Existential bare plurals: From properties back to entities

Post-carlsonian analyses of existential bare plurals have basically followed three lines of investigation, analyzing existential bare plurals as names of kinds, weak indefinites or properties. I show that none of the existing proposals can explain the behavior of bare plurals in Romance languages. I argue that we need to combine a revised version of Diesing's analysis of existential bare plurals with Steedman's analysis of indefinites and with Chierchia's neo-carlsonian analysis of kind-denoting bare plurals. The theoretical novelty of the paper is a denotational definition of Milsark's distinction between weak and strong indefinites. Both types of indefinites denote entities, which nevertheless differ regarding their domain: weak indefinites, and in particular existential bare plurals, refer to sums, which are derived entities that relate to each other by the part-whole relation, whereas strong indefinites refer to primitive entities that belong to an unordered set.