Brian Weatherson has some comments on Jeff King’s new paper on using syntactic evidence in support of semantic theories. Brian uses NPI-licensing in conditionals as his example. I posted the following comment on Brian’s site, which I repeat here:
Your thoughts here are quite on target. One can take distributional/syntactic facts (NPI-licensing in conditional antecedents) as an argument for a semantic analysis (monotonic semantics for conditionals with additional epicycles). But one can also take semantic evidence (apparent entailment patterns) as an argument against a particular analysis of the distribution patterns (against the Fauconnier-Ladusaw theory of NPIs for example). So, there is a tension here between syntax and semantics, which is precisely why it is necessary to always do both of them: you can’t be a semanticist without knowing a whole lot about syntax, and vice versa. On top of that, it is inevitable that one needs to take pragmatics into account. In the end, this kind of inquiry is part of a complex science and there are a lot of moving parts.
The particular fact of NPI-licensing in conditional antecedents has been a major focus of my own work on conditionals, see my two papers:
Counterfactuals in a Dynamic Context (2001) in Michael Kenstowicz (ed.) Ken Hale: A Life in Language, MIT Press. pp. 123-152. preprint: http://web.mit.edu/fintel/www/conditional.pdf
NPI Licensing, Strawson Entailment, and Context Dependency (1999) Journal of Semantics, 16(2), pp. 97-148. preprint: http://web.mit.edu/fintel/www/npi.pdf
What Brian wrote is about a perpetual and recurrent issue in Linguistic theory.
I have added both your comments and Brian’s post to my blog.
July 26th, 2004, at 1:01 pm #