A critical assessment of exhaustivity for Negative Polarity Items

Abstract

In some recent works on negative polarity, exhaustivity is posited as the single defining property of all negative polarity item (NPI) and free choice item (FCI) paradigms. Chierchia (2006; 2013), and Chierchia & Liao (2015) are the best-known implementations of this theory. They stipulate that all NPIs and FCIs must be exhaustified, and posit a covert O(nly) and a syntactic feature [+Σ] to derive exhaustification and licensing respectively. In this paper, I challenge the exhaustivity hypothesis and find it, after careful empirical investigation, to be inadequate to explain the distribution and interpretation of NPIs in Greek, Korean, and Mandarin, which have been described in the literature as non-exhaustive. We also find the theory to be unable to derive the actual distribution of any in nonveridical contexts. Analytically, the problems with exhaustification are twofold. First, the use of covert O(nly) fails to account for why NPIs are licensed. Licensing is a grammaticality condition, and in order to capture it the syntactic feature [+Σ] is stipulated, NPI-licensing thus amounting to checking the [+Σ] feature. The stipulation of [+Σ], without a coherent characterization of its semantics, is a regression to a Klima-esque (1964) syntactic account, and faces precisely the challenges that that account faced. Second, for any variant of the Chierchia system to work for the data discussed here, the system built around it must posit additional ad hoc rules on a case-by-case basis. This produces a system with very little predictive power beyond each specific case because of the ad hoc nature of the rules posited. Our overall conclusion will be that the exhaustivity hypothesis, as formulated in the works discussed here, is a falsified, therefore unnecessary, hypothesis for NPIs.

From Latin QUO(D) VELLES to Romagnol Cvël: A Case of Degrammaticalisation from a Free‐choice Indefinite to the Noun ‘Thing’1

“…This suggests either that widening does not necessarily correlate with strength, or that widening and strengthening (in whatever version) do not predict the right kind of polarity sensitivity» for NPIs, an insufficiency which also extends to an analysis of negative indefinites as intrinsically scalar or exhaustified items (cf. Giannakidou 2018). In other words, as already noted in fn.…”

Unifying Universal and Existential wh’s in Mandarin

“…By assigning singleton alterantives to Mandarin existential wh's, we depart from Chierchia & Liao who take them to activate subdomain alternatives of the full range, including both singleton alternatives and the large ones (see Chierchia & Liao 2015: (45)-(46)). The departure helps avoid the criticisms raised by Giannakidou & Lin (2016) and Giannakidou (2018) against Chierchia & Liao 2015. Essentially, Giannakidou and Lin criticize Chierchia and Liao for positing almost identical semantics for existential-wh's and English any, and this is problematic since the two have different behaviours (compare (24a) and (24b) in the hide and seek scenario (25)).…”

Rethinking negative polarity and free choice in comparatives: A crosslinguistic perspective

“…Strong NPIs have narrower distribution in the scope of a narrow set of licensers which are nonveridical and negative. There are also additional constraints in NPI distribution due to additional sensitivities of NPIs to properties such as modality or epistemic uncertainty, and such NPIs will prefer nonveridical non-negative contexts; free choice items (FCIs) are polarity items of this kind, see Giannakidou 2011Giannakidou , 2018 for more overview discussion and data from a variety of languages including Korean and Mandarin. Some of the core nonveridical environments for NPIs are summarized in Table 1 below, which includes any, the Greek NPI kanenas and the emphatic variant KANENAS which is a strong NPI licensed only with negation and antiveridical operators.…”

“…While both subject to licensing, as we see, there is a meaning difference between the NPI and the FCI which indicates different lexical semantics. The Greek NPI is not scalar but receives a referentially vague reading (translated above as some or other; seeGiannakidou and Quer 2013;Giannakidou and Yoon 2016) Giannakidou 1998Giannakidou , 2001Giannakidou , 2018. describes how this reading differs from the free choice reading: while both express referential indeterminacy free choice is, in addition, exhaustive giving rise a universal-like reading: John is willing to talk to anybody sounds very similar to John is willing to talk to everybody.…”