ABSTRACT
The standard semantics for counterfactuals comes from Stalnaker (1968, 1981), Lewis (1973, 1986) and Kratzer (1977, 1981). Abstracting away from differences between the accounts, the basic semantics is as follows:
A would-counterfactual P □→ Q is true (at w) iff all the closest
P-worlds (to w) are Q-worlds.
Traditionally, the semantics for counterfactuals is thought to have a limited kind of context-sensitivity. That is, most of the time, there is a consistent way in which the closest P-world(s) are selected (relative to the world of evaluation); what counts as the closest P-world does not vary based on the conversational context. There are some exceptions to this. Lewis discusses Quine’s famous case of Caesar in Korea, arguing that in some conversational contexts (1) is true, while in other contexts (2) is true, and this depends on what facts are being held fixed in the context:If Caesar had been in command in Korea, he would have used the atom bomb.
If Caesar had been in command in Korea, he would have used catapults.
Stalnaker (1981) endorses the idea that in some contexts, different worlds count as equally similar due to negligible differences (whereas those differences might not be negligible in other contexts). For example, in a case in which we have a line in the margin of a book that is actually just less than an inch, and are considering counterfactuals about how long it would be if it was more than an inch long, worlds in which the line is a little more than an inch up to worlds where it is 2 or 3 inches might count as equally similar (suppose the margin is 3 inches wide).