Topic: HUMAN SEMANTICS
The Quantificational/Referential Distinction and Negative Polarity
By Daniel Rothschild
There is an interesting class of expressions, including ever, any, and at all, called negative polarity items. They can only be used in certain linguistic contexts. We speak of such contexts as "licensing" the use of these terms. Some standard accounts of which contexts license negative polarity items (henceforth, NPIs) are inadequate. Here I will briefly discuss the problem with these accounts and propose new licensing conditions. Typically NPI's are thought to be licensed only in downward-entailing contexts (DE). I argue that they are rather only licensed in non-upwardentailing contexts. Then I give a semantic characterization of these contexts (non-UE contexts) in terms of domain-sensitivity. This proposal, I take to be roughly in line with some other proposals in the literature [Chierchia, forthcoming].
It turns out that the success of this account, or any account like it, requires examination of various questions about the semantics of noun-phrases. Definite descriptions, particularly, seem to provide a counterexample to my proposal for NPI-licensing. In order to handle this I examine a class of non-Russellian semantics for definite descriptions.
I then argue that NPI's indicate a fundamental semantic distinction between different forms of noun phrases. This distinction is meant to capture the intuitive distinction between quantificational and referential nounphrases. However, which noun phrases count as which is quite surprising.