(1) [the [ syntax [ of [ some [ apparent [ complements [ to N ]]]]]]]
where the arguments of the nominal heads are introduced as complements. I will argue that at least PP arguments of Ns are introduced much higher up in the structure. The main argument is cross-linguistic and relies on the following PP Peripherality generalization:
(2) In a structure where an N is modified by an intersective AP and takes a PP complement, the unmarked order is one where the AP intervenes between N and PP (so we have [PP AP N] and [N AP PP]).
I motivate this generalization mainly on the basis of data from Celtic, Polynesian, Semitic and Romance (all head initial languages), but show how it also holds for a range of head final languages (Indo-Aryan, Dravidian, Japanese, Turkic, etc).
The standard view makes this generalization difficult to derive. What is required is that the surface position of the PP is high. But then the question is why? Standard views of the syntax/semantics interface (as well as of the linearization properties of specifiers) would suggest that the PP moves to a high position (e.g. Kayne 2004). I argue instead that it is Merged in a high position and that its semantics are negotiated directly by functional structure (cf Borer 2005, Ramchand 208 etc for `complements to V). I propose that the linear order of this high PP with respect to the N arises because the syntax allows part of the extended projection of N to be generated as a high left daughter, as well as as a lower right daughter, of the head in whose specifier the PP sits (cf. Brody and Szabolcsi 2003, Adger, Harbour and Watkins 2009). This gives us:
(3) [ the [ [ alternative syntax] [ [ of [ some [ apparent [ complements ]] [ [ to N ] ]] ]]