Non-Monotonic NPI-Licensing, Definite Descriptions, and Grammaticalized Implicatures

Abstract

Non-monotonic Contexts and Negative Polarity ItemsThis paper comes in two parts: the first argues for the thesis that, in English, nonmonotonic contexts generally license NPIs and that standard extensions of the notion of downward-entailment do not adequately explain this.The second part argues that this observation puts pressure on some accounts of scalar implicatures and singular definite descriptions.A downward-entailing context has the property that the replacement of the predicate in the context by a stronger predicate preserves truth.So, for instance, presuppositions aside, the context after "every" in (1) where the NPI "ever" appears is downward entailing.(1) Every person who'd ever been to the bank got infected.It is common ground that weak NPIs, which in English includes "ever," "at all" and non-free-choice "any," are allowed in downward-entailing contexts-intervention effects and locality conditions aside. 1 What is controversial is the status of various environments that, on the surface, appear to be non-monotonic rather than downward-entailing.In English, almost all of these environments license NPIs. 2 I will discuss three proposals for understanding the rather loose licensing conditions of NPIs and argue that none of them adequately capture the facts of non-monotonic licensing.I will suggest that the best account is simply to say that NPIs are licensed in non-monotonic as well as downward-entailing environments, and I will defend this generalization against apparent exceptions.First, here are an assortment of cases where apparently non-monotonic contexts allow NPIs.All I mean by "apparently non-monotonic" is that in each case, intuitively, neither a stronger nor a weaker predicate can replace the original predicate and preserve truth.I do not think all these contexts license NPIs for exactly Many

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 24 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.