be a terribly useful position to hold; secondly I shall argue that usefulness is not an adequate reason for taking a theory to be literally true; and following on from this I shall thirdly argue that regardless of whether usefulness is a reason to believe a theory to be true, a realist position about possible worlds will never be more useful than an instrumentalist position on the matter, and thus should not be chosen in favour of it.
Lewis is correct to suggest that expressing a concept such as 'I would not have won the lottery had I not bought a ticket' in terms of possible worlds, that 'In the possible worlds where I did not buy a lottery ticket I did not win the lottery', can be a useful way of looking at things. The reason that this is useful is that whilst we may instinctively have some idea of what we mean by a claim that if one thing had not happened something else would, we are actually discussing a rather strange concept. If P did happen, then what do we actually mean by the phrase 'If P had not happened'? - since it of course did. The simple fact of the matter appears to be that such a sentence is simply talking about something non-existent - there isn't such a thing as 'that which occurred when P did not', since nothing occurred when P did not, since P did occur. Lewis' answer to this, which seems very compelling indeed is to say that we are discussing the world where P failed to happen, which is just like our world, except for this non-occurrence. This has the advantage of actually talking about something, rather than nothing, and thus allowing claims of this sort to be correct or incorrect according to actual facts.
It isn't exactly clear what Lewis means by 'just like our world' ; does he mean a world like ours in that it has gravity and people, planets and energy, causation and change? If possible worlds cover every possibility then surely there are ones where there are no people, when planets didn't form, and every where gravity doesn't hold. It isn't clear what things we hold to be so intrinsic to our universe that they cannot be done without - is it a simply necessary fact that there should be energy? Could there be a universe without causation? Could a universe stay the same forever (and would time or 'forever' have any meaning there)? Lewis does not answer these questions, but I do not feel that this has to be a problem for him - it could simply be that other worlds are the same in the sense that they contain things, laws, events, etc., and are more properly considered to be 'the same' in the sense that they aren't objects, or occurrences, units of measurement, ideas, emotions, elements, waves, or any other of the things that the universe contains, they hold the same place as the actual world does for us, not the place held by an actual table, for example.
Lewis' conception, therefore, seems to account for the way we talk, and express modal, causal, and various other notions, but I would argue that this certainly isn't the case for all examples. Consider the complement of my example above, 'If I had bought a ticket with my usual numbers, I would have won the lottery'; this doesn't seem to work as well as the original example. Whilst it might be the case in some possible worlds that if I bought a ticket with my usual numbers I'd have won - and presumably the reason I would make such a remark was that I saw the lottery numbers once they had been drawn, and found that they were my usual numbers, but unfortunately I had forgotten to buy a ticket that week - there are clearly many more worlds where I wouldn't have won, even if I had bought a ticket with my usual numbers, namely the worlds where different numbers had been drawn (it is possible, after all), where my usual numbers were different (perhaps my niece had been born on a different date, for example), and even those worlds where the lottery draw didn't occur at all (though of course I'd have still had to buy a ticket). Lewis does respond to this challenge, clarifying his position, and introducing ideas of 'closeness' of possible worlds. He argues that when discussing causation and the like we are talking about what would happen in the nearest possible world to the effect if the cause had changed, so we must look at the world most like our own to discover what would happen if I had bought a ticket, despite the fact that there are worlds less like our own than those in which my buying a ticket would have resulted in a win. He imagined different worlds spanning out from our own, with the closest being those just like ours, and getting less and less similar the further away they are (Presumably, unlike the rest of his theory, this description should be taken metaphorically, rather than as a literal suggestion about how far away things are in some new dimension.)
It is all very well, however, suggesting that we should take into account closeness of worlds, but this does not, in my opinion, help us a great deal. We can step back to a position of probabilities, where we argue that it is in the nearest possible worlds only that the lottery claim would be true, and that other possibilities do exist - and indeed this would seem to be the case, since clearly the claim 'If I had bought a ticket with my usual numbers I'd have won the lottery' doesn't ensure this fact, so we can't rule out other possibilities; but this doesn't seem to help. If we do retreat to probability claims, it seems we have no way of accounting for the fact that it seems more likely that the win-situation should occur, rather than something else entirely, as would occur in the vast majority of other worlds, since there are always going to be far more circumstances under which you lose the lottery than the particular set of events coinciding to allow a win. Lewis' account does not seem to work quantitatively even if it does qualitatively, and I'm not convinced we could ever formulate the correct weightings to make it useful.
Furthermore, even if we can make claims about the nearest possible world, we cannot be at all sure that they are correct. What if extra trip to the newsagent meant I missed the bus and didn't get home in time for the draw, which meant I never claimed my win and it was invalidated? Arguing that in the nearest possible world where I bought a lottery ticket that should be the only thing to have changed (since otherwise there would be a closer one) doesn't seem to work, since it could be a direct result of the ticket buying that I fail to catch my bus, so the one where I buy the ticket and catch the bus would be further away. In all likelihood, this would make the claim 'If I had bought a ticket with my usual numbers I'd have won the lottery' false, but this doesn't stop it being evidence against realistically construing possible worlds and suggesting it allows you to make claims which you should not. These are perhaps unfair arguments about specific instances, and perhaps not particularly damning, but they do suggest that major problems will occur in such a formulation of such notions.
Another problem seems to be that according to Lewis in the actual world the claim that 'necessarily P' is no different to 'possibly P'. Possible world realism accounts for these notions, but in doing so makes them have no relevance to the actual world, or, in fact (presuming, as everyone appears to, we can have experience only of this possible world) to anything we can know. Consider the claim that 'I can't fly'. This is solely a claim about what I can or can't do in the actual world, at the current time, in the current set of circumstances - but if that is the case it doesn't make a blind bit of difference in terms of the actual world whether it is a necessary or possible truth. Effectively (and this is all we need to care about in terms of usefulness, we can ignore literal truth, especially if we're trying to derive such truth from the usefulness) this, in my opinion, means that 'I could have been able to fly, but can't', means the same as 'I necessarily could never have been able to fly', since all that either of these say about things in the real world is that I can't fly. In terms of usefulness, who cares about anything other than the current world? Yet we seem to be of the opinion that the distinction between these two type of facts, necessary and contingent, is an important one and has a bearing on things in the world which we are in.
Other problems appear when we consider the claim that 'There might be possible worlds'. This seems a perfectly reasonable statement and surely everyone who reads Lewis believes this at some point, since he is such a persuasive writer, yet with his formulation it seems unclear what this could mean. Does it mean that 'In some possible worlds there are possible worlds'? Clearly it can't, and this will be a problem for the theory. Another more sophisticated problem of this sort arises from the claim that 'It is possible determinism holds', but surely if determinism holds then there can't be other possible worlds, since determinism states that the way anything is is the way it had to be!
Even if Lewis is correct, and talk of possible worlds is useful, the next step in his argument, where he claims that a theory being useful is a good indicator that it is correct seems very puzzling indeed. I do not really see on what basis he can claim this since there are any number of theories that we might find useful, but that does not mean that they are correct. It is useful to imagine tiny (i.e. of subatomic proportions) gnomes, which pull things and push them about, producing what we at a higher order of magnitude call 'forces', and yet in all likelihood this is false. I think that usefulness is not a good indicator of truth, even if the two things were to frequently go together - though I'm not convinced that they do.
It is said that that which is true is the best explanation, though (as is frequently pointed out in the literature) this tends to be the case only when we have complete knowledge; in a situation where we do not have all the data we frequently find that one explanation is simpler than another that we would use were we to have a fuller set of facts - such as Newtonian gravity, which was the best explanation for the data they had at the time. Yet clearly the topic of possible worlds is one where we're never going to even approach a full set of facts, and thus we're unlikely to find usefulness (of predication, at least) an accurate guide to truth.
I am not sure what else can be said on the subject, since I simply don't see what other sort of justification Lewis could have, so I shall move straight on to my third objection - that any way in which modal realism is useful, modal instrumentalism will be just as useful. The reason that possible worlds come in useful are for clarifications of what we mean by things, and to explain how sentences which involve things that haven't happened or don't exist work. Clearly however, since we are unable to actually observe possible worlds (and certainly could never survey them all), we are not going to use possible worlds to actually discover whether things are true or not, and whether necessarily or only contingently so. Since the purpose of possible worlds is only for illustration purposes, and cannot be used (even if they would answer the questions if they did exist) to actually answer questions put about modal truths, there seems to be no reason in terms of usefulness to view them as anything more than simply illustrative. If we take an instrumentalist position on possible worlds and claim that they are useful fictions, we are able to illustrate the examples that they work well with, without falling into the various pitfalls of taking a realist position, since we need not explain why the examples which these concepts are not useful for do not work, we simply say that possible worlds are not the best illustration for them.
Overall, therefore, I think there are many reasons why modal realism is not useful, and many examples where if modal realism is invoked it produces very strange results, and various that seem contradictory. Furthermore, I don't believe that even if they were useful that would be an adequate reason to believe in their existence, and if we are taking usefulness as the ultimate arbiter of which theory is correct out of those which cannot be decided other ways, there seems little reason to believe that possible world realism outweighs possible world instrumentalism in terms of utility, and thus we should only ever agree to be instrumentalists on the issue - which seems like a very good idea, since in many cases possible worlds illustrate the point very well!