Lewis's On the Plurality of Worlds: Modal Realism
David Lewis's 1986 defense of modal realism—the thesis that possible worlds are as real as the actual world—transforming debates about modality and counterfactuals.
PhilosophyAll chapters Reading tools Notes & comments
The reality of possible worlds › How Could the Morning Have Gone Differently?Full chapter · Sections in this chapter1 / 9 ↓
How Could the Morning Have Gone Differently?
Suppose you left home without an umbrella, expecting rain. The sky stayed clear. Later you say it could have rained. Now imagine you also claim that if it had rained, you would have gone back for your umbrella. These statements differ. The first says that rain was possible. The second picks out what you would have done on the supposition that it rained. The second would be hard to defend if the only way for rain to be possible involved a different history in which you never owned an umbrella. A satisfactory theory ought to explain what each statement requires.
David Lewis's proposal in On the Plurality of Worlds (1986), §§1.1–1.3 starts from a literal reading of possible worlds. It assigns a role to worlds in which it rains and lets us compare ways those worlds differ from ours. For Lewis, the other worlds that make this account possible are as real as the world we inhabit. They exist independently of any act of imagining them, and their inhabitants are parts of them in the way we are parts of ours.
You can grant the usefulness of the umbrella story while withholding belief in Lewis's worlds. That gap is where his argument begins. He needs to show that treating possible worlds as genuine entities gives us enough explanatory benefit to justify believing in them.
Where a World Ends
Lewis calls our world an inclusive whole in §1.1. It includes remote galaxies and events far in the past or future. A boundary drawn around a planet or around a moment picks out a part of the world. To find another world in Lewis's sense, you would need something different from a part that had been overlooked. There is no distance from Earth at which you would leave the world merely by continuing to travel.
A world is maximal with respect to the relations that make its parts belong together. If an object stands in the relevant space or time relations to parts of our world, it belongs to that world, however far away it is. Distinct worlds have no such relations between their inhabitants. Lewis holds that events in different worlds cannot causally affect one another either. The account of what unifies a world is more flexible than a requirement that all worlds have the same spatial and temporal structure as ours. In §1.6, “Isolation”, Lewis considers whether a world whose inhabitants are related in time alone could still qualify.
Lewis's modal realism treats our world as one of a plurality. The other worlds exist in the same sense as ours. They are isolated from it, and they do not come into being when we describe them.
A universe with other physical laws or with particles unlike those that occur here is one of Lewis's examples in §1.1. Calling the worlds concrete leaves open what kind of matter they contain. Lewis finds the contrast with abstract too unclear to explain his thesis. In §1.7, “Concreteness”, he considers several ways of drawing the distinction and notes that some other worlds contain perfect duplicates of things we ordinarily call concrete, such as donkeys and protons.
What “Actual” Picks Out
From here, our world is the actual one. From the standpoint of an inhabitant of a different world, their own world is actual. In §1.9, Lewis explains this as a feature of how the word works. “Actual” resembles “here” in letting the location of a speaker determine what the expression picks out. The relevant location for “actual” is the world in which the speaker uses it.
Two people in different cities can both say that it is raining here without saying it is raining in the same place. The difference in reference helps both speakers make a claim using the same words. Similarly, when people in distinct worlds say that they are in the actual world, Lewis sees no competition over which group has correctly identified a world that exists in an especially full sense.
There is room for ordinary restrictions in his account. At home I may say that there is no coffee while referring only to the coffee available in my kitchen. My words leave open how much coffee there is elsewhere. Lewis's simple example in §1.1 makes the same general point about our ability to narrow the scope of ordinary existence claims. We can limit the domain to our world without having made a claim about what exists when the domain extends across worlds. From that perspective, the thesis that only our world exists is substantive. It denies that the broader domain contains any other worlds.
Possibility, Necessity, and a Comparison of Worlds
For unrestricted possibility about how things could have been, the starting idea is straightforward. It is possible that it rains if it rains at some world. It is necessary that it rains if it rains at every world. These are claims about where a proposition is true. They do not say that rain is an object whose physical presence in another universe we can detect. Lewis's discussion in §1.2 also distinguishes restricted kinds of possibility. When we ask whether something could happen under the actual laws of nature, we have imposed a condition on the worlds relevant to the question. Other worlds could still be relevant to a broader claim about what might have happened under different laws.
The umbrella conditional requires another piece of reasoning. The fact that somewhere it rains and you go back for your umbrella is too weak. It leaves open whether the worlds where you do that are relevantly similar to this one. Requiring you to go back for your umbrella at every rainy world would be too strong, since some otherwise very different rainy worlds do not contain anything suitable to stand in for the person who set out from your home. The analysis asks how the consequent fares in antecedent worlds when compared from the standpoint of the world in which we are evaluating the claim.
Lewis presents an idealization in §1.3, “Closeness”. Suppose there is one relevantly closest world in which the antecedent holds. Then the conditional is true when the consequent holds at that world. The selection depends on the antecedent, the background facts and the context of the discussion. In our example the ordinary conditions of your trip might be more relevant to the match than keeping the pavement dry. Lewis uses a different example involving tailless kangaroos to illustrate why the comparison need not change every feature one can imagine changing.
The assumption of a single best match belongs to the illustration. Lewis explicitly entertains ties and the possibility that a sequence of better matches never reaches a closest antecedent world at all. He also considers the idea that a successful antecedent world can be closer than all the antecedent worlds where the consequent fails. This is a comparative condition. It does not require the successful world to be closest among all antecedent worlds. The essential teaching point is that a 'would' conditional makes a more selective demand than the possibility claim with which we began.
Who Could Have Played the Piano?
Suppose your friends say that you could have been a concert pianist. The claim concerns you, and they mean more than that some person somewhere could play the instrument. If worlds are isolated and their inhabitants do not overlap in the ordinary way, how can another world tell us what you might have done?
Lewis's answer is counterpart theory. A counterpart is a distinct individual who can represent you for the purpose of evaluating such a claim. The choice depends on resemblance in relevant respects. For the career example, a person with a similar early life and different later choices might be a better candidate than somebody who has your eye color and little else in common with you. Lewis gives his own example involving Hubert Humphrey in §4.1, “Good Questions and Bad”. A counterpart's electoral victory is supposed to represent our Humphrey as winning, even though the counterpart is a different person. In his earlier counterpart theory paper, §1, he sets out a quantified language with an explicit relation for this role.
The relation helps explain how a modal claim can be about a particular actual individual. The individual remains the subject of the claim. A counterpart's life helps determine how that claim comes out.
It is worth slowing down over the last point. One objection says that the existence of a pianist elsewhere can tell us nothing about what would have happened to you. Lewis can say that the objection switches between the subject of the claim and what is supposed to make it true. The subject here is you, just as the person represented by a wax figure in his illustration is still Humphrey. An opponent can grant that the theory is intended to work this way and still argue that relevant similarity cannot do the job. Asking whether a particular resemblance is enough puts pressure on the theory without assuming that Lewis has confused two people.
A further wrinkle prevents a different oversimplification. In §4.1 Lewis distinguishes a world sharing one of its inhabitants with another world from a mereological sum composed of parts from different worlds. He signals there that he will allow cross-world sums while treating them as oddities, rather than using them to explain how ordinary individuals could live alternative lives. So the point of counterpart theory is not established by the slogan that no entity can have parts in different worlds. It is an account of how a world represents something concerning a person who is not among its inhabitants.
The Payoff and Its Price
Lewis compares the proposed plurality of worlds with the role that sets play in mathematics in §1.1. A theory can enlarge the range of things it says exist while reducing the number of unexplained concepts it needs to describe them. The aim is to systematize more of our commitments using a smaller or more coherent starting vocabulary. The extra entities are part of the cost of buying that kind of economy.
Consider properties as one example of the possible payoff. If we represent a property by the set of the things in our world that have it, different properties could be represented by the same set. Lewis discusses apparently coextensive properties in §1.5, “Properties”. Suppose all the objects in our example that have one feature happen to have another. If there is a possible individual that has only the first feature, the extended collections across worlds will differ. Turning to all the possible instances lets his account distinguish extensions that would coincide when the domain is restricted to this world.
That step does not establish that any pair of distinct properties will have different extensions. If their instances could never come apart, enlarging the domain does not separate the two sets. Nor does identifying useful sets settle all questions about which properties are natural or explain similarities. Lewis addresses a difference between plentiful set-like properties and natural properties in the same section. What the example shows is how the additional worlds solve a particular problem for an extension-based analysis, under the condition that the two extensions were only accidentally the same.
Lewis treats systematizing benefits as a reason to believe his thesis, while granting that they could be outweighed or achieved more cheaply. In §2.8, “The Incredulous Stare”, he also gives common sense weight. A theory can lose credibility by discarding too many beliefs we already have. Its gains must compensate for that loss. A reader might give greater weight than Lewis does to the common belief that what exists is exhausted by the actual world. That is a disagreement about the costs of a theory and how much credit its benefits deserve.
Could There Have Been Nothing?
One way to test the account is to ask whether there could have been nothing at all. This question concerns absolute absence. A world with no people, or a world with empty space and no matter in it, would still contain something. In §1.6, “Isolation”, Lewis points out that a world understood as a maximal whole of its parts cannot serve as a container from which everything can be removed while the container itself remains. Even a minimal world would have something to make up that whole.
This produces a real consequence for the account of possibility. If there is no world at which nothing exists, then the analysis does not allow that there could have been nothing. A reader who takes the possibility of absolute absence to be clear may count this consequence against Lewis. The disagreement exposes a pressure on the theory of worlds itself. It is easier to see the problem once we distinguish absence of the familiar items in a domain from the absence of any domain members at all. Changing the contents of a domain and eliminating absolutely everything raise different questions.
Could a Representation Do the Same Work?
An alternative can accept talk of possible worlds while insisting that there is just one concrete world. On that approach, the other “worlds” are abstract ways of representing alternatives. We already distinguish a town from a map representing it. Someone could propose that a possible rainy morning is represented by a description rather than consisting of a concrete episode in a wholly isolated world. Lewis examines versions of this strategy in chapter 3, “Paradise on the Cheap?”.
The map analogy catches the motivation but leaves a task to complete. We have descriptions of countless scenarios, including inconsistent ones. If we keep only descriptions of possible situations, we still owe an account of what makes those descriptions qualify. A purely linguistic proposal has to specify its representational rules. It also has to explain whether the language can describe possibilities involving properties for which no actual speaker has an expression. A rival theory may have an answer, perhaps by treating some modal notions as fundamental. That changes which resources we are counting on the two sides. Lewis's question is how far a rival can match the range of analyses without taking on other costs.
Return to the umbrella. Whether the morning might have gone differently is one question. How to reason about what you would have done is another. Whether the best explanation of such reasoning commits us to isolated worlds is a further question. It is possible to think Lewis gave us a valuable way to distinguish the first two while disagreeing with his answer to the third. The book asks how to support that disagreement with an account that can do comparable work.
Further Reading
- David Lewis, On the Plurality of Worlds (1986). The public excerpt of §1.1 is a good place to start. Read §1.3 for the closeness discussion and §4.1 for the example of Humphrey.
- David Lewis, “Counterpart Theory and Quantified Modal Logic” (1968). The initial formal presentation makes clear how a counterpart relation figures in the treatment of modal claims about individuals.
Join the discussionAsk a question or share what you learned.
Keep exploring
Reference
Citation details
Cite or share this article.

Comments
Add to the conversation
A question, a correction, or an example from your own work.
No comments yet. Be the first to share your thoughts!