Abney (1987) argued for an inflectional head above NP, which came to be called D. This inflectional head realizes possessor agreement, is a functional projection which takes a possessor in its specifier, is a landing position for N to D raising (if such raising exists), and takes VP complements in gerundive constructions. It is effectively a parallel to I or T in the clausal domain.
Szabolcsi (1994) argued for a landing site of WH-Movement above the lexical N head, based on evidence from Hungarian which shows a range of WH-words moving to a left-peripheral position within DP. This position is effectively a parallel to the Spec CP position in the clausal domain.
If these parallels between nominal and clausal syntax hold, then it is important to determine how far the parallel extends. The symmetry must break down somewhere; otherwise there would be no syntactic distinction between nouns and verbs at all, and thus no need for two distinct categories. Finding this break in the symmetry might tell us something about how UG divides up things in the world into different lexical categories.
One of the most important early works on VP Ellipsis, Hankamer & Sag (1976), noticed a distinction between VP Ellipsis and many other anaphoric processes. VP Ellipsis needs to find its antecedent somewhere in the surrounding syntactic environment, while other types of anaphors (pronouns, do it anaphors) can find their antecedents in the surrounding non-linguistic context as well.
Sluicing, noticed by Ross (1969) in `Guess who?', was found to adhere to the same syntactic constraints as VP Ellipsis; specifically, there needs to be a valid antecedent somewhere in the preceding syntactic environment to license Sluicing.
There has been a great deal of research devoted to the correct formulation of the identity condition that holds between the antecedent and the ellipsis site in Sluicing and VP Ellipsis. It is agreed in general that the identity condition must hold of some level of representation beyond surface structure, for example, at logical form. I'll explore how the identity condition of Merchant (1999), a condition on LF implemented at PF, generalizes to NP Ellipsis.
The correct identity condition may tell us something about how UG handles redundancy; if we think of ellipsis as the outcome of a drive to say as little as possible and still be comprehensible, then the different kinds of ellipsis and anaphora give us hints as to the information structure of the discourse (see H&S 1984), and ultimately to the design of UG.
Jackendoff (1971) notices a parallel between ellipsis processes in DP(NP) and CP(S). In his account, both Gapping and VP Ellipsis(VP Deletion) have corresponding processes in DP(NP).
Jackendoff argues convincingly that the two ellipsis processes are distinct from each other and consistent across the nominal/clausal domains. He shows that nominal (N') Gapping, like VP Gapping, is limited to coordinate structures, and that NP Ellipsis(N' Deletion), like VP Ellipsis(VP Deletion), is possible in a much wider range of configurations.
It is interesting to note that bare possessors can accompany material right-adjoined to the NP, even in non-coordinated environments.
Notice that even of Mary, which is most likely a semantic argument of picture, is strandable. This is in line with various proposals (Grimshaw, 1990) that non-deverbal Ns do not have any true syntactic arguments. The process at work in these examples may be an analogue of Pseudogapping.
Lobeck (1995) examines the structures I will call NP Ellipsis, and those I will call bare quantifiers. Her account, however, makes several different choices than mine, relying on a null pro in the ellipsis site and strong agreement features. I will return to it for comparison after giving my analysis.
Kizu (1994) presents an account of NP Ellipsis in Japanese, in a government & binding/relativized minimality theory. She considers stranded possessors of the following type:
|John-no||peepa-wa||nagakatta||ga,||[ Mary-no [ e ]]||-wa||mijikakatta.|
|John-GEN||paper-TOP||was-long||though||[ Mary-GEN [ e ]]||-TOP||was-short|
She analyzes both the syntactic facts and the identity condition on these ellipsis processes as parallel to VP Ellipsis and Sluicing. Although her analysis is strongly linked to government & binding/relativized minimality, the analysis hints that this Japanese construction may be the analogue of NP Ellipsis in English.