The syntax of Complementisers
The aim of this thesis is to provide a detailed analysis of certain syntactic properties of Complementisers (C), formulated within the Minimalist framework (Chomsky 1993, Brody 1993). In particular, I discuss three types of syntactic phenomena where the C position is crucially involved. In chapter 1 I provide a brief discussion of the theoretical framework and an overview of the data. In chapter 2 I discuss that-t phenomena. In languages like English subject extraction from a position adjacent to C yields ungrammaticality (the that-t effect). This is due to a violation of the Empty Category Principle (ECP). I argue that an analysis distinguishing between proper and non-proper head governors (cf. Rizzi 1990) cannot hold within the minimalist framework. Assuming that that can be an expletive (Lasnik and Saito 1984, Law 1991a&b), there are two possible chains to be formed: one by moving I-to-C (C is realised as zero) and the other by coindexing C and I (C is realised as that). In long subject extraction, I must move to C so that it c-commands the subject trace. In subject relatives, on the other hand, there is subject short movement. Thus I and C can only form a chain under coindexation. If I moves to C, then a configuration is created where both the Operator and its variable end up in the same minimal domain. Thus the Operator cannot be 'ordered' with respect to its trace and the result is ungrammatical. I call this the Ordering effect. In chapter 3 I discuss factive complements. In particular, I consider factivity as a property of the complement clause, specifically of its C. I argue that C in factives bears some feature specification which: (a) derives the semantics of factive complements and (b) blocks adjunct extraction. I argue that the different locality properties of factive complements in Modern Greek (MG) and English can be captured on the basis of different features on the C head. In particular, MG possesses a special C for factives (pu vs. the nonfactive oti) which, I argue, is characterised as [+definite]. The strong islandhood of MG factives is then attributed to definiteness in the same way that definite NP's are opaque to any kind of extraction. As for English, I assume, following Hegarty (1992b), that C is specified for a F(amiliarity) feature. The operator status of a [+F] C is enough to block adjunct extraction only. Finally, in chapter 4 I discuss the subjunctive. With respect to MG I argue that the empty C triggers movement of the na+V complex (an instance of I-to-C movement) according to the principle of Full Interpretation. I assume, following Manzini (1994b), that the subjunctive I is licensed by a sentential operator and forms a dependency of the (Op,... ,I) type. Epistemic predicates license the subjunctive because they can be implicitly modal (Veloudis 1985). Moreover, I argue that the presence of an expletive T with epistemic modality allows for independent time reference in the na-clause. With respect to the phenomenon of disjoint reference in Romance, I argue that this is due to the presence of an expletive C, while in the Balkan subjunctives coreference is possible due to the availability of I-to-C movement.