Consider the following sequence of conditionals.
(i) If I get a cat, I’ll be happy.
(ii) Of course, if I get a cat and my landlord kicks me out, I won’t be happy.
This is a
This talk is about the theoretical advantages that can be gained when we analyze (i) and (ii) as merely true for the purposes of the conversation.
Variably strict theories of conditionals (Kratzer 1981, Lewis 1973, Stalnaker 1968, and others) assume that (i) and (ii) are acceptable only if they are both strictly true. This can only happen when (i) and (ii) quantify over disjoint sets of worlds: (i) quantifies over the closest worlds where I get a cat, while (ii) quantifies over the closest worlds where I get a cat and my landlord kicks me out, and these two domains need not overlap.
This can’t be the right theory for Sobel sequences with epistemically possible antecedents. This is because variably strict theories only allow the two domains to be disjoint when the antecedents are false (Willer 2017). But intuitively, the sequence (i); (ii) leaves open whether I get a cat, and whether my landlord does anything.
Following Križ (2016), I argue that Sobel sequence conditionals can be felicitous without being true. Instead, they can just be true enough relative to the current Question Under Discussion. The proposal entails that conditionals are exception-tolerant, but the exception-tolerance of conditionals is not so exceptional. If we analyze if-clauses as plural definite descriptions of worlds (Schlenker 2004), the pragmatic principles behind Sobel sequences turn out to be the very same principles that explain exception-tolerance with pluralities of individuals (plural definites) and pluralities of times (temporal adverbs and habituals).