The notion of a possible world is familiar from Leibniz's philosophy, especially the idea – parodied by Voltaire in Candide – that the world we inhabit, the actual world, is the best of all possible worlds. But it was primarily in the latter half of the twentieth century that possible worlds became a mainstay of philosophical theorizing. In areas as diverse as philosophy of language, philosophy of science, epistemology, logic, ethics and, of course, metaphysics itself, philosophers helped themselves to possible worlds in order to provide analyses of key concepts from their respective domains. David Lewis contributed analyses in all of these fields, most famously, perhaps, his possible worlds analysis of counterfactual conditionals (Lewis 1973). But these analyses invoking possible worlds cry out for a foundation: how is all this talk about possible worlds to be construed? Do possible worlds exist? If so, what is their nature?
David Lewis responded boldly: this talk of possible worlds is the literal truth. Lewis propounded a thesis of modal realism: the world we inhabit – the entire cosmos of which we are a part – is but one of a vast plurality of worlds, or cosmoi, all causally and spatiotemporally isolated from one another. Whatever might have happened in our world does happen in one or more of these merely possible worlds: there are worlds in which donkeys talk and pigs fly, donkeys and pigs no less “real” or “concrete” than actual donkeys and pigs.