Strong versus weak negative polarity and two types of exhaustification: the case of exceptive ʔilla in Levantine Arabic

Abstract Exceptive constructions, headed by particles such as ‘but’ and its counterparts in other languages have their truth conditional meaning built into semantic inferences that govern their semantic construction and distributional use (von Fintel 1993. Exceptive constructions. Natural Language Semantics 1. 123–148). Based on an observed analogy between exceptive phrases and other negative polarity items NPIs (see Gajewski 2013. An analogy between a connected exceptive phrase and polarity items. In R. Eckardt, L. Mingya & M. Sailer (eds.), Beyond ‘any’ and ‘ever’ new explorations in negative polarity sensitivity , 183–213. De Gruyter Mouton), a (multidimensional) exhaustification analysis, which has been standardly used to derive the compositional and distributional facts of other NPIs (e.g., Chierchia 2013. Logic in grammar: Polarity, free choice, and intervention . Oxford University Press), has been extended to account for similar facts concerning the NPI nature of exceptives (Gajewski 2008. NPI any and connected exceptive phrases. Natural Language Semantics 16. 69–110; Sauerland and Yatsushiro 2023. Domain size matters: An exceptive that forms strong negative polarity items. In S. Zheng & S. Laszakovits (eds.), The size of things II . Berlin: Language Science Press). This paper presents a novel instance of a puzzle involving the NPI-like behavior of exceptive ʔilla in Palestinian/Jordanian Arabic: exceptive ʔilla has a paradigm of NPI distribution in which the weak NPI-like ʔilla, which combines with universally or universally-interpreted quantified expressions, forms a strong NPI-like expression with a domain broadening effect as occurring in arbitrary pro-subject sentences. Drawing on a theory of NPI distribution which attributes the strong-weak distinction among NPIs to different types of grammatical exhaustification (Zeijlstra 2022. Negation and negative dependencies . Oxford: Oxford University Press), the paper solves the puzzle using a unified exhaustification-based analysis for exceptive ʔilla with its weak and strong NPI-variants being derived by means of two different types of exhaustification: for the default case of weak NPI-like ʔilla , an instance of pragmatic exhaustification applies which typically captures weak NPIs and weak NPI-like expressions such as exceptive ‘but’ in English. As for the strong variant of ʔilla, an instance of syntactic exhaustification applies which captures other strong NPIs, including the strong NPI variant ʔilla that establishes a syntactic Agree relation between an exhaustivity operator and a Focus-marked quantificational domain variable D+ which triggers subdomain alternatives that associate and are quantified over by the agreeing exhaustivity operator in question.