Topic: HUMAN SEMANTICS
MORE QUESTIONS ON
ANANKASTIC CONDITIONALS
By Arnim von Stechow, Sveta Krasikova & Doris Penka
In von Fintel's blog I have already made two questions about the paper mentioned above. Here I would like to add other three. Let us consider the following excerpt:
The first extensive discussion of the "anankastic conditionals" is due to (S?b?, 1986), who discovered that these conditionals are reluctant to a standard modal treatment. A number of recent papers on this topic refreshed linguistic interest in this phenomenon and motivated further research in the semantics of modality. To illustrate the problem, let us look at the sentences we will focus on:(1) a. You have to take the A train if you want to go to Harlem.
b. If you don't take the A train you can't go to Harlem.
c. To go to Harlem you have to take the A train.
This paradigm was brought into light by (S?b?, 1986), who followed (Bech, 1955/57) in assuming the equivalence of the conditional in (1a) and the infinitival construction in (1c). The conditionals in this list are called "anankastic", a term due to (Hare, 1971). They have the special property that the truth of the consequent is the only way that guarantees the truth of the antecedent. Or, the consequent is a necessary condition for the truth of the antecedent. While Hare had in mind constructions like (1a), a better construction to make the semantics clear is (1c) with the clause "you to go to Harlem" as antecedent A and the clause "you take the A train" as consequent C. The truth of C is the only way to entail the truth of A.
As I do not see why those claims would be the case, let me present some of my interrogations:
[1]. Firstly, the sentence (1a) is not you go to Harlem, you take train A. The sentence has an if and the verb to want, which make a lot of difference than simply saying you go to Harlem.
This reminds me of the paper by Fodor and Lepore, Why Compositionality Won't Go Away, posted here, containing a reaction to Paul Horwich?s comments. At least some the passages of von Stechow and ali sound like the Horwichian claim:
The if-clause, which contains "want", adds a further condition, which does not have much impact on the truth condition, if it has any impact at all.(..)
Note that this paraphrase ignores the contribution of "want"(...) In his lecture notes, von Stechow (cf. (Stechow, 2004)) proposes that the "want" in the antecedent is empty at the logical form. (...)
Want does not contribute to the meaning of the sentence.
The following passage from the same article sounds like a contradiction to the claims above:
The presence of want/be to in the antecedent is obligatory, but as S?b? (2001) accurately observes, the subject of want must corefer with the subject of the matrix clause for the anankastic reading to obtain (...). This requirement on coreference/disjoint reference suggests that want/be to, whatever their semantic contribution is, see to it that the necessary referential relations are established. So the presence of these modals is essential for the [anankastic] reading to be available.
What their exact position is? Do they claim that a verb like to want is empty in such constuctions or that its presence is obligatory?
Anyway, in order to ignore the contribution of a verb like to want in the if-clause it is necessary to abolish compositionality and the difference between extensional and intensional contexts. But was the abolition of such notions among the goals of the authors when they wrote the paper?
Now, if one takes the example (5a), it is visible the contrast between adding and not adding the verb to want to the if-clause:
(5a') i. If you want to have sugar in your soup, you should call the waiter.
ii. If you have sugar in your soup, you should call the waiter.
I do not want to be meanie, but perhaps these points should be made clearer in their paper.
[2]. Secondly, why the propostion you have to take the A train would have to be true to make you want to go to Harlem true also? Suppose that you have to take the A train is false. Then it does not mean that you do not want to go to Harlem.
The same applies to other passage (page 3):
Conditionals of the form (1a) are called anankastic conditionals. This is a sort of conditionals with the consequent expressing necessary condition for achieving the goal or wish contained in the antecedent. Thus, the if-clause always has a bouletic/teleological modal expression and the matrix clause an explicit necessity operator. Here is an example from (Bech, 1955/57):(4) Wenn M?ller mit Schmidt verhandeln will/soll muss er nach Hamburg fahren.
`If M?ller wants/is to negotiate with Schmidt he must go to Hamburg'
Sentence (4) means that the only way for M?ller to negotiate with Schmidt is to meet him in Hamburg. Note that this paraphrase ignores the contribution of "want" to which we will return below.
This is not true either. Even if M?ller does not want to negotiate with Schmidt, still (C1) may be the case:
(C1) The only way for M?ller to negotiate with Schmidt is to meet him in Hamburg (Germany).
But assume that (C1) is not true. Assume that rather (C2) is the case:
(C2) The only way for M?ller to negotiate with Schmidt is to meet him in New Hamburg (Brazil).
Now (C2) does not cancel the truth of the clause M?ller wants/is to negotiate with Schmidt.
We could even imagine a third alternative. Assume that Schmidt never talks about work when he is Hamburg. In such case, the sentence (4') below is not awckward at all:
(4)If M?ller does not want to negotiate with Schmidt, he must go to Hamburg.
[3]. Finally, does the classic maxim of deontic logic (CM) guide their assumptions? I get the impression that it does:
(CM) O(A)-->A
If A is obligatory then A is the case
(CM) is indeed what underlies the reasoning below:
(R) (You must take the A train to go Harlem) --> (You take the A train to go to Harlem)
But (CM) is not accepted in deontic logic anymore. And (CM) is not only false in Logic. It is clear to me that (CM) does not apply to human languages either.
To be continued (...)
Posted by Tony Marmo
at 00:01 GMT
Updated: Sunday, 5 December 2004 04:52 GMT
