(University of Stuttgart and University of Wales, Bangor)
THE EXTENDED PROJECTION PRINCIPLE AS A CONDITION ON THE TENSE-DEPENDENCY*(* Earlier versions of this paper were presented at the LAGB Autumn Meeting in Cardiff (September 1996), the Bangor Research Seminar (October 1996, February 1998), the University of Stuttgart (October 1996), the Workshop on Subjects, Expletives, and the EPP in Tromsø (June 1997), and at the Research Seminar at the University of Geneva (November 1997). We are grateful to the audiences for comments. We would also like to thank Bob Borsley, Liliane Haegeman, Ursel Luhde, Gereon Müller, Peter Svenonius and Tarald Taraldsen for useful comments and discussions. This paper forms part of a larger research project supported by the British Academy (Grant No: APN 2989) which we gratefully acknowledge.)
The purpose of this paper is to provide a unified account of two seemingly unrelated phenomena: what we call the Asubject requirement,@ which corresponds to the Extended Projection Principle (EPP) of Chomsky (1982, 1995), and the AV2 requirement@. We argue that they both reduce to properties of T, and, arguably, to a single property of T in connection to the position where T is spelled-out. In particular, the Asubject requirement@ arises when T is realised in the IP domain, while the AV2 requirement@ arises when T is spelled-out in the CP domain. In the first case the features associated with AgrS identify the T-dependency, and when AgrS is absent as in some infinitivals identification comes directly from C. The mode in which the subject requirement is satisfied further depends on how AgrS, i.e. the subject position, is parametrically realised (if at all), that is either as a head or a specifier, depending on the morphological properties of the particular languages. In the second case, identification of the T-dependency follows from the presence of an XP in SpecCP. This is the essence of our account of the V2 requirement. We take identification to refer to PF-recoverability of LF-features as these are instantiated by dependency-formation. The structure of the paper is as follows: in section 1 we discuss the similarities between the subject and the V2 requirement. In section 2 we provide a provisional unification of these two requirements under a T-Criterion. In section 3 we provide a discussion of the relation between C and T and outline our ideas about parametric variation which accounts for the principal differences between V2 languages and non-V2 languages. We next consider in more detail the subject requirement and how it is manifested crosslinguitically, depending on the properties of AgrS. Following the same line of reasoning we consider the obligatoriness of XP-fronting in V2 clauses in V2 languages in connection with the realisation of T in C. Finally we consider the implications of our analysis for a number of other phenomena, such as expletives, embedded V2, VSO orders, Long Head Movement in languages like Breton, concluding with some speculations about the origin of V2.
In this paper, we will investigate the nature of two widely-assumed, but little-explicated, constraints on clause structure. These are (i) the requirement that all clauses have a subject, namely the EPP in its standard sense, as introduced in Chomsky (1982:10), and (ii) the requirement that a declarative C which attracts the inflected verb in a V2 language must have a filled Specifier. Henceforth, we will refer to (i) as the "subject requirement" and (ii) as the "V2 requirement". Neither the subject requirement nor the V2 requirement can be straightforwardly explained in terms of selection or checking. Following the standard idea that selection is connected to theta-role assignment, only argumental subjects can be selected, and so expletive subjects are unaccounted for. Chomsky (1995, Chapter 4) proposes that the subject requirement can reduce to obligatory checking...