March 13, 2004

If You Must

Something in the von Fintel/Iatridou paper on anankastic conditionals ("You must take the A train if you want to get to Harlem")--if you must know, I think I misread an example--made me think of the following semantics for a reading of "If A, you should B":

(1) Given a modal base f and an ordering base g, "If A, you should B" is true if, when A is added to the modal base f, B is true in all the world that maximally satisfy the ordering base g.

What that means, if I'm understanding the Kratzer semantics aright:
According to the Kratzer semantics, a modal is evaluated against two bases; a modal base, which encompasses more or less the facts that are presupposed; and an ordering base, which expresses what is "good" with respect to the modal. To evaluate "Ought p," you look for all worlds compatible with the modal base, see which of those worlds comes out best with respect to the ordering base, and see whether p is true in all those best worlds (there may be more than one).

So (1) translates into English as follows: Take all the things that you presuppose, and presuppose A as well. Then, if things go as well as they can, B will be true.

This doesn't give a good account of anankastic conditionals. Suppose that your actual goal is to be up early tomorrow, and that this goal determines the ordering base--we evaluate what you should do by its conduciveness to your getting up early. Take the anankastic conditional:

(2) If you want to go out with your friends, you should call them.

As an anankastic conditional, (2) is probably true. But on the analysis (1), (2) comes out false; even if we presuppose that you want to go out, the ordering base is your goal of getting up early, and among the worlds in which you want to go out that goal will be best fulfilled in the worlds in which you do not call your friends.

But (1) does seem to be a suitable analysis of another modal:

(3) If you go out, you should come back before 10.

Here, the idea is that, given that you do go out, you will be most likely to get up early tomorrow if you get back before 10. In other words: Of the worlds in which the antecedent is true, the best (with respect to the ordering base) are those in which the consequent is true. That's just what (1) says.

Interestingly, (3) is very close to this:

(4) If you must go out, you should come back before 10.

(4) suggests that it is a bad idea (with respect to the ordering source) for you to go out. (3) is neutral on this question, I think. But what is the role of "must" in (4)? Clearly it's not part of the antecedent; you can't respond "I'm going out even though I don't have to, so your advice doesn't apply."

Perhaps a different ordering source applies to the "must." Given what the addressee wants to do, she must go out; given the ordering source behind (4), she should come back before 10. I don't pretend that this is an answer, I just want to raise the problem?

(And what of "if you must know, I think I misread an example"? There, I think it's plain--it's a biscuit or sideboard conditional. My take on these is that, literally, "if you must know" is an utterance modifier; the information that I think I misread an example is only relevant if you must know. So, literally, it's not asserted unless you must know. But of course your desire to know has nothing to do with the truth of the consequent, so by a sort of implicature the consequent is asserted whether you want to know or not. I may eventually turn this sort of analysis on the anankastic conditionals themselves.)

(BTW: DON'T Google the phrase "sideboard conditional.")

Posted by Matt Weiner at March 13, 2004 03:54 PM