Joint work with  Sven Lauer, University of Konstanz


Anankastic conditionals are conditionals of the form in (1) that express a necessary-means-of relation between the complement of the desire predicate in the antecedent and the complement of the modal in the consequent, e.g., in (1) that taking the A-train is
necessary to go to Harlem (in an optimal way).  Conditionals of the same form need not have the anankastic interpretation, e.g., (2) does not express that trying not to think about chocolate is necessary for eating chocolate in an optimal way, but rather it is
used to give advice to the addressee on how to avoid eating chocolate.

 (1) If you want to go to Harlem, you have to/should take the A-train.

 (2) If you want to eat chocolate, you should try thinking about something else.

How do the various constituent expressions combine in (1) to give rise to the perceived interpretation, linking going to Harlem, rather than the desire to do so, and taking the A-train, and how does that differ in (2)? Existing accounts largely center on changing the semantics of the modal in the consequent and have to ensure that the goal in the antecedent `wins out' against any conflicting goals (Saebo 2001, von Fintel &
Iatridou 2005, Huitink 2005, von Stechow et al. 2006). They all treat either the desire predicate or the whole antecedent as effectively vacuous.

But, as we show, the same kind of compositionality problem arises more generally, with conditionals that do not convey anything about the means to an end, such as (3).

 (3) If you want to use the exemption now, you will have to pay more taxes next year.

Analyses that are tailor-made for anankastics cannot generalize to account for (3).

We argue that once we adopt an adequate lexical semantics for "want", the apparently peculiar properties of anankastic conditionals can be reduced to properties that conditionals are known to have independently. Anankastic conditionals are just what they appear to be: regular, hypothetical conditionals.