Conceptual restrictions on weakly referential constructions: Evidence from modification

In this dissertation I investigated the restrictions on so-called weakly referential constructions such as weak definites, bare predicates and bare singular nouns. Weak definites are definite constructions in which the definite doesn’t refer uniquely, such as in (1):
Context: Piet spent his evening reading The Guardian, De Volkskrant, and Le Monde.
(1) Piet spent his evening reading the newspaper.
Bare predicates are nominal predication constructions in which the indefinite article is absent, like in (2):
(2) Gaaike is kunstenaar.
Bare singular nouns are singular noun phrases that occur without an article (see the Catalan example in (3)).
(3) Té apartament.
has.3sg apartment
‘(S)he has an apartment.’ (I.e. (S)he is an apartment-owner)
These constructions share several properties, one of which is that they don't set up discourse referents in the usual way, but rather refer more abstractly to roles, properties, etc. Hence the term weak referentiality. Not just anything can be expressed using a weakly referential construction: there are certain restrictions, which are related to cultural and contextual back ground knowledge.
I used adjectival modification as a tool to bring together the linguistic and conceptual aspects of these restrictions. Conceptual in the sense that I used adjectival modification to systematically alter the denotation of nouns, linguistic in the sense that I hypothesized that the conceptual requirements would be linked to the lexical semantics of the various types of adjectives I used. The reasoning behind this hypothesis was that based on observations that can be found all over the literature, conceptual stability is a core property of weakly referential constructions -- that is, they refer to things, roles, functions, properties that are perceived as conceptually stable. The notion of stability also emerges when you look at the lexical semantics of certain types of adjectives. For instance, the interpretation of stage-level adjectives is situation dependent. Similarly, the interpretation of evaluative adjectives depends on whose opinion is expressed: these adjectives are judge dependent. The interpretation of kind-level adjectives, on the other hand, doesn't involve any such dependencies. The lexical semantics of color adjectives involves some dependency but is relatively stable compared to that of stage-level and evaluative adjectives. Linking the notions of conceptual stability and semantic stability, I formulated the stability hypothesis:
The more stable the interpretation of an adjective is, the more acceptable it will be in a weakly referential construction.
I tested this hypothesis in several languages, by collecting native speaker acceptability judgments of sentences containing weakly referential constructions on several modification conditions. The results of my work in this dissertation show that the stability hypothesis has been confirmed. The research I presented here provides the first structural empirical evidence that supports the intuition that constructions such as weak definites, bare predicates, and bare singular nouns require the concepts they denote to be stable in some sense. Using modification as a tool, I showed that something as intangible as conceptual requirements can still be traced back to semantics, showing a link between linguistics and the conceptual level.