University of Southern California Dissertations and Theses

The syntax of clitic doubling in modern Greek

(USC Thesis Other)

The syntax of clitic doubling in modern Greek

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be fr'om arty type of conqjuter printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard m arginc, and inq>roper alignment can adversety affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note w iH indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overl^s. Each original is also photographed in one exposure and is included in reduced form at the back of the book. Photograq)hs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6 " x 9" black and white photographic prints are available for aity photographs or illustrations ^>pearmg in this copy for an additional charge. Contact UMI directty to order. UMI A Bell & Howell Information Company 300 North Zeeb Road. Ann Arbor. M l 48106-1346 USA 313.'761-4700 800.321-0600 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. THE SYNTAX OF CLITIC DOUBLING IN MODERN GREEK by Patricia Schneider-Zioga A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Linguistics) December 1994 Copyright 1994 Patricia Schneider-Zioga Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. DMI Number: 9601057 Copyright 1994 by Schneider-Zioga, Patricia All rights reserved. UMI Microform 9601057 Copyright 1995, by DMI Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES, CALIFORNIA 90007 This dissertation^ written by P a t r ic ia S ch n e id er -Z io g a under the direction of h??.... Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillment of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean o f Graduate Studies Date N oysm ber 29, 1994 DISSERTATION COMMITTEE / l \ L I Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ACKNOWLEDGEMENTS I’m very happy to have finally reached the point where I could write this part of my dissertation. I’d like to thank the members of my committee for their patience. Joseph is a remarkable teacher and it was always an intellectually exhilarating experience to attend one of his seminars. He makes syntax exciting. I also want to thank Yen-Hui Audrey Li, who was officially my co-chair up until my draft defense when the fact that she was on sabbatical in Taiwan made it technically impossible for her to remain on my committee. She helped me focus my ideas when I was utterly lost. The influence of Joseph’s and Audrey’s ideas can be seen on every page of this dissertation, which owes especially much to their work on Chinese syntax. Jean-Roger was very helpful. His willingness to discuss various drafts of the dissertation with me has greatly enriched my research and his sense of humor has made graduate school a lot more fun. Maria-Luisa was always willing to meet, even on very short notice and consistently gave feedback. Her encouragement and insistence helped me through some difficult times. I also want to thank my outside member, Ed McCaan, for being so flexible in arranging meetings and for treating me like I belonged in the academic community. II Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. There are a number of past and present fellow students here who have really made a difference for me, and Tm glad I can acknowledge their support, friendship and their consistent willingness to discuss syntax. They should count as committee members because my dissertation and research in general owes much to their insightful comments. I want to especially acknowledge Fusa Katada, Abbas Benmamoun, Liliana Sanchez and Jose Camacho. I’d also like to acknowledge the following past and present fellow students: Pablo Albizu, Nancy Antrim, Hiroshi Aoyagi, Valentina Apresjan, Alfredo Arnaiz, Marc Authier, Robin Belvin, Lourdes Casanova, Lina Choueiri, Daeho Chung, Nigel Duffield, Gorka Elordieta, Abdesslam Elomari, Fred Field, Jon Franco, Heather Goad, Elena Herburger, Roland Hinterholzl, Miao-Ling Hsieh, Wes Hudson, Sue Kalt, Ibtissam Kortobi, Laila Lalami, Stephen Mathews, Karine Megerdoomian, Keiko Miyagawa, Mohammad Mohammad, Norberto Moreno, Franz Mueller- Gotama, Phillipe Ngessimo Mutaka, Debbie Poole, Ljiljiana Progovac, Vai Ramanathan-Abbott, Charlotte Rheinholz, Magdelena Romera, Kevin Russell, Suchitra Sadanandan, Debbie Schlindwein (Schmidt), Ingrid Shyu, Jin-Hee Suh, Aaron Sonnenschein, Carolyn Taylor, Maki Watanabe, Shin Watanabe, Virginia Yip, Madeline Youmans and Ke Zou. Ill Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I’d like to thank the following professors for all that they taught me about linguistics and/or being in academia; Elaine Andersen, Pat Clancy, Bernard Comrie, Ann Dunlea, Murvet Enc, Ed Finnegan, Jack Hawkins, Larry Hyman, Hajime Hoji, Osvaldo Jaeggli, Elinor Ochs, Bill Rutherford and Barry Schein. And I don’t want to forget my undergraduate professors at the U of Minnesota, who got me interested in linguistics in the first place: Jerry Sanders, Michael Kac, Mike Hammond (who taught me to pronounce ejectives and implosives), Larry Hutchinson, and U of Minnesota graduate student Nancy Hedberg. I’d also like to thank the community of Greek linguists (either speaking or doing) and subscribers of the Modern Greek Studies Association (MGSA) electronic bulletin board, who were so helpful with judgements and other feedback. The good people in charge of the Freshman Writing Program have helped so much financially and in other ways and I’ve learned so much from listening to my writing students— I’m glad that they have given me that opportunity. I’d also like to thank the administrative staff in our department: Kathy Stubaus, Ricardo Aldape, Linda Culver, and Laura Reiter. They’ve done so much to make everyone’s lives here easier despite the fact that their own work loads have greatly increased due to cutbacks. IV Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Arjiri, you have been so loving and supportive through all of this and have never flinched from giving me your unsolicited advice and academic criticisms. I realize if Cypriot Greek didn’t have clitic doubling we might not even know each other. And I’m grateful that you never followed through on the threatened divorce when I asked for grammaticality judgements the moment I saw you or at two a.m. I dedicate this dissertation to my two little daughters, Eleni and Sofia: May you each find your own unique voice. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. TABLE OF CONTENTS Acknowledgements ............................................................................................ ii Abstract................................................................................................................ x Preface .............................................................................................................. xii CHAPTER ONE: INTRODUCTION 1.0. Historical C ontext.................................................................................... 1 1.1. Overview of the Dissertation................................................................ 2 CHAPTER TWO: THE LICENSING OF CLITICIZATION IN MODERN GREEK. 2.0. Introduction .............................................................................................. 9 2.1. The Distribution of Clitics................................................................... 12 2.2. The Distribution of the Clitic in P P s .............................................. 14 2.2.1. Prepositions That Can’t Host Pronominal Clitics .... 14 2.2.2. Prespositions That Can Host Pronominal Clitics .... 17 2.2.3. Prepositions and CFCs ........................................................ 19 2.3. Toward an A cc o u n t............................................................................... 27 2.3.1. The Predication Requirement on Cliticization............. 28 2.3.2. The Binding Requirement on Cliticization..................... 31 2.3.3 Previous Approaches From the Clitic Doubling L iterature.................................................................................. 34 2.3.3.1. Disjointness and Interpretation of Clitic-Doubled Expressions ................................. 34 2.3.3.2. Disjointness and Licensing of Clitic Doubling in Lebanese A ra b ic .................. 36 2.4. Concluding R e m a rk s............................................................................. 39 CHAPTER THREE: THE DISTRIBUTION AND STRUCTURAL REPRESENTATION OF CLITIC DOUBLING CONSTRUCTIONS 3.0. Introduction ............................................................................................ 40 3.1. The Phenomenon of Clitic Doubling .............................................. 43 3.2. The Licensing of Clitic D oubling...................................................... 44 3.2.1. The Distribution of Clitic Doubling................................. 44 3.2.1.1. Prepositional Phrases and Clitic Doubling . . . 45 3.2.1.2. Noun Phrases and Clitic Doubling..................... 46 3.2.1.3. Verb Phrases and Clitic D o u b lin g..................... 47 3.2.2. Why Only the Complement of Verbs? .............................. 52 3.2.2.1. The Doubling of Subjects .................................. 67 3.2.2.1.1. Deictic and Interrogative V erbs 68 3.2.2.1.2. Exceptional Case Marking (ECM) Constructions ............................... 72 vi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.3. Clitic Doubling and Locality ............................................................. 75 3.3.1. A Predication Account of C l-d........................................... 76 3.3.1.1. Additional Support For a Predication Analysis....................................................................... 82 3.3.1.2. The Predicate V ariab le......................................... 91 3.3.2. A Movement Account of Cl-d? ......................................... 94 3.3.2.1. An A’-Movement A cco u n t.................................... 94 3.3.2.1.1. Other A’-Movement T e s t s ................... 97 3.3.2.2. The Possibility of A-M ovem ent.......................... 99 3.3.3. A Binding Approach to Cl-d? ......................................... 102 3.4. Concluding R e m a rk s.......................................................................... 104 CHAPTER FOUR; CLITIC LEFT DISLOCATION CONSTRUCTIONS 4.0 Introduction...................................................................................... 106 4.1. The Interpretation of Cl-ded Noun Phrases as "Old Information"............................................................................ 107 4.2. The Interpretation of Clitic Left Dislocated NPs as "Old Information"............................................................................ 113 4.3. The Link between Cl-d and CLLD Constructions..................... 117 4.4. The Clitic as T(opic)-Morpheme................................................... 119 4.5. Topics in CLLDed P ositio n ............................................................. 120 4.5.1. The Relative Ordering of Wh-Expressions and S ubjects.......................................................................... 120 4.5.2. The Relative Ordering of Focused Expressions and Subjects.......................................................................... 121 4.5.3. The Relative Ordering of Focused Expressions and CPs ................................................................................. 122 4.5.3.1. Focused Expressions and Wh-Expressions . . 122 4.5.4. Focused Expressions and Complementizers................ 124 4.5.4. Word Order Accounts From the Literature ................ 127 4.5.4.1. A CP Recursion Analysis.................................... 127 4.5.4.2. Tsimpli’s Proposal of a Fixed Order: CP, FP, I P ................................................................ 131 4.5.4.3. Agouraki’s Proposal of a Free Ordering of CP and F P ........................................................... 134 4.6. A Movement Approach to Relative O rdering............................ 140 4.6.1. The Relative Ordering of Topics and Subjects 146 4.6.2. The Relative Ordering of Topics and CPs .................. 148 4.6.2.1. Topics and Wh-Expressions............................... 148 4.6.2.2. Topics and Complementizers............................ 149 4.6.3. The Relative Ordering of Topics and Focused Expressions........................................................... 153 Vll Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.6.3.1. Topics: Adjoined or in Specifier Position?.................................................................... 154 4.6.4. The Relative Ordering of Topic, Focus and Com plem entizer.................................................................... 160 4.6.4.1. Relative Ordering Requirements and Minimality ....................................................... 165 4.6.5. Overview of Relative Ordering Requirements ........... 170 4.6.6. The Relative Ordering FP, CP and the Distribution of NPIs .................................................... 171 4.7. Empirical Evidence for the Existence of a Topic in Cl-d Constructions ............................................................................ 174 4.7.1. Topics in Cl-d Constructions............................................. 174 4.7.2. Distribution of Clitic Left D islocation.......................... 177 4.8. Concluding R e m a rk s ............................................................................ 180 CHAPTER FIVE: ASYMMETRIES BETWEEN CLLD AND CL-D CONSTRUCTIONS 5.0. Introduction ........................................................................................... 182 5.1. Non-Movement Accounts of CLLD ............................................... 182 5.1.1. Evidence Consistent With the Proposed Structural Representation of Cl-d ....................................................... 183 5.1.1.1. Parasitic G a p s .................................................... 183 5.1.1.2. Weak Crossover E ffects.................................. 184 5.1.1.3. ECM Constructions.......................................... 186 5.1.2. Evidence Conflicting With a Unified A c c o u n t......... 192 5.1.2.1. Some Languages Allow CLLD But Not Cl-d . 193 5.1.2.2. Cl-d is More Restricted Semantically Than C L L D ......................................................... 195 5.1.2.2.1. Animacy .................................................... 195 5.1.2.2.2. Partitives .................................................. 198 5.1.2.2.2.1. Kayne’s Generalization .... 201 5.2. Concluding R e m a rk s ............................................................................ 203 CHAPTER SIX: THE SYNTAX OF CLITIC-DOUBLED QPs AND THE STRUCTURAL REPRESENTATION OF CL-D 6.0. Introduction ........................................................................................... 204 6.1. Constraints On The Clitic Doubling of Certain Specific OPs ........................................................................................... 205 6.1.1. Doubled QP/NP Asym metries................................... 206 6.1.1.1. Bound Pronouns .................................................... 206 6.1.1.2. An Asymmetry with Focused Expressions . . . 210 6.1.1.3. The Impossibility of the Doubling of Reciprocals......................................................... 214 viii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6.1.1.3.1. The Phenomenon ................................. 214 6.1.1.3.2. The A c c o u n t............................................ 215 6.1.2. Against a Raising Approach With Cl-ded Q P s 225 6.1.3. OR by D eletion..................................................................... 230 6.2. Arguments for the Presence of a Variable With Cl-ded QPs . 235 6.2.1. Differences in Local and Non-Local Behavior of Cl-ded Q P s ...................................................... 235 6.2.1.1. Clitic-Doubled QPs and Clausemates............. 236 6.2.1.1.1. Weak Crossover and Cl-ded QPs . . . 236 6.2.1.2. Pronouns In Separate C la u se s.......................... 238 6.2.1.3. Additional Pronoun Binding F a c ts .................. 241 6.2.1.4. An Account of The Absence of WCQ Effects in Cl-d Constructions............................ 245 6.2.1.5. Scope Interpretation of Clitic-Doubled Q P s .............................................. 248 6.2.1.5.1. Wh-Qperator and Cl-QP Interaction 248 6.2.1.5.1.1. An Account Qf The Relative Scope Interpretations .......... 252 6.2.1.5.2. Negation and Cl-QP Interaction . . . 254 6.2.1.5.2.1. An Account of The Scope Relations....................... 257 6.3. Concluding R e m a rk s .......................................................................... 263 CHAPTER SEVEN: CQNCLUSIQN 7.1. Advantages of the Predication A nalysis....................................... 265 7.2. Specificity............................................................................................... 266 BIBLIQGRAPHY ....................................................................................... 269 IX Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ABSTRACT This dissertation investigates clitic doubling constructions in Modern Greek with particular emphasis on understanding their structural representation. The interpretation of clitic-doubled expressions and the distributional asymmetries that exist between doubled NPs and doubled QPs indicate that: (i) clitic-doubled expressions are interpreted like clitic left dislocated (CLLDed) expressions and have essentially the same syntax; and (ii) these two constructions are not related to each other by movement; i.e., a clitic- doubled NP does not move to the position occupied by CLLDed expressions at some more abstract level of representation. Instead clitic- doubled expressions and CLLDed expressions are instances of a predication structure where the CLLDed expression is the subject of the predication and the clitic and doubled expression serve as the predicate variable: (a) Noun PhrasCj [predicate clitiCj-V Noun Phrasej ...] The CLLDed expression is interpreted as a (discourse) topic. The topic is a copy of the doubled expression. Either the topic or the doubled argument may phonetically delete. The difference between a clitic doubling construction and CLLD is simply a matter of whether the topic or the doubled expression is phonetically (but not syntactically) deleted. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. If the CLLDed expression is phonetically deleted, the resulting construction is pronounced like a clitic doubling construction; if the clitic-doubled expression is phonetically deleted, the resulting structure is pronounced like a CLLD construction. These proposals account for what kind of lexical projections allow clitic-doubling, e.g., VPs but not NPs; the impossibility of doubling QPs that contain expressions needing antecedents, as well as the wide scope interpretation of doubled QPs. This approach contrasts with previous approaches in the generative syntax literature which generally suggest either: (a) clitic doubling is the result of the incorporation of a determiner or pronoun related to the doubled expression, or (b) clitic doubling is the result of a rule of agreement between the doubled expression and the relevant morphological host. My study connects clitic doubling and CLLD and suggests that predication is an additional argument building strategy employed by the grammar. XI Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Eg ho dhen exo vghali to sxolio. Lite exo mat hi ghrammata poll a. ksero omos ena ki'ena kanun dhio ke pos ta fo n ien d a ine epta. Toso kairo mazi mu ke dhen exis rnathi ta dhika mu xughia ke ta fisika. I proparallghusa pote dhen perispate otan i lighusa ine makra. -(from a Greek urban folk song) translation: I never finished school. I’m not very literate. But I know that one plus one equals two and that Greek has seven vowels. We’ve been together for so long and you still haven’t learned my ways or the most basic things about me. The antepenultimate syllable never gets an accent when the ultimate syllable has a long vowel. [This rule is cited by the singer as an example of the most basic kind of thing someone should know.] X II Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER ONE: INTRODUCTION 1.0. Historical Context From a generative perspective, our knowledge of the syntax of clitic doubling, where an overt NP and a corresponding pronominal clitic co-occur, owes much to the study of pronominal cliticization in general. In particular, Kayne’s (1975) important work on French syntax framed the discussion of the syntax of pronominal cliticization in such a way that the existence of clitic doubling was an anomaly in need of explanation and therefore, an attractive area of research. Kayne’s interest in accounting for the fact that a pronominal object clitic in French functions like the argument of a verb and as such is in complementary distribution with other nominal objects of the verb led him to the analysis that the clitic pronoun originates in argument position from whence it undergoes movement to its morphological host. Given the correctness of this kind of analysis, we would not expect the existence of a language where a pronominal object clitic and an overt direct object NP co-occur. The actual existence of such languages became a thing of interest in need of explanation. A modified version of Kayne’s analysis of cliticization was able to account for clitic doubling languages. It posited the co-occurrence of a clitic and an empty category in languages like standard French and the 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. co-occurrence of a clitic and an overt NP in languages that allowed doubling. This placed the burden of contemporary research on explaining why languages like French do not allow clitics and overt Noun Phrases to co-occur. More recently, the approach to the study of clitic doubling has shifted. The insight that clitic doubling fulfills a specific discourse function in terms of encoding "old information" (essentially, "specific information" in the sense of Enc (1987, 1991)) was coupled with the recognition that specific NPs have a special syntax (see, for example, Bittner (1986), Diesing (1990), Enc (ibid.), Mahajan (1990), Moltmann (1990)). In addition, it was recognized that many languages reserve syntactic positions just for old or new information, and so the relation between clitic doubling and the syntactic reflexes of specificity has become a focal point for research. It is from this intellectual location that I approach the study of the syntax of clitic doubling. 1.1. Overview of the Dissertation I begin with an investigation of the structural representation of pronominal cliticization and propose that it involves a predication structure where a non-overt pronoun functions as the subject of a predication and the clitic functions as the predicate variable (in Williams’ (1981) sense); 7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (1) prOj [ p r e d X-cl(itic)j proj ] The primary evidence for the proposed structure comes from the observation that the distribution of clitics in Modern Greek is in general quite free, but is peculiarly restricted in Prepositional Phrases. This observation coupled with the observation that clitics can only occur if the minimal maximal projection that contains them constitutes a Complete Functional Complex leads to the conclusion that predication is the relation responsible for these restrictions. In the third chapter I investigate the structural representation of clitic doubling. Standard views of clitic doubling suggest either that it involves some kind of spec/head agreement relation between the clitic and the doubled expression, or incorporation of a pronoun or determiner. In contrast to standard views, I propose that in Modern Greek, Cl-d signals a different syntactic structuring of phrases than that found when non-doubled arguments occur and in fact it is essentially the same structure as proposed for simple cases of cliticization. Specifically, I propose the following representation of clitic doubling: a copy of the clitic doubled expression occurs in clause initial position. This copy is the subject of a predication, with the doubled Noun Phrase and the clitic which doubles the Noun Phrase serving as the predicate variable. The minimal maximal projection containing the clitic and doubled Noun Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Phrase serves as the predicate (cf. latridou’s (1991) proposai for the structure of Clitic Left Dislocation (CLLD)). The following structure is proposed: (2) [ NPj [,p [ip clitiCj....NPj....]]] A PF rule deletes the copy: the copy is syntactically, but not phonetically, present. Initial evidence for the proposed structure of clitic doubling will come from an analysis of the distribution of clitic doubling. I note that we can capture certain facts about the distribution of clitic doubling in a theoretically natural way if we postulate the existence of a structurally higher Noun Phrase that is a copy of the doubled Noun Phrase and thus co-indexed with it. Given this structurally higher NP copy, it can be deduced that an antilocality licensing condition on Cl-d exists such that a Case-marked Noun Phrase must intervene between the postulated NP copy and the doubled Noun Phrase in the relevant domain. The relevant domain is the minimal maximal projection containing the clitic and the doubled Noun Phrase. The relation between the NP copy and the doubled Noun Phrase can best be described by predication. Predication is subject to the following locality requirements: (a) the subject of predication and the predicate must m-command each other (Rothstein (1983)); and (b) the 4 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. predicate must be the minimal maximal projection containing the predicate variable (latridou (ibid.)). A clitic which doubles an argument can be no more deeply embedded with respect to the NP copy than IP, which is the minimal maximal projection dominating the clitic. A doubling clitic cannot be more deeply embedded with respect to the clause initial NP copy, for instance in PP or NP, because this would violate the various requirements on predication (the predicate must be the minimal maximal projection containing the predicate variable, which by hypothesis is the clitic and the NP it doubles). Besides giving evidence in favor of a predication analysis of Cl-d, it is possible to find evidence against other possible accounts such as a movement or Binding approach. In the fourth chapter, I demonstrate that there are additional reasons to believe that a structurally higher coindexed Noun Phrase does exist in clitic doubling constructions. First I demonstrate that doubled Noun Phrases must be interpreted as specific and that, as such, they are related to a structurally higher position that is reserved for specific Noun Phrases; namely, the Clitic Left Dislocation (CLLD) position. Then I attempt to establish the exact structural location of the CLLDed Noun Phrase, which I refer to as the ‘topic’. Once the location of the topic has been established, this knowledge is used to adduce further evidence supporting the proposed structure of clitic doubling 5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. constructions as involving a coindexed Noun Phrase, which we now postulate must be the topic, that is structurally higher than the doubled Noun Phrase: (3 ) NPj ...[ ... clj-V NPj ...] I then suggest that a rule of PF deletion of the clitic doubled Noun Phrase is responsible for CLLD constructions just as a rule of deletion of the topic copy is responsible for clitic doubling constructions. I have suggested that CLLD and clitic doubling involve essentially identical constructions syntactically; yet, in the literature concerning non-movement analyses of CLLD, certain asymmetries between the two constructions have been pointed out. While some aspects of the non- movement accounts of CLLD are consistent with the representation of clitic doubling that I have proposed here, other aspects seem to cast doubt on the idea that the two constructions are essentially identical syntactically. In chapter five, 1 discuss some of the evidence offered in support of non-movement based accounts of CLLD, especially evidence pointing to an asymmetry between the two constructions, and attempt to reconcile them with the proposal I have made here. In chapter six I provide additional support for the proposed structural representation of clitic doubling chiefly through an exploration of the syntax of clitic-doubled QPs. It is demonstrated that 6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. there is a pervasive asymmetry between clitic-doubled QPs and clitic- doubled NPs such that a doubled QP cannot contain an expression that requires an antecedent. An exploration of this issue offers additional evidence in support of the claim that there is not a movement relation between the topic and clitic-doubled position and leads to a discussion of reconstruction effects in clitic doubling constructions. I suggest that when doubled QPs are involved there is no rule of QR operative. Instead, there is a syntactic deletion of the predicate variable that leaves behind a syntactic variable. We can in this way account for the fact that QPs cannot be doubled if they contain expressions that must have antecedents. To see this consider that if there were a syntactic deletion of the predicate variable, binding of any expression contained in the doubled QPwould be impossible because the topic position, where there is a copy of the doubled QP, is structurally higher than anything else in the clause and thus any expression contained in the doubled QP is outside of the c-command range of any clause internal antecedent. Reconstruction effects with reflexive anaphors and bound pronouns contained within doubled NPs in clitic doubling and CLLD constructions are neatly accounted for because the position occupied by the predicate variable is within c-command range of potential Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. antecedents and this predicate variable does not undergo QD for obvious reasons. It can be confirmed that there is a syntactic variable in the predicate variable position in clitic doubling constructions involving clitic doubled QPs when we compare the interaction of doubled QPs with other scope-bearing or anaphoric elements within the clause containing the doubled expression and with other scope-bearing or anaphoric elements in a superordinate or paratactic clause. Finally, we will see that this picture of the syntax of clitic doubled QPs straightforwardly predicts and accounts for the scope interpretation of clitic doubled QPs in their interaction with other QPs. Finally, in chapter seven I point out some of the advantages of this analysis compared to other anlyses of clitic doubling which have been advanced in the literature. I suggest a possible direction for a typology of constructions involving specific Noun Phrases. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER TWO: THE LICENSING OF CLITICIZATION IN MODERN GREEK. 2.0. Introduction The principal issue addressed in this chapter is the structural representation of pronominal cliticization in M(odern) G(reek). Studying the distribution of clitics reveals that cliticization in MG involves predication with a non-overt pronoun serving as the subject of the predication and the clitic functioning as the predicate variable (in Williams’ (1981) sense): (1) PrO j [p red X-cl(itic)j e.c.j ] The conclusion reached here contrasts with standard views in the literature that clitics are simply agreement markers, or incorporated pronouns or incorporated determiners. Clitics are found coindexed with the arguments of verbs, nouns and prepositions. Although cliticization is quite productive, there is a curious restriction: while it is not possible to cliticize the arguments of prepositions which govern reflexive anaphors, it is possible to cliticize the arguments of prepositions that disallow reflexive anaphors (cf. (2a,b) with (3a,b)): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (2)a. *apo-ton from-cl.acc.3s b. apo ton eafto tu from the self his "from himself" (3)a. dhipla-tu next-cl.gen.3s "next to him" b. *dhipla (se)ton eafto tu next (to) the self his We will see that the interaction of two sub-theories of the grammar can account for this restricted distribution of clitics. These sub theories are: (a) the theory of predication, and (b) the theory of binding. Suppose that reflexives are forbidden in the above environments because to appear in those environments would mean to violate principle A of the Binding Theory: a reflexive must be bound in the Complete Functional Complex (CPC) containing it, but there is no antecedent available in those CFCs, so reflexives cannot occur in those environments. Given the above, the data suggest the following generalization: 10 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (4) a clitic is licensed only if the minimal maximal projection containing it constitutes a CFC. The generalization in (4) singles out the domain of ‘minimal maximal projection’ and connects it to CFCs, a domain relevant to Binding; therefore, the generalization reveals that cliticization is sensitive to a certain property of the minimal maximal projection containing the clitic. This sensitivity can be explained if we view cliticization as involving syntactic predication (see the structure in (1)). It can be argued that predication must adhere to these locality requirements: (a) the subject of predication and the predicate must m- command each other (Rothstein (1983)); and (b) the predicate must be the minimal maximal projection containing the predicate variable (latridou (1991)). These requirements indicate that predication observes very strict locality and, in particular, predication is a relation that singles out the domain: minimal maximal projections. We can account for why predication involving pronominal clitics cares about the status of the predicate involved as a CFC given the following antilocality condition: (5) ANTILOCALITY CONDITION ON CLITICS (ACC): a clitic must be free in the minimal CFC containing it. 11 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Under the locality condition on predication that the predicate must be the minimal maximal projection containing the predicate variable, when the argument of a PP is cliticized, the PP must be the predicate. Cliticization is ungrammatical with PPs that do not constitute CFCs because it would violate the ACC: the clitic, bound by the non-overt subject of the predication, would not be free in the minimal CFC containing it (the minimal CFC would be a larger domain such as the entire clause). Cliticization is grammatical with the arguments of PPs that constitute CFCs because the clitic is free in the minimal CFC containing it (namely, the PP). Cliticization with the argument of a verb or noun is licensed because the minimal maximal projection containing the clitic: IP and NP respectively, always constitutes a CFC. Therefore, in these cases, the clitic always satisfies the ACC. 2.1. The Distribution of Clitics Since we are able to deduce the structural representation of pronominal cliticization chiefly by studying its distribution, I begin with a brief overview of the general distribution of pronominal clitics. I will focus on simple cliticization, by which I mean a construction where a pronominal clitic occurs which corresponds to an empty argument position. 12 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The following kinds of arguments can cliticize. The object and indirect object of the verb can cliticize to the verbal complex: (6)a. tin-protimo cl.acc.f.-prefer.Is "I prefer her." b. tis-tilephonisa cl.gen.fem-telephoned.ls "I called her (on the telephone)." The argument/possessor of a noun can cliticize onto the noun: (7) i fotoghrafia-tis the picture-cl.gen.f "the picture of her" "her picture" The argument/possessor of a noun can also cliticize onto an NP internal adjective: (8) to kalo-tis palto the good-cl.gen.f coat "her good coat" The arguments of certain prepositions can cliticize onto the preposition: 13 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (9) brosta-tu (in).front-cl.gen.m "in front of him" 2.2. The Distribution of the Clitic in PPs Pronominal clitics can correspond to the object of a preposition, but not every preposition can host a pronominal clitic. We will see that it is possible to make the following descriptive generalization about the distribution of pronominal clitics within prepositional phrases in MG: (10) within prepositional phrases, clitics and reflexive anaphors are in complementary distribution 2.2.1. Prepositions That Can’t Host Pronominal Clitics The simple prepositions: se ‘to’, jia ‘for’, me ‘with’, and apo ‘from’ can govern the reflexive anaphor, ton eafto tis ‘herself’; but they cannot host pronominal clitics. The following examples illustrate the fact that the simple prepositions listed above can govern a reflexive: (11)a.i Eleni edhose to vivlio s’ton eafto tis the Eleni gave.3s the book to’the self hers "Eleni gave the book to herself." 14 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b.i Eleni to-ethiakse yia ton eafto tis the Eleni cl.acc.n-fixed.3s for the self hers "Eleni made it for herself." c.i Eleni to-aghorase apo ton eafto tis the Eleni cl.acc.n-bought.3s from the self hers "Eleni bought it from herself." Clitics are disallowed in this same environment as the following examples indicate: (12)a. * s(e)-tin in-cl.acc.f b. * yia-tin for-cl.acc.f c. * me-tin with-cl.acc.f d. * apo-tin from-cl.acc.f It is not the clitic’s status as a pronoun which bars it from this environment because pronouns can, in principle, occur here. So, although clitic pronouns are barred in these contexts, it is perfectly grammatical for a strong pronoun to occur as illustrated in the following examples. 15 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (13)a.i Eleni edhose to vivlio se aftin the Eleni gave.3s the book to her "Eleni gave the book to her." b. i Eleni to-ethiakse yia aftin the Eleni cl.acc.n-fixed.3s for her "Eleni made it for her." c. i Eleni dhen pezi me aftin the Eleni not play.3s with her "Eleni doesn’t play with her." d. i Eleni to-aghorase apo aftin the Eleni cl.acc.n-bought.3s from her "Eleni bought it from her." In the following chart, I give an overview of the generalizations for the prepositions we have discussed so far. I will refer to these prepositions as "^e-type prepositions." Cf.: 16 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (14) NOMINAL EXPRESSIONS WITHIN se-type PPs: expression is licensed within se-type PPs: REFLEXIVES CLITICS STRONG PRONOUNS YES NO YES 2.2.2. Prespositions That Can Host Pronominal Clitics In contrast to the behavior of xc-type prepositions, when we consider prepositions that allow pronominal clitics as their objects, we see that they prohibit reflexives as their objects. This is illustrated for the prepositions dhipla ‘next’, and brosta ‘in front’. Cf.: (15)a.o Ksenophon kitai dhipla-tu/*(s’)ton eafto tu the Xenophon is.looking.3s next-cl.gen.rn/(to) the self his "Xenophon is looking next to him/to himself." b. o Ksenophon idhe ena fidhi brosta-tu the Xenophon saw.3s a snake in.front.of-cl.acc.m / *(s’)ton eafto tu /(to) the self his "Xenophon saw a snake in front of him." 17 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. We find clitics in the above environments but not reflexives. Strong pronouns are also found in this environment as illustrated by the following examples. Cf.:* (16) a.o Ksenophon kitai dhipla se afton the Xenophon is.looking.3s next to him "Xenophon is looking next to him/to himself." b. o Ksenophon idhe ena fidhi brosta se afton the Xenophon saw.3s a snake in.front.of to him "Xenophon saw a snake in front of him." The following chart presents a summary of the distribution of the various nominal expressions which are governed by the currently discussed prepositions. I will refer to this type of preposition as "dhipla- type prepositions." Cf.: (17) NOMINAL EXPRESSIONS WITHIN dhipla-iy^e PPs: expression is licensed within rf/z/p/a-type PPs: REFLEXIVES CLITICS STRONG PRONOUNS NO YES YES 'Non-clitics require the presence of an accusative case assigner in this environment. See Terzi (1991) for some discussion. 18 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.2.3. Prepositions and CFCs The fact that pronominal clitics cannot be hosted by prepositions that govern reflexives has nothing to do with binding properties of the clitic since clitics are totally ungrammatical when hosted by these kind of prepositions, regardless of whether we try to interpret them as free or disjoint from some c-commanding antecedent. The fact that there cannot be pronominal clitics which correspond to the arguments of the prepositions which govern reflexives also does not seem to be morphophonological. It is true that most (but not all) of the prepositions which can take reflexives are monosyllabic; however, in case there were a question that the possibly "weak" status of these prepositions could play a role in their inability to host clitic pronouns, it can be noted that the monosyllabic prepositions are not themselves clitics. Moreover, the preposition se can clearly in principle host a clitic: the cliticized definite article. This is illustrated in the following example. (18) pighame s’to spiti tis Elenis went.Sp to’the house the.gen Eleni.gen "We went to Eleni’s house." Moreover, the restriction on the distribution of clitics has nothing to do with whether the prepositional host is morphologically simple or 19 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. complex. Although it is certainly true that clitics are prohibited from occurring as the object of simple prepositions, there are also complex prepositions that cannot host clitic pronouns as the following example illustrates. But precisely these kind of complex prepositions can take reflexives. The following examples indicate that reflexive anaphors and strong pronouns, but not clitics, may occur as a complement of this and similar prepositions. (19)a. i Elena dhen prosexi kanenan [pp ektos apo ton eafto tis] the E. not look.out.for anyone except from the self-cl.gen.f "Elena doesn’t care about anyone except herself." b. i Elena dhen prosexi kanenan [pp ektos apo afton] the Elena not look.out.for anyone except from him "Elena doesn’t care about anyone except him." c. *i Elena dhen prosexi kanenan [pp ektos (apo)-ton/tu]“ the E. not look.out.for anyone except (from)-cl.acc.m/.gen.m "Elena doesn’t care about anyone except him." It is also not what morphological case a preposition assigns that is decisive in determining whether or not that preposition can host a “It doesn’t matter what case the clitic is, a clitic is still ungrammatical in this environment. We will see later that prepositions that license clitics generally govern the genitive case. 20 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. clitic. The reader might have noted that prepositions that govern accusative case disallow clitics, and those that govern genitive case disallow reflexives. But the correlation between what case a preposition governs and whether or not a clitic or reflexive can occur is not perfect. For example the preposition kata ‘against,’ a borrowing from Katharevousa (the so-called "pure language," that was invented for diglossic purposes), governs the genitive case; nonetheless, it allows a reflexive and disallows a clitic as argument. E.g.: (20) mono o Yiorghos milise kata tu eaftu tu/*kata-tu only the G. spoke against the self his/ *against-cl.gen.m "Only George spoke against himself." The above example indicates that the generalization holds that in PPs clitics and reflexives are in complementary distribution— regardless of which case the preposition might govern. To summarize the data, reflexive anaphors and clitics are in complementary distribution within PPs. Moreover, the distribution of pronominal clitics and overt pronouns inside of PP is not identical. When we consider the data more closely, we see that the distribution of clitics in PPs has an intimate connection to "opaque domains" as determined by the binding theory. To see this, consider that where clitic pronouns are banned (namely, with xe-type prepositions), 21 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the PP does not itself constitute a Complete Functional Complex. Instead, the entire clause minimally containing the PP serves as the CFC. We see this by the behavior of pronouns. A strong pronoun found in this environment cannot be co-indexed by any c-commanding NP in the minimal clause containing the PP. I repeat here (13) with indices as (2 1 ): (21)a. i Elenij edhose to vivlio se aftin^/.j the Eleni gave.3s the book to her "Eleni gave the book to her." b. i Elenij to-ethiakse yia aftin^./.j the Eleni cl.acc.n-fixed.3s for her "Eleni made it for her." The Binding Theory requires a pronoun to be free in the CFC containing it. Since an overt pronoun contained within a PP headed by a ^e-type preposition must be disjoint from any c-commanding NP in the minimal clause containing the PP, the PP itself cannot constitute a CFC, instead the clause minimally containing the PP must be the CFC. The behavior of reflexives is consistent with the conclusion that PPs which are headed by prepositions taking reflexives constitute a CFC. The reflexive can be bound by, e.g., the subject of the relevant clause. To illustrate this point, I repeat here (19a,b) with indices as (22): 9 9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (22) a.i Elenij edhose to vivlio s’ton eafto tisj the Eleni gave.3s the book to’the self hers "Eleni gave the book to herself." b. i Elenij to-ethiakse yia ton eafto tiSj the Eleni cl.acc.n-fixed.3s for the self hers "Eleni made it for herself." The Binding Theory requires a reflexive to be bound in the CFC containing it. Since in this example a reflexive contained in the PP can be grammatically bound by an NP outside of that PP, we know that the PP in question cannot constitute a CFC. In sum, PPs headed by xe-type prepositions do not constitute a CFC; an anaphor is bound outside of the PP, a pronoun must be free with respect to the entire clause minimally containing the PP. Moreover, clitics are not grammatical at all in this environment. The above generalization is reversed for the contexts where we do find clitics. When clitics are grammatical within PPs (with dhipla-type prepositions), the PP itself constitutes a CFC. We know that these kind of PPs constitute a CFC because a reflexive anaphor cannot occur as their object. A reflexive is ruled out when it is not bound within its CFC. Since there is no antecedent within a dhipla-lype PP, a reflexive which 23 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is the object of a dhipla-type preposition cannot be bound and thus is ungrammatical.^ As further support for the claim that dhipla-type PPs constitute CFCs, note that clitic pronouns contained within them can be coindexed with any c-commanding NP outside of the PP containing the clitic. Since a pronoun must be free within its CFC, it must be that its CFC is smaller than the entire clause containing the PP. The following examples illustrate the binding possibilities for clitic pronouns: (23)a. o KsenophoUj kitai dhipla-tuj/k the Xenophon is.looking.3s next-cl.gen.m "Xenophon is looking next to ?him/to himself." ^The key point here is that, within PPs constituting CFCs, there are no antecedents to bind a reflexive. The syntax of PPs in this respect contrasts with that of NPs. The behavior of strong and clitic pronouns clearly indicate that NPs constitute a CFC (Since pronouns can be grammatically coindexed with the subject of the sentence and pronouns must be free in the CFC containing them, we can conclude that the NP itself does indeed constitute a CFC), nonetheless, reflexives can occur as arguments of nouns as the following example illustrates: (i) idha tin fotografia tu eaftu mu saw.Is the photograph the.gen self.gen my "I saw the picture of myself." NPs differ then from PPs in that they can apparently provide a covert or implicit antecedent for a reflexive (or alternately allow CFCs to be computed somewhat differently for pronouns versus reflexives). Regardless of the reason NPs and PPs differ in this way, the relevant point is that PPs headed by dhipla-type prepositions do constitute CFCs. 24 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. o Ksenophoiij idhe ena fidhi brosta-tUj/^ the Xenophon saw.3s a snake in.front.of-cl.acc.m "Xenophon saw a snake in front of him/?himself." The same generalization vis-a-vis binding holds for strong pronouns (I repeat here (15a and b) with indices as (24)); i.e., a strong pronoun can be coindexed with a c-commanding NP outside of the PP containing it.:"* (24) a.o Ksenophon^ kitai dhipla se afton^y^ the Xenophon is.looking.3s next to him "Xenophon is looking next to him/to himself." b. o KsenophoUj idhe ena fidhi brosta se aftonjy^ the Xenophon saw.3s a snake in.front.of to him "Xenophon saw a snake in front of him." D/z/p/a-type PPs clearly constitute a CFC as has been demonstrated via an examination of the binding behavior of strong and clitic pronouns in this environment. In addition, the fact that reflexive anaphors are ruled out in PPs headed by <i/r/p/a-type prepositions ''The judgements about the binding possibilities of strong pronouns contained in PPs are not robust, as is also the case for decisions about binding possiblities for pronouns in these same and similar environments in English. I have nothing to say about why this might be the case. 25 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. follows directly from the fact that these kind of PPs constitute CFCs. Since a reflexive in a PP headed by a dhipla-type preposition must find its antecedent within this domain, but there is no antecedent available; the reflexive anaphor is ungrammatically free in the domain where it must be bound and therefore is ruled out. We see that clitics are only licensed when the minimal maximal projection containing the clitic constitutes a CFC. If we assume that the presence of a subject (a subject at least in the formal if not thematic sense) implies the presence of a CFC including that subject, we can conclude that PPs that license reflexives lack subjects since they do not constitute CFCs. In contrast, PPs that license clitics do constitute CFCs, as is obvious from the fact that they create a domain within which pronouns must be free. If we assume that the presence of a CFC implies the presence of a subject in that domain, it must be that these kind of PPs have subjects (at least in the formal sense of having a specifier). If these kind of PPs do have subjects, they cannot be thematic subjects in an argument position otherwise it seems we could have an antecedent for a reflexive contained in this kind of PP and the account for why reflexives are banned in this environment would fail. One possibility might draw on the fact that prepositions like the MG equivalents of near, far, next to, etc., can only be interpreted with respect to a specific 26 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. point of view or reference point for either the subject or the interlocutors. Taking these considerations into account, one plausible conclusion is that something like a point-of-view operator occurs in the putative specifier position of these kind of PPs. The generalizations concerning the distribution of clitics with respect to CFCs are summarized in the following table: (25) NOMINAL EXPRESSIONS WITHIN PP: grammaticality of: REFLEXIVES CLITICS STRONG PRONOUNS PP docs not constitute a • YES NO YES CFC PP constitutes a CFC NO YES YES So, the distribution of clitics within PPs is sensitive to whether or not the PP constitutes a CFC, and clitics are only possible when the minimal maximal projection containing them forms a CFC. 2.3. Toward an Account We will see that the interaction of two sub-theories of the grammar allow us to capture the distribution of the pronominal clitic. These subtheories are: (a) the theory of predication, and (b) the theory of binding. 27 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.3.1. The Predication Requirement on Cliticization The following provides a restatement of the binding theoretic description of the clitic’s distribution: (26) a clitic is licensed only if the minimal maximal projection containing it constitutes a CFC. The generalization above indicates that clitics, although sensitive to CFCs, are sensitive to them in a way that is different from that of nominal expressions as described by the Binding Theory. The clitic’s sensitivity to CFCs must be distinguished from that of nominal expressions like anaphors and pronouns in that the CFC that licenses the occurrence of the clitic must also be the minimal maximal projection containing the clitic; i.e., the CFC and the minimal maximal projection containing the clitic must be identical. Nominal expressions like anaphors and pronouns, on the other hand, must adhere to the Binding principles within the CFC that contains it, but it doesn’t matter if that CFC is also the minimal maximal projection that contains the nominal expression or not. For example, a reflexive such that the minimal maximal projection containing it is a PP, can be bound outside of that PP as we saw was the case with a sentence like the following (I repeat here (11b)): 28 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (ll)b.i Eleni to-ethiakse yia ton eafto tis the Eleni cl.acc.n-fixed.3s for the self hers "Eleni made it for herself." The fact that the reflexive can be bound by the subject of the sentence means that the CFC for the reflexive is much larger than the CFC containing the reflexive. The difference in behavior between clitics, on the one hand, and reflexives and strong pronouns, on the other, can be summarized in the following table which shows the domains where the different expression are sensitive to the notion "subject." Cf.:^ (27) DOMAINS WHERE SUBJECTS ARE NEEDED CFC I MINIMAL MAXIMAL PROJECTION anaphors, | clitics (overt) pronouns ] The generalization that emerges is that the licensing of a clitic is sensitive to the status of the minimal maximal projection containing it as a CFC; there is even a stricter requirement for governing the role of CFCs than that required by the Binding theory where the concept of ^In other cases where we have clitics (IP and NP) we never have a distinction between minimal maximal projections and CFCs; they always form CFCs. Therefore, the distinction we have seen between clitics and other nominal expressions is not detectable in those environments. 29 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CFCs is a primitive. How do we account for this sensitivity to minimal maximal projections when cliticization is involved? A significant piece of this puzzle can be solved if we view the licensing of clitics as involving predication: where a non-overt pronoun is the subject of the predication, the predicate is the minimal maximal projection containing the predicate variable, and the clitic (or: clitic/pro chain) serves as the predicate variable (cf. the treatment of Clitic Left Dislocation (CLLD) in latridou (1991)). The following structure illustrates this proposal: (28) [proj P-clj ...]] The proposal that predication is involved in clitic structures can provide an account for the fact that cliticization cares about minimal maximal projections. A well-known syntactic requirement on predication is that the subject of the predication and the predicate mutually m- command each other (see Rothstein (1983), Schein (1982), and Williams (ibid.)). latridou (ibid.) points out that this is not sufficient to place a locality requirement on the relation between the subject of the predication and the predicate; she suggests it is also necessary to place a locality requirement on the relation between the subject of a predication and the predicate variable. She suggests the relation 30 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. between the subject of a predication and the predicate variable position is regulated by the following locality condition: (29) the predicate must be the minimal maximal projection containing the predicate variable. The requirement in (29) ensures that the minimal maximal projection containing the clitic must function as the predicate; therefore, when clitics are contained in PPs, the PP is predicate and a non-overt pronoun is the subject of the predication. 2.3.2. The Binding Requirem ent on C liticization Given that cliticization involves predication and that predication must adhere to the conditions discussed above, we can answer the question of why the predicates containing clitic pronouns must simultaneously also be CFCs. Domains that are defined in terms of CFCs are well known from the Binding theory. Witness, for example, the Specified Subject Condition (SSC),^ which requires an anaphor to be bound and a pronoun to be free within the domain defined by a specified subject. In am using this term as well as the term ‘Tensed Subject Condition (TSC)’ simply as descriptive statements, they have no theoretical status here. 31 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. a similar manner, there is the Tensed Sentence Condition (TSC) where Tense is argued to act as a subject and thereby define a domain within which anaphors must be bound. The sensitivity of clitics’ to the presence of a subject seems to suggest then that the Binding Theory is indeed at play in these constructions. It is obvious that the clitic (or: clitic/pro chain) is subject to principle B of the binding theory as far as interpretation is concerned (as opposed to licensing). This is illustrated by, for example, the fact that a clitic (or clitic-pro chain) that corresponds to the internal argument of a verb must be disjoint from the subject of the clause that minimally contains it. E.g.: (30) i Elenij tin,^y.j-thavmazi the Eleni cl.acc.f-admire.3s "Elenij admires her^/.j." What is new here is that I want to extend the domain in which this disjointness requirement applies from reference possibilities (free or bound) of the clitic (or the clitic/pro chain) to the licensing of the clitic ’Or alternately, the clitic-pro chain is involved in the binding requirement. 32 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. as well. I propose the following antilocality condition on licensing of the clitic; ANTILOCALITY CONDITION: (31) a clitic must be free in the predicate containing it. Given that cliticization involves predication, when a clitic occurs, a non-overt pronoun is the subject of the predication. Since the predicate must be the minimal maximal projection containing the predicate variable, when the argument of a PP is involved, the PP must be the predicate. PPs that do not constitute CFCs cannot license clitic pronouns despite the fact that the relation between a non-overt pro subject of the predication and the PP predicate would satisfy the known structural requirements on predication. Cliticization is ungrammatical with the arguments of such PPs because the predicate variable, being bound by the non-overt subject of the predication, is not free in the minimal CFC containing it (the minimal clause containing it). This is illustrated schematically in the following example: (32) [ip=crc - prOj [pp,crc se-clj proJ ... ] In contrast, cliticization is grammatical with the arguments of PPs that constitute CFCs because the clitic, although bound by the non-overt 33 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. subject of the predication, is free in the minimal CFC containing it. the following example illustrates this schematically; (33) [jp^cFc ... prOj [pp = cFc dhipla-clj prOj] ... ] 2.3.3. Previous Approaches From the Clitic Doubling Literature Although, a binding approach to the licensing of simple cliticization has not been previously suggested, it has been proposed in the literature on clitic doubling that there is a disjointness effect in doubling constructions attributable to the binding theory. 2.3.3.1. Disjointness and Interpretation of Clitic-Doubled Expressions For instance, in Schneider-Zioga (1990), it is argued for MG that the clitic morpheme itself is subject to principle B of the binding theory. Schneider-Zioga suggests that this disjointness requirement on clitics affects the interpretation of clitic-doubled NPs. She suggests that this requirement can provide an account for the fact that clitic-doubled QPs are interpreted as having essentially widest scope with respect to other quantificational elements in the clause containing the clitic-doubled expression (see Chapter six for a more complete discussion of this phenomenon). 34 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. To see how her analysis works, consider that when clitic-doubling is involved, the clitic and a doubled QP are coindexed by virtue of the doubling configuration. When doubled QPs undergo QR they must raise to a position such that they do not locally A’-bind the clitic morpheme, in keeping with the antilocality requirement on clitics. In the case of doubling a QP that is the object of a verb, the doubled QP can satisfy the antilocality requirement on the clitic by moving out of the CFC containing it (the clause in this case). The following diagram illustrates this analysis: (34)a. * QPj [ clj-V X j ...]] (the clitic is not free in the CFC containing it) b.ok. QPj ...[cfc clj-V X j ...] (the clitic is free in the CFC containing it) Since a clitic-doubled QP must move to such a high position in the clause in order to satisfy the antilocality requirement on the clitic, the doubled QP will end up taking scope over all the quantificational expressions contained within the CFC (the clause). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2,3.3.2. Disjointness and Licensing of Clitic Doubling in Lebanese Arabic Aoun (1991) has proposed that an antilocality effect is at play in accounting for the distribution of clitic-doubled NPs in Lebanese Arabic (L.A.). Moreover, he argues that the same requirement is responsible for the fact that only certain kinds of nominal expressions can clitic-double. In his system, clitic-doubled NPs in L.A. are in a predication relation with a predicate that contains a clitic-pro chain. In support of this, he presents a variety of structural tests that indicate the so-called "doubled" NP is actually not in argument position, but instead occurs in a dislocated position to the right of the predicate in question. Aoun demonstrates that the relation between the argument position and the dislocated position cannot be one of movement; he suggests a predication relation is at play. He proposes the following configuration where the doubled NP which is the subject of the predicate is to the right of the predicate;® (35)...[[pred Pred-clj proj ] la NPj] Consider the following example of clitic-doubling in L.A.: ®ln L.A., in contrast to MG, whenever there is clitic-doubling, the "doubled" NP must be preceded by a dummy preposition (la in L.A.). 36 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (36) Kariim sheef-o la Saamii Kariim saw.3m-cl.3.acc.m to Saamii "Kariim saw Saami." A sentence like (36) has the following representation: (37) [[Kariim'sheef-Oj prOj ]la Saamiij Kariim saw.3m-cl.3.acc.m to Saamii In Aoun’s system, whenever a clitic occurs, it identifies a non-overt pronoun in argument position as illustrated in (37). Since a pronoun is subject to principle B of the Binding theory, it must be free in its CFC. In a clitic-doubling configuration, the doubled NP and the non-overt pronoun are coindexed by virtue of the predication relation that holds in these constructions. Therefore, in order for the pronoun involved in the clitic-doubling construction to be free in its CFC, it is necessary that the relevant projection containing the clitic constitutes a CFC. Aoun notes that a doubled NP can only occur in phrases which constitute CFCs. 37 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Further supporting the idea that an antilocality effect is at play in clitic-doubling constructions, Aoun notes that the L.A. reflexive expression‘ s haalo "self" cannot clitic double: (38)a. Kariim sheef haalo Kariim saw self "Kariim saw himself." b.*Kariim sheef-o laHaalo Kariim saw.cl.acc.m to-self Under Aoun’s analysis, (38b) has the following structure due to the predication involved in clitic-doubling in L.A. He assumes the internal subject hypothesis, so the subject of the clause originates internal to the predicate. (39)...[Kariimj [[tj sheef-Oj prOj ] laHaalOj]] It is not possible for the reflexive anaphor: laHaalo to be bound for the following reasons: The reflexive is coindexed with the c-commanding subject; due to the predication relation between the doubled NP and the ‘ ^See latridou (1988) for a discussion of clitic doubling of the reflexive in MG. In Chapter 5, we will see that the reflexive in MG is licensed in clitic-doubling constructions, in contrast to the reflexive in L.A. 38 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. predicate, the doubled reflexive will also be coindexed with the predicate variable: pro. But, a pronoun must be free in the CFC containing it and it cannot be since the subject is coindexed with it; therefore, it is not possible to clitic-double a reflexive in L.A. 2.4. Concluding Remarks Keeping in mind the previous analyses which impose an antilocality requirement on the clitic and/or the clitic-pro chain, what is unique about the predication and antilocality requirement proposed in this chapter is that it extends beyond clitic-doubling constructions to simple cliticization as well. The end result is that cliticization is structurally very much like clitic doubling, the subject matter to which we now turn. 39 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER THREE: THE DISTRIBUTION AND STRUCTURAL REPRESENTATION OF CLITIC DOUBLING CONSTRUCTIONS 3.0. Introduction The principal issue addressed in this chapter is the structural representation of clitic doubling (Cl-d) in Modern Greek. In Cl-d constructions a Noun Phrase* co-occurs with a coindexed pronominal clitic which agrees in grammatical features with the Noun Phrase in question: (1) tiUj-idha tin Mariuj cl.acc.f-saw.ls the Maria "(Speaking of) Maria, I saw her." Standard views of clitic doubling suggest either that it involves some kind of spec/head agreement relation between the clitic and the doubled expression, or incorporation of a pronoun or determiner. In contrast to standard views, I will propose that, in Modern Greek, Cl-d signals a different syntactic structuring of phrases than that found when non-doubled arguments occur. ' I use the term "Noun Phrase" in a general sense; i.e., I distinguish between NP and DP and QP only when confusion might otherwise arise. 40 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Specifically, I will propose the following representation of Cl-d: a copy of the Cl-ded expression occurs in clause initial position. This copy is the subject of a predication, with the doubled Noun Phrase and the clitic which doubles the Noun Phrase serving as the predicate variable. The minimal maximal projection containing the clitic and doubled Noun Phrase serves as the predicate (cf. latridou’s (1991) proposal for the structure of Clitic Left Dislocation (CLLD)). The following structure is proposed: (2) [ NPj Lp [ip .......clitiCj....NPj....]]] A PF rule deletes the copy: the copy is syntactically, but not phonetically, present. Initial evidence for the proposed structure of Cl-d will come from an analysis of the distribution of Cl-d. The distributional facts are as follows: 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. Cf. (3), (4): (3) a. N-cl b. P-cl c. cl-V (4) *a. N-clj NPj *b. P-clj NPj c. clj-V NPj The data in (3) and (4) support the idea that, in Cl-d constructions, a copy of the doubled expression occurs in the clause initial position and 41 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. that the relation between the NP copy and the rest of the clause is that of predication. First, I note that we can capture certain facts about the distribution of clitic doubling in a theoretically natural way if we postulate the existence of a structurally higher Noun Phrase that is a copy of the doubled Noun Phrase and thus co-indexed with it. Given this structurally higher NP copy, it can be deduced that an antilocality licensing condition on Cl-d exists such that a Case-marked Noun Phrase must intervene between the postulated NP copy and the doubled Noun Phrase in the relevant domain. The relevant domain is the minimal maximal projection containing the clitic and the doubled Noun Phrase. Intervening Case-marked Noun Phrases exist only in clauses. Therefore, we cannot have NP copies immediately dominating PPs or Noun Phrases. A locality condition that governs the relation between the NP copy and the clitic and doubled Noun Phrase ensures that clitic-doubled expressions in PPs and DPs cannot be related to an NP copy in clause initial position. The relation between the NP copy and the doubled Noun Phrase can best be described by predication. Predication is subject to the following locality requirements: (a) the subject of predication and the predicate must m-command each other (Rothstein (ibid.)); and (b) the predicate must be the minimal maximal projection containing the 42 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. predicate variable (latridou (ibid.)). A clitic which doubles an argument can be no more deeply embedded with respect to the NP copy than IP, which is the minimal maximal projection dominating the clitic. A doubling clitic cannot be more deeply embedded with respect to the clause initial NP copy, for instance in PP or NP, because this would violate the various requirements on predication (the predicate must be the minimal maximal projection containing the predicate variable, which by hypothesis is the clitic and the NP it doubles). Besides giving evidence in favor of a predication analysis of Cl-d, it is possible to find evidence against other possible accounts such as a movement or Binding approach. To see this, reconsider the assumption that there is an NP copy in clause initial position when clitic doubling occurs. Now consider the Cl-ded expression. The relation between the postulated NP copy and the Cl-ded expression does not mimic A - movement: we do not find Cl-ded expressions in the same position where we find variables which clearly arise through movement; e.g., movement is possible out of Noun Phrases in MG but Cl-d is not possible in this environment (i.e., there are no doubled arguments of Noun Phrases). 3.1. The Phenom enon of Clitic D oubling The phenomenon of clitic doubling in Modern Greek is illustrated below where a clitic co-occurs with a nominal expression with which it 43 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. agrees in gender, number, case and person (compare (5a) and (6a) with (5b) and (6b))r (5)a. idhes [jv -,, tin kopella]? saw-2 the-acc.f girl "Did you see the girl?" b. tin-idhes tin kopella]? cl.acc.f-saw-2 the-acc.f girl "Did you see the girl?" (6)a. idhes [^p ton Aleksandro]? saw-2 the-acc.m Alexander "Did you see Alexander?" b. ton-idhes [f,T p ton Aleksandro] cl.acc.m-saw-2 the-acc.m Alexander "Did you see Alexander?" 3.2. The Licensing of Clitic Doubling 3.2.1. The Distribution of Clitic Doubling A distinctive property of Cl-d in MG is its distribution in comparison to the distribution of simple cliticization. In MG, clitic "The clitic and the definite article in MG are phonologically identical. 44 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. doubling does not have the same distribution as simple cliticization; i.e., not every that can have a clitic associated with its argument licenses clitic doubling. The environment of clitic doubling encompasses only a proper subset of the environments where simple cliticization occurs. We will see in fact that, in MG, doubling is possible only with the complements of verbs/" 3.2.1.1. Prepositional Phrases and Clitic D oubling It is not possible to clitic double the complements of prepositions. As illustrated in (7) and discussed in the previous chapter, an enclitic, attached to P, can be co-indexed with the object of a preposition (see (7a)). Moreover, an overt NP can occur as object of a preposition (see (7b)). But a clitic and a coindexed overt Noun Phrase cannot co-occur within PP (see (7c)). Cf.: ^ In the final stage of revising this dissertation, it has come to my attention that some speakers of Modern Greek, allow Cl-d in any environment where clitics are found; for example, they accept, at least marginally, doubling of the complements of Ps and of Ns. But even for these speakers, it seems that doubling is best when the complements of verbs are involved. 1 will provide an account for this difference between native speakers in future research. In any case, by far the majority of speakers questioned strongly felt doubling was grammatical only with the complements of verbs. 45 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (7)a. brosta-tu in-front-cl.gen.m "in front of him" b. brosta apo [^p ton Yiorgho] in-front from the.acc.m George "in front of George" c. *brosta-tu apo [^p ton Yiorgho] in-front-cl.gen.m from the.acc.m George d.*brosta-tu [^ptu Yiorghu] in-front-cl.gen.m the.gen.m George 3.2.1.2. Noun Phrases and Clitic Doubling In addition, it is not possible to clitic double the complements of nouns (see (8c)),'* although a genitive clitic pronoun, enclitic to N, can mark the complement of the noun (see (8a)). A full Noun Phrase can occur as well (see (8b)). Cf.: '* Native speakers I consulted find clitic doubling slightly less marginal when deverbal nouns are involved. It is plausible that Cl-d is more acceptable with deverbal nouns than with other nouns by analogy to sentential Cl-d. An example follows: i.*?to plisimo-tu tu Alexandru (apo tin Rani) the washing-cl.gen the A.gen (by the Rani) 46 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (8)a.i fotografia-tis the photograph cl.gen.f "the photo of her" b.i fotografia [^p tis Anulas] the photograph the.gen.f Anula.gen "the photo of Anula" c.*i fotographia-tis [^p tis Anulas] the photograph cl.gen.f the.gen.f Anula.gen 3.2.1.3. Verb Phrases and Clitic Doubling We see that it is possible for pronominal clitics to correspond to the arguments of all of the above lexical categories, but clitic doubling is possible only with the complement of verbs. I illustrate this fact for verbs in indicative mood in (9a,b) where a clitic pronoun, cliticized to the inflected verb and inflected auxiliary respectively and therefore located in (the relevant projection of) l'’, doubles an argument of the verb: (9) INDICATIVE a.tin-idha [„p tin kopella] cl.acc.f-saw-ls the-acc.f girl "1 saw the girl." 47 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b.tin-ixa dhi [„p tin kopella] cl.acc.had.ls see the-acc.f girl "I had seen the girl." In fact, any verb that can host a clitic can clitic double. Since the clitic itself is clearly located with the inflectional categories outside of VP (and not with V° itself) as evidenced by examining complex tenses involving auxiliaries (cf. example (9b) above), for the sake of completeness 1 will illustrate the possibility of clitic doubling for verbs in different inflectional environments. Clitic Doubling is also possible with the arguments of verbs in the subjunctive mood: (10) SUBJUNCTIVE thelo U h j u n c t iv c na tin-episkefto [„p tin pethera-mu]] want.Is subj cl.acc.f-visit.ls the mother.in.law-cl.gen.f "I want to visit my mother-in-law." The same generalization holds for imperatives. Cf.: 48 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (11) IMPERATIVE a. dhiavase-tin! read.singular-cl.acc.f "read it!" b. dhiavase [„p tin efimeridha]! read.singular the.acc.f newspaper "read the newspaper!" c. dhiavase-tin [„p tin efimeridha]! read.singular-cl.acc.f the.acc.f newspaper Modern Greek lacks (morphologically) infinitive clauses. That is, MG lacks clauses which are not inflected for subject/verb agreement or tense. Nonetheless, there are clauses which morphologically appear to be subjunctive but behave syntactically and semantically like infinitives including having the syntactic property of licensing PRO in subject position. See Terzi (1992), and Varlokosta (1992) for discussion. There is clitic doubling of the complements of these infinitive-type clauses: (12) INFINITIVE prospathisa Lrinitivc na to-puliso [„p to vivlio]] tried.Is subjunctive cl.acc.n-sell.ls the book "I tried to sell the book." 49 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The complements of active participles can Cl-d. This is illustrated in (13). Cf.: (13) PARTICIPLE a. [participle vlepondas-tin], apokimithika watching-cl.acc.f fell.asleep.Is "watching it, I fell asleep." b. vlepondas tin tileorasi], apokimithika watching the.acc.f television fell.asleep.Is "watching television, I fell asleep." c.vlepondas-tin [f,;p tin tileorasi], apokimithika watching-cl.acc.f the.acc.f television fell.asleep.Is After surveying the various environments where Cl-d is and is not allowed, we have seen that clitic doubling is only possible with the complement of verbs. This is certainly not a necessary fact about clitic doubling if we look at Cl-d cross-linguistically.^ We see, for example. ^In this way, Cl-d in MG patterns with Cl-d in many of the Cl-d dialects of Spanish such as River Platte Spanish or standard Peruvian Spanish (but, Andean Spanish, for example, patterns differently (J. Camacho and L. Sanchez (p.c.))— with Arabic Cl-d dialects instead— a pattern that is described in the immediately following main text). In standard Peruvian Spanish for example, pronominal clitics are found 50 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. that in Lebanese Arabic (LA) all categories that license clitics, license Cl-d: the complements of Ns, the complements of Ps, as well as the complements of Vs can Cl-d. The following examples from Aoun (1981) illustrate this fact: (14) PREPOSITIONAL PHRASE hki:t mai-o (la Kari:m) talked-I with-him (to Kari:m) "I talked with Karim." (15) NOUN PHRASE shtre:t kte:b-o (la Kari:m) bought-I book-his (to Kari:m) "I bought Kari:m’s book." (16) INDICATIVE CLAUSE she:fit-o (la Kari:m) saw.she-him (to Karim) "She saw Karim." In the next section I provide an account of the restriction of Cl-d to the complement of verbs in MG. associated with the object of verbs and in Noun Phrases associated with the arguments of nouns; Cl-d, however, is possible only with the complement of verbs. 51 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.2.2. Why Only the C om plem ent of V erbs? The empirical observation is that Cl-d in MG has a distribution that is a subset of the distribution of simple cliticization: it can only occur with the complement of verbs. When we consider the domain of clitic doubling, the question arises as to whether to consider the lexical projection— VP or the functional one— AgrP (O) or AgrP (S). It seems most accurate to consider the entire clause (AgrP (S)) for several reasons. For instance, when the complement of a verb is involved, the clitic that doubles the argument is located outside of VP; a more accurate description of the domain that is involved in the licensing of Cl- d should take this fact into account. Moreover, the doubled argument itself is also ultimately located outside of VP (in Agr-O) under the view that Case is assigned in the specifier of the relevant agreement projection.^ This fact too suggests we need to consider Cl-d of the complement of a verb as being a property of the entire clause, not just of VP. Furthermore, it is a priori possible for the external argument of the verb to clitic double, a fact which also requires us to consider the entire clause as the domain of Cl-d. From this perspective it is more ‘ 'Clitic-doubled expressions in MG receive the same morphological case that a non-doubled expression receives; in other words, MG doesn't conform to Kayne’s Generalization, which notes that the doubling of an argument requires the presence of an additional Case assigner. 52 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. accurate to say that the domain that is involved in licensing Cl-d is the clause. What then is special about clauses that licenses Cl-d? One basic property of clauses is that they constitute a CFC. However, the property "constitutes a CFC" isn’t limited to clauses; a number of other categories also form CFCs, for instance. Noun Phrases and certain PPs. Moreover, as discussed in the previous chapter, being a CFC is a precondition to licensing simple cliticization. Therefore, it can’t be the property "constitutes a CFC" that licenses Cl-d in clauses since this property doesn’t uniquely pick out clauses. Another obvious properties of clauses is that they have comps. If having a comp is something that singles out all and only clauses,^ a ^ In MG, all clauses, even participle phrases seem to have comps in contrast to a language like English where participle phrases clearly lack comps. For example, it is possible to have a wh- or focused expression in the initial position of a participle phrase: (i) PARTICIPLE a. [participle vlepondas tin tenia] apokimithika watching the film fell.asleep.Is "Watching the film, I fell asleep." b. [com p tij [participle vlepondas Xj ] apokimithikes what watching fell.asleep.2s c. [com p tin TENIAj [participle vlepondas X j ] apokimithika the FILM watching fell.asleep.Is "Watching the MOVIE, I fell asleep." However, the existence of a comp position in participles is not unequivocally clear because they do not seem to provide a transit site for A’-movement. For instance, questioning the object of a participle requires pied-piping of the entire clause; i.e., it is not possible to extract just the object out of the participle. Focus in participles behaves in the same way: a focused phrase can occur in the initial position of the participle but cannot be extracted long distance from the participle. 53 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. closer look at comps is in order so that we can determine whether this property plays a role in licensing Cl-d. But while it is certainly true that all clauses in MG have comps, this property doesn’t single out just clauses in MG because, in contrast to a language like English, nominal phrases have comps too. Therefore, we can dismiss the presence of a comp as being relevant to the licensing of Cl-d. The following examples help illustrate that Noun Phrases in MG have comps (see Horrock and Stavrou (1987) for a thorough discussion of comps in Noun Phrases in MG); the evidence for this claim resides in the fact that there are positions in Noun Phrases which can be landing and transit sites for A’-movement (Wh-extraction and focus movement have properties of A’-movement in MG as reported in the literature). The examples in (17) illustrate that there is a special phrase initial site in Noun Phrases reserved for wh- and focused expressions respectively: These facts become obvious when the participle is in postverbal position, as below (cf. with (ib,c)): (i)a. *tij apokimithikes [participle vlepondas X j ] what fell asleep.2s watching b.*tin TENIAj apokimithika [participle vlepondas X j ] the FILM fell.asleep.Is watching Although participles allow A’-elements in initial position, they have no A’-transit sites. In this way participles differ from NPs and tensed clauses in MG both of which allow transit through their comp sites. 54 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (17) Noun Phrases a. U m p [np to vivlio tu Xenophon ]] the book the Xenophon.gen "Xenophone’s book" b. [ ,o n ,p tinosj [to vivlio X j ]] whose the book "whose book" c. [comp tu X E N O P H O N j [„p to vivlio X j ]] the XENOPHON.gen the book "XENOPHONE’S book" In (18) we see that it is possible for the wh- or focused phrase to be extracted long distance out of a Noun Phrase. (18) [comp tinoSj/tu X E N O P H O N j [ipes oti dhiavasez [Xj[„p to w h o se/th e X E N O P H O N .gen said.2s that read.2s the vivlio X j ]]]] book LIT.: Whose/XENOPHON’S, you said that you read the book 55 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The data above clearly demonstrate the existence of a cornp position for Noun Phrases in MG and militate against seeing the distribution of Cl-d as having any correlation to the distribution of comps. It is possible to make an empirical generalization about MG that picks out all and only clausal categories: in clauses there are two Case assigning projections (related to the lexical head that the clausal functional categories minimally dominate) in contrast to PP and DP which, in MG at least, have only one Case-assigning projection related to the lexical head they minimally dominate as we will shortly see in more detail.® When we examine the various categories, we see that only V (and via its relation to V, the clause) can be related to two (or more, when relevant) Case assigning projections whereas P and N can only be related to one. Consider first P. Prepositions can, at most, have only one argument expressed. Even though some prepositions (for example, apo ‘from’) govern accusative case, and other prepositions (such as metaksi ® The above generalization concerning Case assignment in clausal versus nominal or prepositional phrases does not require us to be committed to the view that Case assignment happens only via functional projections as opposed to Case assignment also being possible by government when relevant. It would be equally possible to formulate the generalization under various theories of Case assignment. 56 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ‘between’) govern genitive case, we never find prepositions that have more than one argument and accordingly assign more than one case. We do however find apparent examples of prepositions that assign more than one morphological case, albeit not at the same time and while having only one argument at a time.^ Now, under the view that each morphological case is assigned via a distinct Agr-projection, the ability of a lexical item to assign more than one case raises the possibility that that item simultaneously has two distinct Agr-projections and thus are parallel to clauses in this way. The preposition kata ‘toward, according to; against’ exemplifies the kind of preposition described above. Kata can govern either accusative or genitive case. However, as pointed out by Joseph and Philippaki-Warburton (1987), the choice of case is not arbitrary: Accusative case is assigned strictly only when the meaning of kata is ‘towards, according to’. Genitive case occurs only when the meaning of kata is ‘against’. Joseph and Philipakki-Warburton suggest that this clearly indicates that homophony (and, from the perspective pursued here, not two Case-assigning projections) is at play. ^ Later in the text we will see examples of verbs that can assign either accusative or nominative case to their subjects and that allow clitic doubling of their subjects. 57 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Other apparent counterexamples, Joseph and Philipakki- Warburton point out, are in the form of borrowings from Katharevousa (the MG high language) which almost invariably involve fixed expressions (see Joseph and Philipakki-Warburton (ibid.) for examples and some discussion). Given all the above, it seems safe to conclude that each preposition really is related to only one Case assigning projection. Consider next nouns. Nouns clearly are related to only one Case assigning projection and this projection assigns genitive case. Any additional argument(s) needs a preposition for Case assignment (if an additional argument is possible at all). E.g.: (19)a.i fotoghrafies tu Aleksandru the pictures the.gen Alexandros.gen "the pictures owned by Alexandros" "the pictures depicting Alexandros" b. *i fotoghrafies tu Aleksandru tis Ranis^° the pictures the Alexandros.gen the Rani.gen "Rani’s pictures of Alexandros" Of course, this phrase is grammatical on the following bracketing where we are referring to Rani’s Aleksandros as a way of identifying which Aleksandros we are talking about: i. [i fotoghrafies [[tu Aleksandru] tis Ranis] 58 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. i fotografies tu Aleksandru apo tin Rani the pictures the Alexander.gen by the Rani.acc The example in (19a) illustrates the fact that genitive case is assigned in Noun Phrases; the examples in (19b and c) show respectively that: (a) we cannot have two genitive-marked Noun Phrases (see 19b); and (b) that any additional argument of the noun requires a preposition (see 19c) to assign Case. This observation holds for all different kinds of nouns including, for example, deverbal nouns where, in the instance of a noun based on a transitive verb, the direct object receives genitive case and the subject is in a prepositional apo ‘by’ phrase (examples essentially follow Joseph and Philippaki-Warburton (ibid.)):*' (20) to plisimo ton ruxon (apo ton Apostoli) the washing the clothes.gen from the Apostoles "the washing of the clothes by Apostoles" Even when non-overt nominal expressions are involved, there can only be one non-overt nominal expression per Noun Phrase. This is "Note that this kind of noun displays "nominal" behavior: it assigns genitive case to its objects, not accusative (the case that verbs assign to their objects). It is comparable to the English "the hunting of the lions." MG does not have anything comparable to the English "John’s hunting lions." 59 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. expected since non-overt pronominals, like overt Noun Phrases, need Case and only one Case assigning projection is available in Noun Phrses. Consider first non-overt pro: it is not possible to have two non-overt pronouns as evident from the fact that it is not possible to have two pronominal clitics. (Small pro in MG must be identified by X°s clitics or subject/verb agreement.) E.g.: (21) *i fotografies-tu-tis the pictures-his-hers "her pictures of him" "his pictures of her" (22) *to plisimo-ton-tu/tu-ton the washing-theirs-his/his-theirs Consider next PRO. Under some views PRO receives null Case, a Case reserved explicitly for PRO. Consider now whether, in Noun Phrases, null Case can be assigned in addition to other morphological cases (given the existence of this case (see, for example, Chomsky and Lasnik (1991) for discussion)). Note that, when only one argument occurs overtly, it is marked genitive. For example, an argument that can be understood as the subject of a deverbal noun is marked genitive: 60 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (23) to plisimo tu Aleksandru the washing the Aleksander.gen "Alexander’s washing" This can be contrasted to the (morphological) case of a subject when an object is also overtly present where we saw (cf. (20)) that the subject cannot receive Case directly from the noun but instead from a preposition. Although there seems to be no evidence directly bearing on the matter, it seems (since, by hypothesis, PRO Case is a Case assigned like any other) unlikely that a noun would exceptionally have two Cases available (one for the object and one for the subject) just in the situation when PRO were the subject. In contrast to Noun Phrases and PPs, clauses can have more than one Case assigning projection at the same time, a fact which is obvious when we consider that clauses must have subjects. So, in addition to the direct object (which we are considering with respect to Cl-d), clauses always have subjects: either overt Case-marked subjects, or non-overt Case-marked ones such as pro and PRO. Given the above set of facts, the following generalization can be made: 61 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (24) clitic doubling is only possible when the minimal maximal projection that dominates the clitic has a Case-marked subject. In some instances it might appear that the pronominal clitic is so deeply embedded that we cannot accurately characterize the minimal maximal projection dominating the clitic as having a Case-marked subject, instead it would seem that the subject occurs in the specifier position of a structurally higher functional category. The following instances exemplify these more complex cases: (a) negated verbs, (b) verbs modified by the modal/future tense marker tha, (c) when the subjunctive marker (generally agreed to be in the verbal complex and not a complementizer) occurs, and (d) when any combination of negation and modal/future tense marker tha or subjunctive marker occurs. Schematically, this is the situation: (i) clitic +inflected verb (simplest clause) (ii) negation clitic + inflected verb (iii) modal clitic + inflected verb (iv) negation modal clitic +inflected verb However, that there are varying depths of syntactic embedding for the pronominal clitic would only be true if the ordering of these elements in the verbal complex reflect the hierarchical ordering of functional categories in the clause (due to the HMC (head movement constraint— or an equivalent constraint or principle)). In fact, as has been frequently observed for MG, negation, the modal tha, and the subjunctive marker are actually all proclitics on the verbal complex, so their orderings are presumably not regulated by the HMC since clitics don’t seem to obey this constraint. The clitic status of these elements suggests two possibilities, given that these are syntactic (and not simple phonological) clitics: (1) the verbal complex has right-adjoined to the inflectional clitics (e.g.: [ncg [Neg] [ cl [V [TNS]] [AGR]]] ); or (2) the modal and/or negation clitics have cliticized out of a hierarchically lower position and are left-adjoined to the verbal complex (e.g.: [Neg] [ cl [V [TNS]] [AGR]]] ). Under either scenario, the minimal maximal projection dominating the clitic has a Case-marked subject. 62 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. So far, the data has actually only shown that Cl-d is licensed just in case there are two Case-marked Noun Phrases in the same phrase. However, if Cl-d were just sensitive to Case marking in general, namely, to the presence of two (or more) Case-marked Noun Phrases in the relevant minimal X'"''*, then it should be possible to double the subject of a transitive verb but not the subject of an intransitive verb— transitive verbs having two Case-marked Noun Phrases and intransitive verbs only one. However, such a distinction does not exist; it is generally impossible to clitic double subjects in MG.’^ This fact indicates that it is not simply the presence of more than one Case-marked Noun Phrases in a projection that licenses Cl-d, instead the presence of a Case-marked subject is decisive. It seems then that the generalization in (24) is correct: only a phrase that minimally dominates a Case-marked subject licenses Cl-d. Nonetheless, (24) needs a slight amendment in order to fully capture the distribution of Cl-d. Consider the following. It is not just the presence of a Case-marked subject that licenses Cl-d. If this were enough, then it should be possible to clitic double the subject of a clause (regardless of the transitivity of the verb) because in this situation the minimal maximal projection dominating the clitic has a Case-assigned subject. ‘^I discuss exceptions such as Exceptional Case-Marking Constructions (ECM), later in this chapter. 63 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. However the following example, where the subject of an indicative clause cannot clitic double, establishes that, in general, subjects cannot clitic double: (25)* telika o Yiorghos tos-irthe finally, the George.nom cl.nom-arrived "Finally, George arrived." It seems then that the grammar does not allow the Case-marked Noun Phrase (the subject), crucially required for the licensing of Cl-d, to be the Noun Phrase that clitic doubles. The following revised generalization is more accurate than (24): (24)’ clitic doubling is only possible when the minimal maximal projection that dominates the clitic has a Case-marked subject. The subject may not itself double. In sum, the possibility of a phrase having a subject coincides with the possibility of licensing clitic doubling; moreover, the subject itself may not clitic double. And although this statement is essentially accurate, it is not easy to capture its content in a theoretically meaningful way. There is no theoretical place in the grammar for a restriction such as the following: 64 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (26) "must occur in a phrase that has a subject and may not be a subject" Suppose, however, that a clitic-doubled expression is coindexed with a c-commanding expression that is located structurally higher than the doubled expression, as schematically indicated below: (27) NPj ...[ ... clj-V NPj ...] If this is indeed the structure, then the generalization noted in (24)’ can easily be captured in familiar theoretical terms: it is an antilocality requirement. To see this point, consider the following. Given the proposed structure in (27), the obligatory presence of a Case-marked subject in clitic-doubling constructions can be interpreted as fulfilling a requirement that a Case-marked specifier hierarchically intervenes between a clitic-doubled Noun Phrase and a coindexed, c-commanding Noun Phrase. This can be viewed as an antilocality requirement since the clitic-doubled Noun Phrase cannot be too close structurally to the c- commanding expression that is co-indexed with it; instead, the clitic- doubled Noun Phrase can be co-indexed with a c-commanding Noun Phrase only if it is located outside of the domain that consists of a 65 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. projection that has a Case-marked specifier. The required configuration is shown in (28): (28) [ NPj [ip + c litiC j ...NPj...]] The idea that the distribution of clitic-doubled Noun Phrases is subject to some kind of antilocality requirement, as suggested above, is not totally unexpected when we consider various recent analyses of clitic doubling. For instance, Schneider-Zioga (1990) proposes an antilocality requirement on clitics in clitic-doubling constructions to account for the frequently widest scope interpretation of clitic-doubled QPs. Aoun (class lectures (1991)) proposes that an antilocality requirement is at play in clitic-doubling constructions in order to account for what kinds of NPs can clitic double as well as accounting in general for the structure of clitic doubling in Lebanese Arabic (see also Aoun (1993)). And in chapter one we saw that an antilocality requirement can account for the distribution of simple clitics: the presence of an intervening specifier in the minimal maximal projection containing the clitic licenses cliticization. The antilocality requirement on clitic doubling proposed here accounts for the distribution of clitic-doubled NPs; specifically, it predicts which grammatical categories allow doubling, and which grammatical functions can themselves be doubled. 66 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Given that Cl-d, like simple cliticization, involves a co-indexing relation between a superordinate expression and the clitic (or clitic, doubled NP chain), the licensing of simple clitics and the licensing of clitic doubling can both be attributed to an antilocality requirement when clitics are involved; the two constructions differ simply in the choice of what counts as a licensing intervening specifier: for simple clitics it is the structural intervention of any specifier between the clitic (the predicate variable) and the pro (see chapter two); for clitic doubling it is the structural intervention of a Case-marked specifier between the clitic-doubled Noun Phrase chain and a coindexed Noun Phrase in a certain domain. In the next section, I will introduce additional evidence that a Case antilocality requirement is indeed operative with respect to Cl-d. 3.2.2.1. The Doubling of Subjects The generalization that the presence of a structurally intervening Case assigner in a phrase licenses clitic doubling in that phrase poses the expectation that if some construction had an (additional) intervening Case assigner present, then clitic doubling would be licensed for all structurally lower Noun Phrases in that construction. This means that variation could be easily accommodated provided that the licensing of clitic doubling really does vary along these lines. 67 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In fact, there is a construction in MG which appears to have two Case-marked positions available for subjects since subjects in these constructions can occur in either nominative or accusative case and in exactly this situation (where there are two Case assigners for the subject position), it is possible to have nominative clitics and to Cl-d the subject. The existence of such constructions provides additional evidence for the antilocality condition on the distribution of clitic doubling.^'* 3.2.2.1.1. Deictic and Interrogative Verbs There is a very small class of verbs in MG (recently discussed by B. Joseph (1992), and M. Valiouli (1990)) that allow nominative clitics; they are the deictic verb "na" and the locative interrogative verb: "Pu’n." The following examples illustrate these constructions:'^ ''‘The existence of nominative clitics in just these constructions, independently of the question of whether clitic doubling is licensed, is also consistent with the analysis of clitics in Chapter 1, where cliticization is mediated by predication, with the relation between the subject of the predication and its syntactic predicate variable being regulated by an antilocality constraint which requires a clitic to be free in the minimal maximal projection containing it. These nominative/accusative Noun Phrases are sometimes characterized as predicates. Shortly I will show that they should not be considered predicates. 68 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (29) na-tos here.is-cl.nom.m "Here he is!" (30) pu’n-tos? where’is-cl.nom.m "Where is he?" As independently discussed by Joseph, these verbs can optionally take accusative clitics instead of nominative clitics with no change in meaning. Cf.: (31) na-ton here.is-cl.acc.m "Here he is!" (32) pu’n-ton? where’is-cl.acc.m "Where is he?" The Noun Phrase that can bear either nominative or accusative case is best characterized as a subject. There is syntactic evidence that this Noun Phrase cannot be a predicate. Predicates in MG do not allow determiners as illustrated in the following examples: 69 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (33)a.o Yiorghos eghine naftis/*enas naftis the George became sailor/*a sailor "George became a sailor." b.emena me-lene Yiorgo/ *ton Yiorgo’'* me cl.acc.1-call.3p George/*the George But the Noun Phrase used with the deictic or locative verb can never occur without a determiner and therefore doesn’t ever behave like a predicate. Cf.: (34)a.na enas naftis/*naftis/ena nafti/ *nafti there-is a sailor.nom/*sailor.nom/a sailor.acc/sailor.acc "There is a sailor." b.pu’n enas naftis?/*naftis where-is a sailor/*sailor "Where is a sailor?" Names require definite articles except when they are predicates, are in the vocative case or when they are used indefinitely (in which case they require an indefinite article). For example: (i) o Yiorghos/*Yiorghos idhe tin Maria/*Maria the George/*George saw the Maria/*Maria "George saw Maria." (ii) Yiorgho! (pu ise?) George! where be.2s "George, (where are you?)" (iii) enas Yiorghos perase prin apo ligho one George passed.by.here before by little "Someone named George passed by here a little while ago." 70 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Since the Noun Phrase in these constructions doesn’t behave syntactically like a predicate, it seems safe to conclude that it is the subject instead. These constructions allow doubling of the Noun Phrase with either nominative or accusative clitics. Compare (35a,b) (nominative case) with (36a,b) (accusative case): (35)a.na-tos o Yiorghos here.is-cl.nom.m the George.nom "Speaking of George, here he is!" b.pu’n-tos o Yiorghos where’is-cl.nom.m the George.nom "Speaking of George, where is he?" (36)a.na-ton ton Yiorgho here.is-cl.acc.m the George.acc "Speaking of George, here he is!" b.pu’n-ton ton Yiorgho where’is-cl.acc.m "Speaking of George, where is he?" The above set of facts can be captured under the assumption that there are two Case-assigned specs for the subjects of these kind of clauses. Assuming with Koopman (1990) and Koopman and Sportiche 71 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (1991) that Case is assigned via a spec/head relation with a Case- assigning projection, there is one projection for nominative Case assignment and a separate one for accusative Case assignment. Suppose nominative Case is assigned to the clitic-doubled Noun Phrase. Given that the accusative Case-assigning projection occurs even if it does not overtly Case-mark the doubled Noun Phrase (alternately, it Case-marks a non-overt expletive) and that it can be optionally projected structurally higher than the nominative projection, there is an additional Case- marked specifier intervening between the doubled NP/clitic chain and the postulated NP that c-commands and is co-indexed with the doubled NP. That is, the following configuration exists; (37) [ NPj [ip NP[+,_„ C litiCj ....NPj....]] The existence of a configuration such as in (37) fulfills the antilocality licensing requirement on Cl-d and correctly allows for the exceptional doubling of a nominative NP in such instances. 3.2.2.1.2. Exceptional Case Marking (ECM ) Constructions There is another construction in MG where it appears that a subject can have either nominative or accusative case, namely, in ECM constructions involving certain ECM verbs. An example is given in (38): 72 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (38) ksero ton Apostoli/o Apostolis na ine eksipnos know.Ip the A.acc/ the A.nom subj. be.3s smart "We know Apostolis to be smart." Despite the possibility of having either nominative or accusative subjects, it is not possible to double the subjects of these exceptionally Case-marked verbs such that the doubling clitic occurs on the embedded verb: (39)a. *kserume ton Apostoli na tos/ton-ine eksipnos know.Ip the A.acc subj. cl.nom/acc-be.3s smart b. *kserume o Apostolis na tos/ton-ine eksipnos know.Ip the A.nom subj. cl.nom/acc-be.3s smart The reason for this is quite straightforward: In ECM constructions, accusative Case is not assigned internally to the embedded clause. Instead, it is assigned by the superordinate verb via its spec of Agr-O. We see then that there are not two Case-assigning projections internal to the clause containing the (potentially) exceptionally Case-marked subject; instead one Case assigning projection (the nominative one) is internal to the embedded clause containing the (potentially) exceptionally case-marked subject and the other Case-assigning 73 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. projection (the accusative one) is in the superordinate clause. The following example illustrates this: (4 0 ) [ip L„r-o [v p -N P -co p y j [ip clj-V ...]]]] If there is not an additional Case-assigning projection intervening between the NP copy in embedded clause initial position and the doubled Noun Phrase, doubling is not licensed since the anti-locality requirement can’t be met. In fact, an exceptionally Case-marked Noun Phrase can be doubled by a clitic on the superordinate verb: (41) ton-ksero ton Apostoli na ine eksipnos cl.acc-know.lp the A.acc subj. be.3s smart "I know Apostolis to be smart." This is exactly what we would expect given that clitic doubling is only licensed when the anti-locality requirement is fulfilled as it is in (41) with respect to the superordinate clause which contains the doubling clitic. In this situation a Case-assigning specifier (for the subject of the superordinate clause) intervenes between the clitic and the doubled 74 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Noun Phrase, and the NP copy in initial position of the superordinate clau se.T h is configuration is illustrated in (42): (42) NPj [ip clj-V NPj [ip...]] The fact that Cl-d is not possible with a clitic on the embedded verb but is possible with a clitic on the superordinate verb is what we would expect given the structure associated with ECM constructions. 3.3. Clitic D oubling and Locality In the previous sections, we saw that there is an antilocality requirement that must be met in order for clitic doubling to occur. But we haven’t yet addressed the issue of what domain this antilocality requirement holds in other than to note that it is relevant with respect to the minimal maximal projection that contains the clitic and the (potentially) doubled Noun Phrase. It would be interesting to learn why it isn’t possible for Noun Phrases contained within PPs and DPs to meet the antilocality requirement by virtue of being related to an NP copy in ‘^Moreover, the accusative Noun Phrase is not actually the subject of the embedded clause, but rather it is the topic— a fact I have noted and discussed elsewhere (Schneider-Zioga (1992a,b)) and will also touch on later in this dissertation. 75 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. clause initial position. Consider the configurations in (43), which illustrate this possibility: (43)a.*NPj [ip ...[pp P-clj NP^]...] b.^NPj [ip NP, . . . [ „p N-clj NPj]...] In these configurations, the subject of the clause intervenes between the postulated NP copy in clause initial position and the potentially doubled Noun Phrase in the PP or Noun Phrase respectively. Thus the anti locality requirement is, in fact, met. Nonetheless, Cl-d is still ungrammatical in these environments. This is because the antilocality requirement is, itself, subject to a strict locality requirement such that the (doubled) Noun Phrase must be free in the minimal maximal projection containing the clitic and the Noun Phrase. In the next section, I explore why minimal maximal projections are the domain where this Case antilocality requirement is operative. 3.3.1. A Predication Account of Cl-d We can account for the fact that the antilocality requirement operates within the domain of minimal maximal projections (as opposed to another domain such as CFCs or clauses) under the view that a predication relation holds between the NP copy (the subject of the predication) and the minimal maximal projection that contains the clitic 76 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (the predicate), and this relation is a chain relation. Under this proposal, the clitic and the Noun Phrase it doubles form the predicate variable.'® The proposed structure is then very similar to the representation of Clitic Left Dislocation (CLLD) proposed in latridou (ibid.), cf.:"' (44) [ NPj [,p [ip clitiCj....NPj....]]] The relation between the NP copy and the doubled expression behaves unlike either an A- or A’-movement chain as we will see in some detail in following sections. Moreover, as will be discussed, the relation doesn’t mimic binding. So there is evidence in favor of a predication analysis of Cl-d and evidence against a movement or binding analysis of the same. With respect to the properties that identify a construction as involving predication, as pointed out in chapter one, it seems undisputed that predication is subject to a requirement of mutual government/m- discuss the nature of the predicate variable in more detail later in this chapter. latridou’s proposed structure for CLLD is as follows: (i) NPj [^p ... clj-V (prOj) ... ] The main difference between latridou’s proposal for CLLD and my proposal for Cl-d is that the predicate variable in Cl-d is the chain: clitic, doubled NP instead of the clitic (or possibly the clitic, pro chain) proposed by latridou as the predicate variable in CLLD constructions. In effect, my claim is that CLLD and Cl-d are the same construction— I address this point at length especially in chapters four and five. 77 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. command between the subject of the predication and the predicate (cf. e.g., Rothstein (ibid.), Williams (ibid.); see also Schein (ibid.) for related discussion). This explains why (45a) is a possible predication and (45b) is not (from Schein (ibid.)): (45)a.John ate [„p the meatjj rawj b.*John ate [pp at [„p the meatjj] raWj In the grammatical (45a) the subject of the predication c-commands the predicate. In the ungrammatical (45b), c-command fails. This is because the "subject" of the predication is embedded in PP and the predicate is outside of PP. This same mutual m-command requirement holds for predications in Modern Greek too. (46a) illustrates a grammatical case of predication where the subject and predicate mutually m-command each other. (46b) illustrates an ungrammatical predication where the "subject" is embedded in a PP and thus cannot stand in a configuration of mutual m- command with the predicate which is outside of PP: (46)a.i Maria idhe ton Yiorghoj methismenOj the Maria saw the George drunk "Maria saw George drunk." 78 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. *i Maria milise [pp me [„p ton Yiorgho]] methismenoj the Maria.nom spoke with the George.acc drunk.acc "Maria spoke to Georgej drunkj." Moreover, Muysken (1989) points out that predication can involve no more than two elements: a subject and a predicate, and in this way a predication chain contrasts with an A’-chain which can have multiple members of the chain. latridou (ibid.) suggests an additional requirement, namely, a locality requirement on predication such that we must take as the predicate the minimal maximal projection containing the predicate v a ria b le .S h e points out that this kind of requirement would, among other things, capture the fact that predicate variables in relative clauses must occur in the specifier of the predicate phrase"* as opposed to being more deeply embedded, for instance, in the specifier of CP of a CP that is subordinated within the predicate. The following example from English illustrates the kind of ungrammatical relative clause formation "**Muysken (ibid.) makes a similar claim. He suggests that a predication chain cannot cross a maximal boundary. "'The fulfillment of this requirement wouldn’t necessarily be obvious in the overt syntax as witnessed by the existence of internally headed relative clauses (given that they involve covert movement). 79 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. strategy her proposal would rule out, where we see that the predicate variable who cannot be too deeply embedded within the predicate:” (47)* I like the woman [p^^d John said [^p who [;p Mary met]]]. This can be contrasted with the grammatical derivation where the predicate variable occurs in the specifier of the predicate and thus the predicate (the relative clause) is the minimal maximal projection containing the predicate variable: (48) I like the woman [p^cd who John said [^p [jp Mary met]]]. Given the correctness of latridou’s proposal, the assumption that the relation between the topic and Cl-ded NP is mediated by predication gives us the correct result regarding the distribution of clitic doubling; namely that only the complements of Vs can double. Let’s see how this works. Following latridou, we must take as the predicate in a predication the minimal maximal projection containing the predicate variable. I have suggested that the clitic and doubled Noun Phrase form a chain which “ It seems to me that her proposal forces a distinction between control and predication, since in control structures the purported predicate variable "PRO" is too deeply embedded. The conclusion that control and predication are distinct relations is something that Schein (ibid.) argues for and convincingly demonstrates is the case. 80 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. functions as the predicate variable"^ in Cl-d constructions. This means that in the case of a potentially clitic-doubled Noun Phrase in DP, the DP, and only the DP, must be taken as the predicate since it is the minimal maximal projection containing the predicate variable. The mutual m-command requirement on predication requires the NP copy, the proposed subject of the predication, to be in a configuration of mutual m-command with the predicate and therefore, the NP copy must be immediately dominated by the same maximal projection that immediately dominates the predicate. This means the NP copy must be minimally outside DP as follows: (49) XP / \ N P co p y PP If the NP copy were elsewhere such as in clause-initial position, mutual m-command between the subject and predicate would not hold and predication would fail. A relation of mutual m-command between the NP copy and predicate fulfills the requirements on predication. "^By ‘predicate variable’ we mean syntactic predicate variable as in Williams (ibid.). The predicate variable in this sense is the open position in some XP which allows the XP to function as a predicate. 81 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Nonetheless, Cl-d is impossible because, as we saw earlier, the Case antilocality requirement on Cl-d is violated in this situation. Cl-d in PPs fails for the same reason: By assumption, we must take as the predicate the minimal maximal projection containing the predicate variable. In the case of doubling the complement of a preposition, PP is the minimal maximal projection containing the predicate variable (the clitic and doubled Noun Phrase chain). Therefore, the PP is the predicate in the case under consideration. The mutual m-command requirement on predication requires the NP copy to be in a configuration of mutual m-command with the PP predicate. Therefore, in order to fulfill the requirements on predication, the NP copy cannot be in clause initial position. However, if the NP copy is in a configuration of mutual m-command with the PP, as required by the syntax of predication, then the Case antilocality requirement cannot be met since no Case assigning specifiers intervene between the subject of the predication and the potentially doubled Noun Phrase in the PP. 3.3.1.1. Additional Support For a Predication Analysis We see then that a predication analysis of Cl-d in conjunction with the Case antilocality requirement captures the distribution of clitic- doubled expressions in MG. But beyond the fact that a predication analysis can effectively account for the distribution of Cl-d in MG, there 82 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is further, albeit only suggestive evidence, that predication plays a central role in the syntax of Cl-d constructions. This evidence has to do with a widely reported observation in the predication literature that predication is basically impossible when the subject of the would-be predication is an oblique object. For instance, as reported in Schein (ibid.), Pesetsky (1982) notes that small clauses cannot have dative subjects, instead the subject must be accusative; so a verb that assigns dative case to its object can have nothing predicated of that object. This is illustrated by the contrast between (50a), which has an accusative subject of the predication, and (50b), which has a dative subject of the predication: (50)a.Masha peredraznival Sashu p’janym Masha mimicked Sasha.accj drunk.instrumentalj b. *Masha podrazhala Sashe p’janym Masha imitated Sasha.dativej drunk.instrumentalj Simpson (1981) notices the same kind of facts for Finnish, where the subject of a small clause can be accusative, but not oblique. Also for Icelandic, Levin and Simpson (1981) report that the subject of a small clause can be accusative but not dative. They give the following examples: 83 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (51)a.Hann at matinn hraan he ate the meat.acc raw.acc "He ate the meat raw." b. *Hann raendi mig matnum hraum he robbed me meat.dative raw.dative * "He robbed me of the meat raw." We can observe a similar kind of behavior in Quechua (the variety spoken in Cuzco, Peru). Muysken (ibid.) argues that a co-case marking phenomenon in the language, where a Noun Phrase and an X"’“* are morphologically rnarked with the same morphological case, is a predication chain relation."'* He notes that co-case marking is possible with accusative arguments of the verb but not with genitive ones. This contrast is illustrated in (52a) ( = Muysken’s (15a)) versus (52b) ( = Muysken’s (40)): (52)a.[np e; qulqiy-ta] tarirqa-ni [^p llipin-ta]; money.acc found.Is all-acc "I found all my money." “'’Muysken (ibid.) independently proposes that Cl-d involves predication but the actual structure he proposes is quite different from the one proposed here; and, of course, his is a proposal specific to the grammar of Quechua. 84 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. *[[np6| qulqiy-niq] [qpllipin-iq]j kaypi kasquan-ta] yacha-ni money.gen all-gen this be-acc know.Is "I know that my money is all here." In addition, Williams (ibid.) notices that within the verb phrase, predication is only possible with the theme, but not the goal of a verb. He gives the following example to illustrate the ungrammaticality of predicating something of the goal even when the mutual m-command requirement on predication can in principle be met.:^^ (53) John gave Billj the dog% dead^^^.j Similar facts seem to hold for MG. So for instance a small clause within the verb phrase can have an accusative subject, but not a genitive oner^ “^In Williams’ example, there is no obvious nominal morphology indicating whether the goal is accusative or oblique since English does not morphologically distinguish the two cases. It is plausible that the goal is oblique based on Kayne’s (1984) analysis of Double Object Constructions in English. ■ ‘ ’The historically dative case assigned by certain verbs has been replaced by genitive case in standard MG and largely by accusative in certain dialects sucvh as Thessalian MG. 85 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (54)a. o Yiorghos theori tin Maria eksipni the George considers the Maria.acc smart.ace "George considers Maria smart." b. *i Maria milise [tu Yiorghuj ] methismenuj the Maria.nom spoke the George.gen drunk.gen * "Maria spoke to GeorgCj drunkj." There is only one exception that I am aware of in the literature where it is proposed that there is a predication relation involving an oblique object as the subject of the predication and even this exception is probably not really an exception. This case involves clitic doubling in Lebanese Arabic where Aoun (1991) suggests that clitic doubling of a direct object involves predication such that the subject of the predication is a dative Noun Phrase and the VP is the predicate. Aoun’s analysis is illustrated in (55): (55) [ p r e d she:fit-Oj proj] [la Kari:m]j saw.she-him to Karim "She saw Karim." If this is the correct analysis of Cl-d in Lebanese Arabic then we would have an exception to the generalization that it is not possible for an oblique object to function as the subject of the predicate. In fact, it 86 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is possible that it is incorrect to analyze the doubled object la Kari:m as being dative. Aoun indicates in a footnote that, following a suggestion of Ouhalla, the morpheme la is perhaps best analyzed as being related to the focus marker in Lebanese Arabic rather than a dative case marker. The generalization concerning predication noted in the discussion here would seem to support the conclusion that la is indeed not a dative case marker. So even in the case of Lebanese Arabic, it seems that predication does not involve an oblique argument of the verb. As an aside, it should be noted that the generalization concerning the restriction of predication to non-oblique arguments has to do with arguments and predicates related to the verb and not, for instance, objects of nouns or prepositions, or subjects and predicates within DP, all of which seem to freely be able to participate in predication even when the subjects of the predication are morphologically oblique. This point can be illustrated for MG where the object of a noun will be morphologically genitive but it is perfectly grammatical to predicate an adjective of it: (56) idha mia fotografia tu Yiorghuj methismenuj saw.Is one picture the George.gen drunk.gen "I saw a picture of George drunk." 87 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Likewise, we see it is perfectly grammatical to predicate a relative clause of a genitive noun; (57) milisa tis kopelasj puj itan sto Kastello spoke.Is the girl.gen who was in-the Castle (a cafe) "I spoke with the girl who was in "the Kastello." The point here that is relevant to the analysis of the distribution of clitic-doubled Noun Phrases in MG is that predication is in principle possible in the environments that prohibit Cl-d, so the fact that we cannot have Cl-d in those environments can be directly attributed to the Case antilocality requirement on Cl-d and not to any general restrictions on predication. Moreover, the fact that I have proposed a predication analysis of simple clitics in FPs and DPs is not at all inconsistent with the observation regarding the impossibility of predication with obliques that are arguments of the verb. What is directly relevant about the observed ungrammaticality of predication when an oblique argument of the verb would serve as the subject of the predication is the cross-linguistically observed dichotomy between the syntax of clitic-doubled accusative Noun Phrases and clitic- doubled oblique Noun Phrases. For instance Suher (1988) notesfor clitic doubling dialects of Spanish that whereas accusative doubled expressions have a limited distribution in terms of specificity of the doubled 88 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. expression (accusative clitic doubled expressions must be specific), dative doubled expressions have no such restriction. She accounts for this difference in terms of agreement features of the clitics, suggesting that accusative clitics involve [ + specific] agreement features and that dative clitics do not involve [ + specific] agreement features. Aoun (1981), Jaeggli (1980, 1981), and others discuss the differences in extraction between doubled accusatives versus doubled obliques where doubled accusatives are quite limited in terms of extractability but doubled obliques can essentially freely extract. Various proposals have been made to account for these differences. One approach among others was in terms of categorial differences between accusative Noun Phrases (which were analyzed as NPs and thus subject to the ECP) and obliques (which were analyzed as PPs and thus excempt from the ECP). Aoun (ibid.) also discusses the different behavior of clitic-doubled expressions in the loismo and leismo dialects of Spanish Cl-d dialects. The loismo Cl-d dialect doubles direct object Noun Phrases as accusative and the leismo dialect doubles direct object Noun Phrases as oblique."^ " In Spanish, case is morphologically indicated on pronouns, including clitics, but not on nouns. The only way to tell the case of the noun itself is by inspecting the case indicated on the doubling clitic. An accusative clitic presumably doubles an accusative noun and a dative clitic a dative noun. 89 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In the leismo dialects, oblique doubled arguments behave essentially like Noun Phrases which are not doubled (in terms of extractability, etc.). MG displays these same kinds of differences between accusative and oblique doubled Noun Phrases. A detailed discussion of the characteristics of doubled accusatives will follow in later chapters; but essentially, it can be said that doubled oblique Noun Phrases in MG behave like Noun Phrases that have not been doubled at all in contrast to doubled accusatives which have a unique syntax. The different behavior of clitic-doubled oblique arguments of verbs in contrast to doubled accusatives is expected in a predication analysis of Cl-d; it doesn’t need a separate stipulation, it simply follows from independently observed constraints on predication. If accusative clitic-doubled Noun Phrases are structurally predications, oblique doubled arguments of verbs could never in principle be subjects of predications and therefore could not have the same syntax as doubled accusatives. (Of course, this is assuming that the typical argument structure of sentences is not based on predication— a conclusion which does seem to be very widely held.) In the rest of this dissertation I will focus exclusively on the syntax of doubled accusatives, since, in contrast to doubled obliques, its syntax is not well understood and thus it is worthy of further investigation. Although the above observation about the failure of predication with oblique subjects of predication can’t provide more depth to an 90 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. explanation of why oblique case has a special status in Universal Grammar, relating this fact to the syntax of Cl-d does at least provide another dimension to the issue of why there is an asymmetry between the behavior of doubled accusatives and doubled obliques. 3.3.1.2. The Predicate Variable In my analysis of Cl-d, I have proposed that the clitic and doubled (accusative) Noun Phrase chain functions as the predicate variable. In the previous literature on predication, overt and non-overt pronominal elements have been proposed as predicate variables. It has never been the case that an r-expression has been observed to function as a predicate variable. I suggest that the key factor allowing r-expressions to serve as predicate variables in Cl-d constructions is the fact that they can do this only in conjunction with a clitic doubling them. It is not possible for a non-doubled r-expression to serve as a predicate variable. Verifying this claim is difficult with respect to evaluating the predicate-variable status of a Noun Phrase in argument position since we could not syntactically distinguish between r- expressions that function like predicate variables and those that don’t. For instance, any time a predicate variable is ruled out for some syntactic reason, we could not discover this via the distribution of (this kind of) predicate variables since in such a hypothetical case we could 91 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. always have an r-expression which is not functioning as a predicate variable in the same environment but we could not observe the distinction. Nonetheless, later in this dissertation we will see overt examples of the NP copy and in this case, a clitic must always co-occur with the doubled expression in argument position. If a clitic does not co occur, the supposed NP copy has different syntactic properties, ones which are not associated with Cl-d. In these cases, it is obvious that we are dealing with a very different construction not involving predication-^ and it is obvious that Noun Phrases cannot be predicate variables without a clitic doubling them. Generally in predication constructions, the predicate assigns a theta-role to the subject of the predication. The predicate variable does not itself bear a theta role, it simply marks the open place that allows something to function as a predicate. A predication analysis of Cl-d constructions suggests that the NP copy is assigned a theta role in its clause initial position via predication and that the predicate variable (the doubled NP/clitic chain) does not itself have a theta role. This result strongly accords with an intuition expressed in the literature on Cl-d in the early 1980s (see e.g., Aoun (ibid.). Borer (1981), Jaeggli (ibid.)), namely that the clitic in Cl-d constructions somehow interferes -^These are cases of focus movement. 92 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. with what was conceptualized as the government abilities of the verb such that Case or the verb’s theta role could not be assigned to a doubled Noun Phrase. For instance, Aoun (ibid.) accounts for certain extraction facts concerning Cl-d in various Cl-d languages by parametricizing whether or not a clitic in Cl-d constructions absorbed the theta role of the verb. If it did, the doubled expression was analyzed as being an adjunct rather than argument and thus as being inaccessible to extraction. The proposal that the clitic is what allows a doubled Noun Phrase to function as a predicate variable expresses this same insight in slightly different terms. I have proposed that a predication analysis can help account for the distribution of Cl-d in MG by defining the domain within which the Case antilocality requirement must operate. Moreover, 1 have pointed out how the differing behaviors of doubled accusatives versus doubled obliques are expected under a predication analysis of Cl-d of accusatives. But I’d also like to demonstrate that not only is there reason to believe that predication is at play in Cl-d constructions, there is also clear evidence that a movement or binding analysis of Cl-d is incorrect. 93 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.3.2. A M ovement A ccount of Cl-d? 3.3.2.1. An A -M o v em en t A ccount The constraints on the distribution of Cl-d are not like the ones that, for instance, govern the distribution of variables left by unbounded A ’-movement. If it were, clitic doubling would have the same distribution as wh-extraction or the extraction of focused expressions.-^ Suppose we momentarily consider the possibility of A’-movement of the doubled expression to the clause initial position as an analysis of Cl-d constructions to illustrate the (different) distribution of Cl-d if its distribution were regulated by constraints on A’-movement. When we consider the VP complement position, we find variables^” and clitic-doubled expressions have the same distribution: they are both allowed in this environment. The verbal complement position is the most canonical position that a movement-constraining principle like the ECP or relativized minimality designates as a "‘ ^Focus movement clearly behaves like A’-movement, for instance, it licenses parasitic gaps. This kind of behavior for focus movement has been widely reported in the literature on MG. ^'^In all of these cases we are interested in variables that are the result of unbounded movement such as arise through wh-extraction or focus movement; we are not concerned about the distribution of variables left by QR (Quantifier Raising), which have a much broader distribution due to the local nature of QR. 94 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. grammatical extraction site. And so, given that such a principle or theory were to govern the distribution of clitic-doubled expressions, it is not surprising that clitic doubling is allowed in this environment too. Likewise, A’-movement and clitic doubling have the same behavior when we consider their distribution with respect to PPs: neither variables nor clitic-doubled expressions are possible as complements of prepositions: It is impossible to extract wh- and focused expressions out of PPs; when questioning or focusing a Noun Phrase embedded in a PP, it is necessary to pied-pipe the entire PP (cf. (58a,b)). Similarly, it is impossible to have a clitic-doubled expression contained in a PP. (58)a.apo pion pires to vivlio? from whom took.2s the book "From whom did you get the book?" b. *pion pires to vivlio apo/ *pion pires apo to vivlio whom took.2s the book from/ whom took.2s from the book "Whom did you get the book from?" We see then that, with respect to PPs, variables and clitic-doubled expressions have the same distribution. In so far as the prohibition on preposition stranding is an ECP-type effect, it would seem that the same kind of effect were determining the distribution of clitic-doubled expressions. 95 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Up until now we have seen that clitic-doubled expressions and variables have the same distribution; however, when we consider the distribution of variables and clitic-doubled expressions contained within DP, we find that they have differing distributions. It is not possible to double complements and possessors of N. However, it is possible to extract wh- and focused elements out of Noun Phrases in MG (to Noun Phrase initial positions as well as clause initial position— for discussion see, e.g., Horrocks and Stavrou (ibid.)). The following examples illustrate these extraction possibilities: (59)a. tinoSj ipes oti xalases [to vivlio X j ] whose said.2s that wrecked.2s the book "Whose book did you say that you wrecked?" b. tu Xenophonj ipa oti xalasa [to vivlio X j ] the Xenophon.gen said.Is that wrecked.Is the book "It is Xenophon’s book that I said I wrecked." The fact that it is not possible to have Cl-ded Noun Phrases in the same positions where we can have variables which arise through unbounded A’-movement indicates that the distribution of Cl-d does not follow from constraints on A’-movement but instead adheres to a stricter requirement. If Cl-d involved A’-movement we would expect to be able to detect this via the construction’s behavior vis-a-vis movement 96 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. constraints such as the ECP. And as we will see in the next section, an examinination of the construction’s behavior with respect to other phenomenon which are taken to indicate A’-movement, such as WCO and Parasitic Gap phenomenon all lead to the same conclusion, namely that Cl-d does not involve A’-movement. 3.3.2.1.1. Other A ’-M ovement Tests The Weak Cross Over (WCO) configuration is also taken as diagnostic of the presence of a variable. This view is based on the observation that a pronoun cannot easily be coindexed with a variable to its right that the pronoun doesn’t c-command. The following example illustrates the kind of relation that is ruled out by constraints on WCO: (60)*?[[ ... pronouoj ...] ... X j ...] This means that if we find that a non-c-commanding pronoun cannot be coindexed with an expression to its right, then that expression must be represented by a variable in the expression’s argument position at LF. When we consider the behavior of clitic-doubled expressions with respect to the WCO constraint, we see that the doubled expression does not behave as if there were a variable in the position it occupies; that is, it is perfectly grammatical to coindex a clitic-doubled expression with a non-c-commanding pronoun to its left as the following example, which 97 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. indicates that a doubled and non-doubled name behave exactly the same, illustrates: (61)a.[i mitera-tUj] tonj-protim ai ton Arjirij the mother-his cl.msc.acc-prefer.3s the Arjiris "HiSj mother prefers Arjirisj." b.[i mitera-tUj] protimai ton Arjirij the mother-his prefer.3s the Arjiris "HiSj mother prefers Arjirisj." Another test of the presence of a variable is based on the licensing of parasitic gaps. In a parasitic gap construction, the presence of a variable in the overt syntax licenses, in another clause, the occurrence of a gap whose existence is parasitic on the presence of the variable. A parasitic gap construction is illustrated in (62): (62) pio arthrOj taksinomises X j [xoris na dhiavasis e.c.j] which article filed.2s without subjunctive read.2s "Which article did you file without reading?" If clitic-doubled expressions were associated with a variable (necessarily in the overt syntax— perhaps through a process of right dislocation), then a parasitic gap construction should be possible based on a clitic-doubled expression. However, Cl-d cannot license parasitic gaps: 98 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (63) *tOj-taksinomisa to arthrOj [xoris na dhiavaso e.c.j] cl.n- filed.Is the article without subjunctive read.2s *"I filed the articlej without reading e.c.j." This test, which is sensitive to the presence of a variable in the overt syntax, also leads us to the conclusion that no variable is present in Cl-d constructions. (It must be cautioned that this second test only has something to say about the presence of a variable in the overt syntax.) 3.3.2.2. The Possibility of A -M ovem ent It is equally possible logically that clitic doubling has the distribution of A-movement instead of A’-movement. And in fact we find Cl-d essentially wherever we find NP-traces (modulo the Case antilocality requirement being met). Nonetheless, given what we know about A-movement it does not seem plausible to assume that this kind of movement would be involved in the Cl-d constructions. First of all, A- movement is known to be blocked (due to relativized minimality) by an intervening A-specifier. For instance, an intervening subject blocks long distance A-movement in super-raising constructions. If a doubled expression were to undergo A-movement out of PP or DP or even VP, the subject of the clause, an A-specifier, would intervene. Nonetheless, it is not at all clear that A-movement within a clause is restricted by intervening specifiers in the same way as happens with super-raising. For 99 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. instance, an A-specifier, must, in fact, intervene between the subject’s base position and the subject’s derived position when transitive verbs are concerned. But, there are several theory-internal reasons, and one empirical reason to reject an A-movement analysis of Cl-d constructions. Consider first that even though local A-movement can skip one A-spec position without violating anything, it cannot skip more than one specifier. This has the repurcussion that when there is more than one spec in clause initial position, even modulo the expanded domain provided by head movement of the verb, A-movement would not be possible across both the subject and a wh-expression in spec of CP for instance (or a clause initial focused expression, etc.). A-movement would be in principle possible across just the subject, provided that head movement of the appropriate kind also took place. But in fact as we will see in great detail in the next chapter, the NP copy always occupies the structurally highest position in the clause and thus would sometimes have to move across a number of specifiers. I leave a confirmation of my claims here to the reader. The second theory-internal point that militates against an A- movement analysis of the distribution of Cl-d is the fact that A- movement doesn’t originate from a Case-marked position. The doubled Noun phrase is clearly case-marked with the morphological case that is 100 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. appropriate to the direct object position. So there is no motivation for A-movement. There is also an empirical reason to reject an A-movement account of Cl-d and this has to do with the behavior of Cl-d constructions with respect to reconstruction. Mahajan (ibid.) points out that expressions that undergo A-movement don’t reconstruct into their original position. In contrast, as will be discussed at length in chapter six, clitic-doubled expressions have quite a different reconstruction behavior: doubled NPs do seem to reconstruct and doubled QPs don’t reconstruct. This behavior is then quite different from A movement where NPs clearly don’t reconstruct.^* 3.3.3. A Binding Approach to Cl-d? Application of the Binding Theory to Cl-d constructions would also provide an inadequate account of the distribution of Cl-d. To see this consider the following. Suppose the NP copy binds a doubled Noun ^*After considering a movement account of Cl-d we see that the distribution of Cl-d expressions is like a classic ECP problem: we find something only in the context where it is (most) "properly governed" by the verb. Nonetheless, an account of this distribution cannot appeal to the ECP. The MG data indicate that ECP-type effects can’t all be collapsed to constraints on movement since ECP-effects and movement clearly have different domains in MG. 101 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Phrase. Binding is clearly from a non L-related position.^-- The relation between the topic and the Cl-ded Noun Phrase could a priori obey Principle A or B of the Binding Theory with respect to the A’- system (see Aoun (1982) (1985) for the theory of Binding generalized to the A’-system). Suppose a Cl-ded Noun Phrase is subject to principle A with respect to the A’-system. Then, given that they are referential expressions, Cl-ded Noun Phrases should have the same distribution as variables left by Wh-extraction of arguments since such variables obey principle A with respect to the A ’-system and principle C with respect to the A-system. Such a result is clearly contrary to fact since, as already discussed when we considered whether Cl-d obeyed movement constraints, Cl-ded NPs in MG do not have the distribution of variables: ^"Generalized Binding is the theory which generalizes the Binding theory to the A’-system. More recent developments in the Principles- and-Parameters approach to syntactic theory would not dichotomize the grammar with respect to A- or A’-positions; rather the relevant distinction would be L-related versus non-L-related. Whether this kind of dichotomy leads to different results for Generalized Binding theory is an empirical issue which, to the best of my knowledge, has not been explored. For sake of concreteness here, I will refer to A- versus A’- positions as the relevant distinction for Generalized Binding. ^^That the NP copy Is an antecedent in an A-position is unlikely given that the doubled element is an R-expression which obeys principle C: it must be free everywhere. If the NP copy were in an A-position, the doubled Noun Phrase would violate principle C by virtue of it being co indexed to the c-commanding NP copy and thus A-bound by it. 102 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. variables left by unbounded movement can occur in Noun Phrases, but Cl-ded Noun Phrases cannot. Moreover, even if Cl-ded Noun Phrases were not subject to principle C with respect to the A-system; i.e., if they were treated as non-referential, we would still get the wrong result. This is because it is perfectly grammatical in MG to extract non-referential expressions out of Noun Phrases as the following example of extraction of a quantifier indicates; (64) ollaj mu-ipes pos dhiavases [ X j ta vivlia] all cl.acc.lp-said.2s that read.2s the books "You told me that you read ALL of the books." Nonetheless, Cl-d isn’t possible in this context. In light of the above extraction possibilities for which we can provide a binding theoretic account, we would expect to find Cl-ded Noun Phrases in the same contexts only if their distribution also had a binding theoretic account. Suppose that, instead of principle A, Cl-ded Noun Phrases obey principle B with respect to the A’-system, or a similar antilocality requirement. Again, contrary to fact, Cl-ding should be possible in Noun Phrases and the PPs that form CFCs. An element subject to principle B with respect to the A’-system must be locally A’-free (in its CFC), but can be A ’-bound outside the CFC. If the CFCs for pronominal binding 103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. are IP, NP, and certain PPs (recall the different kind of PPs discussed in chapter two),^'’ it should be possible to Cl-d the complements of V, N and the relevant Ps given that they form CFCs when the topic is located outside these domains in clause peripheral position. A topic in clause peripheral position could bind any of these potentially doubled positions and still respect principle B with respect to the A’-system. A binding theoretic account of the licensing of Cl-d seems to ultimately make the wrong predictions about the distribution of Cl-d in MG. After considering movement and binding approaches to Cl-d, we find then additional support for a predication approach to Cl-d since other approaches simply don’t work empirically. 3.4. Concluding Remarks The representation proposed here for the Cl-d of accusative Noun Phrases in MG not only provides an account for the kind of data outlined above and in subsequent chapters, it will also allow us to make a direct connection between two constructions involving clitics: Cl-d and ^'’In determining the appropriate governing categories for A - pronouns we can follow similar considerations as are followed for A- pronouns (recall that anaphors and pronouns are not in exact complementary distribution). 104 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Clitic Left Dislocation (CLLD) constructions. CLLD is illustrated in (65): (65) tin Maria o Yiorghos tin-aghapa the Maria.acc the George cl.acc.f-love.3sg "(Speaking of) Maria, George loves her." In the next chapter I will identify the clause initial NP copy that I have postulated to exist in Cl-d constructions with the topic that occurs in initial position in CLLD constructions. I will suggest that Cl-d and CLLD have virtually identical syntactic representations. I will argue that the main difference between the two constructions is found at the interface level of PF. In CLLD, a PF rule deletes the doubled argument: the doubled argument is syntactically, but not phonetically, present. In Cl-d the reverse process takes place: a PF rule deletes the doubled topic, which is then syntactically, but not phonetically, present. 105 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER FOUR: CLITIC LEFT DISLOCATION CONSTRUCTIONS 4.0 Introduction In the previous chapter we saw how the hypothesis that a structurally higher Noun Phrase is coindexed with a clitic-doubled Noun Phrase helps capture the distribution of clitic doubling in MG. In the following sections we will see that there are additional reasons to believe that a structurally higher coindexed Noun Phrase does exist in Cl-d constructions. First I will demonstrate that Cl-ded Noun Phrases must be interpreted as specific (in the sense of Enc (ibid.)) and that, as such, they are related to a structurally higher position that is reserved for specific Noun Phrases; namely, the Clitic Left Dislocation (CLLD) position. Then I attempt to establish the exact structural location of the CLLDed Noun Phrase, which I will refer to as the ‘topic’, (roughly in the sense of Givon (1976), where no confusion arises. Once the location of the topic is established, this knowledge will be used to adduce further evidence supporting the proposed structure of Cl-d constructions as involving a co-indexed Noun Phrase, which we now postulate must be the topic, that is structurally higher than the Cl-ded Noun Phrase: 106 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (1) NPj clj-VNPj...] 4.1. The Interpretation of Cl-ded Noun Phrases as "Old Information" Cl-d constructions Impose the interpretive requirement that a clitic-doubled expression have a special discourse function: it must be interpreted as "old information;" i.e., essentially as a topic given by the previous discourse. This fact has long been noted in research on MG; cf., for instance, Haberland and Van Der Auwera (1990), Horrocks (1983), Joseph (1983), Kazazis and Pentheroudakis (1976), Mackridge (1985), Philippaki-Warburton (1975, 1985), Stavrou (1985), Schneider-Zioga (1990), Tzartzanos (1946); see also Valiouli (1990). With respect to the "old information" requirement on Cl-ded Noun Phrases, we see that they must be specific in the sense of Enc (ibid.), as discussed in Schneider-Zioga (ibid.). Enc defines specificity in the following way: "[a]n NP is specific if and only if its head is definite." She extends Heim (1982) and Kamp (1981) by suggesting that "a noun which is indefinite introduces a new set into discourse, and a noun which is definite is mapped onto a set that is already in the discourse (p.13)." Enc notes that whether an NP itself is definite or not depends on properties of the determiner; e.g., some determiners are inherently marked as to definiteness (for instance, "the"). The definiteness of the determiner 107 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. determines the definiteness of the Noun Phrase. Determiners are lexically specified as to whether they select a + definite and + singular head. The intuition behind her analysis is that specificity is covert partitivity. Loosely speaking then, specific Noun Phrases represent "old information." Enc indicates that a good syntactic test of specificity is found with there-sentences. which, she claims, cannot contain specific Noun Phrases. This observation provides us with a heuristic device to establish that non-specific Noun Phrases cannot Cl-d. The MG analogue of the there-insertion construction involves the verb exi (literally ‘to have’) in its existential interpretation.^ The following (non-exhaustive) paradigm lists Noun Phrases that can occur in this construction: ^ Exi can only be interpreted as existential when it takes a locative PP as a complement; for instance: (i) exi ena ghaidaro sto kipo have a donkey in.the garden "There is a donkey in the garden." 108 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (2) exi _________________ stin avli (there is in the courtyard) (ena) ghaidaro ((a) donkey) ena (one) alio ghaidharo (another donkey) merikus (some/ a few) kati bukalya (some bottles) toso (so much) tosa luludhia (so many flowers) tetio luludhi (that kind of flower) ap’ ola (of all, i.e.: something of everything) dhen tipota (not...something, i.e.: nothing) dhen katholu luludhia (not at all (any) flowers) pola luludhia (many flowers) ligho kafe sto flizani (a little coffee in the cup) arketo kafe sto flizani (enough coffee in the cup) kamposo kafe sto flizani (quite a bit of coffee in the cup) None of the Noun Phrases listed above can undergo Cl-d. These facts indicate that non-specific Noun Phrases cannot Cl-d. Consider now the (non-exhaustive) paradigm of Noun Phrases that cannot occur in the existential exi construction. Only Noun Phrases that fit this paradigm can Cl-d.; (3) exi ________________ stin avli (there is in the courtyard) *afto/ekino (this/that) *afto to luludhi (this the flower, i.e.: this flower) *opiadhipote luludhia (whatever flowers) *kathe ghata (every cat) *to kathe ghata (the every cat) *i mia ghata (the one cat) *ke i dhio ghates (and the two cats, i.e: both cats) *tin Maria (the Maria) 109 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In addition to Noun Phrases, CPs can also Cl-d in MG. In this case they are interpreted as "factive" when they are Cl-ded; i.e., they are understood as necessarily true. Kiparsky and Kiparksky (1971) suggest that this property can be understood as the sentential equivalent of a specific Noun Phrase. The following examples illustrate the behavior of doubling with CPs: (4) dhen to-fandastike oti les psemata not cl.n-imagined.ls that say.2s lies "I never would have imagined that you tell lies!" (The implication is that you really do tell lies, I just wouldn’t have imagined it if you hadn’t told me.) (5) dhen to-perimena na erthis noris not cl.n-expected.ls subjunctive arrive.2s early "I didn’t expect you to arrive early." (The implication is that you really did arrive early, I just didn’t expect that to happen.) (6) to-anarotieme pics podise ta luludhia cl.n-wonder.ls who watered the flowers "I wonder who watered the flowers." (The implication is that someone really did water the flowers.) 110 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Another way the "old information" requirement is manifested in Cl-d constructions is noted by Philippaki-Warburton (1976, 1985), who points out that the effect of Cl-d is to defocus the doubled NP, which cannot be understood as part of the presentational focus' of the sentence. Philippaki-Warburton notes that this requirement prevents a normal intonational pattern for a sentence with a doubled Noun Phrase. The unmarked intonational pattern in MG is when the sentence final constituent carries the most prominent stress, as is illustrated in (7): (7) o Aleksandhros protimai tin RANI the Alexander prefers the RANI "Alexander prefers RANI." But when Cl-d of the direct object "tin Rani" occurs, main stress must fall either on the verb, or the subject, it may not fall on the doubled Noun Phrase. Cf. (8)a.,b. and c.; (8)a.o Aleksandhros tin-PROTIMAI tin Rani the Alexander cl.acc.f-prefer the Rani.acc "Alexander PREFERS Rani." " For a discussion and definition of the concept "presentational focus" see Rochemont (1986). Ill Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b.o ALEKSANDHROS tin-protimai tin Rani "ALEXANDER prefers Rani." c. *o Aleksandhros tin-protimai tin RANI ^"Alexander prefers RANI." The doubled Noun Phrase is not allowed to be part of the presentational focus as confirmed by the intonational patterns of sentences with Cl-ded Noun Phrases. Along similar lines, Schneider-Zioga (ibid.) points out that Cl-ded Noun Phrases in MG cannot be modified by focus operators such as akoma (even) and mono (only) that "associate" (in the sense of Jackendoff (1972)) with focused constituents.^ For example: (9)a. o Yiorghos aghapai mono/akoma tin RANI the George loves only/even the Rani "George loves only/even RANI." b.*o Yiorghos tin-aghapai mono/akoma tin RANI the George cl.acc.f-loves only/even the Rani ^ The modification by a focus operator is not possible only on the focus reading of the Noun Phrase. If the modified Noun Phrase is not focused, doubling is permissible. 112 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Moreover, Cl-ded expressions cannot be contrastively focused'* although it seems a contrastively focused Noun Phrase could be interpreted as old information. The following examples illustrate the prohibition in MG against doubling focused Noun Phrases: (10)a. o Aleksandros aghapai tin Rani, oxi ton Yiorgho the Alexander loves the Rani, no the George "Alexander loves Rani, not George." b. *o Aleksandros tin-aghapai tin Rani, oxi ton Yiorgho the Alexander loves the Rani, no the George "Alexander loves Rani, not George." In sum, the data indicate that Cl-d is possible only with discourse topics in MG and that this process defocuses the doubled expression. 4.2. The Interpretation of C litic L eft D islocated NPs as "Old Information" There is another construction in the grammar of MG that has precisely these same properties: namely the Clitic Left Dislocation (CLLD) construction where a Noun Phrase corresponding to an internal ■ * In this way MG differs from clitic-doubling languages that do allow contrastively focused Noun Phrases (as well as topics) to Cl-d. An example of such a language is Peruvian Spanish, which allows doubling of focused Noun Phrases as well as topics. 113 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. argument of the verb occurs in left peripheral position. The Noun Phrase is doubled by a pronominal clitic on the verb. The following example illustrates the construction: (11) tin Maria o Yiorghos tin-aghapa the Maria.acc the George cl.acc.f-love.3sg "(Speaking of) Maria, George loves her." As discussed in latridou (ibid.), CLLDed Noun Phrases are interpreted as old information. latridou points out, for instance, that in the following kind of example, the CLLDed Noun Phrase is understood as having been previously mentioned in the discourse as indicated by the fact that the CLLD (see (13)) can answer (12a) but not (12b) ( - latridou’s (9a), (7a), and (7b) respectively): (12)a. Who saw Mary? b. Who did Kostas see? (13) tin Maria o Kostas tin-idhe the Maria.acc the Kostas.nom cl.acc.f-saw.3s We see then that, as with clitic-doubling structures, CLLD constructions involve "old information." Moreover, as with Cl-ded Noun 114 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Phrases, the CLLDed Noun Khrase must be specific. The following example is representative;^ ( 14)a. *arketa provlimata dhen-ta-exo enough problems not-cl.acc.n(pl)-have b.arketa provlimata dhen-exo enough problems not-have.ls And CPs can also CLLD, in which case they are interpreted as factive. This property is also parallel to clitic doubling constructions. The following example illustrates the CLLD of a CP: (15) oti les psemata dhen to-fandastika that say.2s lies not cl.acc.n-imagined.ls "That you lie, I would not have imagined it." Just as with Cl-ded Noun Phrases, a CLLDed Noun Phrase may not bear focal stress; i.e., it must be defocused: ^It is true that CLLD and Cl-d both require the doubled expressions to be old information, but what can clitic double is, in many languages, a subset of what can CLLD. This fact is discussed at length in chapter five. 115 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (16)a. tin Maria o YIORGHOS tin-aghapa the Maria.acc the GEORGE cI.acc.f-love.3sg "Maria, GEORGE loves her." b. tin Maria o Yiorghos tin-AGHAPAI the Maria.acc the George cl.acc.f-LOVE.3sg "Maria, George LOVES her." c. *tin MARIA o Yiorghos tin-aghapai the MARIA.acc the George cl.acc.f-love.3sg "MARIA, George loves her." Like Cl-ded Noun Phrases, a CLLDed Noun Phrase may not be contrastively focused: (17)a. *tin Rani, o Aleksandros tin-aghapai, oxi ton Yiorgho the Rani.acc the A. cl.acc.f-love.3s, not the George.acc b.tin Rani, o Aleksandros aghapai, oxi ton Yiorgho the Rani.acc the Alexander love.3s, not the George.acc "(It’s) Rani, Alexander loves, not George." Nor may CLLDed Noun Phrases be associated with focus operators; again, this is a property shared by the doubled expression in Cl-d constructions: 116 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (18)* akom a/m ono i Rani tin-aghapo even/only the Rani cl.acc.f-love.ls 4.3. The Link between Cl-d and CLLD Constructions The striking similarities in interpretation and constraints on interpretation shared by the two constructions suggest that the two positions of the Noun Phrases in Cl-d and CLLD constructions are syntactically related in some way. To clarify the issue, consider for a moment an analogous situation: the issue of the relation between wh- elements that have moved in the syntax versus those that remain in-situ (due, for example, to the fact that there are two wh-elements in the clause and the language respects the doubly-filled comp filter, or because the language is of a type like Chinese or Japanese which might appear to not have overt wh-movement). Since both kinds of wh- elements (moved and in-situ) are interpreted in the same way, namely, as having scope over a sentence, any theory of the syntax of wh- constructions wants to relate these two positions. Generally one of the following positions is assumed: that wh-elements in-situ move at LF to the same position they would have moved to if they had moved in the overt syntax or, increasingly, wh-movement is being reconsidered as either always involving overt movement (cf. e.g. Watanabe (1991) or as— at least sometimes— involving a Qu(estion)-operator unselectively 117 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. binding an indefinite wh-word (of. e.g., Aoun and Li (1993), Cole and Hermon (1994)). In sum, since both moved and in-situ wh-elements are interpreted as taking wide scope over the sentence, both kinds of elements are assumed to either actually occur in the understood position by LF if not in the overt syntax, or to be related by unselective binding to the position where wh-elements are understood as taking scope. Following the same kind of reasoning, we want to capture the fact that Cl-ded expressions and CLLDed expressions are necessarily understood as the ‘topic’ of a sentence about which the rest of the sentence is the ‘comment’ and we can do this by relating the two positions: In the previous chapter, I suggested this relation is established via predication; in the next chapters, I discuss the relation between the two positions in greater detail. Based on the above considerations, we can take the CLLDed position to be the unidentified structurally higher Noun Phrase that was postulated in the previous sections to exist in Cl-ding constructions. Specific evidence in support of this conclusion is given in the ensuing discussion. 118 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.4. The Clitic as T(opic)-M orphem e The clitic itself can be understood as the topic counterpart to a Qu(estion)-morpheme.‘ " Qu-morphemes are employed by some languages to mark the occurrence of a yes/no and/or wh-question and they occur in the clause that the question operator is understood to have scope over (cf. for instance, Aoun and Li (ibid.), Benmamoun (1992)). The Qu-morpheme does not necessarily mark the syntactic scope of the question operator in the sense that a question operator might occur quite a bit higher structurally than the Qu-morpheme itself or vica versa. The Qu-morpherne cannot be equated to a syntactic scope marker because it doesn’t mark the syntactic scope of the question operator, although certainly, scope is always determined with respect to the Qu- morpheme since the question operator must be in the same clause as the Qu-morpheme. The pronominal clitic can be thought of in a parallel way, as a morpheme which signals the occurrence of a topic (old information) construction (hence: T-morpheme) in the clause containing the clitic. However, the location of the T-morpheme does not structurally correspond to the actual syntactic scope of the topic which evidence ‘ ’In this dissertation I do not explore the syntax of the clitic itself in any detail. 119 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. indicates is structurally much higher than the clitic; it is outside of the minimal maximal projection containing the clitic and structurally higher than any other non-L-related expressions that occur in the same clause. 4.5. Topics in CLLDed Position In order to establish the location of the topic, I will consider the various possible orderings for preverbal specifiers with respect to each other and with respect to the clitic when there is an overt topic, namely in CLLD constructions. In this way we will see that the topic is located structurally higher than any other preverbal expression (such as focused or wh-expressions. Finally I will demonstrate that Cl-d constructions behave as if they also have a topic occupying the same position as topics clearly do in CLLD constructions. 4.5.1. The R elative Ordering of W h-Expressions and Subjects Before investigating the ordering of the topic relative to other preverbal specifiers, it is necessary to explore and understand the seemingly complex ordering requirements between the various possible preverbal specifiers. Consider first wh-questions. MG requires subject/verb inversion with wh-questions in both root and embedded contexts, so it is generally assumed that these constructions involve V" to C" movement. The following representative examples illustrate this: 120 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (19)a.pion idhe i Sophia who.acc saw the Sophia.nom "Who did Sophia see?" b. *pion i Sophia idhe who.acc the Sophia.nom saw (20) U p WH U - V [ip ... ]]] 4.5.2. The R elative Ordering of Focused Expressions and Subjects Focus constructions^ contrast with wh-constructions in that there is no unwavering requirement that the verb occur next to a preverbal focused Noun Phrase. This observation is confirmed by the following example, where a focused Noun Phrase precedes the subject which precedes the verb: (21) to MORO o Yiorghos frondise the BABY the George cared.for.3s "George cared for the BABY." The fact that the verb does not need to be adjacent to the focused expression* (at least not in the overt syntax) indicates that V“ to C° follow traditional practice of capitalizing focused expressions. *In the literature there is some disagreement about the data, with some linguists suggesting that the focused Noun Phrase in fact must 121 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. movement is not obligatory with focus constructions in the dialect under investigation here. I will discuss this point in more detail shortly. Focused expressions must precede the subject (recall (21) above). The opposite ordering, while not ungrammatical, is interpreted as a dislocation of the subject: (22) o Yiorghos, to MORO frondise the George, the BABY cared.for "(speaking of) George, he took care of the baby." 4.5.3. The Relative Ordering of Focused Expressions and CPs 4.5.3.1. Focused Expressions and Wh-Expressions It is well known in the literature on MG (see, for example, Agouraki (1990) and Tsimpli (1990)) that there is an asymmetry between matrix and embedded clauses such that focused expressions and wh- expressions are in complementary distribution in matrix clauses but can co-occur in embedded clauses. The following examples illustrate this contrast. Examples (23) and (24) illustrate that wh- and focused expressions are in complementary distribution in matrix clauses: immediately precede the verb. However, a number of linguists, Philippaki-Warburton for instance as well as latridou find FOCUS S V word order perfectly grammatical; most of the native speakers I consulted agree with these later judgements. 122 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (23)a.ti dhiavaze i Maria what read the Maria.nom "What did Maria read?" b.i MARIA dhiavaze tin efimeridha the MARIA.nom read the newspaper "MARIA read the newspaper." (24)a. *i MARIA ti dhiavaze the MARIA what read b. *ti i MARIA dhiavaze what the MARIA read This complementary distribution can be contrasted to the behavior of wh- and focused expressions in embedded clauses, illustrated in example (25), where we see that they do co-occur; (25) anarotieme to M O R O j pioSy frondize X y X j wonder.Is the BABY who cared.for "I wonder who took care of the BABY." Although wh-expressions and focused expressions can co-occur in embedded clauses, they are not freely ordered with respect to each other. The contrast between (25) above and (26) below indicates that a focused expression must precede a wh-expression.: 123 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (26) *anarotieme pioSy to MOROj frondize X y X j‘ ^ wonder.Is the BABY who cared.for "I wonder who took care of the BABY." The following table summarizes the distribution of wh-expressions and focused expressions in matrix and embedded clauses respectively: (27) CO-OCCURRENCE AND RELATIVE ORDERING RESTRICTIONS ON WH-EXPRESSIONS AND FOCUSED EXPRESSIONS MATRIX CLAUSE EMBEDDED CLAUSE WH: WH > V WH > V FOCUS: FOCUS (SUBJECT) V FOCUS (SUBJECT) V FOCUS & WH; * FOCUS > WH 4.5.4. Focused Expressions and C om plem entizers When we consider the ordering requirements of focused expressions with respect to complementizers, we find a contrast to the ‘ ^The ordering requirements here have nothing to do with nested versus crossing depndencies since the ordering constraints hold regardless of what grammatical function the relevant expressions have as illustrated here where the grammatical functions of focused and wh- expressions are reversed resulting in nested dependencies and the sentence is still ungrammatical: (i) *anarotieme piony o YIORGHOSj frondize X j X y wonder.Is who.acc the George.nom cared.for 124 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. above ordering requirements. Moreover, there are two dialects to consider here: Most speakers I consulted require a focused expression to follow a + or-wh complementizer. The following examples illustrate the required ordering for this dialect with +wh and -wh complementizers respectively: (28)a.me-rotise an ta VIVLIA epestrepsa me-asked if the BOOKS returned.Is "He asked me if I returned the BOOKS." b.[...V [ + wh complementizer [FOCUSed NP [jp NP....]]]] c. *[...V [FOCUSed NP [ + wh complementizer [jp NP....]]]] (29)a.nomizo oti to MORO o Yiorghos frondize think.Is that the BABY the George cared.for "I think that George cared for the BABY." b.[...V [-wh complementizer [FOCUSed NP [jp NP....]]]] c. *[...V [FOCUSed NP [-wh complementizer [jp NP....]]]] There is a second dialect, which contrasts with the above in that it observes the following requirement: as long as CP does not contain a wh-expression, focused expressions and CPs are freely ordered with respect to each other. The following examples illustrate how, in this second dialect, a focused expression can precede a +wh or -wh complementizer (example b. is from Tsimpli (ibid.) ( = her (51c))): 125 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (3ü)a.me-rotise ta VIVLIA an epestrepsa me-asked the BOOKS if returned.Is "He asked me if I returned the BOOKS." b.[...V [FOCUSed NP [ + wh complementizer [jp NP....]]]] (31)a.nomizo to MORO oti o Yiorghos frondize think.Is the BABY that the George cared.for "I think that George cared for the BABY." b.[...V [FOCUSed NP [-wh complementizer [jp NP....]]]] In this second dialect, focused expressions can also follow a +wh or -wh complementizer as is possible in the first dialect too. This possibility is illustrated for the first dialect in (28) and (29). The following chart summarizes the data discussed up till now; (32) WORD ORDER REQUIREMENTS DIALECT ONE DIALECT TWO WH/V: FOCUS/WH; WH > V FOCUS > WH FOCUS/COMP.: COMPL > FOCUS WH > V FOCUS > WH FOCUS > COMPL COMPL > FOCUS 126 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.5.4. Word Order Accounts From the Literature 4.5.4.1. A CP Recursion Analysis There have been several proposals in the literature, all of them made with only dialect two in mind, which are intended to account for ordering facts in embedded contexts. For instance, latridou (ibid.), in a study on CLLD, proposes that MG allows CP recursion in embedded contexts in which case the following configuration is permitted: (33) [ip---V [(.p] [cp2 [........]]] Her proposal of CP recursion is made with the relative ordering of CLLDed Noun Phrases and complementizers in mind, and not focused expressions, but it is easy to imagine how her proposal might be extended to explain the variable ordering of focused expressions and complementizers. For example, the possible pre-complementizer ordering of focused expressions in dialect two can be attributed to the focused expression occuring in CP-1 and the complementizer in the head of CP-2 and vica versa for the post-complementizer ordering of the focused expression. And the fact that wh-expressions must follow focused expressions can be explained via a spec/head "agreement" requirement between a + wh- expression and the verb, which holds of the overt syntax. Such a 127 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. requirement is something which we know to be independently necessary in order to explain the fact that subject/verb inversion is obligatory with wh-questions in root and embedded contexts. The spec/head agreement requirement would force the wh-expression to be in a spec/head relation with the verb and hence structurally lower than the focused expression which we have seen seems to have no such agreement requirement placed on it (recall (21)). Although the ordering facts for dialect two can be captured under a CP-recursion account, several points seem to militate against this kind of analysis. For instance, consider the following data. Focused expressions do not constitute islands for the extraction of wh- adjuncts:’” (34) posj kseris oti AFT! ravun tora pia X j how know.2s that THEY sew.3p now-still "How do you know that THEY still sew?" A possible answer to (34) is: me palies mihanes ‘with old sewing machines’, which confirms that the adjunct is interpreted as originating ‘°This is despite the fact that both kinds of expressions behave like they have undergone A’-movement, a fact which has been widely noted in the literature and will also be briefly discussed here in this dissertation. 128 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in the embedded clause. While the possibility of extraction across a focused constituent might be explained as being possible due to the existence of the additional CP escape hatch in the embedded clause, the CP-recursion hypothesis could not explain the impossibility of extracting wh-adjuncts across wh-expressions (example here from Agouraki (ibid.)( = her (43));" (35) *poSj kseris posi ravun tora pia X j how know.2s how.many sew.3p now still * "HoWj do you know how many people still sew Xj?" The above should, contrary to fact, be perfectly acceptable if an additional CP escape hatch exists in the embedded clause. Moreover, why is it forbidden for there ever to be what would look like a doubly-filled comp in embedded contexts given the existence of two CPs? The impossibility of two wh-expressions occurring in the embedded comps is not problematic for the CP-recursion analysis: it could be explained by a failure of spec/head agreement for the less "Wh-expressions also don’t intercept extracted focused expressions; so, in terms of island creation, the two types of expressions seem to run on parallel tracks, as do CLLDed phrases which are intercepted neither by wh-expressions in weak islands nor focused expressions and the reverse (CLLDed phrases intercept neither wh-expressions nor focused expressions). 129 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. deeply embedded wh-expression (recall that MG requires inversion in the overt syntax with wh-questions in both root and embedded contexts). But why is it impossible to have two focused expressions originate and/or occur in the embedded clause, as illustrated below, given the existence of two CPs which equally accept focused expressions in the embedded clause and the fact that there is no spec/head licensing requirement between verbs and on focused expressions that holds of the overt syntax? Cf.: (36)* [jp V [(.p focused NPj [^p focused NPy [ ...X j ...X y . . . ] ] ] ] If there are two CPs which can equally accept focused expressions, nothing would seem to prevent the configuration in (79), yet it is still ungrammatical. Finally, it is not clear how a CP recursion analysis can account for dialect one.^“ ^ “Moreover, if we consider MG from the perspective of the Minimalist Program, there is a further reason to reject a CP recursion analysis of embedded clauses in MG that is technical. Recall our working assumption that the spec of one of the CPs is the location of the preverbal focused expression. Now consider that the Shortest Movement economy principle requires a lexical item to check its features in the first possible position. For instance, if we are considering A-movement of an XP, then the XP must move to the closest possible position (within the parameters of the domains defined by head movement) where it can check its Case and/or agreement features. Similarly, a wh-expression must move to the first possible specifier where it can check its features. Since wh-expressions and focused expressions can co-occur in MG, with a focused expression being located structurally higher than a wh- expression, it must be that they do not have the same kind of features to 130 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I find the above objections persuasive enough to cause one to favor more the competing analyses of preverbal specifier phrase structure which hypothesize two separate projections: a CP and an FP (Focus Phrase projection). 4.5.4.2. Tsimpli’s Proposal of a Fixed Order: CP, FP, IP Next I’d like to consider and ultimately reject Tsimpli’s (ibid.) proposal, which was also made with dialect two in mind. Tsimpli suggests that matrix clauses in MG have only FPs and no CPs whereas embedded clauses have CPs which select FPs. The following examples illustrate her analysis schematically: (37)a. [fp [ip ... ]] MATRIX CLAUSES b-[fp [ip ...V Lp [fp [...]]]] EMBEDDED CLAUSES Tsimpli’s analysis forces her to allow wh-expressions to occur in FPs (for wh-questions in matrix and embedded contexts) and focused expressions to occur in CPs (when focused expressions precede complementizers). It seems problematic for Tsimpli’s analysis that check since otherwise the structurally lower specifier containing the wh- expression would intercept the focused expression. Since this does not happen, we can conclude that there are two separate projections: one where wh-features can be checked and one where focus-features can be checked. 131 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. focused expressions can precede +wh complementizers (e.g. an ‘if’); I repeat here (30)a. for convenience: (30)a.me-rotise ta VIVLIA an epestrepsa me-asked the BOOKS if returned.Is "He asked me if I returned the BOOKS." As I read her analysis, Tsimpli would assign (30)a. the following structure, with the focused expression occurring in the spec of a +wh CP: (30)a.’[jp me-rotise [^p ta VIVLIA [^. an [fp [jp epestrepsa]]]]] me-asked the BOOKS if returned.Is It seems unlikely, due to feature clash, that a -wh expression (the focused expression) would occur in the spec of a CP that is headed by a + wh-complementizer. Also, Tsimpli must account for the obligatory echo interpretation of a wh-expression which, according to her analysis, occurs in the spec of CP in (38): (38) su-ipan pjon oti tha dhune you-told.3p who that will see.3p "They told you that they will see WHO?" (echo interpretation) 132 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The fact that the wh-expression in (38) can only be interpreted as an echo suggests one of two things: (a) that the head of CP, here -wh, does in fact dictate the nature of the projection, contrary to what must be assumed to allow for facts like in (30) under Tsimpli’s analysis. If this is true, a -wh head cannot have an expression in its specifier that is interpreted as + wh. Any wh-expression in such a specifier position will be interpreted only as focused (as wh-expressions generally seem to be when interpreted as an echo) if it is not interpreted as -rwh (or cannot be interpreted in this way); or (b) that the wh-echo appears in the spec of an FP that is higher than the CP headed by the complementizer. If (a) is true, then Tsimpli has no direct account of (30) where a focused expression appears in the specifier of a + wh-headed CP. And if (b) is true, then there is a different structure than that proposed by Tsimpli. Moreover, if there is a semantic correspondence between the fact that an expression occurs in the spec of FP and the interpretation of an expression as focused, Tsimpli’s analysis, which frequently locates wh- expressions in FP, would not always be compatible with the interpretation of wh-expressions (see Agouraki (ibid.) for discussion of ‘^Moreover, the echo interpretation of the wh-expression in sentences like (38) also seems problematic for a CP recursion story since the wh- expression would be occurring in a spec of CP which isn’t headed by a complementizer and nonetheless be interpreted as an echo-question. 133 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the implausibility of an analysis of wh-expressions as always being focused). It should, however, be noted that Tsimpli can straightforwardly account for dialect one by proposing that, in dialect one, focused expressions are subject to some version of the doubly-filled comp filter and thus cannot co-occur with complementizers. 4.5.4.3. Agouraki’s Proposal of a Free Ordering of CP and FP The third analysis, also made with only dialect two in mind, is that of Agouraki (ibid.). Like Tsimpli, she posits a CP and an FP but she allows them to be freely ordered with respect to each other with licensing requirements on wh-questions and focus regulating the order that actually surfaces. For example, certain licensing requirements on focus derive the fact that focus constructions and wh-constructions are in complementary distribution in matrix clauses: She allows a verb to assign either the feature -t-wh or +focus, but not both. In matrix contexts only a focused expression or a wh-expression can be licensed, but not both. In embedded contexts, the superordinate verb can assign the feature -t-focus to the embedded FP in case the embedded verb assigns -t-wh: a process of "exceptional focus marking," in the same spirit as Exceptional Case Marking. Since CP is a barrier, FP must be structurally higher than CP for the superordinate verb to assign -rfocus. 134 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In this way, an embedded clause can have both a wh-expression and a focused expression. Under Agouraki’s analysis, when a focused expression precedes a complementizer, it is occuring not in the specifier of that CP, but in a structurally higher FP. For example, she would assign (30)a. above, the following structure: (3 0 )a .” [jp me-rotise [fp ta VIVLIA [^p [,. an [jp epestrepsa]]]]] me-asked the BOOKS if returned.Is "He asked me if I returned the BOOKS." She would assign (28)a., which has the reverse ordering between the complementizer and focused expression, the following structure: (2 8 )a .’[jp me-rotise [< ,p [< ,- an [^p ta VIVLIA [jp epestrepsa]]]]] me-asked if the BOOKS returned.Is "He asked me if I returned the BOOKS." Since Agouraki’s analysis was made with only dialect two in mind, it is necessary to expand on some details of her analysis in order to account for the existence of dialect one as well. A plausible proposal might relate the different syntaxes in these two dialects to morphological differences between the two dialects: Suppose first that, in addition to being licensed in the manner Agouraki proposes, all preverbal focused 135 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. expressions must also stand in a spec/head relation with the verb to which they are thematically related at least by LF, if not in the overt syntax. This means that, for example, one of the following configurations would be required at LF: (39)a.[fp focused expression [pV [^p [jp...... ]]] b.[cp [fp focused expression [pV [jp.......]]]] Consider the consequences of this proposal now in embedded contexts. Suppose that in dialect two, the complementizer has both a strong and weak form (in the sense that it can be either an independent word or a clitic/contracted form); this means, that the verb is able to raise to C" in this particular dialect which fulfills the morphological requirement on weak forms that they can’t be independent words.*'' The only additional proviso that would be necessary would be the assumption that adjunction of the inflected V to C **, and its subsequent raising to F", wouldn’t suppress the verbal features necessary to meet the putative '''This view doesn’t force the conclusion that the verb raises in order to fulfill the morphological requirements of the weak complementizer (a situation which would be in violation of the principle of Greed). Instead, it must be that verb raising happens because this satisfies some feature of the verb itself; if a weak complementizer occurs and verb raising hasn’t happened, the derivation would be ungrammatical due to the fact that the weak complementizer remains a free morpheme. 136 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. spec/head requirement between focused expressions and their thematically related verbs. This analysis means that when there is a focused expression in precomplementizer position, the verb raises at LF first to and ultimately to F°. In this way, the grammar satisfies the morphological requirement of the weak complementizer, which must be a bound morpheme, and since no independent word exists to block head movement, the grammar can also fulfill the spec/head agreement requirement between the focused expression and its thematically related verb in the process; thus allowing for precomplementizer focused expressions in this dialect. The following example illustrates the analysis: (40)a.OVERT SYNTAX: [...V [fp focused NP [f. [^p [,. o ti/a n [jp ...V...]]]]]] b.LF: [...V [fp focused NP [f. W+oti/an-^ [,p [,. tj [jp ...]]]]]] In dialect one, in contrast, the complementizer is an independent word. In this way, head movement of the verb is not possible past the complementizer in order for the verb to stand in a spec/head relation with the focused expression. Therefore, only post-complementizer focused expressions are allowed since it is only in this configuration that the verb can stand in a spec/head relation with the focused expression. 137 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. When we consider the relative ordering of wh-expressions and focused expressions, both dialects behave the same because there are no complementizers or other words involved to block head movement; so the verb can always ultimately move through to F". Of course, it is necessary to assume that in wh-constructions, the verb needn’t remain in a spec/head relation with the wh-expression, in order for a wh- expression to be licensed. The only requirement seems to be that the wh- expression must be licensed in the overt syntax.*^ Agouraki’s analysis with the additional details I have proposed captures a great deal of data, yet there is a conceptual reason to reject it and at least one empirical reason. From a conceptual point of view, the aspect of Agouraki’s analysis that assumes functional projections are freely ordered with respect to each other (in the sense that any one can select the other) is problematic. Usually it is assumed that the ordering of projections is ^^The opposite kind of ordering where FP occurs lower than the wh- expression represents the kind of vacuous movement that is ruled out by economy principles. Consider the following structure: (i) * [,p wh [,. V°j [fp focused NP [f. tj [jp ...tj ... ]]]]] Raising the verb past F° to C° in order to produce the spec/head relation that is required to hold between V" and C" in the overt syntax would produce a structure that has a much longer derivation than the following structure that doesn’t involve this kind of vacuous movement: (ii) [^p wh [^. V ® [jp ... tj ... focused NP ... ]]] In (ii), there is simply a VP internal focused expression, as opposed to a focused expression that is licensed in the preverbal specifier positions. 138 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. determined by universal grammar and languages don’t vary along the lines of which functional projections select which other ones and that there should be this kind of variation within the same language as would be the case under Agouraki’s analysis is even more unexpected.'^ From an empirical point of view, it is difficult for the amended version of Agouraki’s analysis to account for certain word orders. For instance, there are some problems with the covert verb-raising hypothesis. Recall that certain speakers require subject/verb inversion in the overt syntax when a focused expression is in initial position; i.e., the focused expression must be followed by the verb for these speakers. This requirement cuts across dialects, so there are speakers in dialect one and dialect two who have this inversion requirement. But if focus/subject-verb inversion is required in the overt syntax for a speaker of dialect two, how do we reconcile this with the fact that it is no longer required in the overt syntax (but only covertly instead) when a focused expression precedes a complementizer (or a wh-expression for that matter)? '^This is the same kind of problem that has arisen in the context of studying the various orderings of inflectional projections: For instance, does negation select agreement in one language and the reverse in the other language? 139 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Furthermore, there are some problems with the ordering of XFs and the complementizer when topics are also considered. For instance, as we will see in some detail later, when a topic, a focused expression and a complementizer occurs, only the topic can precede the complementizer. The focused expression must follow the complementizer. In order to account for this ordering requirement, some additional machinery must be postulated. Furthermore, topics must always be structurally higher than focused expressions, an account of which would seem to require even more machinery. It does not seem easy under Agouraki’s proposal (with the addendum to cover dialect one) to provide an elegant, unified account of preverbal word order facts when more complex word orders are considered. Given the conceptual and empirical concerns discussed above, it might be fruitful to reconsider her approach to the relative ordering of preverbal specifiers in MG in favor of another approach. 4.6. A M ovem ent Approach to R elative Ordering With respect to relative ordering of CP and focused expessions, there is a fourth possible analysis to be considered and that is that the variation in word order is due to movement; i.e., we can consider the possibility that, for example, the focused expression moves to FP and can further move to another non-L-related position; or alternately, the 140 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. complementizer can undergo head movement out of its base-generated position to a higher non-L-related head, etc. This line of thinking looks at variation in word order of expressions that represent discourse functions such as topic and focus and explores it along the lines of research that has studied the relative ordering of subjects and objects with respect to the verb. In other words, this approach considers the relative ordering of expressions that occur in non-L-related positions and attributes variability in ordering to movement within the various non-L- related projections. There are two possible word orders to consider as the base order once we assume with Agouraki that the asymmetry between focus and wh-expressions in matrix and embedded clauses is due to a process of "exceptional focus marking": (a) either the base order is CP FP: (41) ...[(.p [fp [jp .... ]]] ... or, (b) the base order is FP CP: (42) ...[fp [cp [jp .... ]]] ... Any apparent variations in word order between FP and CP such as the complementizer preceding the focused expression or the focused expression preceding the wh-expression would be due to movement of some kind. 141 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Suppose that the base order were CP FP as in (41 ) (repeated here for convenience): (41) ...[(.p [fp [ip — ]]] Consider first what this means for dialect one: when the complementizer precedes the focused expression, there has been no movement, the simple base order is presented. Cf.: (43) ...[ep [c-oti [fp to M ORO [jp .... ]]] ... that the BABY ..... When a wh-expression occurs, the focused expression must precede it and therefore, under the structure in (41), the focused expression must have undergone obligatory movement to some as yet unnamed non-L- related position that dominates CP. For instance: (44)a.base generated order: ...[cp Pios [fp to M ORO [jp .... ]]] ... who the BABY ... b.after movement: ...[.,., to MOROj [ < ,p pios [fp X j [ip .... ]]] ... ... the BABY who ... 142 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The challenge for this particular analysis of the underlying order would then be to explain why movement of a focused expression across CP is required when a wh-expression occurs in the specifier of CP but is impossible when a complementizer heads CP. This is peculiar given that the ungrammaticality or grammaticality of putative movement of focused expressions across complementizers represents the difference between dialect one and dialect two respectively: in dialect two, this kind of postulated movement would be quite generalized and would be the norm for all focused expressions, but we would be forced to say that it is quite restricted in dialect one, in which focused expressions would putatively move only when wh-expressions were present. This seems particularly inelegant in light of the fact that dialect one is by far the more widespread dialect, with many speakers refusing to accept dialect two as being a possible dialect of MG.^^ ‘^When the relative ordering of topics are also taken into account, as they shortly will be, this analysis of the base order would require that topics and focused expressions move past wh-expressions in all dialects and in dialect two, only topics move across complementizers just in case focused expressions are also present. These are constraints which seem quite arbitrary and the need to postulate them further underlines the undesirability of the analysis under consideration in the text. The data concerning topics and a plausible analysis thereof will be discussed shortly in detail. 143 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. If we consider the opposite order of FP and CP as underlying, this inelegance does not arise. Under this second analysis, the base order would be as follows (I repeat here (42)): (42) ...[fp [cp [ip .... ]]] ... Consider first dialect one. The base order in (42) means that a sentence involving a focused expression and wh-expression would reflect the underlying order of the discourse functions, and a sentence involving a focused expression and a complementizer must be derived via head movement of the complementizer. The following examples illustrate the two possibilities: (45) focused expression and wh-expression ...[fp to MORO [cp pios [ip .... ]]] ... the BABY who ... (46)a. focused expression and complementizer, underlying order: •••['.'? [fp I® MORO [cp oti [ip — ]]] ... the BABY that ... b.after head movement of the complementizer: ••■ ['.'■ ' [.'otij [fp to MORO [cp tj [ip — ]]]] ... ... that the BABY ... 144 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Under this analysis, in dialect one, there is no movement of specifiers to the spec position of the unidentified non-L-related projection. Consider now how this second analysis would treat dialect two. This dialect would contrast with dialect one only in that it would be possible in dialect two for specifiers to optionally move to the spec of the unidentified non-L-related position (marked here as "??"): (47)a.focused expression and complementizer, underlying order: [fp to MORO [ep oti [jp .... ]]] ... the BABY that ... b.after head movement of the complementizer: •••[?■ ' [?’Otij [fp to MORO [cp tj [jp .... ]]]] ... ... that the BABY ... c.after movement of the focused expression: to MOROjj [v'Otij [fp tf. [cp tj [jp — ]]]] ... ... the BABY that The optional movement of the focused expression is a stylistic variation of the more basic (i.e., more widely accepted and used) order that is possible in both dialects. It is beyond the scope of this dissertation to determine the exact identity of this projection; it seems sufficient to note that this projection has the stylistic function noted 145 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. here. (Henceforth, I will refer to the projection as "StP" for Stylistic Projection.)'^ I will adopt this fourth analysis of the relative ordering of preverbal expressions and apply it to account for the relative ordering of topics with respect to other preverbal expressions. Again, we will see that there are two dialects: (a) dialect one, which requires topics (as well as focused expressions) to be more deeply embedded than a + or -wh complementizer, and (b) dialect two, which allows topics, like focused expressions, to be somewhat freely ordered with respect to CPs: they can either precede or follow complementizers. In both dialects however, the topic has a relatively fixed order vis-a-vis the focused expression: the topic must invariably precede the FP. 4.6.1. The Relative Ordering of Topics and Subjects If we consider the ordering requirement between a subject and CLLDed NP we see, as pointed out in the relevant literature, that a CLLDed Noun Phrase must precede the subject: "'From a Minimalist perspective, movement is not optional; it is motivated by the need to check some formal feature(s) of the moved constituent. This theoretical approach would suggest that there are certain features at play in this stylistic movement as well as in the raising of the complementizer. Again, a more precise investigation of their nature is beyond the scope of this dissertation. 146 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (48)a. to more o Yiorghos to-frondise the baby, the George cl.acc.n-cared.for.3s "(speaking of) the baby, George cared for it." b.[ CLLDed NP [jp subject NP [p cl-V ]]] (49)* [subject NP [ CLLDed NP [p cl-V....]]]*'' These examples indicate that the CLLDed Noun Phrase is outside of the minimal maximal projection immediately dominating the clitic (which I assume here to be IP given V° to I* * movement).-** The example in (96) does not directly shed light on the relative ordering of the topic with respect to other preverbal specifiers; moreover, it tells us nothing about whether the topic is adjoined or in its own projection. These are all issues which I will address below. '"'If this ordering occurs, then the subject which precedes the CLLDed Noun Phrase is itself interpreted as a dislocation and must be followed by a pause. “**A s discussed earlier in this chapter, the clitic is always in the highest position of (an exploded) INFL, and more complex possibilities such as when negation, auxiliaries and the like are involved behave the same in this regard. 147 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.6.2. The Relative Ordering of Topics and CPs 4.6.2.1. Topics and Wh-Expressions Consider next the relative ordering between the topic and CP in matrix contexts. That the topic can be located structurally higher than CP is established by the following example, which illustrates the required ordering between a CLLDed Noun Phrase and a wh-expression. In this case we see the topic occurs structurally higher than a wh-element; the reverse ordering is impossible. This fact holds for dialect one and two alike. Cf.: (50)a. to vivlio se pion to-edhikse i Maria the book to who cl.acc.n-show.3s the Maria.nom "(speaking of) the book, who did Maria show it to?" b. [ CLLDed NP [ WH [ cl-V .... ]]] (51) * [ WH [ CLLDed NP [ cl-V .... ]]] Since there can be no doubly-filled comp in MG, these facts clearly indicate that the topic occurs outside of CP in this example. Moreover, they indicate that topics are not licensed in precisely the same way as wh-expressions (or focused expressions); if they were, we would find them in complementary distribution in matrix contexts. 148 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In dialect one and dialect two alike, this same ordering constraint holds in embedded contexts with respect to the relative ordering between a wh-expression and a topic: the topic must be structurally higher than the wh-expression. The following examples illustrate this fact: (52)a.anarotieme to moro pios to-frondize wonder.Is the baby who cl.acc.n-cared.for "I wonder who took care of the baby." b. *anarotieme pios to more to-frondize wonder.ls who the baby cl.acc.n-cared.for 4.6.2.2. Topics and Complementizers When a complementizer heads CP, the two dialects behave differently in terms of relative ordering of topic and CP. In dialect one, a topic must follow the + or -wh complementizer (the same behavior as focused expressions, which, in this dialect, also may only follow the complementizer). The examples in (53) and (54) below illustrate this fact: (53)a.anarotieme an to moro to-frondize wonder.ls if the baby cl.acc.n-cared.for "I wonder if he took care of the baby." b.[...V [ + wh complementizer [ TOPIC [jp NP....]]]] 149 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. *[...V [ TOPIC [ + wh complementizer [jp NP....]]]] (54)a.ksero oti to moro to-frondize know.Is that the baby cl.acc.n-cared.for "I know that he took care of the baby." b.[...V [-wh complementizer [ TOPIC [jp NP....]]]] c. *[...V [ TOPIC [-wh complementizer [jp NP....]]]] In dialect two the topic, like the focused expression, is freely ordered with respect to the complementizer. The following examples illustrate for dialect two that a topic can precede a +wh or -wh complementizer: (55)a.anarotieme, to moro, an to-frondize wonder.ls the baby if cl.acc.n-cared.for tenderly "I wonder if he took care of the baby." b.[...V [ TOPIC [ + wh complementizer [jp NP....]]]] (56)a.ksero, to moro, oti to-frondize know.Is the baby that cl.acc.n-cared.for "I know that he took care of the baby." b.[...V [ TOPIC [-wh complementizer [jp NP....]]]] The topic can also follow the complementizer in this dialect. Again, we see that, in this dialect too, the ordering constraints on topics 150 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. pattern with those of focused expressions. In a moment we will return to the relative ordering of specifiers with respect to CP when a focused expression is also involved."^ Let me briefly summarize the ordering requirements discussed so far for these two dialects: The speakers I have consulted who accept these constructions require the comment (the post complementizer material), of which the CLLDed expression is the topic, to be important or heavier in someway informationally (pointed out by Philippaki-Warburton, p.c.) in order for the construction to be natural. Moreover, Philippaki-Warburton suggests that they require a pause both before the topic and before the complementizer as illustrated here: (i) o Janis nomizei, ti Maria, oti dhen tha ti-dhjorisun telika sti trapeza. the Janis thinks, the Maria, that not will cl.acc.f-hire finally in’the bank "Janis thinks that (speaking of) Maria, they will not hire her at the bank finally." However, this construction cannot be analyzed as just involving a parenthetical expression because, as we will see, there are word order restrictions on the positioning of the focused expression in precomplementizer position just in case a topic is already in this position; more specifically, if there is a topic in this precomplementizer position, then there cannot also be a focused expression preceding the complementizer although in principal focused expressions can precede complementizers. In any case this word order is stylistically marked, an intuition which is captured by the account in the text. 151 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (57) WORD O R D ER REQUIREM ENTS DIALECT ONE DIALECT TWO WH/V: WH > V WH > V FOCUS/WH: FOCUS > WH FOCUS > WH FOCUS/COMP.; COMPL > FOCUS FOCUS > COMPL COMPL > FOCUS TOPIC/WH: TOPIC > WH TOPIC > WH TOPIC/COMP.: COMPL > TOPIC TOPIC > COMPL COMPL > TOPIC We see that the ordering of topics patterns with that of focused expressions. Assessment of the above array of facts raises the question of why the word order of topics is restricted vis-a-vis complementizers in the same way as that of focused expressions. It cannot be because topics and focused expressions are licensed in the same way because we have already seen that they are not: topics and focused expressions can cooccur in matrix clauses and this shouldn’t be possible if they were licensed in an identical way. The fact that topics and focused expressions behave essentially the same with respect to their word order restrictions when complementizers are involved will be accounted for in the next section where I consider the relative ordering of topics and focused expressions with respect to each other. 152 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.6.3. The Relative Ordering of Topics and Focused Expressions Consider now the relative ordering between the topic and a preverbal focused expression where we see that the topic must invariably precede and thus occur structurally higher than the focused Noun Phrase. The following examples illustrate this ordering constraint in matrix contexts: (58)a. to moro, O YIORGHOS to-frondise the baby THE GEORGE cl.acc.n-cared.for.3s "(Speaking of) the baby, GEORGE cared for it." b.[ CLLDed NP [ FOCUS NP [cl-V ....]]] c.*[FOCUS NP [ CLLDed NP [cl-V ]]] The same observation holds in embedded contexts too; i.e., a topic must always precede a focused expression. This observation is true for both dialects under discussion: topics are invariably structurally higher than focused expressions. This observation extends to structures where a wh-expression is also involved in addition to the focused expression and topic. Recall that it was previously established that, in embedded contexts, a focused expression must precede a wh-expression and not the reverse. The same 153 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is true when a topic is also considered with the additional proviso that the topic must precede the focused expression: (59)a.anarotieme afto to vivlio stin RANI pios to-edhikse wonder.ls that the book to-the RANI who cl.acc.n-showed.3s "I wonder who showed that book to RANI." b. *anarotieme stin RANI afto to vivlio pios to-edhikse wonder.ls that to-the RANI the book who cl.acc.n-showed.3s 4.6.3.1. Topics: Adjoined or in Specifier Position? The above data suggest that the topic "selects" a focus phrase. That is, the requirement that the topic must always be (immediately) higher than the focus phrase can be captured if there is a selection relation between a topic phrase and a focus phrase. This would mean that the topic is actually in the specifier position of a phrase whose head has the property of selecting a focus phrase.“ If the claim that the head of a topic phrase selects an FP is correct, then the topic cannot be analyzed as being adjoined to a clause ““ This kind of syntactic requirement on the relation between topic and focus such that the topic must be structurally higher than the FP, mirrors the syntax of specific expressions proposed by Diesing (1990) for German, in the sense that a specific expression must be outside of the scope of new information. I discuss Diesing’s proposal later in the text as it pertains to the syntax of CLLD and clitic doubling. 154 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. peripheral position, counter to what has been proposed in the literature on MG (Cf., for instance, latridou (ibid.), and Tsimpli (ibid.)). One argument for the view that the topic is adjoined has been the fact that it is possible to have multiple topics in preverbal position. The following example illustrates this possibility: (60) (o Janis nomize oti) i mitera-su, ton Arjiri (the J. thought.3s that) the mother-yours, the A. ton-kalouse sto spiti-mu poli sichna cl.acc.m-invited to’the house-my very frequently "(Janis thought that) (speaking of) your mother (and) Arjiris, she very frequently invited him to my house." The analogy has then been drawn between the fact that when grammatical functions are considered, it is possible to have only one subject, object, etc. but any number of adjuncts are possible, and the fact that in preverbal non-L-related positions it is possible to have a number of topics, but other discourse functions such as focused expressions and wh-expressions are limited to only one expression per function. Holders of this view have suggested that the topic is behaving as an adjunct in that more than one topic can occur in a clause just as more than one adjunct can occur in a clause. But this analogy is imperfect. Consider that in some languages, there can be two subjects or two objects, but it 155 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is obvious that they are not adjuncts (at least not both occurrences of the subjects or object) and the fact that there is more than one occurrence of subjects or objects does not require the conclusion that they are adjuncts. So without positive evidence, it seems hasty to conclude that topics are in adjoined positions just because there can be more than one of them. Another argument for the view that the topic is adjoined is the fact that its presence does not block selection between a superordinate verb and its complement. For example, a verb such as anarotieme ‘I wonder’, which must select a +wh complement, can do so across a topic as we saw in (52), repeated here for convenience: (52)a.anarotieme to moro pios to-frondize wonder.ls the baby who cl.acc.n-cared.for "I wonder who took care of the baby." But if this is problematic, then the same problem should also exist for FPs since they also don’t prevent selectional restrictions from being satisfied for a superordinate verb as illustrated in (25), repeated here for convenience: 156 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (25) anarotieme to M O R O j pioSy frondize X y X j wonder.ls the BABY who cared.for "I wonder who took care of the BABY." In this example an FP intervenes between a verb that selects a + wh complement and a CP whose specifier is occupied by a wh- expression. Despite this, the selectional restrictions of the verb are satisfied. It is usually assumed for MG that preverbal focused expressions are not adjoined to CP but rather in a specifier position. One motivation for this has been, for example, speakers who require inversion in focus constructions; it would be difficult to account for the requirement that for some speakers a focused expression must in certain contexts be followed by a verb unless the verb and focused expression were required to stand in a spec/head relation. It seems that such a relation could only come about if the focused expression occupies a specifier position and the head of that projection requires raising of the verb to the head position under certain circumstances. Clearly, both constructions (involving preverbal focused expressions and topics) could be analyzed uniformly with regard to their like behavior in contexts of selection; i.e. it is not necessary to conclude that FPs and Topic Phrases do not interfere with selection for different reasons. 157 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. It seems that selectional restrictions can be satisfied across several categories, just in case the intervention of functional categories only is involved. This can be illustrated for selection at least with respect to semantic features. For instance, the following German verb selects a subject that is [ + animal, -human]: fressen ‘to eat (as an animal eats)’. The following sentence exemplifies its appropriate usage: (61) der Hund frisst the.nom.m dog ate "The dog ate." The relevant features to satisfy selection restrictions are in N°, not D°. So we see that selectional restrictions can clearly be satisfied across a functional category, in this case across DP. Since the intervention of a functional category apparently doesn’t interfere with the satisfaction of selectional restrictions, there is no a priori reason to postulate a different structure for topics because selection is possible across them than already exists with focused expressions. A third argument for the view that the topic is adjoined is the fact that topics do not prevent extraction of wh-adjuncts whereas wh- expressions do. The idea behind this is that specifiers will intercept other expressions that go to specifier positions; therefore, under this view, topics cannot be in specifier positions since they do not intercept 158 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the relevant wh-expressions. But we have already seen data that indicate that topics, focused expression and wh-expressions are all on parallel tracks in terms of how they interact with extracted elements. We saw that focused expressions also don’t intercept wh-expressions; nonetheless, there is reason to believe that focused expressions are in specifier positions. It seems then that whether an expression in a non-L- related position intercepts an extracted wh-element or not cannot be decisive in determining its status as adjoined or occupying a specifier position.-^ Moreover, when we consider now further facts about the relative ordering of topic and focused expressions with respect to the complementizer in dialect two, we will see further reason to believe that the topic is in a specifier position and not an adjoined position. In sum, we have seen that the topic is consistently structurally higher than the FP and this can be accounted for if there is a topic phrase (TpP) which selects a FP. This means that the following structure exists in MG: ■^A plausible account of the conditions under which elements in non- L-related positions can intercept other elements in non-L-related positions can be found based on Minimalist assumptions. That is, we can identify potential "interceptors" based on the notion of feature checking and the Shortest Movement economy principle; for example, it must be that focused and wh-expression check different features and therefore, one kind of expression doesn’t interfere with the extraction of another. 159 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (62) underlying order: •••[stp [ipp to moro^ [fp o YIORGHOS [,ppios/oti[jp ]]]] ... the baby the GEORGE who/that .. cl.n-V... We will see further evidence for this structure when we consider the ordering of topics and focused expressions with respect to complementizers, where we will see that the two dialects behave differently. 4.6.4. The Relative Ordering of Topic, Focus and Complementizer In dialect one, the topic must follow the complementizer and immediately precede the focused expression. No other word order is allowed. Cf.: (63)a.ksero oti to moro o YIORGHOS to-frondise know.Is that the baby the GEORGE cl.acc.n-cared.for "I know that GEORGE took care of the baby." b.*ksero oti o YIORGHOS to moro to-frondise know.Is that the GEORGE the baby cl.acc.n-cared.for c. *[...V [TOPIC [ FOCUS [ complementizer [ ...]]]] d.*[...V [FOCUS [ TOPIC [ complementizer [ ...]]]] e.*[...V [TOPIC [ complementizer [FOCUS [ ...]]]] f. *[...V [FOCUS [ complementizer [TOPIC [ ...]]]] 160 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This required relative ordering of topic, focused expression and complementizer is straightforwardly accounted for in dialect one if a TpP selects a FP and there is raising of the complementizer to the highest functional projection (namely StP): (64) order after complementizer raising: [stp [sf otij [tpp to morOk [fp o YIORGHOS [,p t^ [jp ..to^-V.. ]]]] ... that the baby the GEORGE ...cl.n-V... In dialect two, topics, like focused expressions, can freely precede or follow complementizers as long as they obey the requirement that the topic precede the focused expression, and as long as no more than one expression precedes the complementizer. The following examples illustrate these possibilities for dialect two: (65)a.ksero oti to moro o YIORGHOS to-frondise know.Is that the baby the GEORGE cl.acc.n-cared.for "I know that GEORGE took care of the baby." b.*ksero oti o YIORGHOS to moro to-frondise know.Is that the GEORGE the baby cl.acc.n-cared.for "I know that GEORGE took care of the baby." 161 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (66)a.ksero to moro oti o YIORGHOS to-frondise know.Is the baby that the GEORGE cl.acc.n-cared.for "I know that GEORGE took care of the baby." b.*ksero to moro o YIORGHOS oti to-frondise know.Is the baby the GEORGE that cl.acc.n-cared.for "I know that GEORGE took care of the baby." c. *ksero o YIORGHOS oti to moro to-frondise know.Is the GEORGE that the baby cl.acc.n-cared.for "I know that GEORGE took care of the baby." We see then that, although in dialect two topics and focused expressions alike may either freely precede or follow a complementizer, they cannot do this independently of each other if to do so would violate the requirement that topic must precede focus. For example, it is a priori possible for a topic to follow a complementizer and it is a priori possible for a focused expression to precede a complementizer, but they cannot do this in the same clause because this would incorrectly place the topic syntactically lower than the focused expression, the ungrammaticality of which is illustrated in (66c) above. The reverse ordering of topic and focus with respect to the complementizer (illustrated in (66a)) is grammatical, which is expected 162 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. since it still respects the requirement that the topic must be structurally higher than the focused expression. The word order facts are summarized in the following chart: (67) WORD ORDER REQUIREMENTS DIALECT ONE WH/V: FOCUS/WH: FOCUS/COMP.: TOPIC/WH; TOPIC/COMP.; TOPIC/FOCUS: TOP/FOC/WH: TOP/FOC/ COMP.: WH > V FOCUS > WH COMPL > FOCUS TOPIC > WH COMPL > TOPIC TOPIC > FOCUS TOP > FOC > WH COMPL > TOP > FOC DIALECT TWO WH > V FOCUS > WH FOCUS > COMPL COMPL > FOCUS TOPIC > WH TOPIC > COMPL COMPL > TOPIC TOPIC > FOCUS TOP > FOC > WH TOP > COMPL > FOC COMPL > TOP > FOC We can account for the behavior of dialect two in the following way. We have already assumed that a TpP selects an FP and that there is obligatory complementizer raising (cf. (64) repeated here for convenience), which of course captures the relative ordering requirements for complementizer-initial orderings in the same way as in dialect one. Cf.: 163 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (64) order after complementizer raising: [sip L- otij [tpp to morOk [fp o YIORGHOS [,p tj [^p ..to^-V.. ]]]] ... that the baby the GEORGE ...cl.n-V... Consider now the fact that only the topic can occur in precomplementizer position when a focused expression is also involved. This suggests first of all that there is only one position available in precomplementizer position. If there were more than one position, there would be no a priori reason why both the topic and the focused expression could not precede the complementizer at the same time.-"* Moreover, while this ordering fact does not prove whether or not the topic is initially in a specifier or adjoined position, it does indicate that the topic is in a specifier and not an adjoined position when it precedes the complementizer. To see this, consider that if the restriction were on ■ ‘ ‘Recall that it is possible to have multiple topics in precomplementizer position, which might seem to suggest that there are, indeed, more than one positions available preceding the complementizer. However, if the above analysis of complementizer placement is correct, it suggests that multiple topics are best analyzed as not involving topics in multiple specifier positions. Rather it is possible to analyze them as involving one topic in specifier position and the others as adjoined; i.e., they would have a similar structure to double object constructions. Alternately, multiple topics could involve conjunction ((speaking of) George (and) the baby (and)...). The syntax of multiple topic constructions, as well as the syntax of nominative and genitive (the MG equivalent of dative) topics is clearly in need of further investigation, but such an undertaking is outside the scope of inquiry of this dissertation. 164 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. having only one precomplementizer position, the restriction must refer to the specifier of the projection headed by the complementizer: it does not seem possible to put such a constraint on adjoined positions. If there is only one specifier position preceding the complementizer and the topic occupies an adjoined position when it precedes the complementizer, then there is no reason why it would not be able to precede the focused expression in precomplementizer position deriving the incorrect order: topic focus complementizer."^ It seems then that the topic occurs in a specifier position when it occurs in precomplementizer position. 4.6.4.1. R elative Ordering R equirem ents and M inimality The fact that only the topic and not the focused expression can precede the complementizer when both occur suggests that the postcomplementizer topic is also in a specifier position. To see this, consider that it is possible to account for the fact that only the topic and "^There might at first glance appear to be the counter-analysis that StP counts as an argument and there is no adjunction to arguments and therefore there could be no adjunction to StP (and thus no order: topic focus complementizer). Note, however, that if the constraints on precomplementizer word order were due to the distribution of adjunction sites and the hypothesis that topics were to adjoin in precomplementizer positions, then it shouldn’t be possible to derive the grammatical order: topic complementizer. 165 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. not the focused expression can precede the complementizer if we view this as a minimality effect; namely, the topic intercepts the focused expression and prevents it from moving to precomplementizer position in the presence of a topic. If there is no topic, there is no XP to intercept the focused expression; therefore, focused expressions should be able to move to precomplementizer position just in case there is no topic present. If the topic intercepts the focused expression in this way, it indicates that they are minimally both in specifier positions since only expressions of "the same kind," in the relevant way, can block the extraction of another "same kind" of element across it. If topics and focused expressions are the "same kind" of element in the relevant way, why does the topic intercept the focused expression only when short distance movement is involved and not when long distance movement is involved?"^ Recall that when long distance movement is involved, topics, focused expressions and wh-expressions run on parallel tracks so that one kind of expression does not intercept the other. It is possible to provide an account for this discrepancy in “‘ ’Likewise, we must ask why movement from L-related to non-L- related position is not intercepted by intervening discourse functions such as topic, focus, wh-expression, etc. For instance, a wh-expression doesn’t intercept focus movement from its L-related position to the local specifier of FP. This behavior can be accounted for in the same way as the lack of interception with long-distance movement is. 166 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. behavior if we consider not only the position of intervening expressions (i.e., whether or not they are in the same kind of position), but also the nature of the target of movement. In long distance movement, topics, focused expressions and wh-expressions are going to different target positions (spec of TpP, spec of FP and spec of CP respectively); therefore, since they are not competing for the same positions, they do not intercept one another even if they are all in the same kind of positions (in the sense of non-L-related specifier versus adjoined position versus L-related specifier). In short distance movement, in contrast, the topic and the focused expression are competing for the same position and therefore, only the closest one, namely the topic, can move to that specifier position."^ The same kind of interception must come into play in the case of wh-expressions too, which shouldn’t be able to move to the specifier of the StP if a topic or focused expression intervenes. It is, however, not easy to detect whether a wh-expression can move across a focused -^This kind of behavior is quite intuitive from a Minimalist perspective: long distance movement involves the checking of different features (topic, focus and wh-features) whereas short distance movement would involve the checking of the same kind of features (presumably the "stylistic features" in the case under consideration in the text). Therefore, by Shortest Move, focus couldn’t move across topic in short distance movement because they are both moving to check the same kind of features. 167 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. expression or topic to the specifier of StP. This is because such movement would be vacuous given that wh-expressions in MG must stand in a specifier/head relation with a verb. Suppose that a wh-expression moved to the spec of StP. Suppose also that this kind of movement would also trigger verb movement to St° in adherence with the requirement that the verb and wh-expression stand in a specifer/head relation. If wh- movment takes place across a focused expression and topic, we would have the following word order after verb raising: (68)[stpSe pionj[,i.to-edhoseJ,ppto vivlio[fpi MARIA[^p tj [jp...t„...]]]] to whom cl.n-give-3s the book the MARIA.nom "Whom did MARIA give the book to?" This would look superficially the same as a sentence where the topic and focused expression did not occur in preverbal specifier positions but instead occurred internally to IP (I leave unelaborated the derivation of VOS word order— just assuming this is internal to IP): (69)[cp se pion to-edhose^ [jp to vivlio i MARIA]]] to whom cl.n-give.3s the book the MARIA.nom "Whom did MARIA give the book to?" Firstly, as pointed out in footnote 15, a derivation like that in (68) would presumably already be ruled out for reasons of economy with 168 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. respect to verb movement, since the verb must make a longer movement in (68) where there is a much shorter path for movement of the verb, namely, that given in (69) where there is no vacuous movement. Secondly, the derivation in (68) would by hypothesis be ruled out by minimality in that the more superordinate specifiers that have the spec of StP as a target would intercept the more deeply embedded wh- expression. Moreover, as far as empirical evidence is concerned, we can see that at least sometimes, a wh-expression initial word order cannot possibly be due to the kind of vacuous movement under consideration here, but instead must be an instance of a topic and focused expression occurring in situ, internal to IP; that is, there is empirical evidence that topic and focus do sometimes occur internal to IP. It is possible to have the following word order where a wh-expression precedes a focused expression which precedes a topic: (70)[cp se pion [^. to-edhose„ [jp i MARIA t^ to vivlio ]]] to whom cl.n-give.3s the MARIA.nom the book "Whom did MARIA give the book to?" The order in (70) cannot possibly be due to vacuous movement of the wh-expression to the specifier of StP across the topic and focused expression because, in preverbal position, the focused expression must 169 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. always be more deeply embedded than the topic, but here the reverse ordering is apparent. This indicates that at least some word orders with focus and/or topic constructions must involve focused expressions and/or topics that are internal to IP which shows that at least sometimes it would not be possible to have the kind of analysis where everything raises. We see there are a number of theory internal reasons to believe that wh-expressions cannot move to the specifier of StP when topics and/or focused expressions are involved. Moreover, although there is not empirical evidence that proves we can never have the kind of vacuous movement discussed above, we do have empirical evidence that some constructions that are wh-initial cannot possibly involve movement of the wh-expression across preverbal discourse functions. 4.6.5. Overview of Relative Ordering Requirem ents In sum, we can determine the location of the topic and the relative ordering of preverbal specifiers under a theory that foresees distinct projections for topics, focused expressions and wh-expressions. The variation in possible word orders that exist for the two dialects described here is accounted for via a process of complementizer raising and movement of discourse functions within the non-L-related projections in a manner similar (but not identical) to what happens with grammatical 170 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. functions within the L-reiated projections as well as a hierarchy of projections such that Topic Phrase immediately dominates Focus Phrase, which immediately dominates CP. In the next section I will present independent evidence in support of this proposed structural hierarchy. 4.6.6 The Relative Ordering FP, CP and the Distribution of NPIs Negative Polarity Items (NPIs) in MG can be found in both pre- and post verbal position. In post verbal position, the NPI kanenan ‘no one’ is licensed by all the polarity operators. For example, as illustrated in (71), negation licenses a postverbal NPI and so does a Yes/No question: (Vl)a.dhen-idha kanenan not-saw.ls no one "I didn’t see anyone." b.idhes kanenan? saw.2s no one "Did you see anyone?" 171 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In preverbal position, an NPI is obligatorily interpreted as focused. This suggests that it is located in the specifier of FP. This word order is illustrated for a preverbal object in (72):’’ ^ (72) KANENAN dhen-idha no one not-saw.ls "I didn’t see ANYONE." Although NPIs in postverbal position can be licensed by any polarity operator, preverbal NPIs can only be licensed by negation. We see, for example, that Yes/No question formation is incompatible with preverbal NPIs: (73)* KANENAN idhes? no one saw "Did you see ANYONE?" This cannot be attributed to a general incompatibility between focus NPIs and Yes/No questions since a postverbal NPI in a Yes/No question can be focused: -^I don’t discuss subject NPIs in MG in the text, but the generalizations illustrated in the text essentially hold for subject NPIs as well. 172 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (74) idhes KANENAN? saw no one "Did you see ANYONE?" We can account for this asymmetry in the licensing of NPIs given the validity of the following three assumptions: (1) in MG, negation licenses NPIs via a spec/head relationship; (2) the structure of the discourse functions is as proposed earlier: TpP, FP, CP; and (3) Progovac (1988) is correct in claiming that NPIs are licensed via a binding relation to polarity operators in the specifier of CP (modulo the spec/head licensing relation with negation assumed for MG). If these assumptions are correct, a preverbal focused NPI is located in the specifier of FP, which I have proposed to be structurally higher than the specifier of CP. Any polarity operator located in the specifier of CP could not bind an NPI in the specifier of FP due to a failure of c- command in this case. Negation can license a preverbal NPI through a head movement process such that negation moves to F°. The distribution of preverbal NPIs would directly fall out from the phrase structure I have proposed for discourse functions. In the next section I will introduce empirical evidence that the CLLDed Noun Phrase is actually present, playing a syntactic role, in clitic doubling constructions. 173 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.7. Empirical Evidence for the Existence of a Topic in Cl-d Constructions 4.7.1. Topics in Cl-d Constructions Let us now reconsider the question of the existence of a topic in Cl-d constructions in light of the restrictions on the relative ordering between the topic (namely the CLLDed Noun Phrase) and other preverbal specifier positions. We will see that, where there is evidence bearing on the existence of and the location of the topic in Cl-d constructions, it indicates that the topic occurs in the same position as CLLDed Noun Phrases. This evidence is found with constructions which forbid the co-occurrence of certain XPs in preverbal position just in case the clause also involves clitic doubling or CLLD. Recall that CLLD didn’t seem to trigger inversion; i.e., it was possible for a topic to precede a subject— it wasn’t necessary for the subject and verb to invert (in contrast to wh-constructions which triggered inversion in all contexts and, for some speakers, focus constructions triggered inversion). However, it seems that CLLD does, in fact, indirectly trigger inversion. To see this, recall the following; there are speakers for whom it is possible for a focused Noun Phrase to immediately precede a subject. However, even for these speakers, this ordering isn’t possible if CLLD also occurs in which case subject/verb 174 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. inversion is obligatory so that the required word order is; topic, focused expression, verb, subject, etc. This fact is illustrated in (75), which demonstrates the ungrammaticality which results when inversion doesn’t take place, shown schematically in (75b), and (76) which illustrates the accepted ordering involving inversion (namely: topic,focused expression, verb, subject), shown schematically in (76b): (75)a. *to vivliOj stin MARIA o Yiorghos tOj-edhikse the book to-the MARIA the George cl.n-showed.3s b.*[ CLLDed NP [ FOCUS NP [Subject NP [cl-v ....]]]] (76)a.to vivIiOj stin MARIA tOj-edhikse o Yiorghos the book to-the MARIA cl.n-showed.3s the George "the book, George showed it to MARIA." b.[ CLLDed NP [ FOCUS NP [cl-v [Subject NP ....]]]] I have nothing to say along the lines of an account of this indirect inversion phenomenon;’® what is important here is the fact that this word order is obligatory when CLLD takes place. This obligatory "®For instance, it might be interesting to explore whether there is a more deeply embedded level of discourse functions from which topic and focused expressions move so that the inversion phenomenon between topic and focus can be seen analogously to the kind of verb movement phenomenon that is necessary to allow raising of a subject across a direct object, etc. 175 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ordering sheds light on the structure of Cl-d because, if Cl-d requires the presence of a topic in the same position as the CLLDed Noun Phrase, inversion due to presence of the topic should be mandatory in Cl-d constructions in exactly the same environments as in CLLD. This is indeed the case. Cl-d constructions require subject/verb inversion when a focused NP occurs in preverbal position, even for speakers who don’t require subject/verb inversion when just a focused expression is present. This fact becomes especially salient when the ungrammatical doubling example (with the order: focused expression, subject, cl-verb, doubled object) is contrasted with a grammatical non-doubled example (with the order: focused expression, subject, verb, object): (77)a.?*stin MARIA o Yiorghos tOj-edhikse to vivlioj to-the MARIA the George cI.n-showed.3s the book b.stin MARIA o Yiorghos edhikse to vivlio to-the MARIA the George showed.3s the book "George showed the book to MARIA." When inversion occurs, so that a preverbal focused expression is followed by the verb, clitic doubling constructions and constructions without clitic doubling are equally grammatical: 176 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (78)a. stin MARIA tOj-edhikse o Yiorghos to vivlioj to-the MARIA cl.n-showed.3s the George the book "The book, George showed it to MARIA." b.stin MARIA edhikse o Yiorghos to vivlio to-the MARIA showed.3s the George the book "George showed the book to MARIA." We wouldn’t expect Cl-d to be sensitive to the ordering of specifiers much higher than the doubled expression unless the Cl-ded Noun Phrase were directly related to that higher position in some way. The fact (cf. (77), (78)) that Cl-d is indeed sensitive to the ordering of specifiers much higher than the doubled Noun Phrase strongly supports its direct relation to the higher position. 4.7.2. Distribution of Clitic Left D islocation Another kind of fact is consistent with the idea that a topic occurs in a structurally high position in Cl-ding constructions. This set of data relates to the distribution of CLLD which has exactly the same distribution as Cl-d. This can be explained in precisely the same way that the distribution of Cl-d was explained; an antilocality requirement exists which governs the relation between the topic and the clitic/NP chain in a certain domain. 177 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In MG, CLLD is possible only with the arguments of verbs. For instance, CLLD is possible in an indicative clause; (79) ton Aleksandro, o Yiorghos ton-frondizi the Alexander the George cl.acc.m-cares.for "(speaking of) Alexander, George takes care of him." Moreover, CLLD is also posible in subjunctive clauses: (80) thelo to ghramma, i Maria na to-stili want.Is the letter, the Maria.nom subj cl.acc.n-send.3s "(speaking of) the letter, I want Maria to send it." In addition, it is possible to clitic left dislocate an argument of a verb in an imperative clause: (81) to vivlio paCto the book take.cl.n "(speaking of) the book, take it!" Moreover, it is possible to clitic left dislocate an argument of a participle: 178 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (82) tin efimeridha dhiavazondas-tin, apokimithika the newspaper reading-cl.acc.f fell.asleep.Is "(Speaking of) the newspaper, reading it, I fell asleep." We see that, as true for clitic-doubling, CLLD is possible with the complements of verbs. It is, moreover, not possible to clitic left dislocate the object of a preposition, regardless of whether the dislocated Noun Phrase occurs in FP initial or clause initial position. This fact too mirrors the distribution of Cl-d. E.g.: (83)a. *idha ena fidhi [tu Yiorghu/apo ton Yiorgho [brosta-tu ]] saw.Is a snake the.gen.m George/from G. in-front-cl.gen.m "(speaking of) George, I saw a snake in front of him." b. *tu Yiorghu/apo ton Yiorgho idha ena fidhi brosta-tu the.gen.m George/from G. saw.Is a snake in-front-cl.gen.m "(speaking of) George, I saw a snake in front of him." In addition, it is not possible to clitic left dislocate the complements of nouns, regardless of whether the dislocated NP is in pre- Noun Phrase or clause-initial position. E.g.: (84)a. *exasa [tis Ranis [tin fotoghrafia-tis]] lost.Is the Rani.gen the picture.cl.gen.f "(speaking of) Rani, I lost her picture." 179 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b.*tis Ranis exasa tin fotoghrafia-tis the Rani.gen, lost.Is the picture.cl.gen.f "(speaking of) Rani, I lost her picture." This restriction on distribution holds for Cl-d too. We see then that Cl-d and CLLD are found in all and only the same environments in MG. The fact that the distribution of CLLD mirrors that of Cl-d points to a similarity in licensing for the two constructions and is consistent with the hypothesis that a CLLDed Noun Phrase occurs in Cl-d constructions. If CLLDed Noun phrases are subject to the same antilocality requirement that was postulated for Cl-d constructions (by virtue of the fact that the CLLDed expression is the topic that is subject of the predication in Cl-d constructions), then there is a straightforward, unified account for these two constructions. 4.8. Concluding Remarks In this chapter, a variety of facts were introduced which indicated that the clitic-doubled expression is related to a topic which occurs in a structurally higher, c-commanding position; and it was suggested that the topic be identified as the structurally higher c-commanding Noun Phrase evident in Clitic Doubling constructions. It was established that the topic occurs in the specifier position of a phrase which selects an FP. Various 180 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. facts about relative ordering between topics, focused expressions, complementizers and wh-expressions as well as the distribution of preverbal NPIs fall out essentially from this proposal. Finally, additional evidence was introduced which suggested that a topic occurs in Cl-d constructions based on an indirect inversion phenomenon which happens when CLLDed and preverbal focused expressions co-occur. In the next chapter, I will examine or re-examine proposals that have been made with respect to the movement versus non-movement relation between CLLDed expression and Cl-ded expressions in light of the representation of Cl-d that I proposed in chapter three and four. 181 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER FIVE: ASYMMETRIES BETWEEN CLLD AND CL-D CONSTRUCTIONS 5.0. Introduction I have suggested that CLLD and Cl-d involve essentially identical constructions syntactically; yet, in the literature concerning non movement analyses of CLLD, certain asymmetries between the two constructions have been pointed out. While some aspects of the non movement accounts of CLLD are consistent with the representation of Cl-d that I have proposed here, other aspects seem to cast doubt on the idea that the two constructions are essentially identical syntactically. In this chapter I will discuss some of the evidence offered in support of non-movement based accounts of CLLD, especially evidence pointing to an asymmetry between the two constructions, and attempt to reconcile them with the proposal I have made here. 5.1. Non-M ovement Accounts of CLLD It has been argued in the literature, convincingly in my opinion, that CLLD is not related to Cl-d via movement (cf., e.g., Cinque (1990) and latridou (ibid.)). It has also been argued specifically for MG (see latridou (ibid.)) that CLLD in that language does not derive from Cl-d via movement. Consider, in what follows, some of the arguments 182 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. presented by Cinque (ibid.) and latridou (ibid.) that CLLD is a base generated construction. 5.1.1. Evidence Consistent With the Proposed Structural Representation of Cl-d 5.1.1.1. Parasitic Gaps The first set of arguments that we will consider do not conflict with the analysis of Cl-d proposed here, especially with respect to the proposed relation between Cl-d and CLLD.^ Cinque and latridou point out that the complement position to which a CLLDed Noun Phrase is related does not behave like a variable. This means that A’-movement cannot be involved in the relation between CLLD and Cl-d. As an illustration of the claim that variables are not present in CLLD constructions (at least not in the overt syntax), latridou points out that it is not possible for a locally CLLDed Noun Phrase in MG to license a parasitic gap: 'This result is not surprising since these are arguments advanced by latridou, who proposes a predication analysis for CLLD— the main difference between our proposals is in the nature of the predicate variable, which I suggest is the clitic and the doubled Noun Phrase chain as opposed to just the clitic (or: the clitic-pro chain). 183 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (1) *to arthro to-arxiothetisa xoris na dhiavazo the article cl.n-filed.ls without subjunctive read.Is * "The article, I filed without reading." This behavior can be contrasted with the behavior of wh-extraction or preverbal focus constructions which do indeed license parasitic gaps as illustrated in (2a and b) respectively: (2)a. pio arthro arxothetises xoris na dhiavazis which article filed.2s without subjunctive read.2s "Which article did you file without reading?" b. to ARTHRO arxothetisa xoris na dhiavazo the ARTICLE filed.Is without subjunctive read.Is "The ARTICLE I filed without reading." 5.1.1.2. W eak Crossover E ffects Moreover, as Cinque and latridou point out, there are no WCO effects in CLLD constructions. This means that the CLLDed Noun Phrase can grammatically be coindexed with a pronoun that is to the left of and does not c-command the argument position to which the CLLDed Noun Phrase is related. This is illustrated in (3) for MG: 184 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (3) ton YiorghOj tonj-aghapai i mitera-tuj the George.acc cl.acc.m-love.3s the mother-his "(Speaking of) Georgej, hisj mother loves himj." If the CLLDed Noun Phrase had undergone A’-movement out of the argument position, a variable would have occurred in that site and a WCO configuration would have resulted since a non-c-commanding pronoun would be coindexed with a variable to its right. This behavior can be contrasted with that of extracted wh-expressions and preverbal focused expressions, neither of which can be coindexed with a pronoun that does not c-command the argument site to which the wh-expression or focused expression is related: (4)a. *pio pedhij aghapai i mitera-tUj which child love.3s the mother-his *?"Which childj does hiSj mother love?" b. *ton YiorghOj aghapai i mitera-tUj the George love.3s the mother-his *?"Georgej hisj mother loves." These constructions behave as if there is a variable in the argument position to which they are related. 185 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The above data, which consistently indicate that no variable occurs in the argument position to which a CLLDed Noun Phrase is related, strongly support the conclusion that there is not a relation of A’- movement between the doubled Noun Phrase and the CLLDed position. 5.1.1.3. ECM Constructions I’d like to add an additional argument against a movement analysis of CLLD in MG. But I must point out that this argument goes through only in so far as we can essentially equate CLLD and Left Dislocation (LD), a position asserted by latridou (ibid.). The relevant data comes from Exceptional Case Marking (ECM) constructions in MG. This construction contains an accusative case- marked NP which is understood as thematic subject of the clause subcategorized by the ECM verb. An example is given in (5): (5) dhen perimena tin Tzeni] na maloni tin Elena not expect.Is the Tzeni subjunctive scold.3s the Elena "I didn’t expect Tzeni to scold Elena." I suggest, following Schneider-Zioga (1992a,b), that in ECM in MG there is no process of movement from the embedded spec of IP to the superordinate clause, nor is there a process of clausal transparency; instead, I suggest that the following structure is right for ECM in MG, 186 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. where the exceptionally accusative Noun Phrase has raised from a peripheral Left Dislocation position in the embedded clause to the spec of the superordinate VP: (6)[ip [i-[vp NPj [,. V° [,pp tj [,p [jp ]]]]]]] Following latridou (ibid.). Left Dislocation constructions involve a Noun Phrase base-generated in the Left Dislocation position as the subject of a predication and the minimal maximal projection containing the predicate variable (open position) is the predicate. The syntax of ECM constructions seems to support both parts of the hypothesis that I am pursuing; namely, that ECM involves the peripheral Left Dislocation position and that Noun Phrases are base-generated in this position. Initial support for the Left Dislocation hypothesis for ECM comes from intonational patterns in ECM constructions. When a Left Dislocated Noun Phrase precedes a preverbal focused Noun Phrase, there must be a pause (comma intonation) between the dislocated Noun Phrase and the focused one: (7) to moro, TON YIORGHO protimai the baby, the George.acc prefer.3s "The baby, he prefers GEORGE.’ 187 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The same kind of intonational requirement holds in ECM constructions when a focused Noun Phrase precedes the embedded verb: there must be a pause between the exceptionally accusative subject and the focused Noun Phrase. This contrasts with the intonational pattern in a control structure when an object of the superordinate clause precedes a focussed Noun Phrase which is an argument of the embedded verb: with control structures, there is no pause (cf. (8a,b)): (8)a. dhen perimena ton Yiorgho, to MORO na frondisi not expect.Is the G.acc, the BABY subjunctive care.for.3sg "I didn’t expect George to take care of the BABY." b. episa ton Yiorgho to MORO na frondisi persuade.Is the G.acc the BABY subjunctive care.for.3sg "I persuaded George to take care of the baby." Support for the hypothesis that LD, and by extension, CLLD, involves base generation is indicated by the behavior in ECM contexts of certain "case agreement constructions." These are constructions that involve a number of expressions which must have the same overt morphological case as their antecedent. They include: (i) circumstantial adverbs which agree in case with the Noun Phrase they modify; and (ii) the reciprocal, which agrees with its antecedent in case. In (9a) the circumstantial AP is nominative in agreement with the nominative 188 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. subject it modifies. In (9b) the quantificational part of the periphrastic reciprocal is nominative in agreement with its nominative antecedent. In (9c) the quantificational part of the reciprocal is accusative in agreement with its accusative antecedent. (9)a.o Yiorghos efighe eftixismenos the G.nom left happy.nom "George left happy." b. I yinekes voithane i mia tin ali the women.nom help the one.nom the other.acc "The women help each other." c. O Yiorghos sistise tis yinekes tin mia s’tin ali the G. introduced the women.acc the one.acc to the other "George introduced the women to each other." With this widespread requirement of case agreement, we would expect an adverb to be in accusative case when it modifies an exceptionally accusative-marked Noun Phrase since the adverb is subject to case-agreement. Similarly, we would expect that the reciprocal anteceded by the subject of the clause embedded under an ECM verb would be in the accusative case, agreeing with the accusative-marked subject Noun Phrase. These expectations, however, are not borne out. 189 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The reciprocal and adverb take nominative case; violating the case- agreement requirement, as illustrated in (10). (10)a.dhen perimena ton Yiorgho na yirisi melankholik-os/ *-o not expect.Isg [g^^the G.] return.3sg depressed-nom/*acc "I didn’t expect George to return depressed." b. perimena tis jinekes na simpathune i mia/ *tin mia tin ali expect. ls[accthe women]like.3pl the one.nom/ *acc the other "I expected the women to like each other." (lOa-b) do not violate the case agreement requirement if we adopt an ECM structure as in (11) where the Left Dislocated Noun Phrase is base generated in the topic position. I will indicate the position of the predicate variable with the notation ‘e.c.’ for empty category: (11)a.[ipdhen perimena [ton Y. [jp e.c. na yirisi melanxolikos]] not expect.Isg [accise G.] subj return.3sg depressed.nom "I didn’t expect George to return depressed." b. [ipperimena[tis jinekes[ip e.c. na simpathune i mia tin ali]] expect. ls[ap(.the women] like.3p the one.nom/ *acc the other "I expected the women to like each other." In the above structure the predicate variable is in the position to which nominative case is standardly assigned: it is in the spec of an I" 190 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. containing Agr(eement).“ The clause embedded under ECM verbs is inflected for person and number and hence has the capability to assign nominative case to its subject. The subject predicate variable straightfor wardly provides an antecedent with nominative case features for the agreeing adverbs and reciprocal. Under the Left Dislocation proposal, the apparently deviant case agreement facts are shown to simply be another instance of standard case agreement: the adverb/reciprocal agree with the nominative subject predicate variable. And if there is no movement from subject position to topic position we can provide some kind of explanation for the fact that not all the links in a putative chain formed by movement bear the same morphological case: they do not form a movement chain. In so far as subject and object (CL)LD can be viewed as involving essentially the same kind of structure, i.e., both predication, the above facts provide support for the hypothesis that (CL)LDed Noun Phrases “It is not clear if the antilocality licensing requirement, which we have seen plays an important role in licensing Cl-d, is relevant with respect to the licensing of LD of a subject. If it is relevant, this would suggest that the predicate variable (subject) is in postverbal, specifier of VP position, with an expletive occupying the spec of IP. Under this scenario, the expletive and postverbal subject would form a chain for Case assignment purposes. 191 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. are base generated in topic position, as opposed to arriving there via movement? In sum, we see there is substantial evidence that (CL)LD constructions must be base-generated and thus are not derived from Cl-d constructions (or regular overt subjects in the case of LD) via movement. The kind of evidence discussed above is entirely consistent with the structural representation of Cl-d proposed here which is based on a non movement relation between the two positions. 5.1.2. Evidence Conflicting With a U nified Account Although the evidence discussed earlier is consistent with the structural representation of Cl-d I have proposed here, there are a number of facts concerning CLLD that are used to argue for their base generated status precisely because it doesn’t seem possible that a Cl-d ^It is plausible that LD must involve a pro as a predicate variable instead of an overt Noun Phrase. Recall that I suggested earlier that it is precisely the presence of a clitic that allows an overt Noun Phrase to function as a predicate variable in which case, subject Left Dislocation cannot have an overt Noun Phrase as the predicate variable since there is no subject clitic. The conclusion that subject Left Dislocations involves pro as the predicate variable is consistent with latridou’s view that CLLD and LD involve essentially the same structure. Under the view argued for in the text, the two constructions are not identical, but the sole difference between them would be in the nature of the predicate variable. From this perspective, we can generalize a base generation argument for one construction to the other kind of construction, given that both involve predication. 192 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. expression can be the underlying source for a CLLDed Noun Phrase in these cases. I examine this kind of evidence next. 5.1.2.1. Some Languages Allow CLLD But Not Cl-d Cinque points out that a number of languages have CLLD but no Cl-d, e.g., Italian (examples from latridou (ibid.)); (12)a.Gianni, lo-conosciamo Gianni, cl.acc.m-know.lp "Gianni, we know him." b. *lo-conosciamo (a) Gianni cl.acc.m-know.lp (to) Gianni This fact indicates that the putative source of CLLD is never evident in some languages; an obligatory dislocation would have to be postulated. Cinque argues that such an analysis would preclude any meaningful insight into the CLLD construction. This argument need not necessarily be damning against an analysis that relates the two constructions (Cl-d and CLLD). Consider, e.g., the behavior of the wh-adjunct "why" in a language like English. It can never remain in-situ, not even as an echo question. But still, it is typically assumed that it moves out of the underlying position where non-wh reason adjuncts are found; i.e., its non-occurrence in adjunct position is 193 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. not taken as evidence against it originating in that position. In addition, there are languages like Chinese, where there is apparently never wh- movement in the overt syntax; but, typically this is not taken to indicate that there is no covert wh-movement at LF although there are no constructions in the language that overtly show such movement.'* The point is that absence of evidence for a phenomenon in the overt syntax does not necessarily mean that the phenomenon doesn’t occur. There is also a second possibility. And this is that there are two possible constructions for CLLD: (1) the structure proposed in this text, where a doubled Noun Phrase serves as predicate variable; and (2) the structure proposed by latridou, where a clitic (or: clitic/pro chain) serves as predicate variable. It is a priori posible that these two constructions co-exist within the same language or that different languages have either one or the other. Whether or not a doubled overt Noun Phrases could serve as a predicate variable in a language would depend on whether the language meets the necessary and sufficient conditions for having this kind of predicate variable in principle.^ The existence of two separate structures would help capture the following '*But cf. Aoun and Li (1993) for a non-movement analysis of Chinese wh-questions. ^This issue; i.e., the necessary and sufficient conditions that allow a language to have Cl-d in principle, is not explored in this dissertation. 194 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. implication, which, to the best of my knowledge, holds cross- linguistically: if a language has Cl-d, then it has CLLD, but not necessarily the reverse. That is, we cannot with confidence predict anything about the existence of Cl-d in a language just from knowing that the language has CLLD; but if we know a language has Cl-d, we can accurately predict that it has CLLD. If Cl-d languages do not have a representation of Cl-d that is directly connected to CLLD, these implications (or lack of implications) would just be accidental. Or, stated from another perspective, if CLLD in Cl-d languages has the structure proposed here (namely with an overt Noun Phrase as predicate variable), this implication can be easily captured, given that languages that do not have Cl-d have the representation of CLLD proposed by latridou. 5.1.2.2. Cl-d is More Restricted Sem antically Than CLLD 5.1.2.2.1. Animacy A more serious problem for my analysis of Cl-d, which essentially equates Cl-d and CLLD constructions, is posed by the fact, discussed in latridou (ibid.), that, in languages where clitic-doubling is sensitive to the animacy of the doubled Noun Phrase, this sensitivity is relaxed for CLLD. She notes that there are dialects of Spanish that allow clitic- doubling only of animate Noun Phrases. Nonetheless, she points out, inanimate Noun Phrases can be CLLD in these dialects. This means that 195 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. what can Cl-d is a proper subset of what can CLLD. But, if this is the case, how can Cl-d be the source for CLLD? The following examples illustrate the differing behavior of animate versus inanimate Noun Phrases: (13)a. lo-vimos [a Juan] cl.acc.m-see.lp to Juan "We see Juan." b.[a Juan] lo-vimos (14)a. *Io-vimos [el/al iibro] cl.acc.m-see.lp the/to + the book b.[el Iibro] lo-compramos ayer the book cl.acc.m-bought.lp yesterday "(speaking of) the book, we bought it yesterday." This restriction on the Cl-d of inanimates is related to Kayne’s generalization: Cl-d is only possible if an independent Case assigner is available for the doubled Noun Phrase. So, for instance, in the dialect of Spanish under discussion the morpheme a, analyzed as a Case assigner (see Jaeggli (ibid.)), must precede a Noun Phrase in order for that Noun Phrase to be able to double. The morpheme a can only precede animate Noun Phrases— it does not (and cannot) precede inanimate Noun Phrases. Since only animate Noun Phrases have an 196 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. independent Case assigner (in addition to the verb) (contrast (13a) with (14a)) only animate Noun Phrases can Cl-d as expected under Kayne’s generalization/’ ’ ^ latridou’s observations regarding the asymmetries between Cl-d and CLLD indicate that Kayne’s generalization is not relevant for CLLD, in direct contrast to Cl-d (I address the question of why this is so in the next section). The data seem to confirm a possibility alluded to in a previous section of the discussion, namely that there are two structures available for CLLD: one with a doubled Noun Phrase as predicate ‘ ’It is well known that not all Cl-d languages obey Kayne’s generalization. ^We see that the asymmetry between animates and inanimates with respect to Cl-d versus CLLD is clearly not an inherent semantic restriction on Cl-d. So, for example, in a dialect like the Porteho dialect of Spanish, discussed by Suner (ibid.), it is perfectly grammatical to Cl-d inanimates. Suner points out that this particular dialect of Spanish apparently doesn’t obey Kayne’s generalization. Also, the Cl-d of inanimates is perfectly grammatical in MG as is apparent throughout the text. Ingria (1980), among others, has pointed out that MG doesn’t seem to adhere to Kayne’s generalization. ^The kind of account that Kayne’s generalization can provide of the possibility of and the distribution of Cl-d is extremely intuitive under the idea, untenable from a minimalist perspective due to the elimination of government as a primitive relation, that a clitic asorbs the governing abilities and thus the Case assigning abilities (or at least absorbs the Case-assigning abilities) of the X" bearing the clitic, if a clitic absorbs the Case-assigning abilities of an X°, doubling is only possible with Noun Phrases that have independent Case assigners other than the verb (or other relevant X°), since only in this situation can a doubled Noun Phrase be Case-marked and hence grammatical. 197 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. variable and one with pro as predicate variable, and these two kinds of structures for CLLD can exist within the same language. This kind of asymmetry between CLLD and Cl-d does not force us to give up the hypothesis that all Cl-d constructions are connected to CLLD in the way proposed in chapter two. Maximally, we need to say that there are two different kinds of CLLD in the same language: (a) one with a doubled Noun Phrase as predicate variable, and (b) one with a pro as predicate variable whenever Kayne’s generalization comes into play.'^ 5.1.2.2.2. Partitives In addition to pointing out the sensitivity to animacy displayed by Cl-d in some languages, latridou (ibid.) also notes that there are semantic classes of Noun Phrases that can clitic left dislocate but cannot clitic double. She gives the following contrasting examples from MG: ‘ ^It is beyond the scope of this dissertation to explore CLLD with an eye toward whether or not its occurrence entails the presence of a Cl- ded Noun Phrase as predicate variable whenever possible (i.e., whenever its presence is not in violation of Kayne’s generalization) or not or just to what extent the grammar can choose between the two kinds of predicate variables. This problem seems to raise issues similar to those raised by wh-constructions in a language such as French, which allows both in-situ and ex-situ wh-expressions in wh-questions. 198 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (15)a.tria provlimata mono o Kostas ta-elise three problems only the Costas cl.acc.n(p)-solved.3s "Three problems are such that only Costas solved them." b. *mono o Kostas ta-elise tria provlimata only the Costas cl.acc.n(p)-solved.3s If clitic doubling were the source for CLLD, latridou points out, how would it be possible for Noun Phrases that cannot be doubled to occur as CLLDed Noun Phrases: the putative source of CLLD is a proper subset semantically of what can CLLD. To the above kind of case, I can add another kind of example (based on Mackridge (1985)): (16)a.fruta ta-tro sixna fruit, cl.pl-eat.ls frequently "Fruit, I eat frequently." b. *ta-tro fruta sixna cl.pl-eat.ls fruit frequently It can be noted that the semantic classes that can CLLD but not Cl-d are partitive. This fact is more salient in a language such as French that marks this kind of partitivity morphologically. In French, a (clitic) left dislocation comparable to (15a) and (16a) would use a special 199 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. partitive clitic en and a special partitive determiner des as illustrated in the following examples: (17)a. des problèmes, j’en ai résolu trois the.part(itive) problems, l’cl.part-solved three "Three problems are such that I solved them." b.des fruits, j’en-mange the.part fruits, I’cl.part-eat "Fruits are such that I eat them." MG does not have a special partitive clitic, the pronominal clitic can apparently carry this function for CLLD constructions.^” ‘”ln other Cl-d languages I am familiar with such as Lebanese Arabic and Peruvian Spanish, these kind of partitives also cannot be clitic doubled. Perhaps surprisingly, neither can they CLLD; i.e., when these kind of partitives appear in clause initial position, no clitic can co-occur on the verb, instead the partitive expressions are related to a gap as illustrated here for Lebanese Arabic: (i)a. *tleet masheekil Halleet-un three problems solved.Is-cl.acc.p "Three problems are such that I solved them." b. tleet masheekil Halleet three problems solved.Is "Three problems are such that I solved them." (ii)a. *fweeke beekal-un fruits eat.ls-cl.acc.p "Fruit is such that I eat it." b. fweeke beekal fruits eat.Is "Fruit is such that I eat it." To the best of my knowledge, these facts have not been discussed to any extent in the literature and certainly seem to raise intriguing questions 200 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. One a priori possible reason that Cl-d is not grammatical with partitives could be due to semantic restrictions on Cl-d. For instance, it is clear that only specifics can Cl-d. Nonetheless, it is not clear that there is a semantic reason for the prohibition on the doubling of these kind of partitives. Consider for instance that it is possible in MG to double the generic ta fruta, literally; ‘the fruits’; this means that a generic like ta fruta meets the specificity requirement on Cl-d. The partitive fruta would seem to be equally specific since it seems to designate part of the generic type ta fruta ‘fruit’. So the problem isn’t one of semantic constraints on doubling. 5.1.2.2.2.1. Kayne’s G eneralization There is, however, a clear syntactic difference between the two expressions in MG; namely, the fact that the generic, but not the partitive, has an overt determiner. Descriptively, this seems to be a necessary condition on Cl-d in MG: a doubled expression must have an overt determiner (either definite, or specific indefinite); specifics that lack determiners cannot Cl-d. I’d like to suggest that this requirement is a manifestation of Kayne’s generalization in MG, if we slightly reinterpret the generalization to require not the presence of an for comparative studies of grammar. 201 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. additional Case assigner, but to forbid the presence of a non-overt functional category where the grammar expects an (overt) functional category. For instance, in MG, clitic doubling requires a DP headed by an overt D", if there is no overt D°, doubling is impossible. In Peruvian Spanish, doubling requires the functional morpheme a to head a doubled expression; if there is no (overt) a, doubling is impossible. From the perspective advanced here, another way of viewing Kayne’s generalization is to notice that there can be no empty categories embedded within a predicate variable." If Kayne’s generalization is a restriction on predicate variables, it is not surprising that the NP copies in the subject position of the predication are insensitive to Kayne’s generalization since they are obviously not predicate variables. As is true with asymmetries concerning Cl-d versus CLLD of inanimates, the fact that an asymmetry exists in the doubling versus dislocation of partitives does not undermine the proposal that Cl-d constructions involve a topic in CLLDed position. The asymmetry simply indicates that not all CLLD constructions are based on Cl-d constructions. "This does not mean that a predicate variable cannot itself be an empty category, it just means that no subpart of a predicate variable can be empty. 202 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.2. Concluding Remarks In this chapter we have explored some of the asymmetries that exist between Cl-d and CLLD and have seen that the existence of these asymmetries is not incompatible with the analysis of Cl-d proposed here. The data suggest that there is more than one structure involved in CLLD: (a) one with a doubled Noun Phrase as the predicate variable, and (b) one with pro as the predicate variable. But no data suggest that there is more than one structure involved with Cl-d. In the next chapter, I will examine certain asymmetries that exist between doubled .QPs and doubled NPs. Moreover, I will look at the syntax of Cl-ded QPs in general. We will see additional reasons to believe that there is not a movement relation between the argument position and the left peripheral CLLDed position and that there is a predication relation between the topic and the doubled Noun Phrase. 203 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER SIX: THE SYNTAX OF CLITIC-DOUBLED QPs AND THE STRUCTURAL REPRESENTATION OF CL-D 6.0. Introduction In this chapter I provide additional support for the proposed structural representation of Cl-d chiefly through an exploration of the syntax of clitic-doubled QPs. We will see that there is a pervasive asymmetry between clitic-doubled QPs and clitic-doubled NPs such that a doubled QP cannot contain an expression that requires an antecedent. An exploration of this issue offers additional evidence in support of the claim that there is not a movement relation between the topic and clitic- doubled position and leads to a discussion of reconstruction effects in Cl-d constructions and supports the conclusion that the topic is in a specifier position. I will suggest that when doubled QPs are involved there is no rule of QR operative. Instead, there is a syntactic deletion of the predicate variable that leaves behind a syntactic variable. We can in this way account for the fact that QPs cannot be doubled if they contain expressions that must have antecedents internal to the minimal clause containing the clitic-doubled QP. To see this consider that if there were a syntactic deletion of the predicate variable, binding of any expression contained in the doubled QP would be impossible because the topic 204 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. position, where there is a copy of the doubled QP, is structurally higher than anything else in the clause and thus any expression contained in the doubled QP is outside of the c-command range of any clause internal antecedent. Reconstruction effects with reflexive anaphors and bound pronouns contained within doubled NPs in Cl-d and CLLD constructions are neatly accounted for because the position occupied by the predicate variable is within c-command range of potential antecedents and this predicate variable does not undergo QD for obvious reasons. It can be confirmed that there is a syntactic variable in the predicate variable position in Cl-d constructions involving Cl-ded QPs when we compare the interaction of doubled QPs with other scope- bearing or anaphoric elements within the clause containing the doubled expression and with other scope-bearing or anaphoric elements in a superordinate or paratactic clause. Finally, we will see that this picture of the syntax of Cl-ded QPs straightforwardly predicts and accounts for the scope interpretation of Cl-ded QPs in their interaction with other QPs and with negation. 6.1. Constraints On The Clitic Doubling of Certain Specific QPs We noted in chapter 4 that only certain QPs can Cl-d in MG: they must be specific (in the sense of Enc (ibid.)). Nonetheless, meeting this 205 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. requirement of specificity is not sufficient to ensure that a QP can clitic double. For instance, it is not possible to double specific QPs which contain bound pronouns or focused expressions. Moreover, the reciprocal expression: o enas ton alo ‘the one the other’, which behaves like a QP, also cannot be clitic doubled. These facts can be shown to follow from the interaction of the syntax of clitic doubling in MG with the syntax of the prohibited QPs. 6.1.1. Doubled Q P /N P Asymmetries 6.1.1.1. Bound Pronouns A pronoun contained within a non-doubled NP or QP can function as a bound pronoun as long as a c-commanding operator is coindexed with it. In the following sentences we see that a c-commanding wh-subject can bind a pronoun contained within a non-doubled NP or QP respectively: (l)a. pioSj pulise [„p to vivlio tUj] who sold the book his "Who sold his book?" b. pioSj agorase [^p ton kathe pinaka apo to xorio tUj] who bought the every painting from the village his "Who bought every painting of his village?" 206 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The same is true of a pronoun contained within a clitic-doubled NP; it can function as a bound pronoun as long as a c-commanding operator is coindexed with it: (2) pioSj to-pulise [„p to vivlio tUj] who cl.acc.n-sold the book his "Who sold his book?" But, although it is perfectly grammatical to double a QP (see (3)), a pronoun contained within a clitic-doubled QP may not function as a bound pronoun (see (4)) even if there is a coindexed operator which c- commands the pronoun in the overt syntax. The following sentences illustrate these facts: (3) i Tzeni tOj-malose [to kathe pedhi]j the Janey cl.acc.n-scolded the every child "Janey scolded every child." (4) pioSj ton-aghorase [^p ton kathe pinaka apo ton xorio tu^/.j] who cl-acc.m-bought the every painting from the village his "Who bought every painting from his village?" It would be possible to account for this asymmetry between doubled QPs and doubled NPs if doubled QPs, but not doubled NPs were to raise outside of the c-command range of the wh-subject at LF. 207 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. If the doubled QP occurs outside of the c-command range of antecedents internal to the clause, this would prevent a bound pronoun interpretation for a pronoun contained within the doubled QP since c- command by the relevant operator is required in order for a pronoun to receive a bound interpretation/ If such a raising/non-raising distinction exists, it would mean that the following LF representations would exist for doubled QPs and doubled NPs respectively (I abstract away from the presence of the topic copy for the moment); (5)a. [QPj [,p WH, ... clj-V x- ...] b. [,p WHk ... clj-V NPj ...] The above generalization suggests that there is nothing, in principle, ungrammatical about bound pronouns contained within clitic- 'In the pre-minimalist literature there are two different proposals as to the level the bound pronoun requirement holds at: s-structure or LF. Notice that the Cl-d facts are consistent with Minimalist conclusions that there is no s-structure because they indicate that the c-command requirement for bound pronouns is not (and cannot be) imposed at the putative level of s-structure. If it were, there would be no asymmetry between bound pronouns in doubled QPs versus bound pronouns contained in doubled NPs; they should both be licensed since they are both c-commanded by the wh-phrase at the putative level of s-structure. But, if the c-command requirement holds at LF, it is possible to distinguish between the two constructions given that the doubled QP but not the doubled NP undergoes QR or an equivalent rule at LF. 208 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. doubled QPs provided that there is a c-commanding operator available to bind the pronoun in question. Given that the raising of a doubled QP is clause bound and if the operator and bound pronoun contained within the doubled QP are in separate clauses, then as long as c-command obtains between the operator and the pronoun contained within the doubled QP a bound pronoun interpretation should be allowed. This expectation is accurate as the following data indicates where an operator in a superordinate clause grammatically binds a pronoun contained in a Cl-ded QP in the subordinate clause: (6)? pioSj ipe oti i Maria ton-misouse who said that the Maria cl.acc.m-hated ton kathe erasti apo ton parelthon tUj the every lover from the past his "Who said that Maria hated every lover from his past?" If the raising of a clitic-doubled QP is clause bound such that the doubled QP raises outside the c-command range of potential antecedents contained within the embedded clause and adjoins to the subordinate CP, the following configuration results: (7) pioSj ipe oti [^ .p [ton kathe erasti apo ton parelthon-tUjj^ who said that the every lover from the past-his 209 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. [cp i Maria ton-misuse x^.]] the Maria cl.acc.m-hated In this configuration the bound pronoun contained within the raised, doubled QP, although outside the scope of any operator internal to the clause that contains the doubled OP, is still within the c-command range of the operator in the superordinate clause; therefore, a bound pronoun reading is available when an antecedent in the superordinate clause is involved. In sum, the postulation of a raising/non-raising distinction between doubled. QPs and doubled NPs respectively, such that the doubled QP raises just outside of the CP containing it, captures the asymmetric distribution of bound pronouns in the two types of Noun Phrases. 6.1.1.2. An Asymmetry with Focused Expressions An asymmetry similar to that which exists for bound pronouns exists for focused expressions contained within doubled QPs and NPs. The distribution of non-focused expressions is not affected by clitic 210 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. doubling: doubled and non-doubled QPs and NPs alike can contain non focused expressions.- The distribution of focused expressions contrasts with that of non focused expressions in that focused expressions can occur internal to doubled NPs but not internal to doubled QPs. This is illustrated in the following examples: (8)a. tOj-dhiavaza [tu CHOMSKY [to vivIio]]j cl.acc.n-read.ls by CHOMSKY the book "I read the book by CHOMSKY." “Within the Noun Phrase, focused expressions have a different word order pattern from non-focused expressions. Non-focused expressions occur in post nominal positions, whereas focused expressions occur in immediate prenominal position (see Horrocks and Stavrou (ibid.), and Mackridge (ibid.)): (i)a. to vivlio tu Chomsky (non-focused) the book the.gen Chomsky "the book by Chomsky" b.tu CHOMSKY to vivlio (focused) the.gen Chomsky the book "the book by CHOMSKY" (ii)a. kamia paramia apo to Muzaki (non-focused) some saying from the Muzaki (a town) "some saying from Muzaki" b.apo to MUZAKI kamia paramia (focused) from the Muzaki some saying "some saying from MUZAKI" 211 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b.? *tOj-dhiavaza [tu CHOMSKY [to kathe vivlio]]j cl.acc.n-read.ls by CHOMSKY the every book "I read every book by CHOMSKY." It is perfectly grammatical for a non-doubled NP or QP to contain a focused expression. Cf.: (9)a. dhiavaza [„p tu CHOMSKY [to vivlio]] read.Is by CHOMSKY the book "I read the book by CHOMSKY." b. dhiavaza [^p tu CHOMSKY [to kathe vivlio]] read.Is by CHOMSKY the every book "I read every book by CHOMSKY." This asymmetry can be captured in a manner identical to the doubled QP/NP asymmetry with bound pronouns if there is a raising/non-raising distinction for doubled QPs and doubled NPs respectively. To see how this would work, suppose that when there is a focus interpretation there is a focus operator in the clause that binds the focused expression, marked XP j? This is illustrated in the following example: ^Alternately, there is a raising operation of the focused expression to the specifier of FP. 212 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (10) [ O p r (focus operator) [ [np ^ C.C.f ] ] .........]] Recall that there is an FP that is structurally higher than CP but structurally lower than the Topic Phrase (CLLDed position). The natural location for the focus operator would be in the spec of FP. The following example illustrates the proposed location of the focus operator: ( 1 1 ) [fp ^ P f (focus operator) [ip ............................ '^f........... 11 Given that a doubled QP is raised to a position that is structurally higher than the FP, binding of a focused expression within a doubled QP by the focus operator in spec of FP should be impossible due to the fact that the focus operator would be unable to c-command the focused expression. The following example illustrates the configuration under discussion: (12) DQUBLED QP *[[ X P j . [ Q P j ] ] [...Qpr (focus operator) - c l j - V Xj . . . ] ] ] Cl-ded NPs can contain focused expressions since the doubled NP does not raise so the focus operator c-commands the focused expression, allowing a focus interpretation as illustrated in the following example: (13) DOUBLED NP [ - O P f (focus operator) - clj'V [ XPj^ [NPj]] ...]]] 213 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In sum, asymmetries in the behavior of doubled QPs and doubled NPs could be captured if Cl-ded QPS but not Cl-ded NPs undergo LF raising to a structurally very high position.'* We have seen that the raising of a doubled QP would have to be to a position that is outside of the CP or FP containing the clitic doubled QP. Indication that the raising must be to such a high position came from the behavior of Cl-ded QPs which could not contain neither pronouns bound by an operator in the local CP nor focused expressions bound by a focus operator in FP. 6.1.1.3. The Impossibility of the Doubling of Reciprocals 6.1.1.3.1. The Phenomenon There is a third kind of specific expression that cannot be doubled: the reciprocal. We will see that this restriction can be analyzed in almost exactly the same way as the previous examples involving bound pronouns and focused expressions contained in Cl-ded QPs. The fact that reciprocals cannot be clitic-doubled is illustrated in (14): “ ’Notice that this conclusion goes against some proposals in the literature that Cl-ded QPS behave like names and do not undergo raising (cf. Dobrovie-Sorin (1990), etc.). I discuss this in more detail later in this chapter. 214 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (14)a.i kopelles idhan [reciprocal i rni^ tin all] the girls saw.3p the one the other "The girls saw each other." b. *i kopelles tin-idhan [reciprocal i mia tin ali] the girls cl.acc.f-saw.3p the one the other This restriction cannot be attributed to a general prohibition on the clitic-doubling of anaphors, since it is perfectly grammatical to double a reflexive anaphor;^ (15)a.i kopelles idhan ton eafto-tous s’to kathrefti the girls saw.3p the self-cl.gen.f.p in’the mirror "The girls saw themselves in the mirror." b.i kopelles ton-idhan ton eafto-tous s’to kathrefti the girls saw.3p the self-cl.gen.f.p in’the mirror "The girls saw themselves in the mirror." 6.1.1.3.2. The A ccount The more traditional view of the syntax of the reciprocal holds that it is to be treated as an anaphor. Specifically, the reciprocal expression is subject to binding principle A and must be bound by its ^MG lacks reflexive clitics, in contrast to Spanish, for instance. Reflexives in MG are doubled by the pronominal clitic. 215 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. antecedent in the appropriate local domain; it is treated exactly like the reflexive anaphor. The indexing in the following example illustrates this position: (16) [aftes i jinekes]j simpathun [i mia tin ali]j these the women like the one(f.sg.nom) the other(f.sg.acc) "These women like each other." We can refer to this view as the binding analysis of reciprocals. Reciprocals in MG provide good evidence for an opposing view of the representation of reciprocals, which has been advanced by Heim, Lasnik, and May (1991) (HLM). HLM propose that the reciprocal consists of two parts, the pronominal ‘other’ (its language specific equivalent in MG is tin ali) and the quantifier ‘each’ (its language specific equivalent in MG is i mia). Both parts of the reciprocal are subject to the binding theory. HLM suggest that the pronominal ‘other’ must be contraindexed with the so-called "antecedent." They argue that ‘each’ functions as a distributor which at LF raises to the "antecedent" from whence it binds its trace left by the raising process/' In addition, both parts of the reciprocal expression: [[ NP ] eachy], and [ X y other] '’They assume that the index of ‘each’ is transmitted to entire NP. For additional details of their analysis see HLM (ibid.). 216 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (where X y is the variable left by the raising of the distributor ‘each’), are QPs which undergo QR. Their analysis is illustrated in (17) as it applies to English and in (18) for MG: (17) [jp[[The children]j eachy] [,p X y [^p [ X y otherj^ Lp like x^ ]]]] (18) [jp[[i jinekes]j]i miay][ipXy Lp[ X y tin alij^ Lp simpathun x,, ]]]] the women(nom) the one(nom) the other(acc) like "The women like each other." We can refer to this view as the movement analysis of reciprocals. The behavior of the reciprocal expression o enas ton alo supports the movement analysis of reciprocals. On the one hand, the reciprocal behaves like a single constituent. For instance, a reciprocal direct object can either precede or follow the indirect object in which case the reciprocal must move as a unit, it is not possible to move only the distributor part or only the pronominal part of the reciprocal. Cf.: (19)a.i jinekes sistisen i mia tin ali stin Klari the women introduced the one the other to’the Klari "The women introduced each other to Klari." b.i jinekes sistisen stin Klari i mia tin ali the women introduced to’the Klari the one the other "The women introduced each other to Klari." 217 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. *i jinekes sistisen i mia stin Klari tin ali the women introduced the one to’the Klari the other d. *i jinekes sistisen tin ali stin Klari i mia the women introduced the other to’the Klari the one For a more complete discussion of the fact that the reciprocal in MG behaves like a unit see Joseph and Philippaki-Warburton (ibid.). Despite its syntactic behavior as one unit, in terms of case morphology, the reciprocal behaves like two separate units. The reciprocal in MG consists of two elements: (i) i mia ‘the one’, and (ii) tin ali ‘the other’. The first part of the expression (/ mia) must agree in case and gender with its antecedent. The second part of the expression {tin ali) agrees in gender with the antecedent, but not in case; instead, its case morphology reflects the syntactic position it occupies. A paradigm which illustrates the case agreement facts is given in (20) (examples a., b., and d. are from Joseph and Philippaki-Warburton (ibid.)). (20)a.afti i anthropij.^^s^ antipathun [o enasj.case ton alo] these the men dislike the one(m.nom) the other(m.acc) "These men dislike each other." 218 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b.aftes ! jinekeSj.„sc simpathun [i tin ali] these the women like the one(f.nom) the other(f.acc) "These women like each other." c.i Maria sistise tus papadheSj ^ase [ton enaj.casc s’ton alo] the M.introduced the priests(m.acc) the one(m.acc) to’the other(m.acc) "Maria introduced the priests to each other." d.o Yanis sistise tis jinekeSj.^-asc [ti oiiaj.casc s’tin ali] the V.introduced the women the one(f.acc)to’the other(f.acc) "Yannis introduced the women to each other." In examples a. and b. the antecedent occurs in subject position and the reciprocal in object position. The first part of the reciprocal bears nominative case morphology in agreement with the antecedent’s nominative case and the second part of the reciprocal is marked accusative reflecting the syntactic position of the reciprocal. In examples c. and d. the antecedent occurs in object postion. The reciprocal is object of a preposition.^ In these instances the first part of the ^When prepositions take reciprocals as objects, the reciprocal must be separated by the preposition and this might seem to cast some doubt on the claim that the reciprocal behaves like a unit syntactically. But it must be noted that although the distributor is placed in front of the preposition, the reciprocal is otherwise frozen and cannot be moved outside of the PP. 219 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. reciprocal is accusative in agreement with the accusative antecedent. The second part is also accusative since the reciprocal is the object of a preposition which assigns accusative case. Under an analysis that treats the reciprocal as a unit it is not expected that a lexical item can have two different cases. The separation of the reciprocal form in MG into two parts morphologically provides support for an approach that "decomposes" the reciprocal into two distinct parts as the movement analysis does. The existence of two different morphological cases for the two parts of the reciprocal, where the distributor agrees in case with the "antecedent," specifically supports the movement analysis of reciprocals since under this view the distributor must occur in the same Case position as the antecedent. Given the movement analysis of reciprocals, suppose that the QP: [ X y tin ali] were clitic-doubled. Suppose also that a clitic-doubled QP must raise outside of CP as illustrated in the following example: (21)[[Xy tin ali]J,p[ip[[i jinekes] i mia^.jy [jp X y tin-simpathun x^ ]]] Since the doubled QP has raised so high, it is no longer in the scope of the distributor i mia; therefore, the variable left by the raising of the distributor i mia to the antecedent position cannot be bound since it is internal to the reciprocal and the reciprocal is outside of the c-command 220 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. range of the distributor. The representation is ungrammatical due to the presence of a free variable in the doubled QP. The impossibility of clitic-doubling a reciprocal in MG can be traced back to the requirement that a doubled OP raise to a position which is structurally higher than the antecedent of the variable contained in the raised QP. So it is the interaction of two independent requirements: (a) clitic-doubled QPs must raise to a very high position, and (b) the requirement that the distributor in reciprocal expressions raise to the antecedent position, in conjunction with the fact that tin all is a QP, that prohibits the occurrence of clitic-doubled reciprocals. In contrast to the impossibility of having Cl-ded reciprocals, consider now the grammaticality of clitic-doubled reflexives in MG. I repeat (15) here to illustrate the phenomenon: (15)a.i kopelles idhan ton eafto-tus s’to kathrefti the girls saw.3p the self-cl.gen.f.p in’the mirror "The girls saw themselves in the mirror." b.i kopelles ton-idhan ton eafto-tus s’to kathrefti the girls saw.3p the self-cl.gen.f.p in’the mirror "The girls saw themselves in the mirror." The reflexive consists of two parts: the invariably masculine: ton eafto ‘the self’ and the possessive pronoun {tus ‘their’ in (15)) which 221 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. agrees in number, gender, and person with the antecedent. latridou (1988) points out that the actual anaphoric part of the reflexive in MG is the possessive pronoun, and that the clitic that doubles the reflexive agrees with the invariable: ton eafto ‘the self’. Although both reflexives and reciprocals require clause internal antecedents, if reflexives in MG, in contrast to reciprocals, are not QPs, then we immediately have an explanation for the grammaticality of their occurrence in Cl-ding constructions: when clitic-doubling occurs, QPs (and only QPs) raise to a very high position which is outside the c- command range of any possible antecedent for an anaphor, or, in this case, for a subpart of the anaphor. Given that the reflexive expression is not a QP, it will remain within the c-command range of its antecedent since doubled NPs do not raise and nothing prevents its being doubled. The general behavior of the reflexive in MG is entirely consistent with the conclusion that reflexives in MG are not QPs. Katada (1992), in her study of the LF representation of anaphors, points out that anaphors that have operator status cannot be scrambled. She suggests this can be attributed to a requirement that when elements that must occur in an A’-position by LF (namely, operators) occur in non-argument positions at s-structure they must remain in those non-argument positions. (That is, it is not possible to reconstruct and raise again in the same level of representation.) Due to this requirement, she points out, 9 7 9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. anaphors with operator status do not reconstruct, and an anaphor that cannot reconstruct will not have an antecedent to bind it in violation of Principle A of the Binding Theory. And, therefore, it is not possible to scramble such an anaphor to an A’-position. Anaphors in MG can easily be tested against this requirement for their operator/non-operator status. The clause-initial focus position in MG behaves with respect to all diagnostics like an A’-position as has been pointed out earlier. In focus constructions the behavior of reflexive anaphors in MG contrasts directly with reciprocals in that reflexives but not reciprocals can be scrambled to the focus position. Cf. (22) and (23); (22) [ton eafto tisj] antipathi i Mariaj the self hers dislike the Maria "Maria dislikes herself." (23) *[i miaj tin ali] simpathun i jinekesj the one the other like the women The impossibility of the reciprocal occurring in the non-argument focus position in (23) can be attributed to the fact that the reciprocal is a QP occurring in an A’-position and since it is a QP, it may not reconstruct. Since reconstruction cannot occur, the reciprocal, or more properly, the variable left by raising of the distributor, cannot find an antecedent and the expression is ruled out. In contrast, the grammaticality of the 223 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. reflexive occurring in an A’-position in (22) is expected if the reflexive is not a QP and therefore may reconstruct to an A-position at LF. In this way the reflexive will have an antecedent in adherence with Principle A of the Binding Theory. In sum, the fact that reflexives but not reciprocals can be clitic doubled can be traced back to the following points: (a) reflexives are NPs, not QPs, in contrast to reciprocals, which are QPs, (b) there is a raising/non-raising distinction between clitic-doubled QPs and clitic- doubled NPs respectively, and, (c) clitic-doubled QPs must raise to a very high position in the clause, which is outside the c-command range of any clause internal antecedent. Given these differences, no QP of which a subpart requires an antecedent internal to the clause containing it can be clitic-doubled because clitic-doubled QPs are raised to such a structurally high position where c-command by any clause internal antecedent would fail. In contrast, NPs that require an antecedent internal to the clause can be clitic-doubled because they do not undergo raising to the peripheral topic position and therefore remain within the scope of any relevant antecedent. 224 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6.1.2. Against a Raising Approach With Cl-ded QPs In the previous chapters, I provided a number of arguments in support of the hypothesis that Cl-ded expressions are related to the clause initial topic position and that this relation is not mediated by movement, but instead by predication. The apparent raising/non-raising asymmetry that exists between doubled QPs and doubled NPs provides another argument against a general movement approach to Cl-d constructions because maximally only QPs would raise to the topic position. If NPs and QPs both were to blindly raise to the topic position, NPs but not QPs would have to reconstruct again. We know this because doubled NPs behave as if they are in the c-command range of clause internal antecedents. It violates economy principles to raise and reconstruct something at the same level of representation, as would have to be the case with doubled NPs. So it clearly cannot be that there is a general process of raising to the topic position in Cl-d constructions. In what follows, I am going to reject the hypothesis that raising of any kind applies in Cl-d constructions. Specifically, I’d like to suggest that, although the hypothesis that Cl-ded QPs raise to a very high position can help account for a wide array of facts, in fact there is no QP raising and therefore no raising of any kind in Cl-d constructions. 225 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A very strong conceptual reason to reject the QP raising approach is the fact that there is complete redundancy between the position to which the doubled QP must raise and the position independently determined for the QP copy in topic position: if raising of the doubled QP were to take place, it would have to apply in such a way when Cl-ded QPs are involved that the doubled QP would always raise to the topic position instead of to some intermediate position between the argument and the topic position. Recall that the data involving wh-antecedents indicated that the doubled QP must be higher than CP and the data involving focused expressions showed that the doubled QP must be higher still: above FP. This places the doubled QP as high as the topic position, a position that, by hypothesis, the doubled QP copy already occupies. This kind of redundancy complicates the grammar in a completely unnecessary way. Moreover, there are some empirical problems with postulating such a rule of quantifier raising in Cl-d constructions because the rule that is usually assumed to take care of the scope of quantifiers (namely the rule of QR (Quantifier Raising (see, for instance. May (1977)), raises QPs to a much lower position than Tpp. The rule of QR is generally assumed to raise a QP to an A’- position (but see Hornstein (1994) for a different view) that reflects the scope of the QP and in this process creates a variable in the argument 226 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. position. It is usually assumed that the raised QP is adjoined either to the VP or IP containing it. The following examples illustrate the standard assumptions about what happens with respect to adjunction sites when an object QP undergoes QR: (24)a.[,p [jp ... Lp QPj Lp ... X j ... ]]]] b-[cp [ip QPj [ip - [vp - X j ... ]]]] To see how QR applies in the usual case, when non-doubled QPs are involved, consider the following examples. Recall that a non-doubled QP can easily contain a bound pronoun, which means that the pronoun contained in the non-doubled QP can remain within the scope of the wh- subject. The following LF representation must be possible for a non- doubled QP: (25)a.Lp WH LpQPj [vp V X j ]]] b.LpWH [ipQPj[LpV X j ]]]] Since the rule of QR adjoins a QP to either VP or IP, it is obvious that the wh-subject can c-command and therefore bind a pronoun contained in a QP that is not doubled. When non-doubled QPs containing focused expressions are involved, there are two possible structures created by QR, both of which correctly predict binding of a focused expression contained within a non- 227 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. doubled QP by the focus operator. The rule of QR can adjoin the QP containing the focused expression either to VP or to IP. In either case, a focused expression contained within the raised QP is c-commanded by and bound by the focus operator which is located higher than the QP in both cases; namely, in the spec of FP. The following examples illustrate these two possibilities for QPs: (26)a.[fp Q p f (focus operator) [ --- [vp [ [ Q F j ] ] [ v p ^ ] ] ] ] b [cp O P r (focus operator) [ip [ [ Q P j ] ] [ | p L p ^ ] ] ] ] So, as expected, QR applies in such a way as to allow a focused expression to be contained within a non-doubled QP. In contrast, when Cl-ded QPs are involved, it is evident from the distribution of doubled QPs which contain expressions that are dependent on clause internal antecedents that a doubled QP occurs very high syntactically.^ The data suggest that we have the following representation where the Cl-ded QP has raised past the wh-operator, to the QP topic: (2 7 ) [ QPj QP,opj,.j [cp piosy [ip X y [^p ... X j ... ]]]] ^And we will later see that the scope interpretations of doubled QPs lead us to the same conclusion too. 228 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The above representation would be appropriate for the sentence given in (4), which involves a clitic-doubled QP, repeated here for convenience: (4)pioSj ton-aghorase [^ p ton kathe pinaka apo ton xorio tu,^y,j] who cl-acc.m-bought the every painting from the village his "Who bought every painting from his village?" From what we have just seen about QR when non-doubled QPs are involved, the rule of QR dosen’t normally raise a QP so high. In fact, the standard QP adjunction sites of VP and IP are not available when Cl-ded QPs are involved. From the fact that we can’t have a bound pronoun contained within Cl-ded QPs we know that the following configuration cannot exist for Cl-ded QPs: (28) * Q Piopic-j [c p Piosy [ip Xy L p Q P j L p ... X j... ]]]] It is not possible for the clitic-doubled QP to have raised to a VP adjunction site since, otherwise, contrary to fact, it would then have narrow scope with respect to the Wh-operator, which would then c- command it at LF and thus be able to bind a pronoun contained within 229 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the doubled QP.‘ ^ The considerations are similar when we consider and exclude IP as an adjunction site. That the doubled QP is not allowed to adjoin to VP is surprising: this contrasts directly with the behavior of a non-doubled QP which must adjoin to VP in essentially the same context. Moreover, the fact that the doubled QP can and must raise above CP is surprising since this is not at all possible for a non-doubled QP. 6.1.3. QR By D eletion The contrast between doubled versus non-doubled QPs indicates that, if it is operating, QR raises the two different kinds of QPs to very different adjunction sites: a doubled QP must be raised to a much higher position than a non-doubled QP. I’d like to suggest an account of these facts that eliminates redundancy and arbitrariness by introducing a rule of QR based on deletion of the QP in argument position rather than raising of the QP to the topic position. Given the structural representation of Cl-ding I have proposed, where a copy of the doubled expression occurs in the specifier of Tpp, ‘ ^That is, provided that the high topic doesn’t detract from binding possibilities. It seems that the topic doesn’t participate in this way as indicated by the behavior of doubled NPs since they are all equally connected to a topic in the clause initial position. 230 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. OR need not actually move the QP in argument position to the topic position at LF, instead, QR could simply affect the QP in direct object position by syntactically deleting it. The existence of QD can provide an answer for why the doubled QP cannot be interpreted as adjoined to VP or IP: such an option doesn’t exist since there is no movement of the QP. UG forces the deletion of the QP to leave behind a variable, because in order for the QP in topic position to be licensed, there must be a variable for it to bind (the Bijection Principle as well as the general observation that vacuous quantification in the syntax is impossible would require a variable). The fact that QPs must bind a variable makes QP deletion obligatory given that there cannot be the regular kind of QR in Cl-d constructions. Moreover, given that the motivation behind the rule of QR is that quantifiers need to take the appropriate scope with respect to the relevant proposition, raising of the QP in direct object position to any possible adjunction site is redundant when topic constructions are concerned since the presence of a copy of a doubled expression in the topic position already accomplishes this. It is more economical for the grammar to simple delete in topic constructions. It seems that something like ‘shortest move’ is operative because here there is no movement. That is, it is possible to have no movement whatsoever; therefore, the grammar requires this option. 231 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I have suggested that there is a deletion process involved in determining the scope/interpretation of QPs in Cl-d constructions. We will see that the characteristics of this deletion process provide another argument against an analysis of movement from the argument to topic position in Cl-d constructions. We will see that there is a difference between the two processes with respect to how reconstruction works. In wh-movement, a Noun Phrase can be extracted which contains a reflexive anaphor (or other expression requiring an antecedent such as a bound pronoun). This reflexive can be interpreted as being bound by an antecedent internal to the clause containing the site from which the wh- expression is understood to originate. The following example illustrates this phenomenon: (29) [piafotoghrafia tu eaftu tu] petakse o Yiorghos s’ta skupidhia which photo the.gen self his threw the G. to’the garbage "Which picture of himself did George throw in the garbage?" We understand the reflexive anaphor tu eaftu tu as having o Y iorghos as its antecedent although it is not obvious that o Y iorghos c-commands the reflexive in the overt syntax. That is, we understand tu eaftu tu as actually occurring in its original position. It has been proposed in the literature that a phrase such as ‘pia fotoghrafia tu eaftu tu’ reconstructs, or is moved back into the orignal 232 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. argument position from whence it originates (minus the operator itself). That is, everything except the operator reconstructs as the operator must remain in the position that marks its scope. Under this view we would have something like the following representation at LF: (30)[pia x]k petakse o Y.j [x^ fotoghrafia tu eaftu tUj] s’ta skupidhia which photo the.gen self his threw the G. to’the garbage "Which picture of himself did George throw in the garbage?" Under the copy theory of movement, the interpretation of the moved expression comes about in a more economical way. Instead of reconstructing part of the extracted expression, in the sense of moving the expression back into its original site, the expression is never moved in the first place. Instead, a copy of the wh-expression occurs in the specifier of the CP in which it is interpreted and at PF and LF the appropriate phrases or subparts of phrases are deleteed in order to get the correct phonetic form and the correct LF. The following sentence gives an example of wh-movement under the copy theory of movement: 233 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (31) [piafotoghrafia tu eaftu tu] petakse o Y. which photo the.gen self his threw the G. [pia fotoghrafia tu eaftu tu] s’ta skupidhia which photo the.gen self his to’the garbage "Which picture of himself did George throw in the garbage?" With QD we do not get the pattern of reconstruction that we saw for wh-movement. Instead, QD behaves as though nothing has reconstructed. That is, the entire expression must delete, not just the subpart corresponding to the quantifier, but everything contained in the doubled QP must delete too as is evident from the fact that there cannot be expressions within the QP that require antecedents. This behavior of Cl-ded QPS can be traced back to Kayne’s generalization in the form we recast it earlier: a predicate variable may not contain an empty category, as would be the case if just the quantifier, for instance, deleted. The fact that copy and deletion works differently in wh-constructions and Cl-d constructions is not surprising if ‘copy’ doesn’t automatically imply movement. Finally, the fact that Cl-ded QPs are not interpreted as if they are reconstructed supports the conclusion advance in chapter four that topics are in specifier, as opposed to adjoined positions. Chomsky (1994) and others have argued that adjoined expressions must fully reconstruct 234 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. into their original sites. Since Cl-ded QPs are clearly not interpreted as if they have reconstructed, this is another reason to conclude that topics are in specifier position. 6.2. Arguments for the Presence of a Variable With Cl-ded QPs 6.2.1. D ifferences in Local and Non-Local Behavior of Cl-ded QPs We have reached the conclusion that Cl-ded QPs undergo a rule of QD which deletes the doubled QP predicate variable and leaves in its place a syntactic variable. The variable is bound by the QP copy in the specifier of TpP. But the conclusion'that there is a variable associated with a Cl-ded QP contrasts with what has been claimed in much of the previous literature where it has been doubted that variables can occur in clitic-doubling constructions or that there really is an operator/variable structure with Cl-ded QPs (but Cf. Aoun (1991), (1993), and Schneider-Zioga (1990), (1993)). In this section, I will study some of the paradigms which have formed the basis for the claim that Cl-ded QPs seem to behave like names (i.e., like NPs as opposed to QPs). 1 will demonstrate that a cl-ded QP seems to lack an operator/variable structure if we consider its behavior with respect to pronoun binding and scope interaction with operators contained within the same clause as contains the doubled QP. But if we consider the behavior of the Cl-ded QP with respect to 235 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. operators and pronouns that are in separate clauses, we will see that a Cl-ded QP behaves exactly like a non-doubled QP; i.e., it behaves like it does have an operator/variable structure. We will reach the conclusion that a Cl-ded QP does indeed have an operator/variable structure in all cases but the fact that doubled QPs occur in such a structurally high position (specifier of TpP) obscures this. 6.2.1.1. Clitic-Doubled QPs and Effects Clausemates 6.2.1.1.1. Weak Crossover and Cl-ded QPs Perhaps the most persuasive evidence that Cl-ded QPs are like names; i.e., lack an operator/variable structure, is the fact that Cl-ded QPs are able to avoid Weak Crossover (WCQ). Descriptively speaking, a WCQ violation arises where a non-referential expression is (ungrammatically) co-indexed with a pronoun to its left which does not c-command it. In the following sentence, which involves a clitic-doubled QP, there are no Weak Crossover (WCQ) effects although the structural description for WCQ is apparently met: (32) [i mitera tUj] tOj-agapai to kathe pedhij the mother his cl.n-loves the every child *? "HiSj mother loves every childj." 236 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The behavior of Cl-ded QPs differs sharply from that of non-doubled QPs which do give rise to WCQ in essentially the same context. Cf. (32) and (33): (33)* [i mitera-tUj ] agapai to kathe pedhij the mother his loves the every child ?*"HiSj mother loves every childj." The fact that Cl-ded QPs do not induce WCQ has led to the suggestion that Cl-ded QPs differ fundamentally from non-doubled QPs. Qne proposal in the literature (the proposal was made specifically for Cl-d in Romanian, see Dobrovie-Sorin (ibid.)) is that Cl-ded QPs lack operator/variable structure; this is essentially a claim that doubled QPs have the LF syntax of names. WCQ arises when an attempt is made to coindex a non-referential expression with a non-c-commanding pronoun to its left. Therefore, Dobrovie-Sorin’s claim that Cl-ded QPs lack an operator/variable structure can immediately account forwhy there is no WCQ when Cl-ded QPs are involved: there is no non-referential expression (no variable) in these cases. 237 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6.2.1.2. Pronouns In Separate Clauses However, contrary to what an analysis like Dobrovie-Sorin’s would predict, Cl-ded QPs behave differently in local and non-local domains. This is revealed when we examine the behavior of doubled QPs in configurations where the non-c-commanding pronoun is no longer in the same clause but instead occurs in a clause which is superordinate to that which contains the Cl-ded QP. The kind of configuration under consideration is illustrated in the following example: (34) [ip ...[ N pronouoj] ....[jp ... clj-V QP/NPj ... ]] As pointed out in Schneider-Zioga (1990), when a non-c- commanding pronoun occurs in the clause which is superordinate to the clause which contains the Cl-ded QP, it is strongly ungrammatical to coindex a clitic-doubled QP with a non-c-commanding pronoun to its left. This fact is illustrated in (35): (35)* [o dhaskalos tuj] ipe oti to-malosan to kathe pedhij the teacher his said that cl.scolded-3pl the every child *"HiSj teacher said that they scolded every childj" 238 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. If Cl-ded QPs were really like names, the distance of the pronoun from its "antecedent" shouldn’t have any effect on reference possibilities, as the following sentence involving a name illustrates. Cf.; (36) [o dhaskalos tUj] ipe oti (ton-)malosan ton Sakij the teacher his said that (cl.acc.m-)scolded-3pl the Sakis "HiSj teacher said that they scolded SakiSj" A true name, Cl-ded or not, can be quite far from the pronoun with which it is coindexed and the sentence is perfectly grammatical in contrast to the behavior of a clitic-doubled QP. When we consider a pronoun that is not in the same clause as an operator that is to bind it, non-doubled QPs and doubled QPs behave exactly alike: it is not possible to coindex a non-doubled QP with a pronoun contained in a clause which is superordinate to the QP. Cf. (35) and (37): (37) * [Q dhaskalos tUj] ipe oti malosan to kathe pedhij the teacher his said that scolded-3pl the every child * " H i S j teacher said that they scolded every childj" The following table summarizes the facts regarding pronoun binding and Cl-ded and non-doubled QPs when clausemates and non- clausemates are involved: 239 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (38) When Is Coindexing Possible? pronoun in: LOCAL DOMAIN | NON-LOCAL DOMAIN CL-DED NP YES YES CL-DED QP YES NO non-doubled QP NO NO In sum, we see that Cl-ded and non-doubled QPs behave alike with respect to pronoun binding when we consider their coindexation with a non-c-commanding pronoun contained in a superordinate clause: neither allow coindexing. Names, in contrast, can always be coindexed with pronouns that do not c-command them. We can account for the behavior of Cl-ded QPs with respect to pronouns in non-local contexts precisely if we assume that Cl-ded QPs do indeed have an operator/variable structure. It is not possible to coindex a non-referential expression with a non-c-commanding pronoun contained in a clause which is superordinate to that which contains the non-referential expression. This is because in order for a pronoun to be interpreted as bound, it must be c-commanded by the relevant QP or wh-operator. This c-command requirement on bound pronouns is independent of questions of whether or not a particular coindexing between a pronoun and QP would result in WCQ. 240 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Since the rule of QR is clause bound and given that the rule of QD is clause bound, a non-doubled or doubled QP will never c-command a pronoun which is contained in a superordinate clause. So, given that there is an operator/variable structure when we have Cl-ded QPs, we can straightforwardly account for the failure of long distance coindexing between a Cl-ded QP and a pronoun as being due to the fact that the doubled QP is an operator and therefore can only grammatically bind a pronoun if it c-commands that pronoun. We seem to have reached a contradiction then in our generalizations about the behavior of clitic-doubled QPs: Within a local domain, Cl-ded QPs behave like names. With respect to processes that relate the Cl-ded QP to anything outside of the local domain, Cl-ded QPs behave as if they have an operator/variable structure. We will reach the conclusion that the Cl-ded QP does in fact have an operator/variable structure, but this fact is obscured when clause mates of the doubled QP are involved due to the predication structure of Cl-d and the fact that Cl-ded QPs are interpreted in the topic position which is very high within the clause. 6.2.1.3. Additional Pronoun Binding Facts Additional evidence supports the conclusion that Cl-ded QPs have an operator/variable structure. Consider that it is a property of non- 241 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. existential QPs, those that Haik (1984) calls "inherent quantifiers," that they may not be construed coreferentially. Haik defines "inherent quantifier" as follows (ibid. p. 189): "An inherent quantifier is an NP that is not satisfiable by one or more objects of the domain of discourse." An example of an inherent quantifier in English would be something like the QP ‘everyone’. An inherent quantifier like kathe pedhi (‘every child’) cannot be coindexed with a pronoun which is not in its scope. For example, it is not possible to corefer across discourse: (39) *kathe pedhij bike mes’to domatio. proj kathise. every child entered into’the room. sat.down.3s *"Every childj entered the room. Hej sat down." Since kathe pedhi is a QP, it must have scope over a pronoun in order for the pronoun to be interpreted as a bound variable, a possibility which is illustrated in the following example: (40) kathe pedhij ipe oti proj idhe to video, every child said.3s that saw.3s the video "Every child said that he saw the video." Cl-ded QPs behave in just the same way. For instance, a Cl-ded QP cannot be coindexed with a pronoun across discourse: 242 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (41)* i mitera-su tin-prosvale tin kathe kopellaj. proj sadistike. the mother-cl.gen.2s cl.acc.f.-insulted.3p the every girl. got.mad.3s *"Your mother insulted every girlj. Shej got mad." In contrast, a name that is Cl-ded, like a non-doubled name, can easily corefer across discourse: (42) o Apostolis tin-prosvale tin Nikij. prOj sadistike. the Apotolis cl.acc.f.insulted Niki. got.mad.3s "Apostolis insulted Niki.j Shej got mad." Moreover, clitic-doubled QP’s also cannot be associated with a co- varying pronoun in donkey anaphora constructions; this is because the doubled QP is an inherent QP and the inherent quantifier does not c- command the pronoun in question. Doubled and non-doubled universal QP’s behave identically: the sentences are ungrammatical. Cf. (43) and (44): (43) * ean to-vro to kathe avghoj, tha tOj-tighaniso am esos if cl-find-ls the every egg, will cl-fry-ls immediately *"If I find every egg, I will fry it immediately." 243 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (44)* ean vro (to) kathe avghoj, tha tOj-tighaniso amesos if find-ls (the) every egg, will cl-fry-ls immediately *"If I find every egg, I will fry it immediately." In contrast, names (Cl-ded or not) and weak quantifiers can easily corefer to or covary with a pronoun in donkey anaphora contexts (cf. (45a,b)). (45)a. ean vro ena avghOj, tha tOj-tighaniso amesos if find-ls an egg, will cl-fry-ls immediately "If I find an egg, I will fry it immediately." b.ean (ton-)vro ton Yiorghoj, tha tooj-maloso amesos if (cl.acc.m.)find-ls the G., will cl.acc.m-scold-ls immediately "If I run into George, I will scold him immediately." The behavior of Cl-ded QPs is then quite unlike the behavior of names and existential quantifiers in donkey anaphora contexts; instead Cl-ded inherent quantifiers behave like non-doubled inherent quantifiers. The ungrammaticality of clitic-doubled QP’s in donkey anaphora contexts suggests that the doubled QP can’t possibly be referential since the doubled element and the pronoun cannot corefer. 244 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Instead, the data support the claim that Cl-ded QPs have the kind of operator/variable structure a comparable non-doubled QP has. 6.2.1.4. An Account of The Absence of WCQ Effects in Cl-d Constructions When we reconsider the ability of clitic-doubled QPs to escape WCQ violations, we can establish that, at the same time the Cl-ded QP can escape WCQ, it still behaves like an inherent quantifier with respect to (the impossibility of) being coindexed with a pronoun it does not c- command. This is illustrated in the following sentence: (46) i mitera-tUj to-aghapa to kathe pedhij ke pro.j the mother-cl.gen cl.n-love.3s the every child and to-kseri poli kala cl.n-know.3s very good. * "hiSj mother loves every childj and hCj knows it very well." The impossibility of coindexing the doubled QP and the pronoun which it does not c-command and the doubled QP at the same time that it is possible to coindex the doubled QP with a non-c-commanding pronoun to its left strongly suggests that in Cl-d constructions, an operator/variable structure does exist even when WCQ violations are avoided. 245 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. We can account for the WCO facts if what is decisive for WCO is not the presence of a variable to the right of the non-c-commanding pronoun that is coindexed with that variable, but rather the theta-status of the operator/position that c-commands the pronoun. To see this, consider the following: It is well known that a pronoun can be grammatically bound by an operator only if the operator c-commands the pronoun. This captures the fact that an operator in subject position can grammatically bind a pronoun contained in the object Noun Phrase; in this case, the operator and the variable associated with it c-command the pronoun in question. Moreover, the c-command requirement accounts for the ungrammaticality of an inherent quantifier binding a pronoun in a separate sentence (c-command fails). It also accounts for the ungrammaticality of "backward pronominalization" (such as we saw earlier) where a QP that is contained in an embedded clause cannot be coindexed with a pronoun in a superordinate clause. The rule of QR raises that QP to take scope over the embedded clause, but because QR is clause bound, the QP in the embedded clause can never take scope over the superordinate clause. Since the QP can never c-command the pronoun in the superordinate clause, binding will fail. Generally, WCQ falls in between: QR can raise the operator to the clause initial position where it does, in fact, c-command the pronoun in question. But the WCQ configuration differs from the canonical case of pronoun binding in that 246 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the operator, but not the variable, c-commands the pronoun. But, I’d like to suggest, what is important about this fact is not the variable status per se of the variable, but the fact that the variable occurs in a theta-marked position. From this perspective, Cl-d constructions involving doubled QPs pattern with the canonical case of pronoun binding where the operator and its theta-bearing variable c-command the pronoun in question. The following example illustrates the relation between theta marked positions and the bound pronoun in the canonical cases of pronoun binding: ( ‘ ^ '7 ) [ip [|p ^j[ + thcta marked] ^ [np ^ ] ] ] To see how this is similar to the configuration that exists in Cl-d constructions, recall that in predication constructions the subject of the predication, in this case the topic, is assigned a theta role through the predication relation. So although a variable is to the right of the pronoun when Cl-ded QPs are in v o lv e d ,a theta marked position associated with the doubled QP c-commands the pronoun. The following example illustrates this: ‘‘ 'We know that there must be a variable in the doubled argument position because of the asymmetries between doubled NP and doubled QPs that contain expressions requiring antecedents. 247 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ( 4 8 ) [ipp l + thcla marked] [ip [np ^ ] C l - V Xj [.ihda marked] 1] We see then that we can capture the behavior of Cl-ded QPs in WCO contexts if pronoun binding requires c-command, not necessarily by a variable, but by a theta position associated with the operator that binds the pronoun. Under this view of WCO, the fact that Cl-ded QPs can escape WCO, but otherwise behave like they have an operator/variable structure is not contradictory. 6.2.1.5. Scope Interpretation of Clitic-Doubled QPs 6.2.1.5.1. Wh-Operator and Cl-OP Interaction We have just seen that, with respect to WCO when Cl-ded QPs are involved, there is a discrepancy in the behavior of the Cl-ded QP with respect to a pronoun in the same clause and its behavior with respect to pronoun binding outside the clause containing the doubled expression. And I have observed that if one considers the behavior of the Cl-ded QP just in a local domain it seems to behave as if it lacks operator/variable structure; but, if one considers the behavior of a Cl-ded QP with respect to non-clause mates, it becomes evident that it has the same properties that a non-doubled QP possesses. 1 have suggested that the discrepancy can be explained through the predication structure involved in Cl-d constructions where the Cl-ded QP occurs in the structurally high topic 2 4 8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. position. We will reach this same conclusion when we consider the scope interaction of Cl-ded QPs with wh-expressions in a local domain (namely, when both XPs occur in the clause containing the Cl-ded QP) and when the Cl-ded QP and the wh-expression are in separate clauses. Cl-ded QPs have unambiguous wide scope with respect to a wh- element contained within the same clause (see (49a)). The relative scope between the Cl-ded QP and the wh-element is represented schematically in (49b). Cf.: (49)a.pios to-malone to kathe pedhi who cl.n-scolded.3s the every child "Who scolded every child." b. [every childj [who^ [ x^ scolded X j ]]] The behavior of doubled QPs contrasts sharply with that of non- doubled QP objects, which take narrow scope in the same environment (see (50a)). The relative scope between the non-doubled QP object and the wh-expression is represented schematically in (50b). Cf.: (50)a.pios malone to kathe pedhi who scolded.3s the every child "Who scolded every child?" b. [whOk [every childj [ x^ scolded X j ]]] 249 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I demonstrated earlier that it is incorrect to conclude that Cl-ded QPs have the syntax of names; nonetheless, if this were true, it would provide a natural explanation for the wide scope interpretation of Cl-ded QPs in the context we consider here. To see this, consider that Cl-ded QPs, if they lacked operator/variable structure, would not be subject to a rule of quantifier interpretation, which, in interaction with other principles of the grammar, determines the scope of a QP. Without operator/variable structure Cl-ded QPs would behave like names; and it is a property of names that they are understood as having widest scope. But if Cl-ded QPs are treated like names, then we will see that we again face a contradiction when we consider the relative scope interpretation of Cl-ded QPs with respect to wh-operators that are not contained in the same clause. The following data, pointed out in Schneider-Zioga (ibid.), illustrate that the wide scope interpretation of clitic-doubled QPs is clause bound: The example in (51a) illustrates that a Cl-ded QP has narrow scope with respect to a wh-operator that originates in a superordinate clause. (51b) indicates the relative scope interpretations schematically. Cf.: 250 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (51)a.Pios ipe oti to-aghorases to kathe pegnidhi who said that cl-bought.2s the every toy "Who said that you bought every toy." b.[whO|, [ X ,, said that [every toyj [ you cl.bought X j ]]]] In this same context, non-doubled QP objects behave exactly like doubled QPs: they have narrow scope with respect to a wh-operator contained in a superordinate clause: (52)a. Pics ipe oti aghorases to kathe pegnidhi who said that bought.2s the every toy "Who said that you bought every toy." b.[whOk [ X k said that [every toyj [ you bought X j ]]]] In sum, Cl-ded and non-doubled QPs behave the same with respect to relative scope interpretation when we consider their interaction with a wh-operator contained in a superordinate clause: doubled and non- doubled QPs alike take narrow scope. It is only when we consider their relative scope interpretation with respect to a wh-operator in the same clause as the QPs in question, that their behaviors diverge: Cl-ded QPs must take wide scope and non-doubled QPs must take narrow scope. 251 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The following table summarizes the facts regarding relative scope interpretation of Cl-ded and non-doubled QPs in local and non-local domains: (53) Where Is Wide Scope Of The QP With Respect To The Wh- Qperator Possible?" Wh-word in: LOCAL DOMAIN I NON-LOCAL DOMAIN CL-DED OP wide scope narrow scope non-doubled OP narrow scope narrow scope 6.2.1.5.1.1. A n A ccount Of The R elative Scope Interpretations On the basis of the observed asymmetry that exists for the distribution of doubled QPs and doubled NPs, I argued that, when Cl- ded QPs occur, a rule of QD applies such that the predicate variable is deleted and a syntactic variable is left in its place. The variable is bound by the doubled QP topic that occupies the specifier of TpP. In this way I was able to account for the fact that Cl-ded QPs cannot contain expressions that require an antecedent that is internal to the clause containing the Cl-ded QP. Since the TpP is structurally higher than 111 For non-doubled QPs, the chart is menât to refer to object QPs only. 252 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. either IP, CP or FP, a Cl-ded QP will always be outside of the scope of these phrases. The obligatory widest scope interpretation of clitic- doubled QPs, which happens only with clausemates, can be captured in exactly the same way. For the determination of scope, I assume the following principle, proposed by Aoun and Li (1989): (54) The Scope Principle: An operator A has scope over an operator B iff A c-commands B or an A’-element coindexed with B. Since the Cl-ded operator (operator A) c-commands the wh-expression (operator B), the Cl-ded operator has scope over the wh-expression. Since a non-doubled QP is subject to QR instead of QD, its scope will always be narrower in similar contexts since QR cannot raise a QP to a position where it will c-command the wh-operator. Regarding the narrow scope interpretation which comes into play when considering scope of Cl-ded QPs with respect to operators which are not clausemates, consider that it is an empirical fact the rule of QR (and thus by extension QD) is clause bound. Therefore, QR/QD may not affect a position which is outside of the clause containing the QP in question. If QR/QD restricts the raising/deletion of a QP so that it cannot be raised out of the clause containing it, QPs will always have 253 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. narrow scope with respect to an operator in a superordinate clause since scope is determined with reference to c-command. 6.2.1.5.2. Negation and Cl-OP Interaction In previous sections we examined the contrasting behavior of doubled QPs with respect to bound pronouns and to wh-operators in local and non-local domains. We saw for both pronoun binding and relative scope interpretation that Cl-ded QPs have contrasting behaviors in the two different kinds of domains. Moreover, with respect to clausemates the Cl-ded QP appears to behave as if it does not have an operator/variable structure but with respect to non-clause mates it behaves as if it does. In this section, we turn to an examination of relative scope interpretation when negation, an X°, interacts with the Cl- ded QP. Here we find what appears to be a perplexing pattern of behavior in light of the previous generalizations. A Cl-ded QP has unambiguous narrow scope with respect to negation contained within the same clause (see (55a)). The relative scope between the Cl-ded QP and negation is represented schematically in (55b). Cf.; 254 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (55)a.i Maria <i/ie«-to-aghorase to kathe pegnidhi the Maria not cl.n-buy.3s the every toy "Maria didn’t buy every toy." b.[not [Maria [every toyj [bought X j ]]] Surprisingly, in this context, the behavior of Cl-ded QPs does not contrast at all with that of non-doubled QPs; they also take narrow scope with respect to negation (see (56a)). The relative scope between the non-doubled QP and negation is represented schematically in (56b): (56)a.i Maria Jhe«-aghorase to kathe pegnidhi the Maria not buy.3s the every toy "Maria didn’t buy every toy." b.[not [Maria [every toyj [bought X j ]]] In this paradigm we see that Cl-ded and non-doubled QPs behave alike, a fact which supports the claim that Cl-ded QPs have an operator/variable structure. But, nonetheless, an unexpected fact since otherwise Cl-ded QPs have unambiguous wide scope with respect to expressions contained within the same clause, because, through QD, they are interpreted in the topic position. The scopal behavior of non- doubled QPs with respect to negation is not surprising because the rule of QR places them within the scope of negation. 255 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In addition, when we consider the relative scope interpretation of Cl-ded QPs with respect to negation in a superordinate clause, we see that the Cl-ded QP still has narrow scope (see (57)). Non-doubled QPs have the same relative scope interpretation with respect to negation in a superordinate clause as doubled QPs do: they must take narrow scope (see (58)). Cf.: (57) i Maria dhen-ipe oti to-aghorases to kathe pegnidhi the Maria not say.3s that cl.n-bought.2s the every toy "Maria didn’t say that you bought every toy." (58) i Maria dhen-ipe oti aghorases (to) kathe pegnidhi the Maria not say.3s that bought.2s (the) every toy "Maria didn’t say that you bought every toy." The following table summarizes the facts: (59) The Scope of QPs With Respect To Negation negation in: LOCAL DQMAIN | NQN-LQCAL DOMAIN CL-DED QP narrow scope non-doubled QP narrow scope narrow scope narrow scope 256 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6.2.1.5.2.1. An Account of The Scope Relations We have just seen that Cl-ded QPs and non-doubled QPs behave alike with respect to scope interactions with negation. From the analysis of the syntax of Cl-ded QPs developed here, we know that when we have Cl-d, the doubled QP is interpreted in the topic position. The topic position is located syntactically higher than the negation morpheme, which I assume is identical to, or indicates the location of, the negation operator. The following example indicates the location of a topic (CLLDed Noun Phrase) with respect to negation: (60) ton Aleksandro <i/ien-ton-idhame akoma the Alexander not cl.acc.m-see.lp yet "(As for) Alexander, we haven’t seen him yet." Due to the location of the topic in structures like this, it would appear that at LF the Cl-ded QP must (through QD) be outside the scope of negation, a fact which is puzzling given the actual interpretation available in (55). The example in (61) indicates the location, with respect to negation, of the doubled QP after QD; (61) [to kathe pegnidhij[jp i Maria d/ie/z-to-aghorase x- ]] the every toy the Maria not cl.n-buy.3s 257 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. One logically possible solution to this puzzle is to conclude that, despite surface appearances, negation actually occurs higher than the topic and this property of negation accounts for the fact that negation is understood to have wider scope than the topic QP in the cases we are considering: in these cases, negation actually is c-commanding the topic at some level of representation. But at first this hypothesis seems implausible because subject QPs in MG can easily be interpreted outside the scope of negation (see (62)) and subjects are located structurally lower than the topic. The available interpretations are given schematically in (63). (62)a.kathe kopella dhen-dhiavase afto to vivlio every girl not read.3s this the book "Every girl didn’t read this book." (63)a.not[every girl read this book] b.every girl [not read this book] So, although subjects are structurally lower than topics, subjects can be interpreted outside the scope of negation. If, in order to account for the wide scope interpretation of negation when topics are involved, we assumed that negation in MG is actually located higher than the topic, there seems to be a contradiction for an account of the scope of negation with respect to subjects. 258 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. But, upon examination of a wider array of facts, the proposal that negation is actually higher than the topic is not as implausible at it might first have seemed in light of the subject/negation scope facts. Consider the following fact; In constructions involving topics, subject QPs are also required to take narrow scope with respect to negation (see (64) for CLLDed Noun Phrases). The available interpretations are given schematically in (65). Cf.: (64) afto to vivlio, kathe kopella <i/ren-to-dhiavase this the book every girl not-cl.n-read.3s "(As for) this book, every girl didn’t read it." (65)a.not [this book every girl read it] b.*?every girl[this book not read it] For topics deleted in PF the facts are the same (see (66)). The interpretations are given schematically in (67). Cf.: (66) kathe kopella d/ie«-to-dhiavase afto to vivlio every girl not-cl.n-read.3s this the book (67)a.not [this book every girl read it] b. *?every girl [not read this book] The above data indicate that, in constructions involving topics, negation really is located higher than the topic. Recall from the 259 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. discussion of preverbal discourse functions in chapter four that the topic is located in the specifier of TpP which is structurally higher than IP. The following example illustrates the structure involved after QD: (68)to kathe pegnidhij[jpi M.[i.d^en-to-aghorase,,[yptk X j]] the every toy the Maria not-cl.n-bought.3s Since negation is an X" in MG (see Tsimpli (1992)), we would expect that, if it moves, it will move to or adjoin to an X" for reasons of structure preservation. Where the kind of structure in (68) is concerned, given the fact that the topic is understood to be in the scope of negation, a plausible location for negation is in Tp.° Alternately, negation could be located in the head of StP.^“ The following example illustrates the proposed structure with the decision that negation is in the head of TpP: (69) [ipp QPj [tp M g g [ [ip ... [vp - X j . . . ] ] ] ] ] ] The hypothesis that negation in topic constructions is located in Tp“ can be tested by examining constructions involving negation in wh- questions where Cl-ded QPs are also concerned. This kind of example ^ “The exact location of negation would depend on whether a spec/head relation with agreement implied that the specifier is necessarily or only optionally in the scope of negation. The answer to this question awaits further study of the syntax of specifier/head agreement. 260 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. would be illuminating for several reasons. Firstly, the topic is structurally higher than a wh-operator. And, secondly, wh-questions in MG must stand in a specifier/head relation with the verb, which moves to C°. Assuming that the raising of negation is not independent of movement of the verb, we would predict that wh-questions block further raising of negation because they must stand in a specifier/head relation with the (negated) verb. Since the topic is higher than CP, we expect it to be interpreted outside the scope of negation in just this case. Consider in this light the following example: (70) Pios dAzen-to-aghorse to kathe pegnidhi who not-cl.n-bought.3s the every toy "Who didn’t buy every toy?" The sentence is unambiguous and here the Cl-ded QP is indeed interpreted as having widest possible scope with respect to the other operators in the sentence. That is, the following interpretation, which I indicate schematically, is the only possible reading for (70): (71) every toy [ who not buy it] If the hypothesis that negation is located quite high in topic constructions is true, this scopal behavior is expected since the 261 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. formation of a wh-question would block further raising of negation. The following example illustrates the structure involved: (72) to kathe pegnidhi[cppioSy[^.dhen-to-aghoraseJjpXy[^ptk X j ]] the every toy who not-cl.n-bought.3s the every toy In (72) the doubled OP is located outside of the wh-operator and thus interpreted outside of its scope. The doubled QP is also obligatorily interpreted outside of the scope of negation since in wh-questions there must be a specifier/head relation with the (negated) verb and this requirement prevents negation from raising any higher. In sum, the hypothesis that in Cl-ding constructions negation occurs quite high can account for the full range of scope facts I have identified.'^ We see then that the behavior of doubled QPs with respect to negation is surprising, but doesn’t compromise my analysis of the syntax of Cl-ded QPs. '^The behavior of negation in Cl-d and CLLD constructions raises many provocative questions that I have not explored or answered such as why negation must be interpreted so high in Cl-d constructions but not in a regular sentence. The issues surrounding the behavior of negation in this construction are interesting and important, perhaps particularly for the Minimalist Program, but they fall outside the scope of the investigation here and so I leave them for future research. 262 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6.3. Concluding Remarks In this chapter I have studied some aspects of the syntax of Cl-ded QPs. I pointed out the existence of an asymmetry in the distribution of doubled QPs and doubled NPs. The existence of such an asymmetry was shown to provide an additional argument against an analysis of Cl-d that involved movement of the doubled Noun Phrase to the CLLDed position. I have explored some of the empirical ramifications of the property of Cl-d that Cl-ded expressions and in particular, Cl-ded QPs must be interpreted so much higher than non-doubled expressions. This property prevents certain kinds of QPs from occurring in Cl-d constructions; namely, QPs that contain expressions that require clause internal antecedents. Moreover, the fact that Cl-ded QPs are interpreted in the topic position tends to obscure the fact that Cl-ded QPs possess an operator/variable structure just as non-doubled QPs do. This fact becomes clear only when we observe the binding and relative scope interpretation of Cl-ded QPs with respect to pronouns and operators that are not clause mates with the Cl-ded QP. I accounted for the relative scope interpretation of Cl-ded QPs by proposing the possibility of quantifier interpretation via deletion as opposed to raising. Deletion of the QP in direct object position in Cl-d constructions was argued to leave behind a variable which the QP in 263 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. topic position binds as required by the Bijection Principle. Finally, certain peculiar facts regarding the relative scope interpretation of clitic-doubled QPs and negation are explored. It is ascertained that in Cl-d constructions negation is interpreted higher than the expected 1° and this fact about negation in Cl-ding constructions leads to an explanation of the peculiar scope interpretations when negation and clitic doubling is involved. 264 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER SEVEN: CONCLUSION 7.1. Advantages of the Predication Analysis The account of clitic doubling proposed here has empirical advantages over accounts of clitic doubling that posit a process of determiner or pronominal incorporation, or some kind of agreement process as an account of the phenomenon. All of the above analyses can provide some reason for the fact that Cl-ded expressions are specific: (a) an incorporated (definite) D° or pronoun would always be specific in the relevant sense; or (b) if the agreement phrase, or clitic phrase, relevant to clitic doubling is outside of VP, it can arguably be construed as forcing a specific interpretation under the view of Diesing that specific Noun Phrases are interpreted outside of VP. But these analyses do not seem to be able to account for the numerous antilocality effects we see with clitic doubling constructions (as well as simple cliticization) where the clitic-doubled expression is related to a position that is much higher than the clitic itself. Moreover, 1 have provided numerous reasons to believe that it is empirically wrong to analyze clitic doubling via the proposal that there is a movement relation between the clitic-doubled expression and the position where CLLDed expressions occur; maximally such an approach can only capture the fact that both positions require a specific interpretation. 265 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Finally, analyses that posit only a local relation between the clitic and the doubled expression are not led to formulate the kind of questions that reveal fresh generalizations about Cl-d such as the fact that Cl-ded QPs cannot contain expressions that need to find antecedents within the CP which contains the doubled QP. 7.2. Specificity In this dissertation I have attempted to demonstrate that there is a clause initial position in MG that plays a central role in the syntax of Cl-d. I have refered to this position as the topic position, a position that is reserved for specific Noun Phrases or, loosely speaking, discourse topics. That the doubled Noun Phrase is related to such a structurally high position in the clause is not totally unexpected under current views of the syntax of specific Noun Phrases which suggest that it would be natural to treat Cl-ded Noun Phrases as not being located within VP at least at LF, if not in the overt syntax. For instance, Diesing (ibid.), in her treatment of specificity, assumes that the tripartite quantificational structure of Kamp (ibid.) and Heim (ibid.), which consists of an operator, restrictive clause, and a nuclear scope, is syntactically realized. The restrictive clause ( = IP) is the domain of presuppositional interpretation and the nuclear scope (=VP) is reserved for new 266 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. information. According to Diesing then, + specific NPs must be mapped to a position outside of VF in order to be in the quantificational domain of the restrictive clause so that they can be interpreted as specific. For Cl-d constructions, this view suggests that a coindexed Noun Phrase or the doubled Noun Phrase itself must, at least by LF, occur in a position that is much higher than the location of the doubled Noun Phrase in the overt syntax. Diesing’s picture of the behavior of specific Noun Phrases accounts nicely for the phenomenon of direct object scrambling in a language like German. In German, specific direct objects cannot occur internal to the VP but must be scrambled out in the overt syntax. Something of this sort is obviously relevant for the syntax of Cl-d in MG, but Cl-d should not be directly construed as a kind of "LF scrambling" phenomenon. This is because clitic-doubled Noun Phrases occur in a much higher position than would be predicted from a straight application of "LF scrambling." In German, specific direct objects are scrambled so that they occur immediately outside of VP. They are structurally lower than the subject or other preverbal specifiers such as specifier of CP, etc. In contrast, the topic in MG is much higher than the 267 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. subject.* It seems that the syntax of specific expressions is language- specific in that the specific expression is interpreted and/or moves outside of VP (at least for many languages), but how far outside of VP the specific expression moves differs from language to language. *Perhaps the difference between the two languages can be traced back to the presence versus absence of the clitic in constructions involving specificity. Following this intuition, the pronominal nature of the clitic morphology would force the topic to appear outside of the restrictive clause as a kind of antilocality effect such that the topic cannot be "too close" to the nuclear scope. (This kind of behavior was manifested for MG as an antilocality requirement with respect to intervening Case assigned specifiers.) 268 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. BIBLIOGRAPHY Agouraki, Georgia. 1990. On the projection of maximal categories. In U.C.L. Working Papers in Linguistics 2, 182-200. University College, London. Agouraki, Georgia. 1992. Spec-head licensing: the cases of clitic constructions and focalization. Abstract from Workshop on Balkan and Slavic Syntax. Department of Linguistics, University of Ottowa. Aoun, Joseph. 1981. The formal nature of anaphoric relations. Doctoral dissertation, MIT, Cambridge, Mass. Aoun, Joseph. 1985. A grammar of anaphora. Cambridge, Mass.: MIT Press. Aoun, Joseph. 1991. The syntax of doubled arguments. Ms., USC, Los Angeles. Aoun, Joseph. 1993. The syntax of doubled International Journal of Basque Linguistics and Philology, 3:709- 730. Aoun, Joseph and Li, Yen-Hui Audrey. 1989. Scope and constituency. Linguistic Inquiry 20:141-172. Aoun, Joseph and Yen-Hui Audrey Li. 1990. Wh-in-situ: s-structure or LF. Ms., USC, Los Angeles. Aoun, Joseph and Yen-Hui Audrey Li. 1993. Wh-elements insitu: syntax or LF? Linguistic Inquiry 24:199-238. Baker, Mark. 1988. Incorporation: A theory of grammatical function changing. Chicago: University of Chicago Press. Benmamoun, Abbas. 1992. Functional and inflectional morphology. Problems of projection, representation, and derivation. Doctoral dissertation, USC, Los Angeles. 269 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Bittner, M. 1986. On the semantics of the Greenlandic antipassive and related constructions. I nternational Journal of American Linguistics. Borer, Hagit. 1981. Parametric variation in clitic constructions. Doctoral Dissertation, MIT, Cambridge, Mass. Burzio, Luigi. 1981. Intransitive verbs and Italian auxiliaries. Doctoral dissertation, MIT, Cambridge, Mass. Chomsky, Noam. 1981. Lectures on Government and Binding. DordrechtrForis. Chomsky, Noam. 1989. Some notes on economy of derivation and representation. In M IT Working Papers in Linguisitics 10, ed. I. Laka and Anoop Mahajan, 43-74. Department of Linguistics and Philosophy, MIT, Cambridge, Mass. Chomsky, Noam. 1993. A minimalist program for linguistic theory. In The view from building 20. ed. Hale, Kenneth and Samuel Keyser, 1-52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1994. Bare Phrase Structure. MIT Occasional Papers in linguistics 5. Cambridge, Mass.: MIT working papers in linguistics. Chomsky, Noam and Howard Lasnik. 1991. Principles and Parameters Theory. To appear in Syntax: an international handbook of contemporary research, ed. Jacobs, J., A. von Stechow, and W. Sternefeld. Berlin: de Gruyter. Cinque, G. 1990. Types of A-bar dependencies. Linguistic Inquiry Monograph 17, Cambridge, Mass.: MIT Press. Cole, P. and G. Hermon. 1994. Is there LF wh-movement? Linguistic Inquiry 25:239-262. Demirdache, Hamida. 1991. Resumptive chains in restrictive relatives, appositives, and dislocation structures. MIT working papers in linguistics, dissertation series, MIT, Cambridge, Mass. 270 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Diesing, Molly. 1990. The syntactic roots of semantic partition. Doctoral dissertation. University of Massachusetts, Amherst. Dobrovie-Sorin, Carmen. 1990. Clitic Doubling, wh-movement and quantification in Romanian. Linguistic Inquiry 21:351- 398. Enc, Murvet. 1987. Specificity. Ms., USC, Los Angeles. Enc, Murvet. 1991. The semantics of specificity. Linguistic Inquiry 22:1-25. Everett, Daniel. 1987. Piraha clitic doubling. Natural Language and Linguistic T/zeory 5:245-276. Everett, Daniel. 1989. Clitic doubling, reflexives, and word order alternations in Yagua. Language 65:339-372. Franco, Jon. 1993. On object agreement in Spanish, Doctoral dissertation, USC, Los Angeles. Givon, Talmy. 1976. Topic, pronoun, and grammatical agreement. In Subject and topic, ed. C. Li, 149-188. New York: Academic Press. Haberland, Hartmut and Johan Van Der Auwera. 1990. Topics and clitics in Greek relatives. I nternational Journal of Linguistics, Acta Linguistica H a f niensia 22:127-157. Haik, Isabella. 1984. Indirect binding. Linguistic Inquiry 15:185-223. Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. Doctoral dissertation. University of Massachusetts, Amherst. Heim, Irene, Howard Lasnik, and Robert May. 1991. Reciprocity and plurality. Linguistic Inquiry 22:63-102. Hornstein, Norbert. 1994. The grammar of LF: from GB to minimalism. Ms., University of Maryland, College Park. 271 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Horrocks, Geoffrey. 1983. The order of constituents in Modern Greek. In Order, concord and constituency, ed. G. Gazdar, E. Klein, G. K. Pullum, 95-111. Dordrecht: Foris. Horrocks, Geoffrey and Melita Stavrou. 1987. Bounding theory and Greek syntax: evidence for wh-movement in NP. Journal of Linguistics 23:79-108. Hurtado, A. 1984. Clitic chains. Ms., Simon Fraser. latridou, Sabine. 1988. Clitics, anaphors, and a problem of coindexation. Linguistic Inquiry 19:698-703. latridou, Sabine. 1991. Clitics and island effects. Ms., MIT, Cambridge, Mass. latridou, Sabine and Anthony Kroch. 1993. The licensing of CP-recursion and its relevance to Germanic verb-second phenomena. Ms., University of Pennsylvania, Philadelphia. Ingria, Robert. 1981. Sentential complementation in Modern Greek. Doctoral dissertation, MIT, Cambridge, Mass. Jackendoff, Ray. 1972. Semantic representation in generative grammar. MIT Press, Cambridge, Mass. Jaeggli, Osvaldo. 1980. On some phonologically-null elements in syntax. Doctoral dissertation, MIT Cambridge, Mass. Jaeggli, Osvaldo. 1981. Topics in Romance Syntax. Dordrecht: Foris. Jaeggli, Osvaldo. 1986. Three issues in the theory of clitics: Case, doubled NPs, and extraction. In Syntax and semantics 19, The syntax of pronominal clitics. New York: Academic Press. Joseph, Brian D. 1983. Relativization in Modern Greek: Another look at the Accessibility Hierarchy. Lingua 60:1-24. Joseph, Brian. 1992. Deictic verbs and the pro-drop parameter in Greek. Abstract from Workshop on Balkan and Slavic Syntax. Department of Linguistics, University of Ottowa. 272 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Joseph, Brian and Irene Philippaki-Warburton. 1987. Modern Greek. London: Croom Helm. Kamp, J. 1981. A theory of truth and semantic representation. In Formal methods in the study of language, ed. J. Groenendijk, T. Janssen and M. Stokhof, 277-321. Mathematical Centre, Amsterdam. Katada, Fusa. 1990. The representation of anaphoric relations in logical form. Doctoral dissertation, USC, Los Angeles. Kayne, Richard. 1984. Connectedness and binary branching. Dordrecht: Foris. Kayne, Richard. 1975. French Syntax. MIT Press, Cambridge, Mass. Kazazis, Kostas and Joseph Pentheroudakis. 1976. Reduplication of indefinite direct objects in Albanian and Modern Greek. Language 52:398-403. King, Larry. 1984. The semantics of direct object a in Spanish. Hispania 67:397-403. Kiparsky, Paul and Carol Kiparsky. 1971. Fact. In Semantics, eds. D. Steinberg and L. Jacobvits, 345-369. Cambridge: Cambridge University Press. Koopman, Hilda. 1990. Restrictions on spec position. Ms., UCLA, Los Angeles. Koopman, Hilda and Dominique Sportiche. 1991. The position of subjects. Lingua 85:211 -259. Levin, L. and Simpson, Jo. 1981. Quirky case and the structure of Icelandic lexical entries. In Papers f rom the seventeenth regional meeting, Chicago Linguistics Society, 185- 196. Chicago Linguistics Society, University of Chicago, Chicago, 1 1 1 . Lyons, Christopher. 1990. An agreement approach to clitic doubling. Transactions of the Philological Society 88:1-57. 273 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Mackridge, Peter. 1985. T he M odem Greek language. Oxford: Oxford University Press. Mahajan, Anoop. 1990. The A/A’ distinction and movement theory. Doctoral dissertation, MIT, Cambridge, Mass. May, Robert. 1977. The grammar of quantification. Doctoral dissertation, MIT, Cambridge, Mass. Moltman, Frederika. 1990. Scrambling in German and the definiteness/specificity effect, Ms., MIT, Cambridge, Mass. Muysken, Pieter. 1989. Predication chains: Case and argument. Linguistic Inquiry 20:627-645. Pesetsky, David. 1982, Paths and Categories. Doctoral dissertation, MIT, Cambridge, Mass. Pesetsky, D. 1984. Wh-in-situ: movement or unselective binding. Ms., University of Massachusetts, Amherst. Pinar, Pillar. 1992. Clitic doubling and "scrambling" in Spanish. Handout, presented at: Linguistic symposium on Romance Languages. Philippaki-Warburton, Irene. 1985. Word order in Modern Greek. Transactions of the Philological 2:113-143. Progovac, Ljiljana. 1988. A binding approach to polarity sensitivity. Doctoral dissertation, USC, Los Angeles. Rizzi, Luigi. 1990. Relativized minimality. MIT Press, Cambridge, Mass. Rochemont, M. 1986. Focus in generative grammar. Amsterdam: Johns Benjamin Publishing Co. Rothstein, S. 1983. The syntactic forms of predication. Doctoral dissertation, MIT, Cambridge, Mass. Schein, Barry. 1982. Small clauses and predication, Ms., MIT, Cambridge, Mass. 274 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Schneider-Zioga, Patricia. 1990. Clitic-doubling, minimality and scope effects. Ms., USC, Los Angeles. Schneider-Zioga, Patricia. 1992a. A case for AgrP(hrase), Ms., USC, Los Angeles. Schneider-Zioga, Patricia. 1992b. Exceptional case marking constructions in Modern Greek, Ms., USC, Los Angeles. Schneider-Zioga, Patricia. 1993. The syntax of clitic-doubling in Modern Greek. Ms., USC, Los Angeles. Simpson, J. 1981. Finnish essive and translative, Ms., MIT, Cambridge, Mass. Sportiche, Dominique. 1992. Clitic Constructions. Ms., UCLA, Los Angeles. Stavrou, Melita. 1985. I klitiki antonimia stis perioristikes anaforikes protasis me exartisi amesou antikimenou pou isagontai me to pou. In Studies in Greek linguistics, 121-136. Proceedings of the 5th annual meeting of the Department of Linguistics, Aristotelian University of Thessaloniki, 2-4 May 1984. Steriade, Donca. 1980. Clitic doubling in the Romanian wh constructions and the analysis of topicalization. In Papers fr o m the sixteenth regional meeting, Chicago Linguistics Society, 282-297. Chicago Linguistics Society, University of Chicago, Chicago, 1 1 1 . Suner, Margarita. 1988. The role of agreement in clitic doubling constructions. Natural Language and Linguistic Theory 6:391-434. Terzi, Arhonto. 1991. Genitive clitics of prepositions. Ms., Cuny Graduate Center. New York. Terzi, Arhonto. 1992. PRO in finite clauses: a study of the inflectional heads of the Balkan languages. Doctoral dissertation, CUNY, New York. 275 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Tsimpli, lanthi. 1990. The clause structure and word order of Modern Greek. In Ü. C. L. W or king Papers in Linguistics 2, 227- 245. University College, London. Tsimpli, lanthi-Maria and Anna Roussou. 1992. On Negation in Modern Greek. Abstract from Workshop on Balkan and Slavic Syntax. Department of Linguistics, University of Ottowa. Tzartzanos Akhilefs A. 1946. Neoelliniki sintaksis (tis kinis dimotikis). 2nd edition (Volume 1). Athens: OESV. Valiouli, Maria. 1991. Anaphora, agreement and the pragmatics of "right dislocations" in Greek. Doctoral dissertation, Thessaloniki, Greece. Varlokosta, Spyridoula. 1992. Control in Modern Greek. Ms., University of Maryland, College Park. Varlokosta, Spyridoula. 1994. Issues on Modern Greek sentential complementation. Doctoral dissertation. University of Maryland, College Park. Warburton, Irene. 1975. The passive in English and Greek. Foundation of Language 13:563-578. Watanabe, Akira. 1991. Wh-in-situ, subjacency, and chain formation. Ms., MIT, Cambridge, Mass. Webelhuth, Gert. 1989. Syntactic saturation phenomena and the modern Germanic languages. Doctoral dissertation. University of Massachusetts, Amherst. Williams, Edwin. 1980. Predication. Linguistic Inquiry 11:203-238. Zubizarreta, Maria-Luisa. 1993. The grammatical representation of topic and focus: implications for the structure of the clause. In Cuadernos de Linguistica 2, 181-208. Instituto, Universitario Ortega y Gasset, Madrid. Zwicky, Arnold. 1985. Clitics and particles. Language 61:283-305. 276 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The syntax of focus and topic in Mandarin Chinese

PDF

Avant-garde criticism: The criticism of exhaustion

PDF

Definiteness types in Spanish: A study of natural discourse

PDF

In the event of focus

PDF

On the nature of particles in Japanese and its theoretical implications

PDF

Identity in crisis: Modernism and the texts of adolescence

PDF

The accidence and syntax in john Wallis' 1653 "grammatica linguae anglicanae": a translation and a commentary on its alleged relationship to the 1660 port-royal "grammaire generale et raisonnee"

PDF

Representation of focus and presupposition in Japanese

PDF

The technoscape in the modern novel: Aleksandr Solzhenitsyn's "The First Circle" and Ralph Ellison's "Invisible Man"

PDF

Syntactic structures in nominals: A comparative study of Spanish and Southern Quechua

PDF

The syntax of the Chinese BA-constructions and verb compounds: A morphosyntactic analysis

PDF

Preview of the vanishing hero: a study of the protagonists in Jacobean drama

PDF

Morpheme recognition in prosodic morphology

PDF

Genre and ideology in the popular German cinema, 1950-1972

PDF

On the representation and licensing of Q and Q-dependents

PDF

A critical edition of the Escorial manuscript of 'historia de los indios de la neuva espana' of fray toribio de benavente (motolinia) (spanish text)

PDF

Evaluating heaviness: Relative weight in the spoken production of heavy-NP shift

PDF

The phonemics and morphology of Hokkaido Ainu

PDF

Ovid's musomachy

PDF

The erudite romanticism of Juan Eugenio Hartzenbusch in 'Los Amantes de Teruel'

Asset Metadata

Creator Schneider-Zioga, Patricia (author)

Core Title The syntax of clitic doubling in modern Greek

Degree Doctor of Philosophy

Degree Program Linguistics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag language, linguistics,Language, Modern,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-468832

Unique identifier UC11352026

Identifier 9601057.pdf (filename),usctheses-c17-468832 (legacy record id)

Legacy Identifier 9601057-0.pdf

Dmrecord 468832

Document Type Dissertation

Rights Schneider-Zioga, Patricia

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA