When we consider extended projections of different lexical categories, a couple of key questions that arise in syntactic theory are how many layers of structure are projected in an extended domain and whether the number of projections in the extended domain of a particular category is the same or different cross-linguistically. While lexical categories themselves and overt, morphologically free functional elements like the definite and the indefinite article in English are generally taken to project their own phrases in the syntax, it is less clear whether all morphologically bound overt functional elements project separate functional structure, as well as whether in languages that lack overt manifestation of a particular functional head, which has overt manifestation in other languages, a null counterpart of that head is still present in their vocabulary inventory or if the relevant head and the corresponding functional projection are missing altogether. Regarding the nominal domain, one view in the literature (referred to as the NP/DP parameter) is that languages with articles project a DP layer, while those without articles allow bare NPs in the lack of other overt functional heads (Corver 1992; Zlatić 1997; Bošković 2005 et seq; Marelj 2008). Under this view, languages with affixal articles like Bulgarian, Icelandic and Romanian are typologically classified with languages with non-affixal articles like English and assumed to always project a DP.
In this talk, I will discuss a number of phenomena in the adjectival and the nominal domain where languages like Bulgarian sometimes do indeed pattern with English, but crucially sometimes also pattern with languages without articles, which separates them out as a third type of language. Comparing the behavior of a number of languages in the two domains, I will argue that the amount of structure in the extended domain of N and A is parallel within a language, while it can vary cross-linguistically. Thus, languages that require functional structure in the domain of N also require functional structure in the domain of A, but functional structure in these two domains is not required universally. I also address some predictions that this view combined with a contextual approach to phases has with respect to extraction possibilities out of bare NPs and APs in languages like Bosnian/Croatian/Serbian.