https://lling.univ-nantes.fr/medias/photo/image-seminaire-synsem-last_1634558540342-JPG
-
Le 16 mai 2024 de 14:30 à 16:30false false
Nina Haslinger (ZAS)
Title: Homogeneity and exhaustification under embedding
Much recent work in the semantics of definite plurals has focused on the closely related properties of homogeneity, a truth-value gap phenomenon that results in strong interpretations in upward-entailing contexts (i) and weak ones in downward-entailing contexts (ii), and imprecision, a form of context-dependency driven by the implicit QUD (iii).
(i) The switches are on. (salient reading: 'All the (relevant) switches are on.')
(ii) I don't think the switches are on. (salient reading: 'I don't think any of the relevant switches are on.')
(iii) Context: A machine comes with four on/off switches. For the machine to work properly, all of them have to be on. To safely clean the machine, all of them must be off.
A: Is the machine running? B: Well, the switches are on... (salient reading: all of them are on)
A: Can I clean the machine? B: Well, the switches are on... (salient reading: some of them are on)
Experimental work on homogeneity and imprecision in embedded environments (Križ & Chemla 2015, Augurzky et al. 2023) has given rise to a debate as to whether they are sui generis phenomena or a special case of a general exh(austification) mechanism that also accounts for scalar implicatures and free choice. On the one hand, their embedding pattern shows a striking parallelism with that of scalar implicatures, which existing non-exh-based theories (e.g. Križ 2015, Križ & Spector 2021) do not account for. On the other hand, approaches based on extending an existing theory of exh to homogeneity (see e.g. Bar-Lev 2021, Guerrini & Wehbe 2024) do not capture the full range of semantic flexibility found in the experimental literature. For instance, the approach of this type that I would consider most promising -- the pex approach developed in Guerrini & Wehbe (2024) following Bassi et al. (2021) -- does not capture the finding that plurals show much more imprecision under *not every* compared to *no*, and needs to be supplemented with extra assumptions to explain the rarity of "intermediate" truth value judgments (reflecting a homogeneity gap) under *no* compared to other embedding operators, and the observation that there are certain special discourse contexts in which imprecision is available even under *no*.
In this talk, I will first go over the experimental evidence leading to this dilemma and then propose a new way out: Instead of taking an existing theory of exh as my starting point and extending it to definite plurals, I will take a non-exh-based approach to plural homogeneity and imprecision as my starting point for developing a new approach to exh. Specifically, the starting point will be Križ & Spector's (2021) supervaluationist approach, which assumes that i) the compositional system maps a plural sentence to several "candidate interpretations" (CIs) obtained by varying the values of certain parameters, and ii) the truth and falsity conditions of such a sentence in a given context are determined by those CIs that are relevant to the QUD without being overinformative. I combine this theory with a global constraint according to which a relevant CI p can block another CI q if p globally entails q, but its derivation involves less local strengthening than that of q. Given this constraint, which receives independent motivation from other imprecision phenomena, large parts of the subtle projection pattern of imprecision and homogeneity can be accounted for. Since Križ & Spector's framework was not intended to extend to standard exhaustification phenomena, this raises the question of how the parallels between homogeneity and scalar implicatures can be explained. I address this question by providing a new implementation of exh in a supervaluationist setting.
Mis à jour le 07 mai 2024.