Sabine Iatridou and I have a new (rough!) draft paper:
- Kai von Fintel and Sabine Iatridou: 2004. “What to Do If You Want to Go to Harlem: Notes on Anankastic Conditionals and Related Matters”.
These notes deal with a side track in a larger project (which is going to issue in our paper in progress “Anatomy of a Modal”, which is what I will be talking about at UConn on April 9 and at SALT in May — Sabine will present the same material at GLOW in Greece in April), but we thought it had enough independent interest to warrant a separate exposition.
We discuss problems raised by what G.H. von Wright called anankastic conditionals:
bq. If you want to go to Harlem, you have to take the A train.
Kjell Johan Sæbø discusses these in his paper “Necessary Conditions in a Natural Language”:http://vivaldi.sfs.nphil.uni-tuebingen.de/%7Earnim10/Festschrift/Saeboe-8-komplett%20fertig.pdf. We show that his analysis does not quite work. We go through three alternative analyses. Along the way, we address some other questions of interest.
Check it out and please send us any comments you might have.
Update: Matt Weiner has “a post on our paper”:http://mattweiner.net/blog/archives/000146.html.
Update: Matt Weiner has “another post on our paper”:http://mattweiner.net/blog/archives/000150.html.
Update: Matt Weiner has “yet another post on our paper”:http://mattweiner.net/blog/archives/000162.html — This one in particular is very thought-provoking.
Good stuff! I will try to make some comments over at my blog. One little question: Why does (22) come out false on a single modal analysis? It seems as though it should be true through falsity of the antecedent.
March 11th, 2004, at 7:21 pm #(22) is the sentence:
“If jaywalking is illegal in this town, that guy over there has to be punished.”
By a single modal analysis, we mean an analysis where the “if”-clause serves as a restriction to the deontic necessity modal “has to”. So, the modal base would be made up entirely of worlds where jaywalking is illegal. But the ordering source would be based not on what the law says in the worlds in the modal base, but on what the law says in the actual world (where jaywalking is legal). So, the worlds in the modal base that are best according to the laws as they actually are, are such that the guy does not get punished. Hence, (22) is incorrectly predicted to be false.
I assume that you have an analysis in mind where the modal is embedded in the consequent of a material conditional. In a way, that is in fact a double modal analysis, it’s just one where the higher structure is one that has the meaning of a material conditional. That is not the analysis we are rejecting for (22). What we are saying doesn’t work is a single modal analysis.
I realize that most of the paper is as of now very much theory-internal, presupposing — rather than gently introducing — the Kratzerian system for modality. We’ll have to see whether we can fix that.
March 11th, 2004, at 7:42 pm #I have one thing to say, which is based on some empricial observations:
‘If… so’ constructions in natural languages differ from logic ‘a-b’ for reasons that need to be investigated within a clear framework.
As in the case of (A) pressuposes that there is a man called Socrates, sentences like a-b in natural languages usually pressupose that a and b are not both false if the whole sentence is true. Otherwise, being both false the resulting larger sentence is interpreted as false, contrarily to what one gets in formal logic languages. The exceptions to this are ironic comments. Cf (B) and (C) bellow:
(A) Socrates wrote no book. |= Socrates existed.
(B) F(If F(Socrates was Austrian) then F (he spoke only German)).
(C) If Giselle Bundchen is ugly then it rains Pepsi light in Japan. (irony)
Notice that some natural language sentences, which have the if… so format, are interpreted as ‘iff’. Imagine a situation where a father tells his children this:
(D) If you break anything in the house again, you will not have icecream for the rest of the year.
The children in that situation will understand that if they do not break anything, they will have icecream. This interpretation is the normal in everyday life, although it is not what one logician would expect in the case of a plain ‘if… so’ proposition. Accordingly, if the father in the situation comes back and sees nothing broken, but says ‘you will not have ice cream for the rest of the year’, he may act as logicians could predict, but not as any lay person would deem reasonable.
March 12th, 2004, at 8:33 am #Antonio,
you are certainly right about the rather dim prospects for an analysis of natural language conditionals in terms of material implication. But both logic and natural language semantics have moved on well beyond the material conditional analysis. So, you are probably beating a dead horse. (Although there are some, like Frank Jackson, who try to maintain an analysis of conditionals as material implication, supplemented with a pragmatic theory. See Bennett’s recent book “A Philosophical Guide to Conditionals” for convincing argumentation against the Jacksonian approach and also for an overview of current ideas about the meaning of conditionals. He does not however spend any time on deontic conditionals, and certainly not on the kind of conditionals that we treat in our paper).
Our paper is situated within a comprehensive and systematic theory of modality and conditionals (due to Angelika Kratzer), which is quite successful in at least approaching the “real” meaning of natural language constructions. What we deal with are some quite recalcitrant cases that create prima facie problems within the Kratzerian framework.
On your last point: I discuss the frequent interpretation of “if” as “if and only if”, a phenomenon known as conditional strengthening, in a manuscript predictably entitled “Conditional Strengthening: A Case Study in Implicature”.
March 12th, 2004, at 8:43 am #Antonio,
one more thing. There are in fact uses of conditionals that fit the mold of material implication. Peter Suber has some such examples on a nice page about material implication: http://www.earlham.edu/~peters/courses/log/mat-imp.htm.
If you promise me that “If I am healthy, I will come to class”, it would appear that you have not broken your promise (which one might equate with the truth of the conditional), if it turns out that you are sick but drag yourself to class anyway (false antecedent, false consequent). In fact it seems that the promise is only broken (that is, the conditional is only false) if you are healthy but do not come to class — which is precisely what the material implication analysis would predict.
This is part of a family of examples of behavior consistent with the material implication analysis, which in the end need to be accounted for within any more sophisticated analysis that one might want to adopt. See for example, Bob Stalnaker’s old paper “Indicative Conditionals”, where he argues that the inference from “p or q” to “if not p, then q” (clearly valid for material implication), while not logically valid in his semantics for conditionals can be argued to be a “reasonable inference” in a precisely definable sense (involving notions of pragmatic felicity). That is, he argues, why we can conclude from “Either the butler did it or the gardener did it” that “If the butler did not do it, the gardener did”.
March 12th, 2004, at 8:54 am #Yeah, I know it is a ‘dead chicken’, as one says in
Portuguese. But there are still many scholars who insist
in saying that (22) or even (B) above are true because the
antecendent and the consequent are false, as predicted
by traditional logic. (Many more than one can suspect). They even explicitly evoke logic to defend this view.
So I think it is important to remember these things once
in a while.
But thanks for the reference.
Off topic: This is the same case as ‘adverbs’. During
March 12th, 2004, at 11:16 am #the time of the XVIIIth, a French linguist wrote that
‘in spite of its name, adverbs do not modify only verbs’.
I have found dozens of authors that, in the XXth century,
began their papers on adverbs with statements like that.
The worse part is that many SYNTACTIC papers still give
deitics, quantifiers and prepositional phrases as examples
of adverbs.
PS:
I mean, in the case that in a possible world
March 12th, 2004, at 1:12 pm #where jaywalking is legal and jaywalkers are
not punished, so that (22) would be true anyway
under a classic interpretation.
Thanks. That single-modal analysis really does seem hopeless–the condition clearly has to affect the ordering base rather than the modal base.
A historical note: In 1952, in The Language of Morals p. 35, R.M. Hare has a discussion (which I think is quite confused) of imperative inference, and claims that one can infer from
Grimbly Hughes is the largest grocer in Oxford
to
If go to the largest grocer in Oxford, go to Grimbly Hughes
claiming “In English, we write this conclusion in the form:
If you want to go to the largest grocer in Oxford, go to Grimbly Hughes.”
This seems to me like an early citation of something like an anankastic conditional.
And: How do you accomodate sentences of the form “If you want X, do P”? This seems as though it raises the same issues–is your view that “do P,” taken as advice, creates an ordering relation much like “you ought to do P”?
March 12th, 2004, at 3:16 pm #Matt,
Thanks for the Hare reference. We cite the later Hare 1971 but haven’t actually looked at that (we just took it from Sæbø).
We have no informed opinion on imperatives and on conditional imperatives. That’s certainly something one should look at. My friend Paul Portner has some relevant ideas, see his extended handout at http://www.illc.uva.nl/MnM/Activities/esslli03/Portner.pdf.
March 12th, 2004, at 3:32 pm #Sorry for posting this so late, but I only just got around to writing it up.
I think there’s a problem with the idea that “If you want to X, you must Y” should be read as “If you want to X, you must Y [in order to X],” where “If you want to X” is added to the modal base and “You must Y in order to X” designates X as the goal that dominates all others in the ordering base. This is that the addition to the modal base sometimes improperly imports information about the subject’s desires into the epistemic possibilities.
Consider the following case: You and I know that Joe has been considering buying a used car in Harlem, but we do not know whether he has bought the car yet. If Joe has already bought the car, then he can get to Harlem either in the car or on the A train; if he does not have the car, then the A train is the only way for him to get to Harlem. The only reason why Joe would want to go to Harlem is to buy the car. If Joe wants to go to Harlem, he has not yet bought the car.
I say: (1) If Joe wants to go to Harlem, he must take the A train.
It seems that (1) is true only if Joe has not yet bought the car. If he has bought the car, then (if he wanted to go to Harlem, which he wouldn’t) he can either drive or take the A train.
But on your reading of anankastic conditionals, I think, (1) comes out true in all circumstances. By adding “Joe wants to go to Harlem” to the modal base, we exclude those epistemic possibilities in which Joe has not yet bought the car. Then, in all the remaining epistemic possibilities, the worlds in which Joe goes to Harlem are all worlds in which he takes the A train. So the best worlds, ordered by the designated ordering base, are all worlds in which he takes the A train.
In short, once we’ve restricted the epistemic possibilities to those in which Joe wants to go to Harlem, “Joe must take the A train in order to go to Harlem” comes out true.
My suggestion is that, if you want to read “If you want to X, you must Y” is elliptical for “If you want to X, you must Y in order to X,” then you should treat the expanded reading as a biscuit conditional (analogous to “If you want biscuits, there are some in the sideboard”). The antecedent of the conditional expresses the usual precondition for “You must Y in order to X” to be relevant, but does not modify what is actually asserted. (At least, that’s how I think of analyzing biscuit conditionals.) If Joe doesn’t want to go to Harlem, it usually doesn’t matter whether he has to take the A train to get there–he’s not going to try to go.
March 29th, 2004, at 2:28 pm #