The structure of clitic doubling in Modern Greek

Abstract

A. Clitic Doubling (Cl-d) in MG exhibits three interesting properties (1) Cl-d requires the doubled expression to be interpreted as old (1 )a ο Yiorghos tin-aghapa tin Maria (II) There is an asymmetry between Cl-ded NPs versus Cl-ded QPs: Cl-ded QPs may not take antecedents internal to the clause immediately containing the doubled QP Doubled NPs face no such restrictions. (2a,b,c and d) are representative examples: (2) a. i Tzeni to y -malose [to kathe pedhi] y c i Tzeni Uypulise to vivlioj,. b. pioSj ton-aghorase ton kathe pinaka apo to xorio tu^.J d. pioSj to-pulise to vivlio tuj (III) Although Ns, Ps and Vs in MG can host pronominal clitics which correspond to their arguments, only the complements of Vs can be clitic doubled. C. Noticing that NPs in the clause initial position occupied by Clitic Left Dislocated (CLDLed) NPs (see (3)) must also be interpreted as old information, I propose the representation of Cl-d in (4): a topic, which is a copy of the Cl-ded expression, occurs in the clause initial CLLD position; the topic is the subject of a predication, the doubling clitic serves as predicate variable: (3) tin Maria ο Yiorghos tin-aghapa (4) U QP/NP, U U cliticj . QP/NP, . ]]] In Cl-d, a PF rule deletes the topic copy In CLLD, a PF rule deletes the doubled expression in argument position (4) captures that there is a common reason for the old information interpretation in Cl-d and CLLD constructions: the constructions have virtually identical representations. The data in (II) confirm that the relation between the argument position and the CLLDed position cannot be one of movement: this would create a contradiction for the analysis of the NP/QP asymmetries. If only doubled QPs, but not doubled NPs, were to raise to topic position, we could attribute their non-cooccurrence with clause internal antecedents as due to a failure of c-command: doubled QPs raise outside the c-command range of any clause internal antecedent But if doubled NPs were also raised to the topic position, they should not be able to contain bound pronouns or other expressions requiring clause internal antecedents either, contrary to fact It must be that there is not blindly movement from the Cl-ded argument position to the CLLDed topic position: a non-movement analysis is supported instead. Finally, I will demonstrate how the data in (III) support the idea that the relation between the clause-initial topic and the rest of the clause is predication.