The proposition 'if Mary had not left, the meeting would have ended differently' is, in ordinary speech, a sentence whose truth conditions we feel competent to judge. The philosopher who looks closely at the sentence, however, finds it harder to specify than its everyday use suggests. The sentence asserts something about a state of affairs that did not occur — Mary's not leaving — and infers from that non-occurrence a different outcome. The asserting and the inferring are both, on inspection, peculiar.
David Lewis's account, articulated in the 1970s and refined over the next two decades, took the philosophical question seriously and proposed an answer of unusual scope. A counterfactual conditional is true, on Lewis's view, if the consequent obtains in the possible worlds nearest to our own in which the antecedent obtains. The notion of nearness is technical: it ranks possible worlds by their overall similarity to actuality, with priority given to similarity in matters of particular fact, and lesser priority to similarity in matters of general law. The proposal had the merit of supplying clear truth conditions for sentences whose ordinary use had felt unclarified.
The proposal had, equally, costs. To say that other possible worlds enter into the truth conditions of an ordinary English sentence is to commit, on a straightforward reading, to the reality of those worlds — a commitment Lewis himself embraced under the heading of 'modal realism.' The view that possible worlds are as real as the actual world, differing from it only in not being ours, was, and remains, one of the more striking metaphysical positions in twentieth-century philosophy. Its defenders argue that it is the price of an otherwise satisfactory semantics; its critics argue that the price is too high, and that any semantics requiring it must be wrong somewhere.
The dispute has not been settled in the decades since. Various deflationary alternatives — taking possible worlds as abstract objects, as linguistic constructions, as fictions employed only as if they referred to anything real — have been proposed by Lewis's successors; each preserves much of the formal apparatus while paying a noticeably smaller metaphysical bill. None has secured the consensus that Lewis's original proposal also failed to secure. What survives is a clearer picture of what an adequate semantics for counterfactuals would have to provide and, in particular, a clearer picture of the costs each candidate account must pay. The clarity is itself an advance, even if the question remains open.
The primary purpose of the passage is to: