Back when I was planning to write a dissertation on imperatives, I spent a while thinking about R.M. Hare's putative example of an imperative inference. Hare turned his conclusion into something like the anankastic conditionals that Kai von Fintel and Sabine Iatridou discuss and that I've been discussing; and the conclusion I reached wound up in "Why Does Justification Matter?", so it seems timely to blog it.
(How did I get from a dissertation on imperatives to a dissertation on testimony? Once I decided there were no imperative inferences, the most interesting thing about imperatives was the norms they were subject to. I thought that imperatives were subject to many different normative considerations, whereas assertions were clearly subject to only one fundamental normative consideration, that of truth. In order to explain how assertions were only subject to that consideration, I prepared a short account of how truth was the norm for testimony. After I'd written about 180 pages on the epistemology and norms of testimony, I realized I wasn't going to get to imperatives in the dissertation.)
Hare, in The Language of Morals (p. 35), argues that the following is a good imperative inference:
(1) Go to the largest grocer in Oxford.Grimbly Hughes is the largest grocer in Oxford.
[Therefore:] Go to Grimbly Hughes.
and that (1) can be conditionalized:
(2) Grimbly Hughes is the largest grocer in Oxford. [Therefore:] If go to the largest grocer in Oxford, go to Grimbly Hughes.
saying that in English we write the conclusion of (2):
(3) If you want to go to the largest grocer in Oxford, go to Grimbly Hughes.
There's a few odd things about this--aside from the dubious claim about the logical form of (3).
First is that (3) isn't a conditional where you can detach the consequent given the antecedent; it's an anankastic conditional like "You must take the A train if you want to get to Harlem." Hence I thought that a better way to English the conclusion of (2) might be:
(4) In order to go to the largest grocer in Oxford, go to Grimbly Hughes.
Second is that it makes a big difference what kind of imperative (3) is meant to be. It simply makes no sense to think of (2) yielding (3) or (4) if the imperative is taken as what C.L. Hamblin (Imperatives) calls a willful imperative, such as an order or a request. Even if I have authority to issue commands to you, there is nothing in the fact that Grimbly Hughes is the largest grocer that requires me to issue a command conditional on your wants. The imperative has to be an accountable imperative, such as advice, where the oomph behind the imperative comes from goals the addressee already has or should have, not from the imperative itself. (This means just that anankastic conditionals can't have willful imperatives as consequents.)
Third, and what really got me going, is that exactly similar reasoning should lead from (2) to:
(5) If you want to go to Grimbly Hughes, go the the largest grocer in Oxford.
And (5) seems unfelicitous. Whether this infelicity is along the lines of a true-but-misleading assertion or a false one, I can't say (and I'm not sure there's an answer), but it's definitely odd. Note that it's oddity can't be accounted for by temporal factors--your going to Grimbly Hughes just is your going to the largest grocer in Oxford. Nor can it be accounted for by considerations about whether going to Grimbly Hughes constitutes going to the largest grocer in Oxford, or vice versa. Consider:
(6) If you want to go to the largest building on this block, go to Grimbly Hughes.(7) If you want to go to Grimbly Hughes, go to the largest building on this block.
The solution, I think, has to do with what the addressee knows or can find out how to do--or can do.* When considering (3) and (5), it's most likely that the addressee does not know which grocer is largest but can find Grimbly Hughes in the A-Z. Hence, when she hears "Go to Grimbly Hughes" she can carry it out; when she hears "Go to the largest grocer in Oxford" she can't. With (6) and (7), it's most likely that she can tell which is the largest building on this block but doesn't know where Grimbly Hughes is. Hence she can carry out "Go to the largest building on this block" but can't carry out "Go to Grimbly Hughes" without further help.
So there seem to be two dimensions of felicity for advice. The first is how effective following the advice would be toward achieving the goal in mind. (3)-(7) are all perfectly effective; going to the largest grocer in Oxford is sufficient for going to Grimbly Hughes is sufficient for going to the largest building on this block, and vice versa. The second is how easy it is for the addressee to follow the advice. Here there's a big difference between (3) and (5), and (6) and (7). (The example I use in "Why Does Justification Matter?" is that "In order to bowl a strike, knock down all the pins" isn't very useful advice, unless the addressee can do whatever she wants with the ball but doesn't know what a strike is.)
I'm not sure whether this poses a problem for Kratzerian semantics for the anankastic conditional (see von Fintel and Iatridou's paper). Sometimes it'll be felicitous for the speaker to say "If you want to bowl a strike, hit the 1-3 pocket" or "If you want to bowl a strike, you ought to hit the 1-3 pocket," even though the addressee does not hit the 1-3 pocket in all the worlds in which she achieves the designated goal of bowling a strike. Hitting the 1-3 pocket is just the most doable advice available that is reasonably effective with respect to bowling a strike. But it may be that these cases can be accommodated as false but helpful utterances, as in cases where telling the truth would be misleading.
*Rereading the old old draft in which I discussed Grimbly Hughes, I found this sentence: "It is also clear that knowledge how to do something is not just knowledge that something is true." So there!
Posted by Matt Weiner at April 12, 2004 12:59 PMMighty lonely around here. Mighty boring too.
Posted by: Observer at April 13, 2004 05:00 PM