University of Southern California Dissertations and Theses

Issues in the syntax of resumption: Restrictive relatives in Lebanese Arabic

(USC Thesis Other)

Issues in the syntax of resumption: Restrictive relatives in Lebanese Arabic

PDF

Download

Open document

Flip pages

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 from any type of computer 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 margins, and improper alignment can adversely 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 will 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 overlaps. ProQuest Information and Learning 300 North Zeeb Road, Ann Arbor, Ml 48106-1346 USA 800-521-0600 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ISSUES IN THE SYNTAX OF RESUMPTION: RESTRICTIVE RELATIVES IN LEBANESE ARABIC by Lina Choueiri 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) May 2002 Copyright 2002 Lina Choueiri Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 3073760 Copyright 2002 by Choueiri, Lina Georges All rights reserved. ___ ® UMI UMI Microform 3073760 Copyright 2003 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA The Graduate School University Park LOS ANGELES, CALIFORNIA 90089-1695 This dissertation, w ritten by !Jjia_jQhQ.u.£±rJL Under the direction o f h-.ex. Dissertation Committee, and approved by all its members, has been presented to and accepted b y The Graduate School, in partial fulfillment o f requirements for the degree o f DOCTOR OF PHILOSOPHY 7 < m “ " Be a n o f Graduate Studies Date May 10, 2002 DISSER TA TION COMMITTEE Chairperson Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Acknowledgments Years from now, nothing will remain of what has led to this dissertation but the occasional flashback. This is when I will go back to these pages: in order to remember that experience. While I attempt to remedy the future failings of my memory, I would like to acknowledge first all those I will forget to thank in spite of the best of intentions. It gives me great pleasure to be able to express my gratitude to the members of my committee. Their guidance, encouragement, and unfailing support made this work possible. I owe a great deal to Joseph Aoun who taught me in my first syntax class ever. Since then, he has been a constant source of inspiration, a true mentor. Our work together is what prepared me for what was to come after USC. I thank him for sharing with me his love for linguistics. It has been a privilege to learn from Jean-Roger Vergnaud, whose ability to see the big picture is unparalleled. I want to thank him for sharing his vision with his students. His influence on my work may not be visible in the particulars of this dissertation but it has been profound. Although Audrey Li was never my teacher, we had many discussions on the topic of relatives, which were decisive for the arguments presented in this work. I would like to thank her for her warmth and for being interested in meeting with me, even at the worst of times. Barry Schein brings in the (indispensable) element of doubt in my committee. I admire his ability to deconstruct any linguistic argument and I Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. thank him for putting up with my ignorance in semantics. Hagit Borer was not part of my committee, but she might just as well have been. She is a great advocate for students and I deeply appreciate her support. Her impact on this work is unquestionable. USC provided the best environment possible for exchanging ideas, and in that context, I would like to extend my appreciation to Hajime Hoji and Maria Luisa Zubizarreta for many enlightening discussions. The staff, which was there through the years I spent at USC, deserves a special mention: Linda Culver and Kathy Stubaus, thank you. As Student Affairs Coordinator, Laura Reiter was the mediator between students and administration. She has become a supportive friend. I am very grateful for the people I met while I was in LA, colleagues and friends who have made my experience as a student in linguistics one of the most enjoyable and memorable ones. Discussing Lebanese Arabic data judgments with Peter Hallman was always a good learning experience. Most of what I know about Lebanese Arabic Morphosyntax now is the result of those discussions. It made giving data judgments almost bearable. I have also greatly benefited from discussions with Junichi 'Ringe' Hayashishita, Ayumi Ueyama, Maire Nunan, Jairo Nunes, Norberto Moreno, Patricia Schneider-Zioga, Mohammad Mohammad, Ur Shlonsky, Ivy Sichel, Cedric Boeckx, Yael Sharvit, Dominique Sportiche, Luca Storto, Ivano Caponigro, Norbert Hornstein, and Philippe Schlenker. Special thanks iii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. to Jamal Ouhalla for his patience and for the many discussions on relative constructions. Then there are the people I have had the greatest pleasure in knowing. I am glad my long stay at USC allowed me to meet Rachel Walker whose friendship and sound advice I value a great deal. She always knew what to say to make me get some perspective in moments of panic. I still remember a time when Jose Camacho and I were the ghosts that haunted the Linguistics Department at USC; Jose has remained a good friend since those days. The first time 1 met Marcello Modesto, I had the feeling that we had known each other for a long time, and now, I like to think of him as the part of my family that comes from Brazil. Thanks to Antonella Vecchiato and Esti Amorrortu for their friendship, which has survived my constant interruptions to finish Anto's sentences and endured a long road trip through the Southwestern desert heat in August 2000. I cannot imagine my life in LA without the long phone conversations with Karine Megerdoomian, starting from her days working at the USC bookstore on Sundays, trying to kill time, going through her move to New Mexico, and ending with my move to Beirut. Beside their entertainment value, they provided me with motivation and energy to work late, even when all I wanted to do is sleep. My friendship with Zoe Wu started very early on, when I took my first syntax class. If I made it through that class, it was mainly thanks to her. It is thanks to her, that I am writing these words now. Nothing good ever came out of a computer breaking down and I have had my share iv Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of experiences in that area. However, it is thanks to one of those incidents that I met Suzanne Curtin. I have learned many things from her, and without her friendship, I do not know how I could have survived my last year at USC. I have been waiting a long time to be able to express all my gratitude to Zeina, Joe, and their sons Karim and Marwan for being my LA family. I am happy to know that my relation with them does not end with these words. This dissertation is dedicated to my sisters Nada and Leila who suffered through agonizing hours of making data judgments but never heard the end of the story, and to my parents, Salwa and Georges, who have been very patient with me and were more eager to see this dissertation through than anyone else I know. It is now time to turn the page. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table of Contents Acknowledgm ents................................................................................................... ii List of Tables.............................................................................................................ix Abstract....................................................................................................................... x 1. Introduction...........................................................................................................1 1.1 Em pirical issues.....................................................................................................2 1.1.1 Resumptive constructions do not form a natural class.................................6 1.1.2 Two types of restrictive relatives................................................................... 10 1.1.3 Nature of the pronominal resumptive element.............................................14 1.2 Fram ew ork and theoretical iss u e s .................................................................16 1..2.1 Core operations...............................................................................................16 1.2.2 Indirect feature driven movement and the syntax of noun phrases 19 1.2.3 Movement and the nature of resumption.....................................................26 2. Resumption and Island sensitivity in Restrictive Relatives..................... 29 2.1 D istribution of w eak resum ptive elem ents..................................................33 2.2 Island sensitivity in restrictive relatives....................................................... 38 2.2.1 Absence of island sensitivity with weak resumptive elements................... 38 2.2.2 Referentiality: Evidence for island sensitivity.............................................43 2.2.2.1 Relativization of m anner adjuncts...................................................43 22.2.2 Relativization of idiomatic NP c h u n k s.......................................... 51 2.3 A m ovem ent analysis of resum ption............................................................ 65 2.4 Evidence for traces in restrictive relatives in LA........................................ 77 2.4.1 Relativization of subjects o f existential constructions................................ 78 2.4.2 Adjunct relatives............................................................................................ 87 2.5 S u m m ary .............................................................................................................. 92 3. Reconstruction and A-chains in restrictive relatives................................... 95 3.1 Scope reconstruction in LA restrictive relatives......................................... 96 3.1.1 Scope reconstruction with ksll NP 'every NP' and Tdddit NP 'many NP' ..............................................................................................................................96 3.1.2 Scope reconstruction and island sensitivity...............................................I l l 3.1.3 Scope reconstruction, amount quantification, and Principle Ceffects... 126 3.2 Reconstruction and bound variable a n ap h o ra......................................... 137 3.3 Some consequences......................................................................................... 145 vi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.3.1 Principle C and 'anti-reconstruction' effects in restrictive relatives...... 147 3.3.2 A copy theory or a chain-binding approach to reconstruction?...............159 3.3.3 Interpretation of restrictive relatives with WRE in Lebanese Arabic..... 178 3.4 S um m ary............................................................................................................189 4. T he Structure of Resum ptive Restrictive Relatives....................................190 4.1 Background discussion..................................................................................192 4.2 The representation of raising resum ptive restrictive relatives in Lebanese A ra b ic ..............................................................................................195 4.2.1 The external determiner hypothesis........................................................... 197 4.2.1.1 Restrictive vs. non-restrictive relatives...............................................198 4.2.1.2 Amount relatives.................................................................................. 201 4.2.2 The complementation analysis................................................................... 205 4.2.3 DP relatives.................................................................................................. 208 4.2.3.1 Morphosyntactic properties o fy a lli....................................................210 4.2.3.2 (p-features ..............................................................................................216 4.2.4 The categorial nature of the moving element............................................ 225 4.2.5 The (in-)definiteness of the relativized noun phrase.................................229 4.2.6 The trigger for movement............................................................................229 4.3 Structure of base-generated resum ptive restrictive re la tiv e s...............253 4.4 The categorial nature of the resum ptive restrictive relative clause in Lebanese A ra b ic ..............................................................................................262 4.5 A problem for the raising analysis.............................................................. 266 4.6 The final p ic tu re ..............................................................................................271 5. M ovem ent and the N ature of R esum ptives..................................................272 5.1 H ebrew free relatives .................................................................................... 273 5.2 Italian clitic left dislocation re v isite d .........................................................278 5.3 Two analyses of clitic doubling .................................................................. 282 5.4 W eak resum ptive elements in Lebanese Arabic as agreem ent m arkers ..............................................................................................................................292 5.4.1 Subject and object agreement morphology.................................................304 5.4.2 Canonical chains and resumptive chains...................................................317 5.4.3 Clitic doubling in Lebanese Arabic............................................................. 335 5.5 Further Consequences.................................................................................... 340 5.5.1 Strong pronouns in resumptive constructions......................................... 340 5.5.2 Note on the interpretation of resumptive constructions........................... 353 5.6 C oncluding rem ark s....................................................................................... 358 vii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. References Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. List of Tables Table 1. Third person non-subject clitic p ro n o u n s............................................... 34 Table 2. Third person strong pronouns ................................................................. 35 Table 3. Independent dem onstrative pronouns in LA .................................... 219 Table 4. H ebrew accusative pronouns ................................................................. 276 Table 5. Clitic left dislocation in Ita lia n ................................................................ 278 ix Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abstract ISSUES IN THE SYNTAX OF RESUMPTION: RESTRICTIVE RELATIVES IN LEBANESE ARABIC In this dissertation, I examine the mechanisms that underlie the formation of long A-dependencies in resumptive restrictive relatives (RRRs). I argue that such dependencies between an antecedent and the related resumptive site may be derived in one of two ways: either via Move or via base-generation. The former display the properties in (1), whereas the latter have the properties in (2). (1) a. Island sensitivity b. Reconstruction effects c. Same range of interpretations for resumptive as for A-traces d. Resumptive element must be a (complex) agreement affix (2) a. No island sensitivity b. No reconstruction effects c. Fixed interpretation for resumptive: definite pronoun d. Resumptive element can be weak or strong Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The antecedent and the resumptive site form a chain, in RRRs derived by movement, but not in base-generated RRRs. The properties in (1) and (2) fall out naturally under this proposal. The behavior of Indefinite RRRs, which display only the properties in (2), further supports this dual analysis. Those constructions are shown to pose a problem for a raising-only analysis of relativization. The proposal makes correct predictions with respect to the interpretive properties of RRRs and cases of feature mismatch between the antecedent and the resumptive element. I provide a detailed proposal linking resumption, movement, and agreement morphology in Lebanese Arabic. Resumptives in the context of movement are the result of an agreement relation between a noun phrase and the strong head that selects it. Not only does the analysis account for the distribution of agreement morphology in Lebanese Arabic, it accounts for the difference between strong pronouns and clitics with respect to resumption. Generally, the dissertation provides answers to questions relating to the nature of resumptive elements and their role in the syntax of A-constructions by arguing that those elements do not form a natural class in triggering movement effects and that their morphological properties play an important role in determining the availability of those effects. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 1 Introduction In this study, I examine in depth several issues in the syntax of long A- dependencies of the sort represented in (1), where what is called the antecedent is resumed by a pronominal element occurring within the sentence. (1) Antecedent ...pronominal element; Descriptively speaking, the antecedent in (1) generally appears in a sentence peripheral position in constructions such as clitic-left dislocation (CLLD), ivh- questions, or relative constructions. The following sentences illustrate the representation in (1) in various constructions in Lebanese Arabic (henceforth LA).1 (2) a. Sami zaarat-o Laila b-l-mastajfa CLLD Sami visited.3sf-him Laila in-the-hospital 'Sami, Laila visited him in the hospital.' 1 The resumptive elements will be marked in bold throughout. 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. ?ayya mariid zaarat-o Laila b-l-mastajfa Wh-Question which patient visited.3sf-him Laila in-the-hospital 'Which patient did Laila visit in the hospital?' c. l-mariid Hi zaarat-o Laila b-l-mastajia Relative the-patient that visited.3sf-him Laila in-the-hospital 'The patient that Laila visited in the hospital' I will be mainly examining the mechanisms that underlie the relation between the antecedent and its corresponding resumptive pronominal in (1) and the role that the nature of the pronominal element plays in establishing this relation. 1.1 Empirical issues The empirical domain of investigation is mainly that of restrictive relative constructions, but, as we shall see, the analysis provided extends to other A- constructions involving resumption. The interest in restrictive relative constructions lies mainly in that they raise questions of descriptive adequacy with respect to the analysis of resumption. New generalizations regarding the distribution of resumptive elements and their interpretive properties will have to be accounted for by any analysis aiming at adequately describing the phenomenon of 2 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. resumption. A case in point relates to the observation made in Doron 1982 and examined in Sells 1985 that resumptive pronouns cannot have de dicto readings. (3) a. Dani yimca et ha-ija fe hu mexapes t Dani will-find the-woman that he seeks t 'Dani will find the woman that he seeks.' b. Dani yimca et ha-ifa je hu mexapes ota Dani will-find the-woman that he seeks her 'Dani will find the woman that he seeks.' Doron notes that (3a) has a 'de dicto' reading that (3b) does not have. That is, in (3a), Dani need not have a particular woman in mind, and so, (3a) can be uttered felicitously in a context where Dani is searching for someone with some particular properties, only one of which is being a woman. (3b) cannot be used in such a situation; it has only the interpretation where there is a particular woman, e.g. Paul's sister, that Dani is looking for. This interpretation is also available in (3a). The difference between (3a) and (3b) is attributed to the distinction between gaps and pronouns in their interpretations. In Lebanese Arabic, the equivalent to the sentences in (3) is (4). Lebanese Arabic does not have the gap option for relativization of direct objects. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (4) Sami rati ylee?e l-mara lli fam bifattij Talay-a Sami fut. find the-woman that Asp. look.3sm for-her 'Sami wiU find the woman that he is looking for.' (4), unlike the Hebrew sentence in (3b), is ambiguous in its interpretation. It has aU the readings available for (3a). This observation leads us to re-examine the connection between D-linking and resumption. However, we can immediately observe that in the case of relativization, the presence of a resumptive pronominal in LA does not impose a D- linked reading on the antecedent.2 2 1 understand D-linking as it is defined in Pesetsky 1987. Thus, in uttering a sentence like (i), (i) Which man did you see? the speaker indicates that both she and the hearer have a set of men in mind. In that sense, which phrases are always D-linked. They do not co-occur with the phrase the hell, which forces non-D-linked readings, as illustrated in the contrast between what and which phrases below. (ii) a. What the hell book did you read? b* Which the hell book did you read? 4 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This certainly cannot be the case in the relativization of idiomatic NP chunks (5), which are not referential and hence cannot be D-linked: (5) l-ratta Hi ?axad-a Sami Tabukra feedit-o the-nap that took.3sm-her Sami morning helped.3sf-him 'The nap that Sami took in the morning helped him.' The noun ratta translated here as 'nap' has in fact no meaning in isolation in Lebanese Arabic and is always non-referential in the context of the idiom. It can clearly however be resumed by a pronominal element, as seen in (5). Another aspect of the same problem deals with the unavailability of resumption in certain contexts, even when a D-linked reading is available. Thus, questions However, which phrases need not always be related to resumptives in questions and what distinguishes the two versions of the sentence in (iii) cannot simply be D-linking. (iii) ?ayya kteeb ?riite(-i) which book read.2sf(-it) 'Which book did you read?' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. involving kam NP 'few N F cannot involve resumption, even when a D-linked reading is forced and the «)/i-phrase is interpreted as a partitive:3 (6) kam waatiad (mn ha-l-wleed) ?aasasit(*-un) l-m'Tallme? few one (of this-the children) punished.3sf(-*them) the-teacher.fs 'H ow many (of the children) did the teacher punish (*them)?' 1.1.1 Resumptive constructions do not form a natural class From the vast literature dealing with resumptive constructions, one can readily conclude that, descriptively speaking, those constructions do not form a uniform class. In some languages, like Swedish for instance, resumptive pronouns occur only in positions where gaps resulting from A-movement are prohibited. A resumptive pronoun must therefore appear in the subject position following an undeletable complementizer (7). 3 It is not that kam NP 'few NP' cannot be related to a resumptive element; in relative constructions, resumptive elements are perfectly acceptable: (i) 1-kam walad yalli badda t?aasas-un l-m'Tallme ray biin the-few child that want.3sf punish.3sf-them the-teacher.fs absent.p 'The few children that the teacher wants to punish are absent.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (7) a. ett fiirslag som som vi inte kan avgiira om det fungerar utan att a proposal that we cannot decide if it works without pruva trying 'A proposal that we cannot decide if it works without trying/ b* ett fiirslag som som vi inte kan avgiira om fungerar a proposal that we cannot decide if works utan att priiva without trying 'A proposal that we cannot decide if it works without trying.' In other languages, resumptive pronouns alternate with traces of A-movement, as illustrated in the LA sentences in (8). (8) a. ?ayya Tadiibe Tazamtuw-a ( Ta-l-tiafle which author.fs invited.2p-her to-the-party 'Which author did you invite to the party?' b. ?ayya Tadiibe Tazamto ?a-l-hafle which author.fs invited.2p to-the-party 'Which author did you invite to the party?' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In some cases, the resumptive constructions exhibit properties of A-traces: as indicated by Georgopoulos (1984), Belauan resumptive pronominals license parasitic gaps (9). (9) a kukau a kalli [el kirem [el melkiokl er ngiii er u'ei er taro food Comp 2s-must Comp Imp-cook P 3s P before P a 'omekelii et ]] eat-Pf 'Taro is food that you must cook before you eat.' Lebanese Arabic resumptive constructions present a case where, within the same language, resumptive constructions do not form a natural class. Resumptive constructions in this language, and specifically definite restrictive relatives exhibit two sets of contrasting properties, listed in (10) and (11): (10) Apparent resumption a. Island sensitivity b. Reconstruction effects for bound variable anaphora, principle C, and scope c. Availability of de dicto reading for the antecedent4 4 But see chapter 3 for a slight modification of this generalization. 8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (11) True resumption a. No island sensitivity b. No reconstruction effects c. Only de re reading available for the antecedent Following a distinction made in Aoun et al. 2001,1 will refer to constructions exhibiting the set of properties in (10) as cases of apparent resumption and to constructions that display the set of properties in (11) as cases of true resumption. In each given set, the properties seem to cluster together, indicating that one analysis is to be given to account for them. Building on work by Aoun and Benmamoun (1998), Aoun and Choueiri (1996) and (2000), and Aoun et al. (2001), I provide, to account for the contrast between (10) and (11), an analysis in terms of the availability of movement for generating resumptive constructions, briefly sketched in (12). (12) a. Movement available A ntecedent... Resumptive-Antecedent b. Base generation A ntecedent... Resumptive-pro; The properties in (10) and (11), like island sensitivity or the lack thereof, the availability of reconstruction effects, and the interpretive characteristics of 9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. resumptive elements, can be directly accounted for if they are analyzed as the result of (12a) and (12b), respectively. This is what I set out to do in chapters 2 and 3. 1.1.2 Two types of restrictive relatives The availability of the base generation strategy accounts for the absence of principle C effects under reconstruction that we observe in relative clauses. In that sense, relative clauses differ from wh-questions: whereas A-movement in wh-questions cannot void principle C violations (modulo the argument/adjunct asymmetry), this seems possible in restrictive relatives, as was observed by Munn (1994). Thus we have the contrast in (13). (13) a. najaro l-mu?eebale ssuhufiyye la-r-ra?iis lli published.3p the-interview the-magazine to-the-president that ?alto ?8nno sa?alu-n < Tann-a ?a-1-television said.2p that asked.3p-/n'»/ about-it on-the-TV 'They published the magazine interview of the president that you said that they asked him about it on TV.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b.* Tan ?ayya mu?eebale sutiufiyye la-r-ra Tiis ?alto about which interview magazine to-tlie-president said.2p ?enno sa?alu-it that asked.3p-/n'm 'About which magazine interview of the president did you say that they asked him?' In (13a), the name embedded within the external head (i.e. the antecedent) of the restrictive relative can be coreferent with a pronorm inside the relative clause itself, even when this pronoun c-commands the relativization site (marked by the presence of a weak resumptive element). No principle C violation is observed. This is in contrast to what we observe in wh-questions, as indicated by the ungrammaticality of (13b). If there is preference for reconstruction under A- movement (Chomsky 1995), the unacceptability of (13b) can be attributed to the inescapability of a principle C violation. The same reasoning carries over to (13a) under the assumption that restrictive relatives can be generated by movement. However, the acceptability of (13a) under the relevant reading can be explained by appealing to an alternative derivation for the relative clause, namely the one in (12b). On a related note, the argument/adjunct asymmetry observed in wh-questions with respect to Principle C effects under reconstruction raises issues with respect to the 11 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. structure assigned to restrictive relative clauses (Lebeaux 1991). On one hand, there have been many recent analyses supporting Kayne's 1994 complementation structure for restrictive relatives (see Bianchi 1999,2000, Safir 1999, Sauerland 1999, Schmitt 2000, and others), in which the relative clause (a CP) is generated as the complement of a determiner head (D), as illustrated in (14): (14) [ d p D [cp relative clause] ] On the other hand, the adjunct/argument asymmetry observed by Lebeaux (1991) seems to require an adjunction analysis for restrictive relatives. (15) a.* Which claim that Mary had offended Johns did hei repeat? b. Which claim that offended Johns did hes repeat? The observation is that Principle C blocks coreference between John and he in (15a); it does not do so in (15b). Chomsky 1995, following Lebeaux's analysis of the facts proposes that the contrast in (15) is due to the following two assumptions: (16) a. There is a preference for reconstruction {Preference Principle).5 b. Adjuncts are joined late in the derivation (Late Adjunction). 5 See Safir 1999 for arguments that support the Preference Principle. 12 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The assumption in (16a) indicates that in all cases of A-movement, the moved element is, ceteris paribus, interpreted in its base position, and ensures the reconstruction of the restriction of the zo/i-phrase to the object position of repeat in (15a). Late adjunction serves to distinguish between the derivation of (15a) and that of (15b): it states that adjuncts do not have to be inserted in the base, so that in (15b) the adjunct relative clause need not have a copy in the object position of repeat. The contrast between (15a) and (15b) can be reduced to the fact that the name John occurs in a noun complement in (15a) but inside an (relative clause) adjunct in (15b). If this analysis is indeed on the right track, it raises serious questions about Kayne's structure in (14), suggesting that a traditional analysis of restrictive relatives, one where the relative clause is generated as an adjunct to the noun phrase it modifies (17), is preferable. (17) [dp D [np [n p N] RC ] ] Chapter 4 examines the structure of resumptive restrictive relatives and discusses how (17) can be reconciled with Kayne's structure in (14). The two sets of properties listed in (10) and (11) further serve as diagnostics to distinguish between two types of restrictive relatives: Definite relatives and Indefinite relatives. Definite relatives are typically those whose external head is introduced by a definite determiner and is followed by the relative marker yalli (18a). The external 13 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. head of the indefinite relative involves a weak determiner and no relative marker (18b). (18) a. zirna l-beet lli Jtaru-u ?ahl-ik visited.lp the-house that bought.3p-it parents-your 'We visited the house that your parents bought.' b. zirna beet Jtaru-u ?ahl-ik visited.lp house bought.3p-it parents-your 'We visited the house that your parents bought.' Indefinite relatives exhibit aU and only the properties of true resumption (11), and hence involve no movement. Definite relatives on the other hand may display properties of true resumption (11) as weH as those of apparent resumption (10), and therefore two possible derivations may be posited for those constructions, one that involves movement and the other base generation. The base generation strategy helps to account for cases where a mismatch in features exists between the antecedent and the resumptive pronoun. 1.1.3 Nature of the pronominal resumptive elements It is observed in the literature (see Borer 1984, Bianchi 1999, among others) that in resumptive constructions involving movement, the resumptive element can only be 14 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. a clitic or a complex agreement morpheme, but not a weak or a strong pronoun.6 A case in point is that of Modern Hebrew free relatives where the relativized position can be occupied by an affix-like pronominal element (19a), but not an accusative pronoun (19b). (19) a. kaniti ?et ma Je hexlatet ?alav bought-I ACC what that decided-you on-it 'I bought what you decided on.' b.* kaniti ?et ma Je ra?it ?oto bought-I ACC what that saw-you it 'I bought what you saw.' This generalization is shown to be true of Lebanese Arabic resumptive constructions as well.7 On the other hand, when movement is prohibited, strong pronouns as well as full lexical DPs are able to serve as resumptive elements (Aoun et al. 2001). (20) illustrates the generalization with strong pronouns and epithets inside a wh-island. 6 1 am using Cardinaletti and Starke's (1994) distinction between strong, weak, and clitic pronouns here. 7 Examples that seemingly contradict this generalization can also be found in Lebanese Arabic (see Aoun et al. 2001). I will discuss such examples in chapter 5. 15 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (20) Wli-islands a. ha-l-muttahame badkun taTrfo miin bifakkir ?anno this-the-suspect.sf want.2p know.2p who think.3sm that hiyye/ha-l-mal*Tuune harabit she / this-the-damned ran. away-3sf 'This suspect, you want to know who thinks that she/the damned one ran away.' 1.2 Framework and theoretical issues I have so far outlined the generalizations that any analysis dealing with resumption and resumptive constructions has to contend with. In this section, I present key concepts of the theory on which I will base my analysis of resumption in A- constructions in general, and in restrictive relatives in particular. The analysis is couched within the Principles and Parameters approach to Generative grammar, and in particular in its extensions in the Minimalist Program (Chomsky 1998,2000). 1.2.1 Core operations Language is a mapping between a set of features F and a set of expressions EXP. F is made available by Universal Grammar (UG) and expressions are generated by the 16 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. computational procedure for human language C hl, also a component of UG. The operations that enter into C h l are Merge and Agree. Move, in this framework is a composition of Agree and Merge. Merge takes two syntactic objects a and p and forms K(a,P) from them. Pure Merge, Merge that is not part of Move, is required of arguments: they must be first introduced in the computation in their 0-position. Agree establishes a relation, e.g. agreement, between a lexical item LI and a feature a in a local domain. Move, as already mentioned, combines Merge and Agree, thus when agreement is established between LI (the Probe) and a (the Goal), P(a), a phrase determined by a (not necessarily its maximal projection) is merged to LIP, a projection headed by LI. EPP features are uninterpretable features that require Merge and are therefore the triggers for movement. Though they are non-semantic in that they are not interpreted on their host, the configuration they establish has effects for interpretation, as is the case, for example, with the [+wh] feature on the wh-words in English. The choice of Move over Agree follows from the presence of EPP features, in the absence of pure Merge (when expletives are available, for instance). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Match is simply understood as identity in the choice of features, not their value.8 It is necessary that matching hold of Probe and Goal in order for them to agree. Furthermore, in order for a matching relation to induce Agree, the Goal must be in the domain D(P) of the probe and satisfy the locality conditions, as stated in (21) (Chomsky 1998). (21) a. D(P) is the sister of P b. Locality reduces to "closest c-command" In order for a Probe to agree with a Goal, the Goal must be in the c-command domain of the Probe and there is no G', such that G' is in D(P), matches the Probe, and the Goal is in D(G'). The derivation of EXP, the set of expressions in a language, proceeds by phase. Phases are determined by subarrays of the lexical items chosen from the original array selected for the derivation. It is generally assumed that vP and CP are phases. They have the two properties listed in (22): 8 This is necessary to explain certain cases of intervention that fall under the Minimal Link Condition (Chomsky 1995). The intervening element does not always bear features with values identical to those of the moving element. 18 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (22) a. They have a certain degree of phonological independence b. They can be reconstruction sites. Phases affect the process of Move in the following way: A phrase deeply embedded in a phase cannot be extracted from that phase. The Phase Impenetrability Condition (PIC), stated as in (23) derives that effect. (23) Phrase impenetrability condition The domain of the head H of a strong phase is not accessible to operations at ZP (the next strong phase), but only H and its edge (i.e. its specifier or elements adjoined to HP). Therefore, at the end of a cycle, operations cannot look into phases below their heads. The PIC thus requires that A-movement targets the edge of every phrase, thus obeying strict cyclicity. It ensures that A-movement obeys "shortest move". 1.2.2 Indirect feature driven movement and the syntax of noun phrases If movement requires Merge and only EPP features can enforce the application of this operation, then we are led to say that cases of overt A-movement must involve an EPP feature at every phase level to ensure that the operation obeys PIC. This is what Chomsky (1998) calls Indirect Feature Driven Movement. To illustrate how 19 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. derivations proceed under PIC, consider the following configuration, which is basically the one for a wh-question: (24) [cp C[+wh)... [vp SU v [vp V O B [+ w h] ] ] ] Supposing along with Chomsky 1998 that CP and vP are phases, and that the EPP feature of C is the [+wh] feature, then at the point where the derivation reaches the CP phase in (24), the object, which has remained in situ, is invisible for agreement relations with C. At the CP level, only v and its specifier position are accessible for the computation, and the derivation crashes because of the presence of the uninterpretable [+wh] feature on the object, which thus remains unchecked. The alternative configuration, which yields a converging derivation, generates an EPP feature on the vP, which triggers movement of the object (OB) to the vP periphery: (25) [c p C|+wh]. . . [O B [+w hi [vp SU v [v p V OB(+W h j ] ] ] ] Due to PIC, only the configuration in (25) can lead to the final one (26), where OB is merged with CP to eliminate its uninterpretable [+wh] feature. (26) [cp O B [+ w h] -C [+ w h] • • • [O B (+ w h] [vp SU v [vp V OB[+W h ] ] ] ] ] 20 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. An important thing to note is that the movement of OB to the vP periphery in languages like English is not for agreement or case assignment purposes. Direct objects in English can remain in situ and Agree with the verb in that position. (27) [ v p S U v [Vp V OB]] In (27), the direct object (OB), which occurs in the domain of v with which it shares matching cp-features, satisfies closest c-command and therefore can establish an agreement relation with that v. OB is therefore assigned case without moving to the vP periphery. Following ideas of Reinhart (1997) and Fox (1995,1999), Chomsky 1998 suggests (28) to regulate the availability of optional rules, such as movement triggered by optional EPP features. (28) Optional rules can apply only when they yield a new outcome. In the context of wh-movement, v is assigned an EPP feature to ensure that the wh- word reaches CP in order to form an operator-variable chain. Indirect feature driven movement interacts very closely with the distribution of participle agreement in French (and Romance) and of object agreement in Lebanese 21 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Arabic. The generalization that accounts for the facts regarding participle agreement in French is stated in Sportiche 1992 and it extends as is to object agreement in Lebanese Arabic.9 Briefly, participles in French and verbs in LA agree with a direct object when the latter precedes the verbal element. The paradigm below illustrates the generalization: (29) a. Quelle maison as-tu construit(-e)? French which house have-you built(-fs) 'Which house have you built?' b. ?ayya bineeye < Tammart(-a)? Lebanese Arabic which building.fsbuilt.2sm(-fs) 'Which building have you built?' (30) a. J'ai construit(-*e) cette maison French I-have built(-*fs) this house 'I built this house.' b. tTammart(-a) 1-bineeye Lebanese Arabic built.ls(-fs) the-building.fs 'I built the building.' 9 1 am assuming here that weak resumptive elements in Lebanese Arabic are agreement markers. I will defend that position in chapter 5. 22 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Under the assumption that overt agreement morphology is the reflex of the elimination of EPP features on v, the generalization about the distribution of overt object agreement in French and Lebanese Arabic be easily accounted for. In order to eliminate the EPP feature on the verb/participle, the direct object must move and therefore the final configuration will have the direct object preceding the verb/participle. As suggested by Chomsky (1998), the availability of movement to the vP periphery (as a result of the EPP feature generated there) is regulated by economy considerations which amount to the following: In French and Lebanese Arabic, v will bear an EPP feature only if the direct object undergoes subsequent movement. We thus have an account for the unavailability of agreement morphology in the contexts where no movement occurs (30). In recent developments in the Minimalist Program (see Chomsky 1998, 2000), the notion of "checking" has been reduced to deletion under Agree, and therefore not all agreement relations necessitate movement of the Goal to the specifier of its Probe. However, the specifier-head relation still has an important theoretical status in defining the operation of Move. Thus PIC states that only the head of a phase and its specifier(s) are accessible for the computation from the next phase level. Overt agreement morphology manifests this relation. However, it is not in every case of A-movement of a direct object that an overt agreement morpheme shows up on the verb/participle. This is indicated by the optionality of object agreement in sentences like (29). I argue that although the 23 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. presence of an EPP feature on v is a necessary condition for overt agreement morphology to be realized on the verb/participle, it is not a sufficient one. Only norm phrases that involve the highest nominal functional projection CP can induce overt agreement on the verb/participle. Based on Cardinaletti and Starke 1994,1 propose the following structure for noun phrases: (31) [cp [ip [NumP .. . [n p ] ] ] ] The CP projection can be filled by dummy elements like the preposition la- 'to', which introduces doubled elements in clitic doubling constructions or by (strong) determiners like kdll 'every', ?ciyya 'which', and the definite determiner. Like in the corresponding clausal projection, the nominal CP also hosts features that are discourse related, like [±specific], for instance. The assumption that only noun phrases involving a CP agree with the verb accounts for all cases of object agreement that can be observed in the context of clitic-left dislocation, wh- questions, and restrictive relatives. Furthermore, it accounts for the cases where object agreement is not possible. This is, for example, the case of how many questions in LA as was observed in (6) (repeated below): (6) kam waatiad (mn ha-l-wleed) ?aasasit(*-un) 1-m‘ Tallme? few one (of this-the children) punished.3sf(-*them) the-teacher.fs 'How many (of the children) did the teacher punish (*them)?' 24 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. kam NP 'few NP' has a configuration that involves at most an IP structure: (32) [ip kam [NumP [np ] ] ] Unlike kail 'every', ?ayyn 'which', and the definite determiner, kam 'few' cannot be separated from its noun head by any other nominal heads: (33) a. ?ayya/k9ll/l- ?arba? wleed which/every/the four children 'W hich/every/the four children' b. kam ?arba< i ‘ wleed many four children 'Many four children' kam 'few' can only be associated to a singular noun and hence is generated in the closest projection to the NP. Cases like (29) where agreement is optional, I take, will involve an ambiguous analysis of the noun phrase undergoing movement: a CP when there is overt agreement morphology on the verb and one without a CP projection, when there is no agreement. 25 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.2.3 Movement and the nature of resumption Analyzing resumptive elements in Lebanese Arabic as agreement morphemes that do not occupy the tail position in an A-chain, we have an explanation as to why strong pronouns cannot be resumptive elements in A-constructions that involve movement (Borer 1984, Bianchi 1999). Strong pronouns are full noun phrases that are assigned structural case and therefore compete for the same position with the noun phrase they are intended to resume. In a configuration like (34), which is intended to represent the base structure for a resumptive construction involving a strong pronoun (SP), (34) *A-position ... [vp SU [v p [SP OB] ] ] whichever one of the two elements that moves to [Spec, vP], whether it is the strong pronoun (SP) or the direct object (OB), will leave the other one behind inside vP. The derivation will crash, not because the stranded element is inside the vP, but because it will not have eliminated its uninterpretable case feature. The cp-features on v will have been eliminated by the element that moves to [Spec, vP], where case will be assigned by the verb. As a consequence of this analysis, we expect resumptive elements that occur in the context of movement to be either clitics, which I will analyze adopting Sportiche's 1992 analysis, or agreement affixes, as in LA. 26 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Such an account provides an explanation for Sells' 1985 distinction between anaphorically-bound pronouns and operator-bound pronouns. Only the latter can be resumptives according to Sells. Sells (1985) follows Chao and Sells 1983 in noting that English does not have resumptive pronouns, because pronouns in English can only be anaphorically bound (and not operator-boimd).1 0 Thus, we have the following contrast between the resumptive pronoun in (35a) and the bound pronoun in (35b): (35) a.* I'd like to meet every linguist that Mary couldn't remember if she had met him. b. Every linguist thinks that Mary remembers him. The pronoun in (35b) can be interpreted as a variable bound by the quantifier phrase every linguist, because, the latter binds it from its subject A-position, and not 1 0 According to Sells 1985, pronouns that occur inside islands, as in (ia) in English, are intrusive pronouns and not resumptive pronouns. Intrusive pronouns, unlike resumptive pronouns, occur only inside islands and cannot be related to quantificational elements. (i) a. I'd like to meet the linguist that Mar)7 couldn't remember if the she had seen him before, b. *I'd like to meet every linguist that Mary couldn't remember if she had seen him before. However, Prince 1990 shows that resumptive pronouns in certain varieties of oral English have a wider distribution that previously thought. 27 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. from the A-position it occupies after QR. Such an option is not available in (35a) and since English pronouns cannot be operator bound, then the unacceptability of that sentence is accounted for. By making the basic distinction between operator-bound pronouns and anaphorically-bound pronouns, Sells 1985 is committed to saying that the difference between resumptive pronouns and other types of pronouns is a difference in nature. Yet, it is surprising that resumptive pronouns are interpreted just like other pronouns.1 1 It is also surprising that no language seems to realize resumptive pronouns as a separate set from regular pronouns. Therefore, it is preferable to seek a different interpretation for the difference observed between (35a) and (35b). English pronouns are strong pronouns and therefore, like strong pronouns in other languages, we do not expect them to be resumptive elements in contexts that involve movement, for the reasons that we have just reviewed. In bound anaphora contexts, the pronoun and the antecedent each occupy their own case position and each DP is case licensed. 1 1 Sells (1985) goes to great lengths to show that himself. 28 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 2 Resumption and Island Sensitivity in Restrictive Relatives Crosslinguistically speaking, the resumption strategy, i.e. the strategy whereby a pronominal element heads an A chain, instead of a gap/trace, is not a uniform strategy. Namely, it has been observed that, in certain languages, resumptive pronouns (RPs) have the distribution of traces of A-movement. Thus, RPs trigger weakcrossover effects, as illustrated in the Irish example in (1), and they can license parasitic gaps, as shown in the example from Belauan given in Georgopoulos 1984 (2). (1) * Sin an fear1 ar dhuirt me leis an bhastardi gur that the man comp said I with the bastard Comp+COP cheart ei achaitheamh isteach i bpriosun right him throw-INF into in prison 'That is the man that I said to the bastard that he should be thrown in prison.' 29 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (2) a kukau a kalli [el kirem [el melkiokl er ngiii er u'ei er taro food Comp 2s-must Comp Imp-cook P 3s P before P a 'omekelii ei ]] eat-Pf 'Taro is food that you must cook before you eat.' In other languages, RPs are in complementary distribution with traces of A- movement. That is, RPs occur where traces of A-movement cannot occur. Swedish provides a clear example of such a case. In Swedish, a resumptive pronoun must appear in the subject position following an undeletable complementizer. We therefore have the contrast in (3): (3) a. ett fiirslagsom som vi intekan avgiira om det fungerar a proposal that we cannot decide if it works utan att priiva without trying 'A proposal that we cannot decide if it works without trying.' b* ett fiirslagsom som vi intekan avgiira om fungerar a proposal that we cannot decide if works utan att priiva without trying 'A proposal that we cannot decide if it works without trying.' 30 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In the latter cases, RPs can be typically found in islands, which prohibit A-extraction (see Suher 1998 and references therein):1 (4) Spanish ique libro me dijiste (tu) que no recuerdas donde *(lo) which book to.me told you that no remember.2s where *(it) pusiste? put.2s 'Which book did you tell me that you don't remember where you put (it)?' 1 English is said to be a language that marginally accepts sentences such as (6) in the text and (ia) (see Kroch 1981): (i) a. the guy who they don't know whether he wants to come or not b* the guy who they don't know whether wants to come or not Only when the alternative with a gap is ungrammatical ((ib) illustrates a COMP-trace effect) will a sentence like (ia) be deemed acceptable by a native speaker of English. However, Prince (1990) shows that RPs in English have a wider distribution than previously thought. Thus, RPs in English can be found outside island contexts, where they alternate with traces of A-movement: (ii) He's very' good at those gold leaf letters you put (them) on from the inside. 31 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (5) Irish na danta sin nach bhfuil fhios again cen ait ar the poems dem C.neg is knowledge at.us what place C cumadh *(iad) were.composed *(them) 'Those poems that we do not know where they were composed' (6) the settlement that Caroline asked when we would get *(it) This state of affairs has led to the traditional analysis of resumptive pronouns as A-bound pronouns, i.e. pronouns that occur in the variable position of an A-chain and whose antecedent is base-generated in its surface A-position (see Chao and Sells 1983, McCloskey 1990). In what follows, I examine the resumptive strategy in Lebanese Arabic restrictive relatives, a pervasive strategy in those constructions. It will appear that resumption in those constructions is selectively sensitive to islands, a fact that will lead us to the conclusion that the generation of restrictive relatives with resumptive elements may involve movement (see Aoun and Choueiri 1997, Aoun and Benmamoun 1998, Aoun, Choueiri, and Homstein 2001, Cinque 1990, Ouhalla 2001, Bianchi 1999, Kayne 1996). 32 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.1 Distribution of weak resumptive elements Descriptively, a weak resumptive element generally appears inside Lebanese Arabic restrictive relatives in all positions except subject positions and adjunct positions. Examples of restrictive relatives with weak resumptive pronouns are given below: (7) Direct Object mna'Trif 1-mara lli rati y?eebal-a Sami know-lp the-woman that will meet-her Sami 'We know the woman that Sami will meet/ (8) Indirect Object 1-mwazzaf lli ?aaluu-l-o ?anno fii ?adraab ma i • 1 the-employee that told.3p-to-himthat there strike neg. ?83a come.3sm 'The employee that they told that there was a strike didn't come.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (9) Oblique tidam a 1-masrahiyyeet lli xabbarna 731111- 1111 Karim saw .lp the-plays that told.3sm about-them Karim 'We saw the plays that Karim told us about/ (10) Noun Complement tidam a 1-masratiiyye lli byaTrif Karim mmassliin-a kall-un saw.lp the-play that know.3sm Karim actors-it all-them 'We saw the play that Karim knows its actors all.' In the sentences in (7-10), the relativization site is occupied by a pronominal element that resumes the relative clause head noun. This resumptive element is the same as the third person pronoun that appears in non-subject argument positions: (11) a. Table 1. Third person non-subject clitic pronouns Masculine Feminine Singular - 0 -a Plural -un b. Laila beesit-o/ -a/ -un Laila kissed.3sf-him/ -her/ -them 'Laila kissed him /her/them .' 34 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. Laila seefarit ma?-o/ -a/ -un Laila traveled.3sf with-him/ -her/ -them 'Laila traveled with him /her/them .' d. Laila staTaarit kteeb-o/ -a/ -un Laila borrowed.3sf book-him /-her/ -them 'Laila borrowed his/her/their book/ When relativizing from a subject or an adjunct position however, no resumptive element appears in the relativization site.2 The sentence in (12) illustrates a case 2 The strong pronouns that may appear in subject positions within relative clauses (see (i)) have a distribution radically different from that of weak pronouns. (i) Table 2. Third person strong pronouns Masculine Feminine Singular huwwe hiyye Plural hanne When strong pronouns occur in relativization contexts, they are generally optional: (ii) lafna l-leefib lli (huwwe) rabiti bi-Wimbledon saw .lp the-player that (he) won.3sm in-Wimbledon 'We saw the player who won in Wimbledon.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. As seen in (ii) above, they also cooccur with agreement morphemes. In that sense, strong pronouns parallel lexical NPs in their distribution, since the latter also cooccur with subject-verb agreement morphemes: (iii) Karim rabih Karim won.3sm 'Karim won.' Another difference that can be noted between strong and weak pronouns is that strong pronouns can be dislocated, and in those contexts they double a weak pronominal element: (iv) a. huwwe/Karim, hkiit maT-o mbeerih he/Karim talked.lswith-him yesterday 'Him/Karim, I talked with him yesterday.' b. hkiit maT-o huwwe/*Karim mbeerih talked.lswith-him he/*Karim yesterday 'I talked with HIM yesterday.' As described also in Cardinaletti and Starke 1994, strong pronouns, but not weak pronouns, are compatible with c-modifiers (adverbs that modify the whole noun phrase). This provides the only context where strong pronouns can be found in complement positions, as can be seen in (v). (v) Jaft bas huwwe/Karim saw .ls only he/Karim 'I saw only him/Karim.' 36 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. adjunct relativization and the ones in (13) illustrate various cases of relativization from subject position. (12) rati nuusal Ta-l-fiafle see*Tt Hi bitballij l-musii?a fut. arrive.lp to-the-party hour that begin.3sf the-music 'We will arrive at the party the time that the music begins.' (13) a. sam'Tiyyit ssaliib l-?ahmar ?addamit hdiyye la-l-fiakiime that helped.3sf wounded the-war 'The organization of the Red Cross presented a gift to the doctor that (she) helped the wounded of the war.' b. (kail) l-katub Hi keen fii Ta-ttaawle saaro (all) the-books that was in-it on-the-table become.3p Ta-rraff halla? on-the-shelf now '(All) the books that there were on the table are now on the shelf.' Having noted these differences, which clearly suggest a different treatment of weak and strong pronouns, I leave the discussion of strong pronouns until Chapter 5 (see also Aoun, Choueiri, and Homstein 2001). 37 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.2 Island sensitivity in restrictive relatives 2.2.1 Absence of island sensitivity with weak resumptive elements In general, weak resumptive elements can occur inside islands in LA restrictive relatives. This is usually the case when the external head of the restrictive relative is related to a weak pronominal element in direct object, indirect object, oblique, or genitive position. The paradigms in (14-16) illustrate the generalization for wh-islands, adjunct islands, and relative clauses, respectively. (14) wh-islands a. mnaTrif 1-mara Hi *T am titsee?alo miin rati y?eebal-a know-lp the-woman that asp. wonder.2p who will meet-her 'We know the woman that you are wondering who will meet (her).' b. 1-mwazzaf lli baddkun ta'irfo miin ?al-l-o ?anno i i the-employee that want.2p know.2p who told.3sm-to-him that fii ?adraab ma ?a3a there strike neg. come.3sm 'The employee that you want to know who told (him) that there was a strike didn't come.' 38 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. fidarna l-masrahiyyeet lli tsee?alto miin xabbarna saw.lp the-plays that wondered.2p who told.3sm Tartn-un about-them 'We saw the plays that you wondered who told us about (them).' d. fiderna 1-masrahiyye lli Tam titsee?alo ?9za byaTrif saw.lp the-play that asp. wonder.2p whether know.3sm Karim mmassliin-a kall-un Karim actors-it all-them 'We saw the play that you are wondering whether Karim knows its actors all.' (15) Adjunct islands a. mnaTrif 1-mara lli fallayto ?abl ma y?eeb8l-a know-lp the-woman that left.2p before Comp meet.3sm-her Karim Karim 'We know the woman that you left before Karim met (her).' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. 1-mwazzaf lli fallayto ?abl ma t?uulu-l-o ? 9n n o fii the-employee that left.2p before Comp tell.3p-to-him that there ?adraab ? 8 3 a strike come.3sm 'The employee that you left before telling (him) that there was a strike came/ c. hdarna l-masrahiyyeet lli fallayto lamma xabbar-na saw .lp the-plays that left.2p when told.3sm-us ‘ Tann-un Karim about-them Karim 'W e saw the plays that you left when Karim told us about (them).' d. hdarna 1-masrahiyye lli thammasto la?anno byaTrif saw .lp the-play that excited.2p because know.3sm Karim mmassliin-a kall-un Karim actors-it all-them 'W e saw the play that you were excited because Karim knows its actors all.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (16) Relative clause a. mnaTrif 1-mara lli tikiito xnaT rra 3 3 eel lli rati know-lp the-woman that talked.2p with the-man that will y?eebal-a meet-her 'We know the woman that you talked to the man who will meet (her).' b. 1-mwazzaf lli bta*Trfo 1-mara lli ?aal8t-l-o the-employee that know.2p the-woman that told.3sf-to-him ?anno fii ?adraab ma ? 8 3 a that there strike neg. come.3sm 'The employee that you know the woman that told (him) that there was a strike didn't come.' c. h d arn a l-m asrahiyyeet lli btaTrfo ssah aafe lli s a w .lp th e-plays that k n o w .2 p the-journalist that Xabbarna Tann-un told.3sm-us about-them 'We saw the plays that you know the journalist that told us about (them).' 41 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. d. "hdarna 1-masrahiyye lli byi^ab-kun ssatiaafe lli saw.lp the-play that please.3sm-you.p the-journalist that bya*Trif mmassliin-a kall-tm know.3sm actors-it all-them 'We saw the play that you like the journalist that knows its actors all/ The facts above seem to suggest an analysis of resumptive elements as A-bound pronouns with a base generated antecedent, an analysis that has been defended by several authors for various Arabic dialects (see Fassi Fehri 1978, Wahba 1984, Demirdache 1991, Elomari 1992, Ouhalla 1999, Shlonsky 1992).3 Under this analysis, restrictive relatives in LA would involve a pronoun in the relativization site, which resumes the base-generated external head of the restrictive relative. In what follows, I will present evidence to show that restrictive relatives generated with a resumptive element are sensitive to islands and therefore may involve movement in their derivation. 3 Borer 1984 adopts this analysis for Hebrew restrictive relatives involving resumptive pronouns and McCloskey 1990 adopts it for resumptive constructions in Irish. 42 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.2.2 Referentiality: Evidence for island sensitivity 2.2.2.1 Relativization of manner adjuncts The first piece of evidence comes from the relativization of abstract norms like sarTa 'speed', fajeefn 'courage', nazeeha 'integrity', narvaze 'nervousness', and tiibit ?alb 'kindness', when they are used as manner adverbials. These abstract nouns are introduced by an obligatory preposition, as illustrated in (17): nazeeha 'integrity' sarTa 'speed' (17) byijtiyil Sami *(bi-) < narvaze 'nervousness' works.3sm Sami *(with-) /njee ?a ' courage' tiibit ?alb 'kindness' V Relativization of such adjuncts involves obligatory resumption, as shown in (18) .4 4 The preposition bi 'with' which introduces the abstract NPs in (17) is realized asfi(y) with a weak pronominal element. I will abstract away from this alternation. 43 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (18) a. ssarTa lli Jtayal fiy-*(a) Sami hiyye 1-matluube the-speed that worked with-*(it) Sami she the-required 'The speed with which Sami worked is what is required.' b. ?a 3 abat-ne JJa3 eeTa lli ha3am fiy-*(a) pleased.3sf-me the-courage that attacked.3sm with-*(it) zzalame Tala 1-harame the-man on the-thief 'The courage that the man leaped with on the thief pleased me.' c. bta^ab-ne tiibt l-?alb/ nnazeeha lli biTaamil please.3sf-me goodness the-heart/ the-integrity that treat.3sm fiy-*(a) l-?asteez tleemiz-o with-*(it) the-teacher.m students-his 'The kindness/integrity that the teacher treats his students with pleases me.' d. nnarvaze lli byitike fiy-*(a) Sami maT 1-zbuneet the-nervousness that talk.3sf with-*(it) Sami with the-clients b*(ithafjal-un drives-away.3sf-them 'The nervousness that Sami speaks with to the clients drives them away.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. All the prepositional phrases containing the resumptive element inside the relative modify the verb, expressing manner. The resumptive element can be related to the relative clause external head across sentence boundaries. Again, the absence of these resumptive elements yields unacceptable sentences, as indicated in (19): (19) a. ssar'Ta lli ftakarto ?anno Sami byiJtiYil fiy-*(a) hiyye the-speed that thought.2pthat Sami works with-* (it) she 1-matluube the-required 'The speed with which you think that Sami works is the required one/ b. < T a 3 ab 8 t-ne ]Ja3 eeTa lli ?alto ?anno ha3 am pleased.3sf-me the-courage that said.2pthat attacked.3sm fiy-*(a) zzalame fala 1-bar ame with-*(it) the-man on the-thief 'The courage that you said that the man leaped with on the thief pleased me.' 45 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. bta^ab-ne tiibt l-?alb/ nnazeeha lli please.3sf-me goodness the-heart/ the-integrity that ?annaTtuu-ne ?anno biTaamil fiy-*(a) l-?9steez convinced.2p-me that treat.3sm with-*(it) the-teacher.m tleemiz-o students-his 'The kindness/integrity that you convinced me that the teacher treats his students with pleases me.' d. nnarvaze lli x abbartuu-ne ?anno byitike fiy-*(a) the-nervousness that told.2p-me that talk.3sf with-*(it) Sami m a*T 1-zbuneet rah bithajjal-un Sami with the-clients fut. drives-away.3sf-them 'The nervousness that you told me that Sami speaks with to the clients will drive them away.' Interestingly however, the resumptive elements corresponding to the manner adverbs cannot be related to the relative clause external head across island boundaries. When the NP corresponding to the manner adverb is relativized from a position inside an adjunct clause, a w/i-clause, or a relative clause, the resulting sentences are unacceptable, although a resumptive element appears in the relativized site (20-22). 46 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (20) Adjunct clauses a. * ssarTa lli btinbasto la?anno Sami byiftiYil fiy-*(a) h iy y e the-speed that pleased.2pbecause Sami works with-*(it) she 1 -matluube i the-required 'The speed with which you arepleased because Sami works is the required one.' b. * Ta3 abat-ne H a3ee^a Hi tfee3 a?to lam m aha3 am pleased.3sf-me the-courage that surprised.2p when attacked.3sm fiy-*(a) zzalame * 1 3 1 3 1-tiarame with-*(it) the-man on the-thief 'The courage that you were surprised when the man leaped with on the thief pleased me.' c. * bta^ab-ne tiibt l-?alb/ n-nazeeha lli nbasatto please.3sf-me goodness the-heart/ the-integrity that pleased.2p la?anno biTaamil fiy-*(a) l-?asteez tleemiz-o because treat.3sm with-*(it) the-teacher.m students-his 'The kindness/integrity that you were pleased because the teacher treats his students with pleases me.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. d. * n-narvaze lli btiddeeya?o la?anno byihke fiy-*(a) the-nervousness that bothered.2p because talk.3sf with-*(it) Sami ma? 1-zbuneet ratibithajjel-un Sami with the-clients fut. drives-away.3sf-them 'The nervousness that you are bothered because Sami speaks with to the client will drive them away/ (21) Wh-islands a. * ssor'Ta lli btaTrfo miin byijtivil fiy-*(a) hiyye the-speed that know.2p who works with-*(it) she 1-matluube i the-required 'The speed with which you who works is the required one.' b. * ?a3 ab8 t-ne Jja3 eeTa lli btaTrfo miin ha3 am pleased.3sf-me the-courage that know.2p who attacked.3sm fiy-*(a) Tala 1-harame with-*(it) on the-thief 'The courage that you know who leaped with on the thief pleased me.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. * b ta^ab -n e tiibt l-?alb/rmazeeha lli btaTrfo miin please.3sf-me goodness the-heart/ the-integrity that know.2p who bi'Taamil fiy-*(a) t-tleemiz treat.3sm with-*(it) the-students 'The kindness/integrity that you know who treats the students with pleases me.' d. * nnarvaze lli btaTrfo miin byihke fiy-*(a) maT the-nervousness that know.2p who talk.3sfm with-*(it) with 1 -zbuneet rah bithajjal-un the-chents fut. drives-away.3sf-them 'The nervousness that you know who speaks with to the clients will drive them away.' (22) Relative clauses a. * s-sarTa lli bta'Trfo 1-mwazzaf lli byijtiyil fiy-*(a) the-speed that know.2p the-employee that works with-*(it) hiyye 1-matluube she the-required 'The speed with which you know the employee who works is the required one/ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. * 'Tasabat-ne f-ja3 ee?a lli bta'Trfo z-zalame H i pleased.3sf-me the-courage that know.2p the-man that ha3am fiy-*(a) Tala l-harame attacked.3sm with-*(it) on the-thief "The courage that you know the man who leaped with on the thief pleased me.' c. * bta'Tsab-ne tiibt l-?alb/ n-nazeeha lli bta'Trfo please.3sf-me goodness the-heart/ the-integrity that know.2p l-?asteez lli bi'Taamil fiy-*(a) t-tleemiz the-teacher that treat.3sm with-*(it) the-students 'The kindness/integrity that you know the teacher who treats the students with pleases me.' d. * n-narvaze lli bta'Trfo J-JaYYiil lli b yih k e the-nervousness that know.2p the-employee that talk.3sfm fiy-*(a) ma'T 1-zbuneet rati bithajjal-im with-*(it) with the-clients fut. drives-away.3sf-them 'The nervousness that you know the employee who speaks with to the clients will drive them away.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The data concerning the relativization of abstract nouns like sarTa 'speed', fajeeTa 'courage', nazeeha 'integrity', narvaze 'nervousness', and tiibit ?alb 'kindness' provides the first indication that the presence of resumptive elements in LA may not always coincide with the absence of island sensitivity. If we take island sensitivity to be a diagnostic for movement, then the presence of resumptive elements in LA restrictive relatives cannot be said to always coincide with the absence of movement. 2.2.2.2 Relativization of idiomatic NP chunks Another piece of evidence to show that resumptive constructions in LA are selectively sensitive to islands comes from the relativization of idiomatic NP chunks. I will consider two different types of idioms, both of the form V(erb)+N(oun)P(hrase). In the first type, the NP takes on a figurative meaning in the context of a light verb, and in the second type, it is the verb that takes on a figurative meaning in the context of an NP that belongs to a restricted list of lexical items. The expression equivalent to take a nap in English is ?axad Yatta in Lebanese Arabic. That this expression is an idiomatic expression can be seen from the fact that the norm Yatta, which is a nominalization from the verb Yatt 'to sleep', has no independent meaning in Lebanese Arabic. 51 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (23) a * 1-Yatta ba*Td d-duhr bitfiid the-nap after the-noon help.3sf 'The nap in the afternoon helps.' b. n-nawme ba?d d-duhr bitfiid • i the-sleep after the-noon help.3sf 'The nap in the afternoon helps.' (24) * w a sa f-l-e 1-hakiim Yatta ba?d d-duhr prescribe.3sm-to-me the-doctor nap after the-noon 'The doctor prescribed a nap in the afternoon.' As shown by the unacceptability of (23a) and (24), the noun Yatta cannot take on the meaning 'nap' when it occurs by itself. In fact it has no meaning at all and the sentences are totally unacceptable. This interpretation of Yatta as 'nap' is contextualized and conditioned by the presence of the light verb ?axad 'to take' in (25):* 5 The idiomatic meaning 'to take a nap' appears also (and only) in the expression Yatt Yatta where the idiomatic NP chunk appears in the context of a cognate verb. The generalizations concerning ?axad Yatta in the text apply to the idiomatic expression Yatt Yatta. 52 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (25) Sami ?axad Yatta ba?d 1-Yada Sami took.3sm nap after the-lunch 'Sami took a nap after lunch/ Interestingly, the idiomatic NP chunk Yatta cannot be pronominalized: (26) * Sami ?axad Yatta b a ld 1-Yada bas Laila m a Sami took.3sm nap after the-lunch but Laila neg. ?a x a d it w a 1 id e/-a took.3sf o n e /-it 'Sami took a nap after lunch but Laila didn't take one/it.' The unacceptability of (26) indicates that the idiomatic meaning take a nap is not available when the NP Yatta is replaced by a(n) (indefinite) pronoun: the contextualized occurrence of one instance of the idiomatic NP chunk Yatta with i i which the (indefinite) pronoun can be anaphoric is not enough to salvadge the sentence. However, the NP Yatta can be relativized, and in such a case, a resumptive element is required in the relativization site: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (27) a. 1 -Yatta yalli ?axad-*(a) Sami baTd 1-Yada the-nap that took.3sm-her Sami after the-lunch feedit-o helped.3sf-him 'The nap that Sami took after lunch helped him.' b. 1-Yatta yalli wasaf-l-e 1-hakiim ?anno the-nap that prescribed.3sm-to-me the-doctor that ?eex9d-*(a) baTd 1-Yada feedit-ne take.3sm-it after the-lunch helped.3sf-me 'The nap that the doctor prescribed to me to take it after lunch helped me.' As we saw in cases of relativization of manner adjuncts, the relativization of the idiomatic NP chunk Yatta is sensitive to islands, even in the presence of resumptive elements in the relativization site: 54 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (28) a. Adjunct Clause * 1-Yatta yalli narvazit Zeina la?anno ?axad-a Sami the-nap that upset.3sf Zeina because took.3sm-her Sami ?abl 1-Yada feedit-o before the-lunch helped.3sf-him 'The nap that Zeina was upset because Sami took after lunch helped him / b. Wli-Clause * 1-Yatta yalli badda taTrif Zeina ?aza ?axad-a the-nap that want.3sf know.3sf Zeina whether took.3sm-her Sami ?abl 1-Yada feedit-o Sami before the-lunch helped.3sf-him "The nap that Zeina wants to know whether Sami took before lunch helped him.' b. Relative Clause * 1-Yatta yalli btalrif Zeina ssabe lli ?axad-a the-nap that know.3sf Zeina the-boy that took.3sm-her ?abl 1-Yada feedit-o before the-limch helped.3sf-him 'The nap that Zeina knows the boy that took before lunch helped him.' 55 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The relativization of idiom chunk NPs provides us then with an additional context where the presence of a weak resumptive element does not coincide with the absence of island sensitivity. Other types of idiomatic expressions provide evidence to the same effect. A case in point is that of the idioms formed with the verbs ?akal 'eat' or tafma 'feed', which in the relevant idiomatic meaning occur with any of the following NPs: ?atle 'a beating', kaff'a slap', Xa^ta 'a punch', darbe 'punch', bahdale 'reprimand, insult', Tayta 'reprim and'. (29) a. ?akal r?atle 'beating' eat kaff 'slap' < Xabta 'punch' darbe v . • 'punch' 'take a beating' b. ?akal eat bahdale 'reprimand, insult' layta 'reprimand' 'be reprimanded' 56 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (30) a. x ta^ma y r?atle 'beating' feed kaff 'slap' < Xabta 'punch' ^darbe 'punch' 'x gave y a beating' b. x ta*Tma v feed bahdale 'reprimand, insult' *?ayta 'reprimand' 'x gave y a reprimand' It is clear that both in the case of ?akal 'eat' and that of taTma 'feed' in (29) and (30), the verbs are used figuratively and not in their literal meaning. In their non-literal meaning the verbs do not select an argument that relates to food or edible things, but an argument from a fixed list of lexical items. This suppression of the selectional restriction of the verbs Takal 'eat' and taTma 'feed' reduces the meaning of those verbs to take and give, respectively. The expressions Takal kaff 'take a beating' and taTma kaff'give a beating' are considered idiomatic expressions because their meaning is conventionalized: it arises only in the context of NPs taken from the lists in (29) and (30). Thus, if kaff 57 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 'slap', for example, were to be replaced by baz?a 'a spit' (31), the result is unacceptable. (31) a * ?akal Karim baz?a ate.3sm Karim spit b.* Nada ta'Tmit Laila baz?a Nada fed.3sf Laila spit Similarly, if the verbs ?akal 'eat' and taTma 'feed' were to be replaced by their semantic equivalents ?axad 'take' or Tata 'give' in (29) and (30), the result does not yield the intended idiomatic readings: (32) a * ?a x a d ?atle 'beating' take kaff 'slap' i Xabta 'punch' ^darbe 'punch' 'take a beating' 58 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b * ?axad take bahdale 'reprimand, insult' Tayta 'reprimand' 'be reprimanded' (33) a * x Tata y give ?atle 'beating' kaff 'slap' Xabta 'punch' darbe 'punch' 'x gave y a beating' b.* x Tata y give bahdale 'reprimand, insult' Tayta 'reprimand' 'x gave y a reprimand' The non-substitutability of the verbs ?akal 'eat' and taTma 'feed' in the intended idiomatic interpretation is another indication that the expressions in (29-30) are indeed idioms. The idiomatic expressions ?akal kaff'take a beating' and taTma kaff'give a beating' share with the idiom ?a% ad yatta 'take a nap' the property that the idiomatic reading 59 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. does not arise if the idiomatic NP chunk is replaced by a pronoun, be it an indefinite or a definite one: (34) a.* Nada 3 arrabit ttaTme Nabil ?atle mbeerih bas Laila Nada tried.3sf feed.3sf Nabil beating yesterday but Laila she that managed.3sf feed.3sf-him one/it 'Nada tried to give Nabil a beating yesterday but it was Laila who managed to give one/it to him.' b.* Laila bittiaddir Tayta la-t-tlemiiz kail yom w L. prepare.3sf reprimand to-the-students every day and Nada feed.3sf one/it to-the-employees too 'Laila prepares a lecture for the students every day and Nada gives them one too.' In light of the facts in (34) it is interesting to note that the idiomatic NP chunks can be relativized and that, in such cases, the idiomatic NP chunk, which heads the restrictive relative clause, must be related to a weak resumptive element in the relativization site: 60 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (35) a. 1-bahdale yalli smaTto ?anno taTmat-o *(yeeha) Nada the-beating that heard.2p that fed.3sf-him *(it) Nada la-Nadixn ma rati yinsee-ha bi-tiayeet-o to-Nadim neg. fut. forget.3sm-it in-life-his 'The beating that you heard that Nada gave to Nadim, he will not forget it in his life.' b. smaTto b-l-bahdale yalli taVmat-o *(yeeha) Nada heard.2p in-the-reprimand that fed.3sf-him *(it) Nada la-Nadim bi-sabab haadis s-siyyara to-Nadim in-reason accident the-car 'You heard of the reprimand that Nada gave to Nadim because of the car accident.' (36) a. l-?atle yalli smaTto ?anno ?akal-*(a) Nadim min the-beating that heard.2p that ate.3sm-*(it) Nadim from 1 -mTallme ma rah yinsee-ha bi-tiayeet-o the-teacher.fs neg. fut. forget.3sm-it in-life-his 'The beating that you heard that Nadim took from the teacher, he will not forget it in his life.' 61 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. smoTto b-l-bahdale yalli ?akal-*(a) Sami man heard.2p in-the-reprimand that ate.3sm-*(it) Sami from ?amm-o mother-his 'You heard of the reprimand that Sami got from his mother/ Any analysis of the distribution of weak resumptive elements must be able to account for the fact that a weak pronominal element cannot corefer with an idiomatic NP chunk (34), but can resume one in the context of relativization. As expected by now, although the idiomatic external head of the relative clause is related to a resumptive element within the relative clause itself, this relation must obey island conditions: (37) Adjunct clauses a* l-?atle yalli zfelto la?anno ?akal-a Nadim ma the-beating that upset.2p because ate.3sm-it Nadim neg. rati yinsee-ha bi-hayeet-o fut. forget.3sm-it in-life-his 'The beating that you were upset because Nadim took it, he will not forget it in his life.' 62 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b.* smaTto b-l-bahdale yalli narvaz Nadim lamina heard.2p in-the-reprimand that angry .3sm Nadim when taTmat-o yeeha ?amm-o ?addeem 1-kall fed.3sf-him it mother-his in-front the-all 'You heard about the reprimand that Nadim was angry when his mother gave it to him in front of everybody.' (38) Wh-claiises a.* sma'Tna b-l-kaff yalli badda taTrif 1-mudiira heard.2p of-the-slap that want.3sf know.3sf the-principal.fs ?aza ?akal-o Sami man l-?isteez bi-s-saff whether ate.3sm-it Sami from the-teacher in-the-class 'We heard of the slap that the principal wants to know whether Sami took it from the teacher in class.' b* xabbarit-na Nada Tan 1-bahdale yalli btaTrfo miin told.3sf-us Nada about the-reprimandthat know.2p who ta'Tme-e yeeha la-Nadim bi-sabab 1-tiaadis fed.3sm-him it to-Nadim in-reason the-accident 'N ada told us about the reprimand that you know who gave it to Nadim because of the accident.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (39) Relative Clauses a.* 1-bahdale yalli btaVrfo 1-mTallme yalli taVmot-o the-reprimandthat know.2p the-teacher.fs that fed.3sf-him yeeha la-Nadim bi-sabab 1-tiaadis ma rati yinsee-ha it to-Nadim in-reason the-accident neg. fut. forget.3sm-it bi-hayeet-o in-life-his 'The reprimand that you know the teacher that gave it to Nadim because of the accident, he will not forget it in his life/ b* xabbarit-na Nada Van l-?atle yalli byaVrfo told.3sf-lp Nada about the-beating that know.3p ttlemiiz s-sabe yalli ?akal-a the-students the-boy that ate.3sm-it 'Nada told us about the beating that the students know the boy that took it.' The facts discussed so far indicate that, for cases of relativization of certain adjuncts and idiomatic NP chunks, resumption obeys island conditions, such as the wh- clause constraint, the CNPC, and the adjunct clause constraint. That is, resumption in LA restrictive relatives appears to be selectively sensitive to islands depending on the nature of the relativized head. 64 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.3 A movement analysis of resumption Regarding the issue of island sensitivity in LA restrictive relatives involving weak resumptives, the picture looks as follows: (40) a. When relativizing from the complement position of V, N, or P, the relation between the relativized head and the resumptive element inside the relative clause violates island constraints, b. When relativizing adjuncts or idiomatic NP chunks, the relation between the relativized head and the resumptive element inside the relative clause obeys island constraints. One remark is in order here about the parallelism between adjuncts and idiomatic NP chunks, which has been discussed in the literature by Rizzi (1990).6 This author observes that whereas idiomatic NP chunks can be wh-moved on a par with other selected complements, extraction of idiomatic NP chunks from wh-islands deviate from that of complements. Thus we have the contrast between (41) and (42): (41) a. What headway do you think [t [you can make t on this project ]] ? b. What project do you think [t [ you can make headway on t ]] ? 6 Rizzi (1990) credits Cinque for the observation about the non-extractability of idiomatic NP chunks from strong islands. 65 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (42) a.* What headway do you wonder [how [PRO to make t on this project ]] ? b.? What project do you wonder [how [PRO to make headway on t ]] ? Rizzi further remarks that in Italian as well, relativization of idiomatic NP chunks out of wh-islands yields an unacceptable outcome. The minimal constrast between the expression prestare i soldi 'to lend money' and the idiomatic expression prestare attenzione 'to pay attention' illustrates this effect. (43) a. I soldi che ho deciso di prestare a Gianni sono molti Italian 'The money that I decided to lend to Gianni is a lot.' b. L'attenzione che ho deciso di prestare a Gianni & poca 'The attention that I decided to pay to Gianni is few.' c. I soldi che non ho ancora deciso a chi prestare sono molti 'The money that I haven't decided yet to whom to lend is a lot.' d.* L'attenzione che non ho ancora deciso a chi prestare & poca 'The attention that I haven't decided yet to whom to lend is few.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Thus, extraction of idiomatic NP chunks from wh-islands is strongly disallowed; it is on a par with the extraction of adjuncts:7 7 Additional evidence that idiomatic NP chunks in LA differ from selected arguments comes from the fact that a typical object of the verbs ?axad 'take', ?akal 'ate', or taTma 'fed' can be questioned using the wh-words fu 'what' or ?ayya NP 'which NP', as illustrated in the (a) versions of (i-iii) below: (i) a. fu / ?ayya bluze ?axadte what/ which shirt took.2sf 'what/which shirt did you take?' b. kam vatta ?axadte 1-yom how.many nap take.2sf the-day 'How many naps did you take today?' (ii) a. Ju/ ?ayya gato ?akalte w hat/ which cake ate.2sf 'What/which cake did you eat?' b. kam ?atle ?akalte ?inte w zriire how.many slap ate.2sf you and young 'How many slaps did you take when you were young?' (iii) a. ju / ?ayya tabxa taTmat-kun Salwa what/ which dish fed.3sf-you.p Salwa 'What/which dish did Salwa feed you?' 67 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (44) a.?? Who do you wonder whether we believe we can help t? b * How do you wonder whether we believe we can help Bill t? (45) Italian a.? Che problem non sai [come [potremo resolvere 11] ] ? 'Which problem don't you know how we could solve?' b * Come non sai [che problema [ potremo risolvere 11 ] ] ? 'How don't you know which problem we could solve?' (46) Lebanese Arabic a.?? ?ayya mijkle baddak ta'irif [kiif [rati ntiill 11 ] ] ? which problem want.2sm know.2sm [how [fut.solve.lp 11 ] ] 'Which problem do you want to know how we will solve?' b. kam ?atle taTmat-kun Salwa ?into w zraar how.many slap fed.3sf-you Salwa you.p and young.p 'How many slaps did Salwa give you when you were young?' When the verbs ?axad 'take', ?akal 'ate', or taTma 'fed', which enter into the idiomatic expressions Tnxad ratta 'take a nap', ?akal ?atle 'take a slap', or taTma ?alle 'give a slap', occur with the wh-word fu 'what', the idiomatic meaning is not available anymore. The only possible way to question the idiomatic NP chunk in those expressions is by using kam 'how much/many', as illustrated in (ib-iiib). 68 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b * kiif baddak taTrif [?ayya mijkle [rati rihill 11 ] ]? how want.2sm know.2sm [which problem [fut.solve.lp 11 ] ] 'How do you want to know which problem we will solve?' The argum ent/ adjunct asymmetry with respect to island sensitivity observed in LA restrictive relatives involving weak resumptives (recall (40)) is a typical property of A-dependencies created by movement. To account for it, I propose the following: whereas a relativized argument can establish its connection to the relativized site via binding, an adjunct (or an idiomatic NP chunk) can only establish its connection to the relativized site via movement. That is, whereas the relativized argument can bind a pronoun inside an island, the relativized adjunct or idiomatic NP must be related to a trace of A-movement inside the island: (47) a. Antecedent..........[island [Argument pro ] ] b.* Antecedent [island [Adjunct/Idiomatic NP t ] ] As mentioned earlier, (47a) is a typical representation given for resumptive constructions which are not island sensitive; the antecedent is related to a pronoun in the variable position of an A-construction. I am suggesting that such a representation is not available in cases of relativization of adjuncts (or idiomatic NP chunks), even in the presence of a weak pronominal element in the relativization 69 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. site. Since islands prohibit A-extraction, the ungrammatically of sentences with the representation in (47b) is explained.8 This leads to the conclusion that the appearance of a weak pronominal element in restrictive relatives in LA is not always an indication of the absence of movement in the generation of those constructions. If this analysis is to be pursued, several questions need to be answered: (1) Why is the binding strategy (47a) unavailable for the relativization of adjuncts and idiomatic NP chunks, (2) What happens outside island contexts, i.e. is the movement strategy available for argument relativization in non-island contexts?, and (3) What is the status of the weak "resumptive" element in restrictive relatives in LA? Regarding question (1), what manner adjuncts and idiomatic NP chunks have in common is that they are non-referential noun phrases, and I understand referential noun phrases to be noun phrases that are associated with a presupposition of 8 Note that the argument for movement involving idiomatic NP chunks does not depend or rely on how these elements are interpreted. Specifically, the adjacency requirement imposed on the idiomatic NP chunk and the verb it occurs with does not play a role here. AH it is important to observe is the contrast in acceptability between cases where the idiomatic NP chunk is related to a weak resumptive element outside island contexts and those where the weak resumptive element is inside an island. 70 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. existence.9 A characteristic property of referential noun phrases is that they can easily enter coreference relations with pronouns in discourse: (48) a. ?arrit 1-m'i'allme 1-kteeb la-ttleemiiz. w ?ana read.caus.3sf the-teacher.fs the-book to-the-students and I kameen ?riit-o too read-it 'The teacher made the students read the book. And I too read it.' 9 It is obvious that definite noun phrases carry such a presupposition, but some sentences with indefinite noun phrases can carry such a presupposition as well (Partee 1972): (i) John succeeded in marrying a girl his parents didn't approve of. Cinque (1990) proposes that the only elements that can undergo long wh-movement are those that are used strictly referentially, referentiality being equated with the ability to refer to a specific member of a preestablished set. Clearly, Cinque's notion of referentiality implies a presupposition of existence. As Cinque (1990) himself points out, his definition of referentiality recalls Pesetsky's (1987) notion of D- linking. In fact, D-linking implies referentiality in this case as well. 71 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. Sami Tam bi-fattij Tala seeTa dahab w Laila Tam Sami Asp. look.3sm on watch gold and Laila Asp. bi-tfattif Talay-a kameen look.3sf on-it to 'Sami is looking for a gold watch and Laila is looking for it too.' c. A: ?ayya kteeb ?iryo ttleemiiz? Which book read.3p the-students 'Which book did the students read?' B: ma baTrif, basl-mTallme ?aalit ?9nno ma Neg. know.ls, butthe-teacher.fs said.3sf that neg. tiabbu-u like.3p-it 'I don't know, but the teacher said that they didn't like it.' A definite noun phrase, an indefinite noun phrase interpreted referentially, and a which phrase can all be coreferent with a pronoun in discourse. We have already observed that idiomatic NP chunks cannot be coreferent with pronouns in discourse (the relevant examples are repeated in (49)): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (49) a* Nada 3 arrabit ttaVme Nabil ?atle mbeerih bas Laila Nada tried.3sf feed.3sf Nabil beating yesterday but Laila she that managed.3sf feed.3sf-him one/it 'N ada tried to give Nabil a beating yesterday but it was Laila who managed to give one/it to him.' b.* Laila bifhaddir Tayta la-t-tlemiiz kail yom w L. prepare.3sf reprimand to-the-students every day and Nada feed.3sf one/it to-the-employees too 'Laila prepares a lecture for the students every day and Nada gives them one too.' (50) * Sami ?a%ad Yatta baTd 1-Yada bas Laila ma Sami took.3sm nap after the-lunch but Laila neg. ?axadit wahde/-a took.3sf one/-it 'Sami took a nap after lunch but Laila didn't take one/it.' The same generalization holds of the manner adjuncts discussed in section 2.2.2.I. Thus, whereas, the first part of the sentence in (51) is totally acceptable in isolation, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the whole discourse in (51), with the relevant interpretation of the pronoun in the second sentence, is unacceptable: fnazeeha 'integrity' sdrfa 'speed' byiftiYil Sami bi- < narvaze 'nervousness'. w kail works.3sm Sami with fn^eefa 'courage' and all Jiibit ?alb 'kindness' JJaYYiile byijthrlo fiy-a kameen worker.p work.3p with-it too 'Sami works with integrity/speed/nervousness/courage/kindness. And all the workers work with it too/ My claim here is that only referential NPs, which can corefer with pronouns in discourse, can be resumed by a weak pronoun in island contexts.1 0 This provides us 1 0 Cinque (1990) argues that the ability of an NP to corefer with a pronoun in discourse is correlated with the ability to clitic-left dislocate this NP in Italian, clitic-left dislocation being a resumptive construction. Note that the claim I make in the text is different from that of Cinque's in that the correlation 1 establish between referentiality and the ability to be related to a resumptive pronoun involves island contexts only. My claim is therefore consistent with the ability to clitic-left dislocate 74 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. idiomatic NP chunks and measure phrases in Italian (i) and the fact that neither expression can be related to a pronoun in discourse (ii). (i) a. 70 chili, non li pesa. '70 kilos, he does not weigh them.' b. Giustizia, non la fara mai. 'Justice, he will never do it.' (ii) a. Speaker A: Speaker B : b. Speaker A: Speaker B : Io peso 70 chili 'I weigh 70 kilos.' *Anch'io li peso 'Even I weigh them.' Fara giustizia. 'He will do justice.' *Anch'iola faro. 'I will do it too.' The proposal seems to run into problems however when it comes to clitic-left dislocation of bare quantifiers. In that context, Cinque notes a contrast between quantifiers that are specific and those that are non-specific. As (iii) illustrates, non-specific quantifiers are not compatible with resumptive clitics: (iii) a. Qualcuno *(lo) troveremo Someone *(him) will-find 'Someone (specific), I will find him.' 75 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. with the reason why the binding strategy is not available for the relativization of abstract nouns and idiomatic NP chunks (Question (1)). Giving full answer to question (2) (What happens outside island contexts, i.e. is the movement strategy available for argument relativization in non-island contexts?) is the object of chapter 3, and a full answer to question (3) (What is the status of the weak "resumptive" element in restrictive relatives in LA) is given in chapter 5. But anticipating the result of the discussion there, the conclusion will be that weak resumptive elements in LA restrictive relatives are the overt realization of an agreement relation between a head and a noun phrase with a null D head. This will b. Qualcuno (*lo) troveremo Someone (*him) will-find 'Someone (or other), I will find him.' This seems to be incompatible with the data in (i) where a non-referential (non-specific) noun phrase has been clitic-left dislocated. I will follow Demirdache 1991 and assume that the unacceptability of (iiib) as a clitic-left dislocated construction has to do with the assumption that (non-specific) quantifiers must undergo QR and are subject to The Scope Constraint given in (iv): (iv) The Scope Constraint Do not adjoin a QP any higher than the lowest IP in which it originates. Under the uncontroversial assumption that the clitic in (iiib) marks the left edge of IP, the QP will not be able to take its scope without violating the scope constraint in (iv). 76 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. account for the fact that they seem to be able to cooccur with full noun phrases as well as null pronouns. 2.4 Evidence for traces in restrictive relatives in LA The discussion so far leads to two possible representations for restrictive relatives involving weak resumptive elements in Lebanese Arabic: one where the relativized head is related to a trace of A-movement inside the relative clause itself (52a)1 1 , and one where the relativized head is related to a null pronoun inside the relative clause, a pronoun whose features are identified by the weak resumptive element (WRE) it co-occurs with (52b). (52) a. Movement Strategy A ntecedent...................... WRE -t; b. Binding Strategy Antecedent;...................... WRE -pro; Since my claim has been that a movement strategy exists for restrictive relatives alongside a binding strategy, we expect that we can see further evidence for traces 111 am abstracting away from what it is exactly that moves in those constructions. I postpone the discussion on that issue to chapters 3 and 4. 77 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in those constructions. In a representation like (52a), the presence of a trace is obscured by the presence of the weak resumptive element. Therefore the expectation is that if one can isolate cases where movement has to be posited, but a weak resumptive element does not appear, those cases will be acceptable in LA. This kind of data is found in two different contexts: cases of relativization from the subject position of existential predicates and relativization of time adjuncts. 2.4.1 Relativization of subjects of existential constructions Consider the following sentence: (53) (kail) 1-katub lli keen fii Ta-ttaawle saaro (all) the-books that was in-it on-the-table become.3p ‘ Ta-rraff haUa? on-the-shelf now '(All) the books that there were on the table are now on the shelf.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. First, it must be noted that neither the particle fii (lit. 'in it') nor the PP predicate 'on the table' bear the necessary agreement features to identify a null subject pronoun.1 2 1 2 Null pronominal subjects in LA are generally licensed in the context of agreement morphemes. In the presence of the obligatory agreement morpheme on the verb, a subject can be dropped in matrix (iia) or embedded clauses (iib-c). (i) a. A: Ween Karim/ Laila/ 1-wleed where Karim/ Laila/ the-children 'where is/are Karim/Laila/the children?' (ii) a. B : Tam bi-?eebil/ bi-t-?eebil/ bi-?eebl-o 1-mudiira Asp. meet.3sm/ meet.3sf/ meet.3p the-principal.fs '(He/She/They) is/are meeting with the principal.' b. B : b-tiftikir 1-mTallme ?anno Tambi-?eebil/ bi-t-?eebil/ bi-?eebl-o think.3sf the-teacher.fs that asp. meet.3sm/ meet.3sf/ meet.3p 1-mudiira the-principal.fs 'The teacher thinks that (he/she/ they) is/are meeting with the principal.' c. B : rati tizTal 1-mTallme la?anno Tambi-?eebil/ bi-t-?eebil/ fut. upset.3sm the-teacher.fs because asp. meet.3sm/ meet.3sf/ meet.3p the-principal.fs 'The teacher will be upset because (he/she/they) is/are meeting with principal.' 79 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The sentences above illustrate the case of third person null pronouns, but the paradigm could be extended to include first and second person pronouns as well. From this discussion, one might surmise that the gap in the relativized subject position in (iii) is an instance of pro-drop, the null pronominal element being licensed by agreement on the verb seeTad 'help'. In fact, the data is consistent with such an analysis. The gap in subject position can appear in island contexts (iva-c). (iii) 3 amTiyyit ssaliib l-?afimar ?addamit hdiyye la-l-tiakiime H i organization the-cross the-red presented.3sfgift to-the-doctor.f that seeTad-it 3 arha 1-fiarb helped.3sf wounded the-war 'The organization of the Red Cross presented a gift to the doctor that (she) helped the wounded of the war.' (iv) a. Adjunct Island 3am< Tiyyit ssaliib l-?ahmar ?addamit hdiyye la-I-hakiime H i organization the-cross the-red presented.3sf gift to-the-doctor.f that t?a3 3abna la?anno/lamma seeTadit 3artha 1-harb surprised.lp because/when helped.3sf wounded the-war 'The organization of the Red Cross presented a gift to the doctor that we were surprised because/when (she) helped the wounded of the war.' 80 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Second, although a gap that corresponds to the subject of an existential construction can be found in embedded clauses (54), it cannot occur in island contexts (55). b. Wh-island 3 amTiyyit ssaliib I-?ahmar ?addamit hdiyye la-l-fiakiime H i organization the-cross the-red presented.3sf gift to-the-doctor.f that byaTrfo l-kall miin seeTadit wa?t 1-harb know.3pthe-all who helped.3sf time the-war 'The organization of the Red Cross presented a gift to the doctor that all know who (she) helped during war times.' c. Relative Clause 3amTiyyit ssaliib l-?atimar ?addamit hdiyye la-l-hakiime Hi organization the-cross the-red presented.3sf gift to-the-doctor.f that hakyo ssahaafiyye Tan 3artia 1-harb H i seeTadit-un talked.3p the-reporters about wounded the-war that helped.3sf-them 'The organization of the Red Cross presented a gift to the doctor that the reporters talked about the wounded of the war that (she) helped.' 81 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (54) a. (kali) 1-katub lli Tam t?uulo ?anno keen fii (all) the-books that asp. say.2p that was in-it Ta-ttaawle leezim yinhatto Ta-rraff on-the- table should put.3p on-the-shelf '(All) the books that you are saying that there were on the table should be put on the shelf.' b. 3 ammaTt (kail) 1-katub lli Tam t?uulo Tanno keen collected.ls (all) the-books that asp. say.2p that was fii Ta-ttaawle ■ i in-it on-the-table 'I collected (all) the books that you are saying that there were on the table.' (55) a. Adjunct island * (kail) 1-katub lli narfazto la?anno keen fii Ta-ttaawle (all) the-books that upset.2p because was in-it on-the-table leezim yinhatto Ta-rraff should put.3p on-the-shelf '(All) the books that you were upset because there were on the table should be put on the shelf.' 82 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. Wh-island * (kell) I-katub lli baddkun btaTrfo miin ?aal ?anno (all) the-books that want.2p know.2p who said.3sm that keen fii Ta-ttaawle leezim yinhatto Ta-rraff was in-it on-the-table should put.3p on-the-shelf '(All) the books that you want to know who said that there were on the table should be put on the shelf/ c. Relative clause * (kail) 1-katub lli baddkun taTrfo ssabe lli ?aal (all) the-books that want.2p know.2p the-boy that said.3sm ?anno keen fii Ta-ttaawle leezim yinhatto Ta-rraff that was in-it on-the-table should put.3p on-the-shelf '(All) the books that you want to know the boy who said that there were on the table should be put on the shelf.' In the presence of a gap that obeys island conditions, we have evidence that movement is indeed available for the generation of restrictive relatives in LA.1 3 1 3 Carlson 1977 refers to those types of constructions with the name amount relatives (AR). In AR, the relativized NP can only occur with a limited range of determiners or quantifiers, as the contrast in (i) and (ii) illustrates. 83 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (i) a. T h e people there w ere at that time only lived a few decates. b. That's a l l there is . c. {Wh a t, T h a t} lig h t th ere i s in this p a in tin g is q u ite diffuse. d. A n y beer there may b e left in that cooler is mine. e. E v e ry lion there is eats meat. (ii) a. *{Five, M ost, Several, M a n y} m en th ere w e re h ere d isa g re e d . b. *{Some, E a c h , A} man there w a s disagreed. Furthermore, wh-forms are banned from ARs. Thus, (iii) which involves a wh-relatve in place of a that relative or a zero relative contrasts sharply with (ib): (iii) *That's all which there is. However, as Grosu and Landmann (1998) point out, those relative clauses that involve relativization of the subject of existential constructions do not always have an amount reading (iv). (iv) I took with me every book that there was on the table. The sentence in (iv) does not mean that the speaker took as many books (from the library) as there were books (on the kitchen table). A sentence like (v) sounds odd despite the availability' of the amount reading. (v) Yesterday’, I spent the whole day drinking the wine that they spilled at the party. 84 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This conclusion is further confirmed by the fact that whereas a strong pronoun may optionally appear in the context of pro-drop, as illustrated in (56), it cannot appear in the sentences in (57), which involve relativization of the subject of an existential construction: (56) 3 am'Tiyyit ssaliib l-?ahmar Taddamit hdiyye la-l-ftakiime organization the-cross the-red presented.3sf gift to-the-doctor.f lli ?aalo ssahaafiyye Torino (hiyye) seeTadit 3 arha that said.3pthe-reporters that (she) helped.3sf wounded 1 -tiarb the-war 'The organization of the Red Cross presented a gift to the doctor that the reporters said that (she) helped the wounded of the w ar/ In the LA sentences given in (53-54) in the text, an amount reading does not adequately describe the interpretation of the sentences either. Those sentences imply an identity of substance (the actual objects) as well as an identity of amounts. I will therefore abandon the term amount relative in the text. 85 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (57) a. (kail) 1-katub 1 1 1 'Tam t?uulo ?anno keen fii (*hanne) (all) the-books that asp. say.2p that was in-it (*them) Ta-ttaawle leezim yinhatto la-rraff on-the-table should put.3p on-the-shelf '(All) the books that you are saying that there were (them) on the table should be put on the shelf.' b. 3 ainmac i ‘ t (kali) 1-katub Hi ‘ Tam t?uulo ?anno keen collected.ls (all) the-books that asp. say.2p that was fii (*hanne) Ta-ttaawle in-it (*them) on-the-table 'I collected (all) the books that you are saying that there were (them) on the table.' The subject position of existential constructions triggers definiteness effects and since personal pronouns are definite, the unacceptability of the sentences in (57) with a resumptive strong pronoun is not surprising. 86 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.4.2 Adjunct relatives Interestingly, adjunct relatives expressing time provide further evidence that the movement strategy is indeed available for restrictive relatives in LA.1 4 -1 5 (58) rah nuusal Ta-l-hafle seeTt lli bitballij l-musii?a fut. arrive.lp to-the-party hour that begin.3sf the-music 'We will arrive at the party the time that the music begins.' 1 4 However, this is only the case with adjunct relatives expressing time. Manner and place adjunct relatives require a weak resumptive element in the relativization site, which is attached to a designated preposition. 1 5 Larson (1985) observes, for English, that the headed relative clauses that have a trace in an adjunct position instead of a stranded preposition or a relative adverb are those whose head belongs to the class of bare-NP adverbs, NPs that can function adverbially without an accompanying preposition. This generalization also accounts for the difference noted between relativized time adverbs and relativized manner adverbs in LA. Only the former can occur as bare NPs: (i) a. waslo (bi-)see'Tata arrived.3p (in-)hour-this 'They arrived at this hour.' b. waslo (Ta-) sseeTa Tajra arrived.3p (on-) the-hourten 'They arrived at ten.' 87 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The gap in (58) can be identified as a trace of movement: the distribution of gaps in time adjunct relatives obeys subjacency and is sensitive to island conditions. The sentences in (59) and (60) are put forth to illustrate this point. (59) a. rah nuusal ^a-l-hafle seeTt lli btiftikro (?? ?anno) fut. arrive.lp to-the-party hour that think.2p (?? that) rah bitballij l-musii?a 0 fut. begin.3sf the-music 0 'We will arrive at the party at the time that you think (that) the music will begin.' b.?? rah na3e Ta-l-hafle seeTt lli bti?tin‘ T o ?enno fut. come.lp to-the-party hour that convince.2p that leezim tballij l-musii?a 0 should begin.3sf the-music 0 'We will come to the party at the time that you are convinced that the music should begin.' In (59), the time adjunct seeTa 'hour' can easily modifiy the higher verb btiftikro 'you think' inside the relative clause: under that reading, the time of arrival to the party coincides with the time the hearers think the music will begin, and not the time when the music will actually begin. The latter interpretation -which implies that 88 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the adjunct modifies the lower verb bitballif 'begins'- is only available in the absence of the complementizer ?snno 'that'. When the embedding verb does not allow complementizer deletion, as in (59b) bti?tin?o 'you are convinced', the reading that is readily available is the one where the adjunct modifies the higher verb: the time of arrival to the party coincides with the time the hearers are convinced that the music should start. This is what the double question marks indicate in (59b). Similarly, all the sentences in (60), which illustrate the relativization of the time adjunct seeTa 'hour' from various island contexts, are acceptable only under the reading where the adjunct modifies the higher verb inside the relative. The reading that forces the time adjunct to be interpreted inside the island is unavailable. (60) a. Wh-island. * wsolna Ta-l-hafle seeTt lli i sa?al Sami ?aza arrived.lp to-the-party hour that ask.3sm Sami whether keeno fallo ?ahl-o 0 were.3p left.3p parents-his 0 'We arrived to the party at the time that Sami asked whether his parents were gone.' 89 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. Adjunct island * wsalna Ta-l-hafle seeTt lli narfaz Sami la?anno arrived.lp to-the-party hour that upset.3sm Sami because keeno ?ahl-o fallo 0 were.3p parents-his left.3p 0 'We arrived at the party at the time that Sami got upset because his parents were gone/ c. Relative clause * wsalna Ta-l-fiafle seeTt lli keeno 1-bolisiyye Tam arrived.lp to-the-party hour that were.3p the-policemen Asp. yil?ato zzalame lli darab Maria 0 catching.Sp the-man that hit.3sm Maria 0 'We arrived to the party at the time that the policemen were catching the man that hit Maria.' The behavior of time adjuncts observed in (60) is typical in contexts of adjunct long- extraction. The facts above replicate what happens in questions: W7z-extraction of a tim e adjunct yields subjacency effects and is sensitive to islands. 90 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (61) a. ?eemta rati bitballij l-musii?a when fut. begin.3sf the-music 'When will the music begin?' b. ?eemta btiftikro (?? ?snno) rati bitballij l-musii?a 0 when think.2p (?? that) fut. begin.3sf the-music 0 'W hen do you think (that) the music will begin?' c.???eemta rati tiTtinTo ?anno leezim tballi) l-musii?a 0 when fut. convince.2p that should begin.3sf the-music 0 'When will you be convinced that the music should begin?' (62) a. Wh-island * ?eemta sa?al Sami ?aza keeno fallo ?ahl-o 0 when ask.3sm Sami whether were.3p left.3p parents-his 0 'W hen did Sami asked whether his parents were gone?' b. Adjunct island * ?eemta narfaz Sami la?anno keeno ?ahl-o fallo 0 when upset.3sm Sami because were.3p parents-his left.3p 0 'When was Sami upset because his parents were gone?' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. Relative clause * ?eemta keeno 1-bolisiyye 'i'am yil?ato zzalame when were.3p the-policemen Asp. catching.3p the-man lli darab Maria 0 i that hit.3sm Maria 0 'W hen were the policement catching the man that hit Maria?' As it turns out then, relativization of time adverbials parallels their extraction in the context of questions. The data in (59-60) can then be accounted for under the hypothesis that adjunct relatives involving gaps are generated by movement and, more specifically, that the gaps in those constructions are A'-traces. 2.5 Summary The picture looks as follows now: Restrictive relatives in LA can be given two possible representations ((52) is repeated below in (63)): (63) a. Movement Strategy Antecedent;..... b. Binding Strategy A ntecedent..... 92 WRE -ti WRE -proi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (63a) is available for sentences involving the relativization of non-referential noun phrases, which cannot be coreferent with pronouns, as well as the relativization of subjects of existential constructions and bare NP adverbs. All these constructions were shown to be sensitive to island conditions. (63b) is available for those sentences involving the relativization of referential noun phrases, which is never sensitive to islands.1 6 1 6 See Aoun 1986, Rizzi 1990, and Cinque 1991 for different approaches to A-extraction that incorporates the notion of referentiality of the extracted NP. For Rizzi 1990, the contrast between (ia) and (ib) below lies purely in the referentiality of the extracted wh-NP. (i) a.?? [Tayya siyyaara]i baddik taTrfe [kiif]j sallati h tj? which car want.2sfknow.2sf [how] fixed.3sm 'Which car do you want to know how he fixed?' b.* [kiif]j baddik taTrfe [Tayya siyyaara]i sallati h tj? how want.2sf know.2sf which car fixed.3sm 'How do you want to know which car he fixed?' In both sentences there are two chains headed by wh-elements and whose tail is a trace. The difference lies in the contrast between referential NPs and non-referential NPs: to be brief, whereas a referential NP can relate to its trace via binding of a referential index, a relation that is not subject to locality conditions, a non-referential NP can only relate to its trace via antecedent government, which is subject to strong locality conditions. This explains the unacceptability of (ib), since the relation between the adverbial wh-phrase fcii/'how' and its trace is intercepted by the wh-phrase ?ayya siyyaara 'which car'. In (ia) both the chain headed by hi/'how' and the one headed by Tayya siyyaara 'which car' are well- 93 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. formed, but the marginality of the sentence is the result of a subjacency violation. Note that in the analysis defended in the text the distinction between referential NPs and non referential NPs correlates with a difference in the nature of the variable they bind: a non referential NP can only bind a trace. This distinction is needed to account for the lack of subjacency effects in resumptive constructions and the contrast between (ia) and (ii): (ii) ?ayya siyyaara baddik taTrfe kiif sallafi-a t? which car want.2sfknow.2sf how fixed.3sm-it 'Which car do you want to know how he fixed it?' 94 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C hapter 3 Reconstruction and A-chains in Restrictive Relatives In the generative literature dealing with the syntax of restrictive relative clauses (see, e.g. Chomsky 1977, Carlson 1977, Cinque 1982, Borer 1984, Vergnaud 1985, Safir 1986, Browning 1987, Kayne 1994, Bianchi 1999, Ouhalla 1999), there is generally a consensus on the presence of movement in the generation of those constructions. One point at which the various analyses differ is the nature of the moving element: in some analyses it is a null operator coindexed with the external head (EH) of the relative via predication, in others it is the EH of the relative clause itself that is moved from the relativization site. In this chapter I provide evidence supporting the latter analysis for restrictive relative involving weak resumptive elements (WRE). The facts dealing with the reconstruction of the EH inside restrictive relative clauses in LA argue for treating the EH as part of the chain whose tail is the relativized position. Assuming reconstruction to be the property of movement chains (Chomsky 1995,1998), this will lead to the conclusion that the movement strategy is indeed available in the presence of WRE outside islands (see question (2) in Chapter (2)). 95 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.1 Scope reconstruction in LA restrictive relatives 3.1.1 Scope reconstruction with kail NP 'every NP' and Tiddit NP 'many NP' The following paradigm provides evidence for scope reconstruction in restrictive relatives in Lebanese Arabic: (1) a. daharit Nada tzabbit j-jarjaf yalli keeno haattiin-o Ta went-out.3sf Nada fix.3sf the-tablecloth that were.3p putting.3p-it on kali taawle b-l~3 nayne every table in-the-garden 'Nada went out to straighten the tablecloth that they had put on every table in the garden.' b. badda 1-mudiira tatTarraf Tala 1-mara yalli kail want.3sf the-director meet.3sf on the-woman that every mwazzaf 3 eeb-a maT-o Ta-l-tiafle employee.ms brought.3sm-her with-him to-the-party 'The director wants to meet the woman that every employee brought with him to the party.' 96 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. ma< i7uule yha??i?o maf 1-kam fiaaris yalli keeno possible question.3p with the-few guard that were.3p mwa?fiin-un ?adeem kali bineeye stood.3p-them in-front every building 'It is possible that they question the few guards that they stood in front of every building/ d. mudiir t-tahriir rah yi?ra l-ma?aaleen yalli kail director the-editing fut. read.3sm the-article.dual that every satiaafe katab-un Tan 1-tiarb ?abl ma i reporter wrote.3sm-them about the-war before neg tinjir-un l-3 ariide publish.3sf-them the-newspaper 'The editor-in-chief will read the two articles that every reporter wrote about the w ar before the newspaper publishes them.' To appreciate how scope reconstruction works in Lebanese Arabic restrictive relatives, let us focus on (la). Imagine the following context: Nada is hosting a dinner party in her garden. Before the dinner guests arrive, the wind blows and damages the table set up. Nada tries to put things back in order. In such a context, the most plausible reading for (la) is the distributive reading, the one where there is 97 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. a different tablecloth on each table in the garden, and Nada straightened each one of them. It is true that another reading is available for this sentence; a reading where one tablecloth covers all the tables in the garden, the non-distributive reading that is, but the pragmatic context clearly favors the distributive reading in this case. In fact, this description of the data carries over to all the sentences in (1). In all those sentences the distributive reading is pragmatically favored by the context. The sentences in (lc) and (Id) also show that the external head of the relative need not be in the singular for the distributive reading to obtain. Those sentences were provided to ensure that the external head of the relative clause is not interpreted as a kind noun in the distributive reading. To account for the availability of the distributive reading in each of (la-d) two analyses can be put forth: (i) either the universal quantifier phrase raises and takes scope over the external head itself (henceforth, long QR), or (ii) the distributive reading is the result of the universal quantifier taking scope over the relativized position, which is coindexed with the EH.1 1 Indirect Binding (Haik 1984, Doron 1982) contexts illustrate such a case: 98 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. While one may argue that the analysis in (i) may indeed be at work in some instances of what appears to be scope reconstruction (see footnote (1)), there is ample evidence to favor the latter analysis as the more descriptively adequate one. First, the availability of the distributive reading is sensitive to the position of the relativized site with respect to the universal quantifier: if the relativized site falls within the scopal domain of the universal quantifier, then the distributive reading becomes available, otherwise, only the non-distributive reading is available. (i) I-mara H i [kail ra3 3 eel]j Tazam-a waslit ?abI-[o]j the-woman that every man invited.3sm-her arrived.3sf before-him 'The woman that every man invited arrived before him.' In (i) a pronoun in the matrix clause prepositional phrase is interpreted as bound by the universal quantifier phrase inside the relative clause. This correlates with the fact that the universally quantified NP distributes over the relative clause external head. As the argument of this section develops, I will show that the cases illustrating scope reconstruction inside relative clauses differ from cases of indirect binding. 99 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (2) a. badda 1-mudiira titTarraf Tala 1-mara yalli btiftikir want.3sf the-director.fs meet.3sf on the-woman that think.3sf ?anno kail mwazzaf rafiyi3 e bi-ta?m Ta-l-hafle that every employee fut.come.3sm with-suit to-the-party 'The director wants to meet the woman that thinks that every employee will come in a suit to the party.' b. mudiirt t-tafiriir rati t?eebil s-safiaafiyyeen yalli editor.fs the-editing fut. meet.3sf the-reporter.dual that fakkaro ?anno kali ma?aal Tan 1-tiarb rati thought.3p that every article about the-war will yiT3ib-a please.3sm-her 'The editor in chief will meet the two reporters that thought that every article about the war will please her.' Both sentences in (2) have only the reading where the head of the relative takes scope over the quantifier phrase inside the relative. Notice that in both cases the relativized site corresponds to the higher subject position in the relative clause whereas the quantifier is in the embedded subject position. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. It can easily be demonstrated that the universal quantifier in the embedded clause cannot take scope or distribute over another quantifier in the subject position of a higher clause: in the sentences in (3) the subject of the higher clause does not co- vary with the universal quantifier, which occurs in an embedded subject position. (3) a. btiftikir mara ?enno kail mwazzaf rah yi3 e bi-ta?m think.3sf woman that every employee.ms fut.come.3sm with-suit Ta-l-hafle to-the-party 'A woman thinks that every employee will come in a suit to the party.' b. ftakaro sahaafiyyeen Torino kail ma?aal Tan 1-harb thought.3p reporter.dual that every article about the-war rah yiT3 ub mudiirt t-tahriir fut. please.3sm director.fs the-editing 'Two reporters thought that every article about the war will please the editor in chief.' Therefore one can account for the absence of the distributive reading in (2) by appealing to the unavailability of long distance Q(uantifier) R(aising), and to the fact that the relativized site doesn't fall in the scopal domain of the universal quantifier. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. It must then be the case that in sentences like (la) and (lc), where the relativized site does not fall within the c-command domain of the quantifier phrase in its surface position (4a-b), this site is within the scopal domain of the universal quantifier at LF (5a-b). (4) a. keeno haattiin kali Jarfaf * T a taawlt-een b-l-3 nayne were.3p putting.p every tablecloth on table-dual in-the-garden They had put every tablecloth on two tables in the garden.' b. keeno mwa?fiin kali tieeris ?addeem bineeyt-een were.3p stood.p every guard in-front building-dual 'They stood every guard in front of two buildings.' (5) a. keeno haattiin JarJaf-een Ta kail taawle b-l-3nayne were.3p putting.p tablecloth-dual on every table in-the-garden 'They had put two tablecloths on every table in the garden.' b. keeno mwa?fiin kam tieeris Taddeem kail bineeyee were.3p stood.p few guard in-front every building 'They stood a few guards in front of every building.' The availability of the distributive reading in (4) and (5) indicates that (i) the direct object position c-commands the the PP complement position and (ii) the direct 102 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. object postion falls within the scopal domain of the universal quantifier in the PP complement position in (5). It has been noted in the literature that the domain in which quantifiers like loll 'each/every' take scope is 'larger' than their s-structure c- command domain (see, e.g. May 1985, Aoun and Li 1993). The availability of the distributive reading in (5) is consistent with that conclusion. Again, in the case of (la) and (lc), as was the case for (lb) and (Id), long distance QR does not adequately capture the availability of the distributive reading. We can eliminate the option of long distance QR by embedding the universal quantifier phrase further inside the relative clause, thus taking it further apart from the external head of the relative clause.2 If the relativization site is within the scopal domain of the quantifier, the distributive reading remains available, as illustrated in (6). 2 Recall that, as illustrated in (3), even after undergoing quantifier raising the universal QP cannot scope over the subject of the embedding verb. 103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (6) a. kowyit Nada j-Jarjaf yalli ?alto ?onno leezim yhittu-u ironed.3sf Nada the-tablecloth that said.2pthat should put.3p-it ?a kali taawle b-l-3nayne on every table in-the-garden 'N ada ironed the tablecloth that you said they should put on every table in the garden.' b. m a^u u le yfia??i?o ma? 1-kam haaris yalli talab possible question.3p with the-few guard that asked.3sm r-ra?is ?anno ywa?fuw-un ?adeem kail bineeye the-president that stand.3p-them in-front every building 'It is possible that they question the few guards that the president asked that they make them stand in front of every building.' Both sentences in (6) allow the distributive reading and as we have observed in (5), the scopal domain of the universal quantifier in the embedded prepositional phrase includes the direct object of the embedded verb, namely, the relativization site. Another type of evidence that can be adduced against long distance QR as an account for scope reconstruction phenomena in restrictive relatives in LA comes from the behavior of other types of quantifier phrases. A case in point is that of the quantifier phrases Tidd.it NP 'many NP'. 104 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Tiddit NP 'many NP' has different scopal properties from the universal quantifier phrase kail NP 'every/each NP'; whereas a universal quantifier occurring within a prepositional phrase may arguably take scope over a clausemate subject (7a), this is not the case for Tiddit NP 'many NP' in (7b): (7) a. wee?fiin TarbaT tiarras ?addem kali bineeye standing.p four guards in-front every/ each building 'Four guards are standing in front of every building.' b. wee?fiin ?arbaT tiarras ?addem Tiddit bineeyeet standing.p four guards in-front many buildings 'Four guards are standing in front of many buildings.' The sentences in (7) constitute minimal pairs in that the only difference between them lies in the quantifier used inside the prepositional phrase. The scopal domain of Tiddit NP 'many NP' corresponds roughly to its s-structure c-command domain.3 With this in mind, consider the following contrasting sentences: 3 The qualification is needed because in a sentence like (i), Tiddit NP 'many NP' can take scope over the direct object. Of course, the final version of the statement depends on the s-structure position of the direct object in such sentences: 105 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (8) ma??uule yfia??i?o ma*r l-Tarba? tiarras yalli wee?fiin possible question.3p with the-four guards that standing.p ?addeem Tiddit bineeyeet in-front many buildings 'It is possible that they question the four guards that were standing in front of many buildings/ (9) a. badda Nada tjuuf 1-badle yalli Tiddit mwazzafiin want.3sf Nada see.3sf on the-suit that many employees lbsuw-a * T a 1-hafle wore.3p-her to the-party 'Nada wants to see the suit that many employees wore to the party/ (i) keeno mwa?fiin ?arbaT hirraas ?addeem Tiddit bineeyeet were.3p standing.3p four guards in.front.of many buildings 'They stood four guards in front of many buildings.' 106 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. m udiirt t-ta?iriir rah ti?ra l-ma?aaleen yalli director.fs the-editing fut. read.3sf the-artide, dual that wasso Tiddit sahaafiyye Talay-un commissioned.3p many reporters for-them 'The editor-in-chief will read the two articles that they commissioned from m any reporters.' Only sentences such as (9a-b) allow a wide scope reading for the QP Tiddit NP 'many NP' inside the relative. The distributive reading is not available for the sentence in (8). (8) has only a reading where the external head of the relative takes scope over Tiddit NP 'many NP'. More specifically, when the QP Tiddit NP 'many NP' is found in subject or object position and the external head is relativized from a lower position (9a-b), the quantifier phrase Tiddit NP 'many NP' can take scope over the external head. However, when Tiddit NP 'many NP' occurs in the complement position of a prepositional phrase, it cannot take scope over the external head when the latter is relativized from a subject position (8). Those facts relate to the contrast that can be observed between the absence of a wide scope reading for Tiddit NP 'many N P in (7b) and its availability in the sentences in (10). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (10) a. Tiddit mwazzafiin lbso badle Ta 1-ftafle many employees wore.3p suit to the-party 'Many employees wore a suit to the party.' b. wasso Tiddit sahaafiyye Tala ma?aleen commissioned.3p many reporters for article, dual 'They commissioned two articles from many reporters.' The inability of Tiddit NP 'many NP' to take inverse scope in (7b) indicates that the scopal domain of the quantifier phrase does not contain the subject position. This is what explains the absence of the wide scope reading in (8): Tiddit NP 'many NP' cannot take scope over the relativized site in subject position and therefore cannot scope over the external head of the relative clause. The crucial example is (9b), which corresponds to (10b): since Tiddit NP 'many NP' cannot take scope over a clausemate subject, long distance QR cannot explain the availability of the wide scope reading of the quantifier; although Tiddit NP 'many NP' cannot take scope over the subject, it is able to take scope over the external head of the relative. This can only be explained by the fact that the relativization site in (9b), like the PP complement in (10b), is in the scopal domain of the quantifier phrase Tiddit NP 'many NP'. 108 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. We have thus evidence that it is the relativization site that determines the scope interactions possibilities between a quantifier phrase inside the relative clause and the external head of that relative. First we have shown that even when we prevent a universal quantifier from directly taking scope over the external head of the relative clause, this quantifier can still distribute over the external head as along as the relativization site falls within its scopal domain. Then, by choosing a quantifier like Tiddit NP 'many N F whose inverse scope possibilities are very restricted, we observed that this quantifier phrase could still take scope over the external head, but only in cases where the relativization site is in the scopal domain of that quantifier.4 4 The fact that one can vary the quantifier phrases inside the relative and still get the reading where those quantifier phrases take scope over the external head of the relative clause is an indication that what I have called so far Scope Reconstruction is different from what Sharvit (1999a) refers to as Indirect Binding (see footnote 1). In fact, when Tiddit NP 'many NP' occurs inside a relative clause, it cannot bind a pronoun in the matrix clause: (i) 1-mara lli (?8lto ?8nno) Tiddit rjeel Tazamuw-a waslit ?abl-nn the-woman that (said.2p that) many men invited.3p-her arrived.3sf before-them 'The woman that you said that many men invited arrived before them.' The only way the matrix pronoun inside the prepositional phrase can be related to the quantified NP inside the relative clause is when the latter has the group reading. In that case, the external head of the 109 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. relative clause takes scope over the quantified NP and there is no binding relation between the quantified NP and the matrix pronoun inside the prepositional phrase. Furthermore, as Sharvit (1999a) correctly points out, a negative quantifier cannot replace the universal quantifier either in the context of Indirect Binding: (ii) 1-mara lli [wala r933eel]i Tazam-a waslit ?abl-[o]j the-woman that no man invited.3sm-her arrived.3sf before-him 'The woman that no man invited arrived before him.' Not only is the pronoun in the matrix prepositional phrase interpreted as a free variable in (ii), the distributive reading is absent as well. The external head of the relative clause must be interpreted as referring to a single woman in this case. In fact, negative quantifiers differ from quantified noun phrases like Tiddit NP 'many NP' in that they do not interact in scope with the external head of the relative clause in which they occur: (iii) 1-mara lli wala ra3 3eel Tazam-a dallit b-l-beet the-woman that no man invited.3sm-her stayed.3sf in-the-house 'The woman that no man invited stayed at home.' Even in the absence of a pronoun in the matrix clause, the external head of the relative clause in (iii) cannot be interpreted in the scope of the negative quantifier (see also footnote 9). 110 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.1.2 Scope reconstruction and island sensitivity So far the picture looks as follows: a quantifier phrase inside a restrictive relative can take scope over the external head of that relative clause by having the relativization in its scopal domain. We have seen that an analysis that involves the quantifier taking scope directly over the external head itself (long distance QR) is not descriptively adequate and cannot account for the range of data examined. One more step is required to show that the phenomena we have observed in the previous section are indeed due to scope reconstruction, i.e. to a 'lowering' of the external head to a position c-commanded by the quantifier phrase. In this section, I provide the evidence needed to reach this conclusion. Siding with Aoun and Choueiri 1997 and Aoun and Benmamoun 1998,1 will argue that the availability of scope reconstructed readings is correlated with the availability of movement in the generation of the restrictive relative constructions in LA. This will lead to the conclusion that the nature of the element occupying the relativized site is relevant for the ability of a quantifier phrase inside the relative clause to take scope over the external head of that relative: (11) a. EH [Relative clause.... QP .... WRE -Copy/ trace ] b. EH [Relative clause • • • • QP • • • ■ WRE-pro] 111 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In a representation like (lib), although the relativized site appears in the scopal domain of the quantifier phrase, the reading where the QP has scope over the external head of the relative is not available. This reading is only available in sentences that can have the representation in (11a). Within minimalist assumptions (Chomsky 1995), we can say that in order for a quantifier phrase inside a relative clause to be able to take scope over the external head of the relative, it must c- command a copy of that external head (this is equivalent to 'lowering' of the external head into the relative clause). Therefore we can conclude that the phenomenon described in section 2.1.1 is indeed one of scope reconstruction (see Romero 1998, Fox 1999). We have seen that in cases like (lb-d) and (6a-b) (the relevant examples are repeated below), the distributive reading is not only available, but contextually more salient. (1) b. badda 1-mudiira tatlarraf Tala 1-mara yalli kail want.3sf the-director meet.3sf on the-woman that every mwazzaf 3eeb-a maT-o Ta-l-hafle employee.ms brought.3sm-her with-him to-the-party 'The director wants to meet the woman that every employee brought with him to the party/ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. d. mudiir t-tahriir rati yi?ra l-ma?aaleen yalli kali director the-editingfut. read.3sm the-article.dual that every sahaafe katab-un ?an 1-tiarb ?abl ma % reporter wrote.3sm-them about the-war before neg tinjir-un l-3 ariide publish.3sf-them the-newspaper 'The editor-in-chief will read the two articles that every reporter wrote about the war before the newspaper publishes them/ (6) a. kawyit Nada J-JarJaf yalli ?alto ?anno leezim yhittu-u ironed.3sf Nada the-tablecloth that said.2pthat should put.3p-it 7a kali taawle b-l-3nayne on every table in-the-garden 'Nada ironed the tablecloth that you said they should put on every table in the garden/ b. ma??uule yfia??i?o ma? 1-kam tiaaris yalli talab possible question.3p with the-few guard that asked.3sm r-ra?is ?anno ywa?fuw-un ?adeem kail bineeye the-president that stand.3p-them in-front every building 'It is possible that they question the few guards that the president asked that they make them stand in front of ever}7 building/ 113 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Interestingly, when the relativized site occurs in an island context, the wide scope reading for the quantifier occurring inside the relative disappears:5 (12) Adjunct clauses a. badda 1-mudiira tatTarraf Tala 1-mara yalli naravazto want.3sf the-director meet.3sf on the-woman that upset.2p la?anno kali mwazzaf 3 eeb-a maT-o Ta-l-tiafle because every employee.ms brought.3sm-her with-him to-the-party 'The director wants to meet the woman that you were upset because every employee brought her with him to the party.' 5 The absence of the wide scope reading may be argued to be the result of the inability of the quantifier phrase to scope outside the island in (12-14). However, recall that we have already argued that the wide scope reading for the quantifier phrase inside the relative is not the result of long QR. In fact, the quantifier need not QR outside the island to take scope over a clausemate object, as shown by the availability of the distributive reading in sentences like (i) below: (i) rah yruu? 1-hawa ?abl ma yhitto JarJaf Ta kail taawle fut. calm.3sm the-wind before neg. put.3p tablecloth on every table 'The wind will calm down before they put a tablecloth on every table.' 114 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. mudiir t-tabriir ?aree l-ma?aaleen yalli sta?aal director the-editing read.3sm the-article.dual that resigned waziir baTd ma kail sabaafe katab-un Tan minister after neg. every reporter wrote.3sm-them about 1 -barb the-war 'The editor-in-chief read the two articles that a minister resigned after every reporter wrote them about the war/ c. kawyit Nada Harjaf yalli ?alto ?anno rab yruu? ironed.3sf Nada the-tablecloth that said.2pthat fut. calm.3sm 1-hawa ?abl ma ybittu-u Ta kail taawle b-l-3 nayne the-wind before neg. put.3p-it on every table in-the-garden 'N ada ironed the tablecloth that you said that the wind will calm down before they put it on every table in the garden.' 115 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. d. maS'Tuule ytia??i?o m a*? 1-kam haaris yalli sma^to ?anno possible question.3p with the-few guard that heard.2p that narvaz r-ra?is la?anno wa??afuw-un ?adeem kali upset.3sm the-president because stood.3p-them in-front every bineeye building 'It is possible that they question the few guards that you heard that the president was upset because they made them stand in front of every building/ (13) Wh-clauses a. leezim yitzabbat f-jarjaf yalli sma'i'to ?anno Nada badda should fixed.3sm the-tableclolth that heard.2p that Nada want.3sf ta*Trif miin hatt-o 'Tala kail taawle know.3sf who put.3sm-it on every table 'The tablecloth that you heard that Nada wants to know who put it on every table should be fixed.' 116 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. maTTuule yfia??i?o ma? 1-kam haaris yalli smaTto ?anno possible question.3p with the-few guard that heard.2p that r-ra?is baddo ya?rif miin wa??af-un ?adeem the-president want.3sm know.3sm who stand.3p-them in-front kail bineeye every building 'It is possible that they question the few guards that you heard that the president wants to know who made them stand in front of every building.' c. mudiirt t-tatirir rati ta?ra l-ma?aaleen yalli director.fs the-editing fut. read.3sf the-article.dual that sma?to ?anno wazir d-deexliyye baddo ya?rif heard.2p that minister the-interior want.3sm know.3sm ?aza kail satiaafe katab-un ?an 1-harb whether every reporter wrote.3sm-them about the-war 'The editor-in-chief will read the two articles that you heard that the minister of the interior wants to know whether every reporter wrote them about the war.' 117 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (14) Relative Clauses a. leezim yitzabbat J-Jarfaf yalli sma?to ?anno Nada badda tahke m a? J-jaYYiil y a lli h att-o ‘Tala k a il taaw le talk.3sf with the-worker that put.3sm-it on every table 'The tablecloth that you heard that Nada wants to talk to the worker who put it on every table should be fixed/ b. ma( i?uule yha??i?o ma'T 1-kam haaris yalli sma?to ?anno possible question.3p with the-few guard that heard.2p that r-ra?is baddo yatike ma? l-mas?uul yalli the-president want.3sm talk.3sm with the-person-in-charge that wa??af-un ?adeem kail bineeye stand.3p-them in-front every building 'It is possible that they question the few guards that you heard that the president wants to talk to the person in charge who made them stand in front of every building.' 118 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. badda 1 -mudiira taHarraf ‘ Tala 1 -mara yalli btiftikro want.3sf the-director meet.3sf on the-woman that think.2p ?anno rah tkuuno b-l-h afle yalli kali mwazzaf t i that fut. be.2p in-the-party that every employee.ms 3 eeb-a maT-o Talay-a brought.3sm-her with-him to-it 'The director wants to meet the woman that you think that you will be at the party where every employee brought her with him.' What the sentences in (12-14) indicate is that when the external head is related to a WRE inside an island, a universal quantifier inside the relative clause is prevented from taking scope over the external head. The only reading available for sentences like the ones in (12-14) is one where the external head takes scope over the universal quantifier. The following representation encodes this generalization: (15) EH [Relativeclause •••• [island ••• Q P... WRE-pro ...]...] 0 K EH > QP, * QP > EH A similar argument can be constructed for the quantifier phrase Tiddit NP 'many NP'. In sentences like (9a-b), which I repeat below, it was observed that the quantifier inside the relative can also take wide scope with respect to the EH of the relative. 119 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (9)a. badda Nada tfuuf 1-badle yalli Tiddit mwazzafiin want.3sf Nada see.3sf on the-suit that many employees lbsuw-a 7 a 1-hafle wore.3p-her to the-party 'N ada wants to see the suit that many employees wore to the party.' b. mudiirt t-tatiriir rah ti?ra l-ma?aaleen yalli director.fs the-editing fut. read.3sf the-article.dual that wasso Tiddit sahaafiyye Talay-un commissioned.3p many reporters for-them 'The editor-in-chief will read the two articles that they commissioned from many reporters.' However, when the relative clause external head is related to a WRE inside an island, the wide scope reading for the quantifier Tiddit NP becomes unavailable. Again, when the relativized site is occupied by a pronoun, only the reading where the external head of the relative clause takes wider scope is available: 120 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (16) A djun ct clauses a. badda Nada tjuuf 1-badle yalli naravazto la?anno ‘ Tiddit want.3sf Nada see.3sf the-suit that upset.2p because many mwazzafiin labsuw-a Ta-l-tiafle < t employees wore.3sm-it to-the-party 'Nada wants to see the suit that you were upset because many employees wore it to the party/ b. mudiirt t-tahriir rati ti?ra l-ma?aaleen yalli director.fs the-editing fut. read.3sf the-article.dual that sta?aal waziir lamma wasso Tiddit resigned minister when commissioned.3p many satiaafiyye Talay-un reporters for-them 'The editor-in-chief will read the two articles that a minister resigned when they commissioned them from many reporters.' 121 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (17) W li-clauses a. badda Nada tfuuf 1-badle yalli badkun taTrfo *Ta-?ayya want.3sf Nada see.3sf the-suit that want.2p know.2p to-which fiafle Tiddit mwazzafiin lebsuw-a i i party many employees wore.3sm-it 'N ada wants to see the suit that you want to know to which party many employees wore it/ b. mudiirt t-tafirir rati ta?ra l-ma?aaleen yalli director.fs the-editing fut. read.3sf the-article.dual that wazir d-deexliyye baddo yaTrif ?9za wasso minister the-interior want.3sm know.3sm whether commissioned.3p Tiddit satiaafiyye Talay-un many reporters for-them 'The editor-in-chief will read the two articles that the minister of the interior wants to know whether they commissioned them from many reporters.' 122 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (18) Relative Clauses a. badda Nada tjuuf 1-badle yalli rah tkuuno want.3sf Nada see.3sf the-suit that fut. be.2p b-l-?isti?beel yalli Tiddit mwazzafiin njabaro yilbsuw-a fi-i in-the-reception that many employees forced.3p wear.3p-it in-it 'Nada wants to see the suit that you will be at the reception that many employees were forced to wear it in it/ b. mudiirt t-tafirir rati ta?ra l-ma?aaleen yalli director.fs the-editing fut. read.3sf the-article.dual that tiake wazir d-deexliyye m at l-mas?uhin yalli talked.3sm minister the-interior with the-responsible.p that wasso ‘ iiddit satiaafiyye Talay-un commissioned.3p many reporters for-them 'The editor-in-chief will read the two articles that the minister of the interior spoke with the authorities that commissioned them from many reporters.' To allow the reader to appreciate the contrast between the sentences in (9) and those in (16-18) let me expand on the minimal contrast between (9b) and (16b) for instance. Consider the following scenario: 123 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (19) From a pool of reporters, many were chosen to write two articles on one given topic (as opposed to one article, for example). It will be up to the editor-in-chief to choose which article will be published in the newspaper. In order to do that, she will read each of the two articles that those reporters will submit. In (9b), under the context developed in (19), it is clear that the editor-in-chief will be reading more than two articles before making her decision. This interpretation corresponds to the wide scope reading of the quantified phrase Tiddit NP 'many NP'. It is this reading that is absent in (16b). The only reading available in that sentence is one where there are only two articles for the editor-in-chief to read and many reporters have collaborated on the writing of those two articles. The facts discussed in this section suggest a parallelism between the availability of the two strategies in (20) for the generation of relative clauses in LA and the availability of scope reconstructed readings in those constructions (the results are summarized in (21)). (20) a. Movement Strategy E H .......................WRE-EH b. Base-generation Strategy E H .......................WRE-pro 124 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (21) a. QP>EH E H Q P WRE b. *QP>EH E H [isla n d Q P WRE It is only in the contexts where the movement strategy is available that a quantifier phrase inside a relative clause can take scope over the EH of that relative. This leads to the conclusion that this phenomenon is due to the reconstruction of the EH to a position that falls in the scopal domain of the quantifier phrase giving rise to what I have termed scope reconstructed readings. These readings are unavailable in island contexts (21b). This is due to the fact that in those contexts, the relative clause external head is related to a base-generated resumptive pronoun without forming a chain (20b). The external head of the relative in (21b) is thus interpreted outside the scopal domain of the quantifier. We are now in a position to provide an answer to the question raised in Chapter 2 as to the availability of the movement strategy outside island contexts in cases where the external head is referential.6 Since reconstruction is available in contexts like (21a) and since reconstruction is the property of movement chains, we can 6 Recall that in cases where the external head is non-referential, the movement strategy is the only strategy available. 125 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. conclude that the movement strategy is indeed available for the relativization of referential NPs outside of island contexts. This is what explains the wide scope reading of the quantifier phrase in (21a). 3.1.3 Scope reconstruction, amount quantification, and Principle C effects The ambiguity of sentences like the following is well known (see e.g., Longobardi 1987, Kroch 1989, Rizzi 1990, Cinque 1990, Heycock 1995, Fox 1999): (22) How many linguists did Bill decide to hire? One reading is the non-referential reading in which quantification is over amounts. This reading does not presuppose the existence of actual linguists applying for a job. It is enough for Bill to have made a decision about a number of linguists to hire, perhaps for the years to come. The other reading is the referential reading which presupposes the existence of a set of actual job applicants and the speaker is asking for the cardinality of that set. The two readings can be attributed to two different LF representations: it is argued that the non-referential reading involves reconstruction of the quantifier restriction to its base position (23a); the referential reading does not (23b). 126 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (23) a. [WH Q]? Bill decided to hire [Q linguists] b. [WH Q]? [Qx: linguist (x)] Bill decided to hire x Fox (1999) proposes to show the possible readings of the question in (22) by considering contexts where the two readings have different answers. Consider, for instance the following scenario: (24) Bill needs to hire linguists for a project he is working on. He decides to hire 50 linguists. However after a long day of interviews, he only finds 40 people who impress him and decides to hire them. For the remaining 10 he forms a search committee to look into the matter. Under the scenario presented in (24), there are two different answers for the question in (22). One answer is 40 and corresponds to the referential reading represented in (23b). The other answer is 50 and corresponds to the non-referential reading represented in (23a). Notice that under the representation in (23b), which corresponds to the reading where the existence of the actual job applicants is presupposed, the description linguist is not attributed to Bill, but to the speaker: it is outside the scope of decide and has a de re reading. On the other hand, under the representation in (23a), the description linguist is attributed to Bill and is within the scope of decide. This 127 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. distinction can help provide a clue to distinguishing the two possible readings of a question like the one in (25): (25) How many criminals did Bill recognize in the line-up? Consider the following two contexts: (26) a. The police is investigating a robbery. Bill was unknowingly a witness to that robbery: he saw the robbers as they were leaving in their car, which he noticed because it was making loud noises. However, he doesn't know that the passengers of that car were criminals. He is asked to pick the individuals he saw in that car in a line-up. b. The police is investigating a robbery. Bill was a witness of that robbery as he noticed the robbers leaving the premises with their guns and loot. He is asked to pick the robbers he saw in a line-up. Under the context in (a), the representation of the question corresponds to (27a), since it asks for the number of individuals who are criminals, such that Bill recognized those individuals. Bill therefore need not know that the individuals he picks out are actually criminals. Under the context in (b) however, the representation of the question corresponds to (27b), and the description criminal is attributed to Bill. 128 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (27) a. [WH Q]? [Qx: criminal (x)] Bill recognized x in the line-up b. [WH Q]? Bill recognized [Q criminals] in the line-up Interestingly, the ambiguity discussed above in the context of wh-questions can be found also in relative clauses: (28) 1-mfiaasib baddo yTeebil 1-kam tebbax yalli the-accountant want.3sm meet.3sm the-few cook that Tarrarit Nada t[aYYal-un b-l-matTam decided.3sf Nada hire.3sf-them in-the-restaurant 'The accoimtant wants to meet the few cooks that Nada decided to hire for the restaurant.' If the two readings available for the interpretation of kam tebbax 'few cooks' are indeed a matter of scope and if they are to be accounted for in terms of reconstruction, we expect these two readings to interact with principle C and they do (Heycock 1995, Romero 1998, Fox 1999, Sauerland 1998). Consider the following paradigm: 129 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (29) a. 1-kam kteeb yalli Picasso?arrar yi?raa-hun the-few book that Picassodecided.3sm read.3sm-them jammaT-un b-l-?uuda gathered.3sm-them in-the-room 'The few books that Picasso decided to read, he gathered them in the room.' b. 1-kam kteeb yalli Picasso?arrar yi?raa-hun the-few book that Picassodecided.3sm read.3sm-them ba*Td ma tiada katab-un still negsomeone wrote.3sm-them 'The few books that Picasso decided to read, noone has written them yet.' As seen in the case of (28) the relative clause head kam kteeb 'few books' has two possible interpretations and the contexts provided in (29) disambiguate the two readings. In one case (29a), the presence of the verb jamma? in the perfective presupposes the existence of actual books that Picasso decided to read: this is what Heycock calls the referential reading. This interpretation corresponds to a wide scope reading for the DP kam kteeb 'few books'. The narrow scope reading with respect to ?arrar is the one available in (29b). Since the books haven't been written yet, their existence cannot be presupposed. This interpretation corresponds to what 130 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Heycock calls the non-referential reading: Picasso has made a decision to read a certain number of books but hasn't found any title to put on his reading list. It has been shown in the context of how many questions, that this ambiguity is indeed a matter of scope reconstruction and that it is sensitive to principle C. The same holds in the context of relative clauses: under the non-referential reading, coreference between a name inside the EH and a pronoun inside the relative yields principle C effects.7 Only the referential reading bleeds condition C effects.8 This is illustrated in the sentences below: 7 Kroch (1989) argues that the referentiality requirement, which Rizzi (1990) uses to explain the resistance of wh-amount quantifiers to long movement, is a semantic and a pragmatic one. Thus the unacceptability of the sentence in (i) is due to the fact that its presupposition in (ii) is highly improbable in any given discourse. (i) * How much money was John wondering whether to pay? (ii) There was a sum of money about which John was wondering whether to pay it. In other words, questions involving long movement of wh-amounts are "unusable under most discourse circumstances" (Kroch 1989, p.8). This explanation, argues Kroch, has the advantage of accounting for the contrast between (i) and (iii) below: (iii) How many points are the judges arguing about whether to deduct? 131 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (iii) can be uttered felicitously in the context of a sport event where fixed point deductions are specified on various mistakes. Note that the presupposition that corresponds to the question in (iii) is plausible in this context as well: (iv) There is a number about which the judges were arguing whether to deduct that number. It is difficult to see how Kroch's account can carry over to relative clauses involving the relativization of amount quantifiers, which seem to display the same resistance to long movement. Furthermore, under the semantic/pragmatic account, the interaction with Principle C of the binding theory is not expected. 8 The absence of principle C effects in the context of restrictive relatives has been discussed in the literature (see van Riemsdijk and Williams 1981, Munn 1994, Safir 1999, Lebeaux 1992). We will take up this issue in section 2.3. It suffices to say however that cases where reconstruction of the head has to be posited do exhibit principle C effects (see also Sauerland 1998): (i) a* 1-kam kteeb * T a n ?8ssit tiayeet Picasso, yallipro, ?aal ?anno keen the-few book about story life Picassoi that pro, said.3smthat was fii b-l-maktabe x taf° in-it in-the-library disappeared 'The few books about the story of Picasso's life that he said that there were in the library disappeared.' 132 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (30) a. 1-kam kteeb Tan ?assit Tiayeet Picasso yalli ?arrar the-few book about story life Picassothat decided.3sm yi?raa-hun jammaT-un b-l-?uuda read.3sm-them gathered.3sm-them in-the-room 'The few books about the story of Picasso's life that he decided to read, he gathered them in the room.' b.* 1-kam kteeb Tan Tassit i i Tiayeet Picasso yalli ?arrar the-few book about story life Picasso that decided.3sm yi?raa-hun baTd ma Tiada katab-un read.3sm-them still neg someone wrote.3sm-them 'The few books about the story of his life that Picasso decided to read, noone has written them yet.' b. 1-kam kteeb Tan Tassit ftayeet-o, yalliPicasso, ?aal ?8nno keen the-few book about story life-his, that Picasso, said.3smthat was fii b-l-maktabe x taf° in-it in-the-library disappeared 'The few books about the story of his life that Picasso said that there were in the library disappeared.' I take it that reconstruction is forced in the context of existential sentences like (i) (see Heim (1987)). One can easily account for the contrast between (ia) and (ib) above in terms of a principle C violation in (ia) but not in (ib), under reconstruction. 133 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. As expected, the sentences in (31) where the name appears inside the relative are both acceptable under the coreference reading. (31) a. 1-kam kteeb Ian Tessit hayeet-o yalli Picasso the-few book about story life-his that Picasso Tarrar yi?raa-hun jammal-un b-l-?uuda decided.3sm read.3sm-them gathered.3sm-them in-the-room 'The few books about the story of his life that Picasso decided to read, he gathered them in the room.' b. 1-kam kteeb *ian ?assit tiayeet-o yalli Picasso the-few book about story life-his that Picasso ?arrar yi?raa-hun baTd ma tiada decided.3sm read.3sm-them still neg someone katab-un wrote.3sm-them 'The few books about the story of his life that Picasso decided to read, noone has written them yet.' The availability of the two readings discussed in the context of amount quantification can be shown to be sensitive to islands as well: When the 134 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. relativization site occurs within an island, the ambiguity disappears and only the referential reading is available (see e.g., Kroch 1989, Rizzi 1990, Cinque 1990): (32) a. Adjunct Clause 1-mhaasib baddo y?eebil 1-kam tabbax yalli fa tab it the-accountant want.3sm meet.3sm the-few cook that open.3sf Nada 1-matTam ?abl ma ?arrarit tJaYYal-un fi-i Nada the-restaurant before neg. decided.3sf hire.3sf-them in-it 'The accountant wants to meet the few cooks that Nada opened the restaurant before she decided to hire them for it/ b. Wh-Clause 1-mhaasib baddo y?eebil 1-kam tabbax yalli baddkun the-accountant want.3sm meet.3sm the-few cook that want.2p ta*Trfo ?aza Nada ?arrarit tJaYYal-un b-l-matTam know.2p whether Nada decided.3sf hire.3sf-them in-the-restaurant 'The accountant wants to meet the few cooks that you want to know whether Nada decided to hire them for the restaurant.' 135 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. Relative Clause 1 -mfiaasib baddo y?eebil 1-kam tabbaxyalli fa tab it the-accountant want.3sm meet.3sm the-few cook that open.3sf Nada l-mat?am yalli ?arrarit tjaw al-un fi-i Nada the-restaurant that decided.3sf hire.3sf-them in-it 'The accountant wants to meet the few cooks that Nada opened the restaurant that she decided to hire them for it.' The sentences in (32) are to be contrasted with the one in (28) where both the referential reading, as well as the non-referential reading, is available. This is an anticipated result in view of the representation given for restrictive relative clauses involving WRE in islands and the discussion in Chapter 1 regarding the relativization of non-referential NPs. If a relativized external head is related to a WRE inside an island, then it must be the case that this WRE co-occurs with a null pronoun. The latter cannot resume a non-referential NP and thus the interpretation of the sentences in (32) is accounted for. Under the two proposals I have argued for so far -namely, that there are two strategies for the generation of restrictive relatives and that scope interaction between a quantifier phrase inside the relative and the EH of the relative is the result of the movement strategy, the behavior of amount quantifiers in relativization can be easily explained without further assumptions. This provides further support for the analysis defended here. 136 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.2 Reconstruction and bound variable anaphora The logic of the analysis is clear so far. The availability of scope reconstruction in restrictive relatives correlates with the availability of movement: if the external head is related to a (null) resumptive pronoun inside an island, it cannot be interpreted within the scope of a quantifier inside the relative clause itself and it cannot have a non-referential reading. When movement is possible, the scope reconstructed reading of the external head is available, as well as the non-referential reading. The presence of Principle C effects and island sensitivity show that the phenomena observed so far are indeed the property of movement derivations. This section presents an additional argument from bound variable anaphora. Consider the following sentences:9 9 It has been noted (see Sharvit 1999 and footnote 2) that negative quantifiers, unlike other types of quantifiers which may appear inside relative clauses, cannot take scope over the external head of the relative clause itself. However, as (33b) illustrates, a negative quantifier can bind into the external head of a relative clause, indicating that the availability of scope reconstruction and reconstruction for binding do not go hand in hand for negative quantifiers (see Szabolcsi 1997 for a similar observation regarding the exceptional behavior of negative quantifiers). 137 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (33) a. Taleemit fatis-[o]i yalli ?arrarit 1-mTallme ?anno grade exam-[his]ithat decided.3sf the-teacher.fs that tfee3 i? [kali talmiiz], fiy-a rati tatlaT bukra surprise.3sf [every student], in-it fut. come-out.3sf tomorrow 'The grade of his exam that the teacher decided to surprise every student with will be ready tomorrow/ b. Ja3 areet 3 naynt-[o]i yalli Nada btaTrif ?anno trees garden-[his]i that Nada know.3sf that [wala 3 nayneete]i byahmal-un ballajo yzahhro no gardner neglect.3sm-them started.3p bloom.3p 'The trees of his garden that Nada knows that no gardner neglects started to bloom.' In both (33a) and (33b) the pronoun contained within the external head can be bound by the quantificational phrase within the relative clause. Notice that in both instances the quantifier inside the relative clause occurs in an embedded clause. We have already seen that a universal quantifier embedded inside the relative clause cannot take scope over the EH unless the latter reconstructs inside the relative 138 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. clause.1 0 Negative quantifiers like wala j nayneete 'no gardner do not take inverse scope (Szabolcsi 1997). Therefore the availability of the bound reading in both (33a) 1 0 In (i) below, the universal quantifier cannot take scope (distribute) over the matrix subject. Thus, in (33a), the bound reading of the pronoun within the external head cannot be the result of long QR. (i) ?arrarit mTallme ?anno tfee3i? [ksll talmiiz], bi-< Talcemt-[o], decided.3sf teacher.fs that surprise.3sf [every student], in-grade-[his], 'A teacher decided to surprise every student with his grade.' Furthermore, an account that assumes that the reading where the quantifier inside the relative takes scope over the external head is derived by long QR would predict that sentences like (iia-b) would display WCO effects, as is the case in (iiia-b). This is however contrary to fact. (ii) a. Taleemit fafis-[o], yalli fee3a?it 1-mTallme [kail talmiiz]i grade exam-[his]i that surprised.3sf the-teacher.fs [every student] j fiy-a talTit mbeerih in-it came-out.3sf yesterday 'The grade of his exam that the teacher surprised every student with came out yesterday.' b. Ja3 areet 3 naynt-[o]i yalli[wala 3 nayneete]j byahmal-un balla]o trees garden-[his]j that no gardner neglect.3sm-them started.3p yzahhro bloom.3p 'The trees of his garden that no gardner neglects started to bloom.' 139 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and (33b) can be traced to the availability of reconstruction of the EH inside the relative. In the sentences above, the relativization site falls in the c-command domain of the quantifier and therefore, the pronoun contained within the external head can be bound.1 1 (iii) a.?? Taleemit fahs-[o]i fee3a?it [kail talmiiz]i grade exam-his surprised.3sf every student 'His grade surprised every student.' b.?? [a3 areet 3naynt-[o]i ma Ta3 abo [wala 3 nayneete]j trees garden-[his]i neg. please.3p [no gardnerjj 'The trees of his garden pleased no gardner.' 1 1 Thus, when the quantifier inside the relative does not c-command the relativization site, the bound variable reading becomes unavailable: (i) a* Taleemitfahs-[o]i yalli?arrarit 1-mTallme Ta-?asees-a ?anno grade exam-[his]j that decided.3sf the-teacher.fs on-basis-its that tfee3 i? [kali talmiiz], bi-hdiyye tlTit mbeerih surprise.3sf [every student]jwith-gift come-out.3sf yesterday 'The grade of his exam on the basis of which the teacher decided to surprise every student with a gift came out yesterday.' 140 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This analysis is confirmed by looking at relativization from island contexts, where reconstruction is unavailable. Just as in the case of scope reconstruction, bound variable readings obtained through reconstruction are sensitive to islands: when the relativization site falls inside an island, the bound reading of the pronoun contained in the external head is no longer possible. b* Ja3areet 3 naynt-[o]i yalli?arrarit Nada Ta-?asees-un ?anno trees garden-[his]i that decided.3sf Nada on-basis-them that [wala 3 nayneete]i byistfii? zawde ballajo ymuuto no gardner deserve.3sm raise started.3p die.3p 'The trees of his garden on the basis of which Nada decided that no gardner deserves a raise started to die.' The sentences in (i) are unacceptable under the relevant reading indicated by coindexing the quantifier phrases and the pronoun within the external head. The only reading available for those setences is the one where the pronoun inside the external head is interpreted as a free variable that picks up its antecedent from the discourse. Even if the EH were to reconstruct inside the relative clause in (i), the bound reading of the pronoun within the EH would still be unavailable since the relativized site does not fall in the scopal domain of the quantifier phrase inside the sentence. 141 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (34) A dju n ct clauses a.* Taleemit fahs-[o]i yalli fallit 1-mTallme ?abl ma grade exam-[his]ithat left.3sf the-teacher.fs before neg. tfee3i? [kali talmiiz]i fiy-a tlTit mbeerih surprise.3sf [every student], in-it came-out.3sf yesterday 'The grade of his exam that the teacher left before she surprised every student with it was ready yesterday.' b.* Ja3 areet 3 naynt-[o], yalli nbasatit Nada la?anno trees garden-[his]i that was.pleased.3sf Nada because [wala 3nayneete]i byahmal-un ballajo yzahhro no gardner neglect.3sm-them started.3p bloom.3p 'The trees of his garden that Nada was pleased because no gardner neglects them started to bloom.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (35) Wh-clauses a.* ‘ Taleemit fahs-[o]i yalli Tam titsee?al 1-mTallme grade exam-[his]ithat Asp. wonder.3sf the-teacher.fs ?eemta bitfee3 i? [kail talmiiz], fiy-a tlTit when surprise.3sf [every student]i in-it came-out.3sf mbeerih yesterday 'The grade of his exam that the teacher wonders when she will surprise every student with it was ready yesterday.' b* Ja3 areet 3 naynt-[o]i yalli Tam titsee?al Nada leej trees garden-[his]i that Asp. wonder.3sf Nada why [wala 3 nayneete]i byihtamm fiy-un ballajo ymuuto no gardner cares.3sm for-them started.3p die.3p 'The trees of his garden that Nada wonders why no gardner cares for them started to die.' 143 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (36) R elative clauses Taleemit fahs-[o]i yalli btaTrfo 1 -mT allme yalli rah grade exam-[his]ithat know.3p the-teacher.fs that fut. tfee3 i? [kali talmiizji fiy-a tlTit mbeerih surprise.3sf [every student]i in-it came-out.3sf yesterday 'The grade of his exam that you know the teacher that will surprise every student with it was ready yesterday/ b.* Ja3 areet 3 naynt-[o]i yalli btaTrfo 1-mara yalli trees garden-[his]i that know.2p the-woman that ?aalit ?8nno [wala 3 nayneete]i byahmal-un ballajo said.3sf that no gardner neglect.3sm-them started.3p yzahhro bloom.3p 'The trees of his garden that you know the woman who said that no gardner neglects them started to bloom.' In all the sentences above, the pronoun contained within the external head must refer to an antecedent in the discourse; it has the interpretation of a free variable and cannot be related to the quantifier inside the relative clause. 144 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.3 Some consequences Restrictive relatives with weak resumptive elements in LA do not correspond to a unified class of representations. I have argued for two different representations so far: One involves movement (37a), in which case the external head of the relative clause is related to a copy/trace of movement in the relativization site. The other representation (37b) involves no movement and the external head in that case is related to a null pronoun in the relativization site, which is identified by a weak resumptive element.1 2 Thus it can be concluded that, in Lebanese Arabic, the presence of a WRE does not necessarily mark a base generated construction. (37) Movement strategy a. E H ............... WRE-EH Base generation strategy (Resumption) b. E H ................ WRE-pro We have seen that there are different contexts that disambiguate between these two different representations. In constructions involving relativization of non- 1 2 According to Shlonsky 1997, all Semitic 'clitics' are agreement markers. Here 1 adopt this analysis without argument on grounds of simplicity: it allows a unified analysis of weak resumptive elements in LA as agreement markers identifying either a trace of movement or a null pronoun. I defend this position in Chapter 5. 145 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. referential NPs, we can only have the representation in (37a), as evidenced by the island sensitivity that those constructions display (see Chapter 2). The acceptability of the relativization of referential NPs from island contexts is accounted for by the availability of the base generation strategy (37b). This chapter has provided further evidence for the representation in (37a). The scope reconstruction facts and their interaction with Binding Theory Principle C, as well as the ability of a quantifier inside the relative clause to bind a pronoun inside the external head of the relative have lead us to the conclusions summarized in (38): (38) a. The movement strategy is available for referential NPs alongside the base generation strategy. b. This movement is available outside island contexts. c. The element that moves is a copy of the external head itself. In what follows I examine three consequences of the analysis defended so far. First, I show how an analysis that assumes two different strategies for the generation of restrictive relatives accounts for the 'anti-reconstruction' effects with respect to Principle C that can be observed in those constructions (see Munn 1994, Safir 1999, Sauerland 1998). Second, in light of the facts regarding the effects of movement, which are discussed in this section, I contrast the copy theory of movement with the chain-binding account (see Barss 1986). Finally, I take up the issue of the interpretation of restrictive relatives in WRE in LA and I show how the dual 146 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. behavior of WRE in LA accounts for an interesting contrast in interpretation between LA restrictive relatives and Hebrew restrictive relatives. 3.3.1 Principle C and 'anti-reconstruction' effects in restrictive relatives One challenge to the movement analysis of restrictive relatives and in particular to the copy theory approach of the movement analysis is the observation that restrictive relatives, unlike wh-questions, do not exhibit reconstruction effects with respect to principle C of the Binding Theory. Munn (1994) observes that restrictive relatives1 3 do not exhibit an argument/adjunct asymmetry with respect to Principle C. Thus, in wh-questions, a name contained within the moved wh-phrase can obviate principle C only if it is contained inside an adjunct (39b), but not if it is inside an argument (39a). In restrictive relatives, no principle C effects can be observed (40) with arguments. (39) a.* [Which claim that John; was asleep] was hei willing to discuss [xohich claim that Johrii xoas asleep] ? b. [Which claim that John* made ] was he; willing to discuss [xohich claim] ? 1 3 as well as appositive relatives. 147 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (40) a. I have a report on Bobi's division he, w on't like, b* Which report on Bobi's division will hei not like? Following an analysis by Lebeaux (1992), Chomsky (1993) provides an account of the contrast in (39) based on three assumptions:1 4 (i) movement is copying, (ii) late adjunction is possible, and (iii) there is a preference principle for reconstruction. The two assumptions in (i) and (iii) account for the unacceptability of (39a). As indicated in (39a), movement of the wh-phrase leaves a copy in the base position, and the preference principle requires that the restriction of the wh-operator be interpreted in that position. This yields a principle C violation, as the name John will be c-commanded by the matrix subject pronoun he. In the context of (39b), the name is contained within an adjunct and by the assumption in (ii), the relative clause adjunct need not be present in the base position, as indicated in (39b). This allows the name John to escape a principle C violation. The facts concerning the relative clause in (40a) and the contrast between (40a) and (40b) seem to be questioning a raising analysis of restrictive relatives. If the same conditions on movement apply in relativization and wh-questions, we expect that (40a) would be unacceptable, as is (40b). But as Munn (1994) observes, this is 1 4 The contrast in (39) is attributed to van Riemsdijk and Williams (1981). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. contrary to fact.1 5 Munn's observation carries over to LA restrictive relatives involving WRE: (41) a. najaro l-mu?eebale ssutiufiyye \a-r-ra?iis lli published.3p the-interview the-magazine to-the-president that ?olto ?onno sa?alu-w Tann-a Ta-t-tele vision said.2p that asked.3p-/zzm about-it on-the-TV 'They published the magazine interview of the president that you said that they asked him about it on TV.' 1 5 See also Safir 1999, Sauerland 1998, and Sportiche 1999. Sportiche (1999) offers examples from French, which support Munn's observation: (i) a* Quelles photos de Jean veut-il vendre which photos of Jean want.3sm-he sell 'Which pictures of Jean does he want to sell?' b. Les photos de Jean qu'il veut vendre the photos of Jean that-he want.3sm sell 'The pictures of Jean that he wants to sell.' 149 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. najaro l-mu?eebale ssuhufiyye la-r-ra?iis lli published.3p the-interview the-magazine to-the-president that ?slto ?8nno sa?alo 1-mwazzafiin ?8ddeem-o *rann-a said.2p that asked.3p the-employees front-/izm about-it 'They published the magazine interview of the president that you said that they asked the employees about it.' c. najaro l-mu?eebale ssuhufiyye bi-maktab r-ra?iis pubUshed.3p the-interview the-magazine in-office the-president lli ?slto ?anno sa?alu-w Sann-a that said.2p that asked.3p-/zim about-it 'They published the magazine interview in the office of the president that you said that they asked him about it.' In aU the sentences in (41), the name embedded within the EH of the restrictive relative can be coreferent with a pronoun inside the relative clause itself, even when this pronoun c-commands the relativization site (marked by the presence of a WRE). No principle C violation is observed. This is in contrast to what we observe in wh- questions: 150 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (42) a* Tan ?ayya mu?eebale suhufiyye la-r-ra?iis ?alto about which interview magazine to-the-president said.2p ?anno sa?alu-u that asked.3p-him 'About which magazine interview of the president did you say that they asked him?' b. Tan ?ayya mu?eebale suhufiyye la-r-ra?iis ?alto about which interview magazine to-t)ie-president said.2p ?anno sa?alo 1-mwazzafiin ?addeem-o i i that asked.3p the-employeesfront-/zzrzz 'About which magazine interview of the president did you say that they asked the employees in front of him?' c. Tan ?ayya mu?eebale suhufiyye bi-maktab r-ra?iis about which interview magazine in-office the-president ?alto ?anno sa?alu-zz said.2p that asked.3p-him 'About which magazine interview in the office of the president did you say that they asked him?' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In the context of wh-questions, the argument/ adjunct asymmetry with respect to principle C surfaces in Lebanese Arabic, as it does in English (39) or French (see footnote 15). The sentences in (42) are equivalent to those in (39). In an analysis of restrictive relatives that does not assume the EH to be part of the same chain as the relativized site, it is easy to see how the contrast between restrictive relatives and wh-questions (see (40)) can be accounted for. If the EH does not participate in the same chain as the relativized site, the pronoun inside the relative clause, which c-commands the relativized site, cannot c-command the name embedded within the EH. Hence the construction does not incur a Principle C violation. Although the absence of principle C effects such as what we have in (40a) and (41) might lead one to question the raising analysis of relative clauses, we have observed in the course of this chapter that relative clauses do indeed display principle C effects in the context of reconstruction. Such are the cases of relativization of amount NPs. The relevant paradigm is repeated below: 152 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (43) a. 1-kam kteeb Tan Tassit tiayeet Picasso yalli ?arrar the-few book about story life Picassothat decided.3sm yi?raa-hun jamma*f-un b-l-?uuda read.3sm-them gathered.3sm-them in-the-room 'The few books about the story of Picasso's life that he decided to read, he gathered them in the room.' b * 1-kam kteeb Tan Tassit tiayeet Picassoyalli ?arrar yi?raa-hun ba?d ma tiada katab-un read.3sm-them still negsomeone wrote.3sm-them 'The few books about the story of his life that Picasso decided to read, noone has written them yet.' What (43) illustrates is the fact that when the quantifier kam kteeb 'few books' needs to be interpreted within the scope of the verb ?arrar 'decide' (non-referential amount reading), coreference is impossible between the name Picasso and the pronominal subject within the relative clause. When the non-referential reading is favored by the linguistic context, as is the case in (43b), unacceptability results. Thus, in the context of reconstruction, principle C effects arise in restrictive relatives. 153 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Furthermore, in contexts where movement is the only strategy available for generating a relative construction, a principle C violation can be avoided only in cases where the name inside the external head is in fact embedded inside an adjunct. That is, the argument/adjunct asymmetry observed in wh-questions shows up again. A case in point is that of the relativization of the subject of existential constructions:1 6 1 6 Recall that these constructions are sensitive to islands, as illustrated in (ia) for adjunct clauses and (ib) for wh-clauses: (i) a* kali suura la-Taylt-o narvaz Karim la?anno keen fii every picture to-family-his upset.3sm Karim because was.3sm in-it b-l-ma3alle 3abar-un yfiiluw-a in-the-magazine forced.3sm-them remove.3p-it 'Every picture of his family that Karim was upset because there was in the magazine, he forced them to remove it.' b* kail suura la-Taylt-o sa?al Karim ?aza keen fii every picture to-family-his asked.3sm Karim whether was.3sm in-it b-l-ma3alle Teed 3abar-un yjiiluw-a in-the-magazine retum.3sm forced.3sm-them remove.3p-it 'Every picture of his family that Karim asked whether there was in the magazine, he then forced them to remove it.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (44) kali suura la-^aylt-o sami? Karim ?anno keen fii every picture to-family-his heard.3sm Karim that was.3sm in-it b-l-ma3 alle 3 abar-un yjiiluw-a in-the-magazine forced.3sm-them remove.3p-it 'Every picture of his family that Karim heard that there was in the magazine, he forced them to remove it/ If the positions of the pronoun and the name are inverted in (44), the sentence becomes unacceptable under the coreference reading (45a). (45) a.* kali suura la-'iaylt Karim pro samiT ?anno every picture to-family Karim pro heard.3sm that keen fii b-l-ma3 alle 3 abar-un yjiiluw-a was.3sm in-it in-the-magazine forced.3sm-them remove.3p-it 'Every picture of Karim's family that (he) heard that there was in the magazine, (he) forced them to remove it.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. kali suura ?eddeem beet Karim pro samiT Tenno every picture in-front house Karim pro heard.3sm that keen fii b-l-ma3alle 3 abar-un yfiiluw-a was.3sm in-it in-the-magazine forced.3sm-them remove.3p-it 'Every picture in front of Karim's house that (he) heard that there was in the magazine, he forced them to remove it.' More interesting though is the fact that in the context of relativization of the subject of existential constructions, the argument/ adjunct asymmetry with respect to principle C effects surfaces again. When the name Karim is embedded in the PP adjunct attached to the EH (4b), coreference with the matrix pronoun subject inside the relative clause is possible. The contrast in (45) parallels the ones in (39) and (41) involving wh-questions. Therefore giving up the raising analysis will lead to the question of how to account for the selective availability of principle C effects in restrictive relatives and for their correlation with the availability of reconstruction.1 7 1 7 As several authors have observed (see Munn 1994, Safir 1999, Sauerland 1998), the relation between the EH and its copy in the restrictive relative is more indirect than the relation between the wh-phrase and its copy in wh-questions. In fact, as Sauerland points out, the chain formed by the EH and its copy 156 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Within the analysis I have argued for so far, the contrast between restrictive relatives and wh-questions can be accounted for in a straightforward manner. has properties of A-chains. In fact, A-chains do not display principle C effects, except under reconstruction. Thus, whereas standard A-movement fails to bleed condition C (i), standard A-movement always bleeds condition C (ii) (the examples are Fox's (1999) (93a) and (94) respectively): (i) *Which argument that Johni is a genius did hei believe? (ii) Every argument that Johni is a genius seems to him, to be flawless. However reconstruction in A-chains feeds condition C and thus we observe principle C effects under reconstruction (the examples are Fox's (51a-b)): (iii) a. [A student of David's,] seems to hiim t to be at the party. (3 > seem; *seem > 3) b. [A student of his;] seems to David, t to be at the party. (3 > seem; seem > 3) The difference between the non-reconstructed reading and the reconstructed reading is given by Fox as follows: when the indefinite NP has wide scope over the verb seem, David has thoughts about a particular student of his, that he is at the party. The reconstructed reading, where the verb seem has scope over the indefinite noun phrase, obtains in a context where David overhears a conversation at the party, which prompts him to conclude that there is a student of his at the party. When the name David occurs within the raised constituent (iiia), the reconstructed reading does not obtain, as it would yield a principle C violation, under coreference between David and the prepositional complement him. When the order of the pronoun and name are reversed (iiib), both readings become available. 157 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Keeping with Chomsky's 1995 account for the argument/adjunct asymmetry in wh- questions, the absence of such phenomenon in restrictive relatives is due to the availability of the base-generation strategy to derive those constructions. According to what we have seen so far, the relative clause in sentence (41a) (which I repeat below) has two possible representations: One where the external head is related to a copy inside the relative clause (46a) and one where it is related to a null resumptive pronoun (46b) (irrelevant details omitted). (41) a. najaro l-mu?eebale ssufmfiyye la-r-ra?iis lli published.3p the-interview the-magazine to-the-president that ?alto ?9nno sa?alu-w Yann-a Ya-t-television said.2p that asked.3p-/izm about-it on-the-TV 'They published the magazine interview of the president that you said that they asked him about it on TV.' (46) a* [ eh l-mu?eebale ssufmfiyye la-r-raPns] [Relative clause yalli the-interview the-magazine to-the-president that W-u P+WRE [E H c o p y mu?eebale ssuhufiyye la-r-raPns] V-him P+WRE interview magazine to-the-president 158 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. [eh l-mu?eebale ssufmfiyye \a-r-ra?iis\ [Relative clause yalli the-interview the-magazine to-the-president that W-n P+WRE pro V-hini P+WRE pro The prepositional phrase attached to the head noun being a complement, it must be present at the base position (the relativized site). The representation where the null pronoun occupies the relativized site yields the coreference reading and voids the principle C violation. When the movement strategy is the only strategy available (45), restrictive relative constructions behave unsurprisingly like wh-questions. 3.3.2 A copy theory or a chain-binding approach to reconstruction? So far I have argued that in cases where restrictive relatives are generated by movement (47a), the external head of the relative construction participates in the same chain that includes the relativization site. The chain in (47a) -I have assumed implicitely, is a canonical chain derived by movement. This assumption accounts in a straightforward manner for the scope reconstruction in restrictive relatives in LA, for the correlation between the availability of scope reconstruction and the surfacing 159 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of Principle C effects, as well as for the bound variable readings of pronouns inside the external head related to quantifier phrases inside the relative clause. (47) a. E H ...............WRE-EHC o p y b. E H ...............WRE-pro In the restrictive relatives that are base-generated (47b), no such chain exists; the external head is related to a null pronoun in the relativization site, which is identified by a WRE. It is the presence of this pronoun that forces the EH in base generated restrictive relatives to be referential. Another account for the data involving reconstruction effects in resumptive restrictive relatives is available in the literature (see, e.g. Demirdache 1991), based on earlier analyses of relativization and going back to Chomsky 1977. According to Demirdache's account, all relative constructions are (null) operator constructions, with the LF representation in (48): (48) EHi [Relative Clause Opi..................... ti ] The relativization site in (48) is occupied by the copy/trace of a wh-moved operator. In sentences like (49a), the overt operator is who(m). 160 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (49) a. The man who(m) I saw on the street was wearing red shoes, b. The man that I saw on the street was wearing red shoes. Sentences like (49b) are assumed to involve a null operator, which undergoes 'overt' wh-movement, as those constructions exhibit the same island effects displayed by constructions like (49a). Island sensitivity is illustrated below using complex NP islands: (50) *The man who(m)/ that Mary heard the claim that John sold t red shoes. According to Demirdache (1991), resumptive relative constructions are just cases of in situ relativization. The resumptive pronoun is in fact an operator that undergoes wh-movement at LF to satisfy its need for a variable. Resumptive relatives and non- resumptive relatives differ in their surface structure representation. The surface structure representation of restrictive relatives involving resumptives is given in (51). (51) EHi [ Res-Opi ] 161 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. At LF, the resumptive pronominal operator raises and the construction can be give the representation in (48).1 8 Restrictive relatives in English further differ from their counterparts with resumption in that the former, but not the latter are sensitive to islands. 1 8 With this account Demirdache (1991) tries to capture a complementarity between relative clause operators and resumptive pronouns. Thus, she observes that constructions that make use of the resumptive strategy in relativization, do not involve relative operators. While this might be true of restrictive relatives in LA, this is not the case in free (headless) relatives (ia) or wh-questions (ib): (i) a. miin ma baddak twassl-o Ta-l-mataar leezim who Comp want.2sm accompany.2sm-him to-the-airport must yihdar ?a-l-wa?t be.ready.3sm on-the-time 'Whoever you want to accompany to the airport must be ready on time.' b. ?ayya tilmiiz baTatu-u Ta-l-museeba?a which student send.3p-him to-the-contest 'Which student did they send to the contest?' Interestingly, only Payya-NP 'which-NP' and miin 'who' but not Ju 'what' can be related to a WRE in wh-questions in LA. A similar contrast can be found in English relativization: which-NP and who can appear as relative operators but not what. 162 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Demirdache attributes this distinction to a difference between 'overt' and 'covert' movement: only 'overt' movement must obey island conditions.1 9 Demirdache's account for the reconstruction effects one observes in restrictive relatives assumes an operation of chain composition (Chomsky 1982) whereby the coindexing between the external head and the operator in (48) allows the external head to be interpreted in the position of the variable in the (Op, t) chain. It can be easily seen that such an analysis could handle the data concerning reconstruction that were discussed in this chapter. Whereas it follows directly from the copy theory of movement that the external head of the relative and the element occupying the relative are identical, chain composition is an operation that serves to establish this identity. Now the question remains as to how does the chain composition approach to A- chain formation in restrictive relatives avoid principle C violations outside the contexts that require reconstruction of the external head, such as in (52): 1 9 One needs to keep in mind that 'overt' here means that the movement took place pre-Spellout; the element that undergoes overt movement in this case can be phonetically null. 163 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (52) a. The pictures of John that he likes best are the ones hanging on this wall, b. suwar Karim yalli bifubb-un ?aktar Ji hanne pictures Karim that likes.3sm-them more thing them l-mTal!a?iin Ta-ha-l-heet the-hanging.3p on-this-the-wall 'The pictures of Karim that he likes best are the ones hanging on this wall/ Browning (1987) suggests that the operation of chain composition is optional, and therefore (52) is just a context where chain composition has not applied. The correlation between the availability of reconstruction and principle C effects is then the result of the operation of chain composition. So far then, it looks like the two approaches to reconstruction and A-chain formation, namely the copy theory approach and the chain composition approach are empirically indistinguishable. However, this is not exactly the case. Recall that in contexts where the movement strategy is unambiguously available for the generation of restrictive relatives, an argument/ adjunct asymmetry with respect to principle C effects surfaces. That was the case in the relativization of the subject of existential constructions ((45) is repeated below): 164 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (45) a * kail suura la-Taylt Karim pro sami'T ?anno every picture to-family Karim pro heard.3sm that keen fii b-l-ma3 alle 3 abar-un yfiiluw-a was.3sm in-it in-the-magazine forced.3sm-them remove.3p-it 'Every picture of Karim's family that (he) heard that there was in the magazine, (he) forced them to remove it.' b. kail suura ?addeem beet Karim pro samiT ?anno every picture in-front house Karim pro heard.3sm that keen fii b-l-ma3 alle 3 abar-un yjiiluw-a was.3sm in-it in-the-magazine forced.3sm-them remove.3p-it 'Every picture in front of Karim's house that (he) heard that there was in the magazine, he forced them to remove it.' Under the copy theory of movement I adopt in this thesis, the difference between (45a) and (45b) was mainly the result of the option of late adjunction, which allows the adjunct to be absent from the base position of the external head inside the relative clause. Under the chain composition analysis, this contrast cannot be made: in fact, in both (45a) and (45b) the external head is not present in the base position inside the relative clause. Chain composition in both cases will yield a principle C violation, which is contrary to fact. 165 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Before concluding the discussion on the contrast between the copy theory account of reconstruction effects in restrictive relatives and a chain composition account, I would like to examine a proposal by Cecchetto and Chierchia (1998) (henceforth C&C) that both analyses are needed to explain the facts about clitic left dislocation (CLLD) in Italian. It will be shown that the empirical support C&C give for their proposal does not warrant the conclusion that chain composition is needed alongside a movement analysis, and that the latter approach can indeed handle the facts they discuss. As Cinque (1990) argues, CLLD in Italian is a resumptive construction that involves a fronted phrase related to a sentence internal clitic pronoun, as illustrated in (53). (53) a. In palestra, Leo (ci) va volentieri To the gym Leo there goes with pleasure 'To the gym, Leo likes to go.' b. Maria, Leo *(la) incontra spesso Maria, Leo her meets often 'Maria, Leo often meets/ Only when an object phrase is clitic left dislocated (CLLDed) is the clitic obligatory. In all other cases of CLLD, the clitic is optional. C&C further note a contrast between NP CLLD and PP CLLD with respect to reconstruction. Whereas both Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. constructions are sensitive to islands (54), and display connectedness with respect to principle C (55), only NP CLLD shows scope interaction between a CLLDed QP and a QP inside the sentence, as the contrast between (56a) and (56b) indicates. (54) a.* In palestra, ho visto Leo prima che (ci) andasse To the gym (I) have seen Leo before that (he) there go (SUBJ) b.* Maria, ho visto Leo prima che la incontrasse Mara, (I) have seen Leo before that (he) her meet (SUBJ) (55) a.* A casa di Leo, pro (ci) va volentieri To the house of Leo (he) there goes with pleasure b* II libro di Leo, pro l'ha letto volentieri The book of Leo (he) it has read with pleasure (56) a. *V3 and 3 V In qualche cassetto, Leo ci tiene ogni carta importante In some drawer, Leo there keeps every important paper 'Every important document Leo keeps in some drawer.' b. 3V and V3 Qualche compito di fonologia, Leo lo assegna a ogni studente Some phonology problem Leo it gives to every student 'A phonology problem, Leo assigns to every student.' 167 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Thus PP CLLD presents mixed properties of a movement strategy, namely, island sensitivity and connectedness, and a base-generation strategy. C&C suggest that two analyses are needed to account for the facts of CLLD in Italian. According to these authors, NP CLLD is the result of movement from the argument position the NP originally occupies in the sentence. Thus the availability of scope reconstruction in those constructions (56b) is the result of the presence of a copy of the CLLDed NP in the base position. As C&C note, in this case, principle C effects go hand in hand with scope reconstruction. C&C take the absence of scope reconstruction in PP CLLD as an indication that in those constructions, the PP does not move from a sentence internal position. The PP in CLLD is base-generated in a topic projection (below the complementizer) coindexed with a clitic within the sentence. This clitic is said to undergo movement to the head of the topic projection in order to be interpreted. Thus, in long distance cases, the PP is base-generated in the topic projection of the clause containing the clitic and then moves to its surface, sentence-initial position.2 0 Thus, a sentence like (56a) can be given two 2 0 As pointed out in C&C, a dislocated PP can occupy overtly the topic position of an embedded clause: (i) Sono sicuro che, in palestra, Leo (ci) va volentieri (I) am sure that to the gym Leo there goes with pleasure 'Going to the gym I am sure that Leo likes' 168 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. representations (56b) and (56c), which correspond to the surface representation and the LF representation, respectively: (57) a. In palestra, Leo (ci) va volentieri to the gym, Leo there goes with pleasure 'To the gym, Leo goes with pleasure.' b. [toPp In palestra [ip Leo ci va [vp •••]]] c. [lopp In palestra ci va [ip Leo tv+ci [vp • • • ]]] To account for the principle C effects in the context of PP dislocation C&C appeal to the concept of cJiain-binding, adapting ideas developed in Barss 1986. A CHAIN is defined as in (58). (58) A CHAIN <Pi,..., (3 n > is a sequence of nodes sharing the same 0-role such that for any i, 1 < , i < n P i c-commands and is coindexed with pi+i This definition subsumes the notion of chain formed by movement, but also extends the definition to cases of PP CLLD, where the PP composes a CHAIN with the (clitic, t) chain. The gist of the proposal concerning chain binding is given in (59). (59) When a phrase YP c-commands a link XP of a CHAIN, it counts for the purposes of Binding Theory as if it c-commands every link of the chain. 169 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Whereas (59) is far from being a formal definition, it provides us with a working definition on the basis of which we can evaluate C&C's proposal. In (55a) (repeated below), the clitic is c-commanded by the preverbal (null) subject. The clitic and the sentence initial PP form a CHAIN under the definition given in (58). Thus, the null subject counts as if it c-commands the sentence initial PP containing the name Leo and a principle C violation results. (55) a.* A casa di Leo, pro (ci) va volentieri To the house of Leo (he) there goes with pleasure The absence of scope reconstruction in (56a) (repeated below) is then accounted for by appealing to the fact that chain-binding does not interact with scope. (56) a. In qualche cassetto, Leo ci tiene ogni carta importante 3V but *V3 In some drawer, Leo there keeps every important paper 'Every important document Leo keeps in some drawer.' The distributive reading in (56a) can only be obtained via scope reconstruction and since the dislocated PP is base-generated in its surface position, it cannot reconstruct to a position c-commanded by the universal quantifier. 170 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Before I examine the adequacy of this analysis with respect to the data presented, a remark is in order about the mechanism of chain binding. The fact that chain binding interacts with principle C (and Binding Theory, in general) but not with scope reconstruction is an artifact of how this mechanism is set up to work: it is built into the definition. Basically, it stipulates that chain binding is sensitive to the Binding Theory but not to quantifier scope. Therefore it can be said that the chain binding mechanism encodes the PP CLLD facts, but does not explain them. In what follows, I pursue an analysis of PP CLLD which adopts C&C's insight that CLLDed PPs and CLLDed NPs differ in that the NPs need an IP internal trace to be assigned case, whereas PPs are free from such a requirement. This difference results in the NPs being generated in an IP internal position, whereas the PPs are generated, as C&C suggest, in the topic projection of the minimal clause containing the clitic to which they are related. The PPs then undergo movement to the sentence periphery. Thus island sensitivity can still be analyzed as the result of the movement of the PP from its base-generated position to its surface position. This analysis provides an explanation for an observed parallelism between scope interaction and principle C effects in PP CLLD, which C&C fail to mention. What C&C do observe is that a preverbal universal quantifier phrase does interact in scope with a CLLDed PP (60). All the cases that illustrate how PP CLLD cannot 171 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. obviate principle C effects involve a preverbal null subject interacting with a CLLDed PP, as illustrated in (55a). (60) 3V and V3 In qualche cassetto, ogni ragazzo ci tiene la foto della fidanzata In some drawer, every boy there keeps a picture of his girlfriend 'A picture of his girl friend, every boy keeps in some drawer/ As the authors also note, a VP internal quantifier phrase does not interact in scope with (56a) and cannot bind into (61) a CLLDed PP. (61) *Nella casa che I 'aveva ospitato, ci ho risistemato [ogni studenteji In the house that him had hosted, (I) there have put again each student Since the topic projection where the PP is generated dominates the IP projection, it is expected that a quantifier inside the PP will not interact in scope with another quantifier inside the IP. In sentences like (56a) above or (61), the IP internal quantifier phrase cannot take scope over the quantifier inside the PP (even after QR has applied) and hence scope reconstruction is impossible in (56a) as is the bound variable reading in (61). 172 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. But what C&C fail to observe is that in certain cases where the name inside the CLLDed PP corefers with a pronoun inside IP we fail to observe PC effects, as illustrated below:2 1 (62) A casa della famiglia di Leo;, Maria (ci) ando con luii con piacere To the house of Leoi's family, Maria goes with him with pleasure That is, as in NP CLLD, in PP CLLD scope interaction and PC effects seem to go hand in hand: a CLLDed PP interacts in scope with and can be bound by a preverbal subject. On the other hand, a CLLDed PP does not interact in scope with and cannot be bound by a non-subject. This generalization finds a straightforward explanation in the analysis proposed here: If CLLDed PPs are generated in a topic position within the minimal clause containing the corresponding clitic, and preverbal subjects are topicalized (as C&C assume),2 2 we expect to see scope 2 1 It is important to mention that the sentence corresponding to (62) that doesn't involve PP CLLD is unacceptable under the relevant coreference reading: (i) * Maria ando con luij a casa della famiglia di Leo, con piacere Maria goes with him to the house of Leoi's family with pleasure — Alternatively, one may assume that CLLDed PPs are generated adjoined to the maximal projection (presumably IP) containing the corresponding clitic. This means that they would be in a mutual c- command relation with preverbal subjects, which would also account for the observations made in the text. 173 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. interaction between the preverbal subject and the CLLDed PP, as well as PC effects. Since CLLDed PPs do not have a trace that is IP internal, they do not interact with non-subjects, neither for scope nor for binding purposes. Thus, the difference between NP CLLD and PP CLLD is reduced to one contrast, namely, the obligatory presence of an IP internal trace for NPs, but not for PPs, without having recourse to the mechanism of chain binding.2 3 This analysis I propose further explains the absence of argum ent/ adjunct asymmetry with respect to principle C effects in the following examples from C&C: 2 3 Note that the absence of a distributive reading in C&C's (9) (repeated in the text as (56a) and here in (i)) cannot be taken as evidence that in PP CLLD, scope reconstruction does not go hand in hand with Principle C effects. (i) In qualche cassetto, Leo ci tiene ogni carta importante 3V but *V3 In some drawer, Leo there keeps every important paper 'Every important document Leo keeps in some drawer.' If as C&C assume, PPs do not have IP internal traces and the PP in (i) forms a CHAIN with its corresponding clitic, then a direct object QP will not c-command any link in that CHAIN. If the clitic is attached to the verb in I, and the QP is in a VP adjoined position after QR, the clitic would still be outside the c-command domain of the quantifier phrase at LF. 174 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (63) a. *Nella scheda che Leoi ha preparato, proi ci ha messo i nostri CV In the file that Leo has prepared, (he) there has put our CVs. b. *NeH'affermazione che Leoi e' un ladro, proi ci vede molta malevolenza In the claim that Leo is a thief, he sees much ill will. Since a preverbal subject can be topicalized so that the CLLDed PP falls in its c- command domain, it comes as no surprise that the sentences in (63) are unacceptable under the intended coreference reading.2 4 2 4 C&C note that the PP CLLC paradigm contrasts with the NP CLLD paradigm, where the argument/adjunct asymmetry shows up along with anti-reconstruction effects. Thus, a name contained within a CLLDed NP and coindexed with the pronominal subject in IP internal position will escape a principle C violation only if it occurs inside an adjunct (ia), but not inside a complement (ib): (i) a. La scheda che Leo, ha preparato, pro, l'ha messa sulla nostra scrivania The file that Leo has prepared, (he) it has put on our desk, b. *L'affermazione che Leoj e'un ladro pro, la ha contestata con forza The claim that Leo is a thief, (he) it has contested forcefully Under the analysis I am proposing in the text for the generation of PP CLLD, this contrast remains unaccounted for. If all preverbal subjects can be topicalized, we don't expect that (ia) is acceptable under the indicated coreference reading. After consulting with several informants, the picture seems to look a bit more complex: whereas all the informants I have consulted with have supported C&C's judgements concerning Principle effects in PP 175 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. We can thus dispense with the mechanism of chain binding to account for reconstruction effects in Italian PP CLLD.2 5 CLLD, the opinion seems divided with respect to the absence of argument/adjunct asymmetry in 'anti-reconstruction'. Whereas one of the informants agrees with the judgments reported for (63a-b) in the text, the other three find a contrast between (63a) and (63b) that parallels the contrast between (ia) and (ib) (I report their judgments in (ii) below). (ii) a. Nella scheda che Leo, ha preparato, proi ci ha messo i nostri CV In the file that Leo has prepared, (he) there has put our CVs. b. ‘Nell'affermazione che Leoj e' un ladro, pro, ci vede molta malevolenza In the claim that Leo is a thief, he sees much ill will. This seems to favor an analysis of PP CLLD involving movement, as suggested in footnote 22. In that case, the contrast between (iia) and (iib) would result from the combined assumptions that there is preference for reconstruction and adjuncts can be introduced late in the derivation. The same analysis carries over to the contrast in (i). 2 5 There is an additional argument against chain-binding adapted from Lebeaux 1992 and which involves another resumptive construction in Lebanese Arabic: wh-questions. As Lebeaux points out, in the chain binding view, all positions in the chain are equivalent to each other and the chain itself behaves like a single entity in relation with an element outside this chain. Under this view, it is not expected that certain chain links will behave as if they enter into a chain-binding relation with a given element outside this chain, while other links behave as though they are not in such a relation with the same element. However, this is exactly what we observe in the following paradigm (adapted from Fox 1999): 176 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (i) a. ?ay y a kteeb pro talabu-u m in T a ly a keeno Tarlabiyyt t-tleemiiz w h ich book (they) ask ed -it fro m A lia w ere most the-students rah yriduul-A yee fut. retum .3p-H ER it 'W h ich book th a t th e y o rd ered from A lia w e re m o st s tu d e n ts g o in g to re tu rn to h e r? ' b.* ? ay y a kteeb pro talabu-u m in T a ly a p ro k e e n it rah taTte w h ich book (they) ask ed -it fro m A lia (she) w a s fut. give.3sf Tarlabiyyt t-tleemiiz m o st th e-stu d en ts 'W h ic h book th a t th ey o rd ered fro m A lia w a s sh e g o in g to g iv e m o s t stu d e n ts? ' c. ?ay y a kteeb pro talab u -u min-A k een it T a ly a ra h taTte w h ic h book (they) ask ed -it from-HER w a s A lia fu t. give.3sf Tarlabiyyt t-tleemiiz m o st th e-stu d en ts 'W h ic h book th a t th ey o rd ered fro m h e r w as A lia g o in g to g iv e m o s t stu d e n ts? ' U n d e r a ch ain c o m p o s itio n /c h a in b in d in g a p p ro a c h to resu m p tiv e co n stru ctio n s, the th ree sen ten ces ab o v e h av e th e re p re se n ta tio n s below , w h e re irrele v an t details h av e b een o m itte d a n d th e fs m ark th e p o ten tial sites o f m ov em en t: (ii) a. [wh-phrase .... p r o n o u n j... r-expressionk ....]■ [cp O P rpi ...[ip .. Q P ,... t ' i .... p ro n o u n k ... t i ... ] b * [wh-phrase p r o n o u n j... r-expressionk — ]i [cp O P rpi ... [ip.. p ro n o u n k t ' i .... Q P j h ... ] C. [wh-phrase .... p r o n o u n j... p ronounk [cp O P rr ...[ ip .. r-e x p re ssio n k t ' i .... Q P j t i ... ] 177 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.3.3 Interpretation of restrictive relatives with WRE in Lebanese Arabic Doron (1982) and Sells (1984) discuss an interesting contrast observed in Hebrew restrictive relatives, between those that involve gaps (63a) and those that involve resumptive pronouns (63b).2 6 (64) a. Dani yimca et ha-ija fe hu mexapes t Dani will-find the-woman that he seeks t 'Dani will find the woman that he seeks/ The contrast between (iia) and (iib) indicates that the bound reading of the pronoun contained within the wh-phrase is available when the QP that binds it is in the IP internal subject position, and the pronoun coreferential with the R-expression is in the IP internal object position, but not vice versa. Due to this contrast, it cannot be assumed that the wh-phrase, which composes a non-canonical chain with the resumptive operator-variable chain (OP, t', t), can stand for any link in the latter chain, because each of these links stands in a different relation to principle C. There is only one position where the wh- phrase can be interpreted without triggering a principle C violation and that is f' in (iia). Both the positions of t and t' in (iib) result in a PC violation, hence the unacceptability of the sentence. When the orders of the R-expression and the pronoun coreferring with it are switched (c), no principle C violation ensues and the bound reading becomes available again. It is clear from the discussion of the paradigm in (i) that a chain-composition/chain-binding approach to reconstruction in resumptive constructions lacks descriptive adequacy. 2 6 In Hebrew restrictive relatives, the resumptive strategy' alternates with the gap strategy. More specifically, only gaps appear in the highest subject inside a relative clause, gaps are optional in object positions, and resumptive pronouns are obligatory inside prepositional phrases and within NPs. 178 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. Dani yimca et ha-ija Je hu mexapes ota Dani will-find the-woman that he seeks her 'Dani will find the woman that he seeks.' Doron notes that (64a) has a 'de dicto' reading that (64b) does not have. That is, in (64a), Dani need not have a particular woman in mind, and so, (64a) can be uttered felicitously in a context where Dani is searching for someone with some particular properties, only one of which is being a woman. As Sells (1984) further notes, (64a) can be used to describe a situation where Dani is looking for a woman who is the same height as his wife, in order to check the length of a new dress he is buying for her. (64b) cannot be used in such a situation; it has only the interpretation where there is a particular woman, e.g. Paul's sister, that Dani is looking for. This interpretation is also available in (64a). Sells (1984) goes on to argue that the distinction to be made between (64a) and (64b) is not to be characterized in terms of the availability of a 'de dicto' reading, but in terms of a contrast between 'concept' and 'individual' readings. The typical example illustrating the concept interpretation is the following (Sells 1984 attributes the example to Barbara Par tee): (65) He got the bicycle that I wanted. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In this example, the bicycle need not be a particular one, or necessarily a kind of bicycle, or even a bicycle with specific properties. The sentence in (65) has an interpretation where the speaker's desire goes towards anything to which the property of being a bicycle applies. That is what the concept reading corresponds to. This interpretation is usually available in intensional contexts, such as the object of verbs like seek, want, etc. The concept interpretation can be distinguished from a de dicto interpretation in the following manner: whereas a norm phrase with a de dicto reading can be related to a pronoun in a discourse context (66a), a noun phrase with a concept reading cannot be coreferent with a pronoun in a discourse context (66b). (66) a. I am looking for the fellow who lives in Apt. 3 and Bill is looking for him too. b. I want a bike and Bill wants it too. As Partee (1972) observes, the personal pronoun in (66b) can only substitute for a referential noun phrase, that is, the personal pronoun forces the reading where Bill must have a particular bike in mind that he is looking for. In that context the NP a bike cannot have a concept reading. This is not the case in (66a), where the pronoun him substituting for the definite description the fellow who lives in Apt. 3 can have a non-referential interpretation. That is (66b) can be uttered in a context where the 180 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. speaker (and Bill as well) is looking for the person, whoever that may be, who lives in Apt. 3.2 7 To illustrate that restrictive relatives with resumptives in Hebrew can indeed give rise to a de dicto reading, Sells (1984) provides the following paradigm: (67) a. anioxal ha dag je itpesu t ba-xaka I will-eat the fish that will-be-caught t with-the-rod jel-xa of-you 'I will eat the fish caught with your rod.' b. ani oxal ha dag \e itpesu oto ba-xaka I will-eat the fish that will-be-caught it with-the-rod of- Jel-xa you 'I will eat the fish caught with your rod.' 2 7 It seems to me that the 'concept' reading Sells (1984) identifies corresponds to the non-referential reading of indefinite noun phrases. The difference between a concept reading and a de dicto reading could be reduced to the presence of a presupposition of existence and uniqueness induced by the presence of the definite article in the latter case but not in the former. At this point, I am unable to pursue this discussion any further (but see Chapter 4). 181 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. As Sells observes, in neither example must there be a particular fish that the speaker commits herself to eating. In both sentences in (65) the definite description containing the relative clause falls in an opaque context (under the modal will), which in this case, favors the de dicto interpretation. Using a similar argument, Sells (1984) distinguishes the concept reading from an 'individual-kind' interpretation.2 8 Thus, individual-kind terms can be the 2 8 Sells (1984) points out that resumptive pronouns allow us to distinguish two types of kind terms: the individual-kind terms such as those illustrated in (68) in the text. These correspond to what Carlson 1977 refers to as kind terms. The other type of kind terms is the concept-kind. A concept-kind term differs from an individual-kind term in that, in the former case, there isn't any individual that can be viewed as representative of that kind. A concept-kind term cannot enter into anaphoric relations with a personal pronoun (ia) and cannot be resumed by a pronoun (ib) (the examples are from Sells (1984)): (i) a. We are looking for a person to teach Montague Grammar and Autosegmental Phonology. ! Unfortunately he or she does not seem to exist. (The'!' indicates that the discourse continuation is inappropriate) b* ani lo xojevet Je nimca et ha mu?amad (e anaxnu be?emet crixim oto I not think that we-will-find the candidate that we really want him 'I do not think that we will find the candidate that we really want.' (ib) contrasts with (ii) which involves a gap in place of a resumptive pronoun: 182 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. antecedents of personal pronouns in discourse (68a) and can be resumed by pronouns in restrictive relatives (68b). (68) a. Dani is looking for the Crested Grebe. It inhabits the marshes here in April and May. b. ha xaya Je dani mexapes otahi nefoca me?od the animal that Dani seeks it is widespread very 'The animal that Dani seeks is very widespread/ By choosing the kind predicate ividespread, the kind interpretation is forced in the external head of the relative. What (67b) shows is that the resumptive pronoun is able to pick up on the kind interpretation, as the pronoun in (67a) can participate in anaphora with kind terms. Going back to the original contrast observed in (64) (repeated below), (ii) ani Io xojevet Je nimca et ha mu?amad Je anaxnu be?emet crixim t I not think that we-will-find the candidate that we really want 'I do not think that we will find the candidate that we really want.' 183 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (64) a. Dani yimca et ha-ija je hu mexapes t Dani will-find the-woman that he seeks t 'Dani will find the woman that he seeks.' b. Dani yimca et ha-ija Je hu mexapes ota Dani will-find the-woman that he seeks her 'Dani will find the woman that he seeks.' the reading that (64b) lacks can be more adequately characterized as being the concept reading. This can be further confirmed in the following contrast also discussed in Doron 1982 and Sells 1984: (69) a. kol gever yimca et ha-ija Je hu mexapes t every man will-find the-woman that he seeks t 'Every man will find the woman that he seeks.' b. kol gever yimca et ha-ija Je hu mexapes ota every man will-find the-woman that he seeks her 'Every man will find the woman that he seeks.' The sentence in (69a) is three-ways ambiguous; it has the three readings given in (70): 184 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (70) a. There is a particular individual woman (e.g. Brooke Shields) that every man is looking for and will find. b. Each man is looking for a woman particular to that man (e.g. Sam is looking for Susie and Jay is looking for his mother and Will for Anne ...). c. Each man is looking for a woman with certain properties, but does not know who such a woman might be (e.g. Sam is looking for a woman the same size as his wife, Jay needs a woman who can milk goats, and Will is looking for someone to act in his movie...). (68b) lacks the third reading (69c), which is the reading that corresponds to the concept interpretation of the external head. The availability of the reading in (69b) for the resumptive construction in (68b) indicates that the distinction between a resumptive and a gap relative is indeed not one between a dere and a de dicto reading. Under a dere reading of the external head, the distributive reading is not expected to be available. Given the distinction between concept and individual interpretations of restrictive relatives in Hebrew, interesting predictions can be made with respect to the analysis I have argued for in LA restrictive relatives with WRE. Under the assumption that WRE are agreement markers that may either cooccur with a trace/copy of movement (71a) or with a (true) resumptive pronoun (pro in (71b)), we only expect 185 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. those cases that clearly display the resumptive strategy to lack the concept interpretation. (71) a. Movement strategy E H i............WRE-EHi b. Base Generation (or resumptive) strategy E H ;............WRE-pro, This prediction is fully borne out by the data in LA. The LA equivalent to the Hebrew restrictive relative in (64b), which has only the individual reading, is in fact ambiguous between the concept reading and the individual reading. Thus (72) corresponds in fact to the Hebrew restrictive relative involving a gap in (64a). (72) Omar rah ylee?e 1-mara yalli Tam bifattij Talay-a Omar fut. find.3sm the-woman that Asp. look.3sf on-her 'Omar will find the woman that he is looking for.' As expected, the sentence which embeds the restrictive relative in (72) under a universal quantifier is three-ways ambiguous; (73) has the three interpretations given in (70). 186 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (73) kail ra3 3 eel rati ylee?e 1-mara yalli Tam bifattij every man fut. find.3sm the-woman that Asp. look.3sm Talay-a on-her 'Every man will find the woman that he is looking for/ The sentences in (72-73) can be analyzed as involving the movement strategy (71a) for the generation of the restrictive relatives and therefore, the relativization site in those sentences is occupied by a trace/copy of the external head. Therefore they pattern with restrictive relatives involving gaps in Hebrew. When the relativization site is clearly occupied by a pronoun, the concept reading becomes unavailable in LA. Thus (74a) contrasts with (72) and (74b) contrasts with (73). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (74) a. Omar rati ylee?e 1-mara yalli btaTrfo miin < Tam Omar fut. find.3sm the-woman that know.2p who Asp. bifattij Talay-a maT-o look.3sm on-her with-him 'Omar will find the woman that you know who is looking for her with him / b. kali ra3 3 eel rati ylee?e 1-mara yalli zTilte every man fut. find the-woman that upset.2sf la? anno ‘ Tarn bifattij ?alay-a because Asp. look.3sm on-her 'Every man will find the woman that you were upset because he is looking for her.' Both sentences in (74) involve relativization from an island context (zo/i-island in (74a) and Adjunct island in (74b)). (74a) has unambiguously the reading where there is a particular woman that Omar is looking for. The concept interpretation is unavailable for this sentence. This is also the case in (74b). The readings available for (74b) are the ones equivalent to (70a-b), that is (74b) is ambiguous between the reading where there is one woman (e.g. Brooke Shields) that every man will find (70a) and the reading where there is one particular woman for each man, such that this man will find her (70b). The relativization site in those sentences, I have argued, 188 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is occupied by a pronoun, which, we have seen cannot resume an NP with a concept reading. This accounts for the readings that (74a-b) lack. 3.4 Summary In this chapter, I have provided evidence for a copy theory approach to movement inside restrictive relatives with WRE in LA. The evidence for this analysis can be gleaned from the availability of scope reconstruction in those constructions, its interaction with principle C of the Binding Theory, and the availability of bound variable anaphora readings. I have also shown that a copy theory of movement is more descriptively adequate than a chain composition/chain-binding account for reconstruction. The availability of a movement strategy helps to account for the contrast between the interpretation of Hebrew resumptive restrictive relatives and that of LA restrictive relatives with WRE. The availability of a base generation strategy alongside the movement strategy accounts for the absence of principle C effects (anti-reconstructions) in restrictive relatives. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 4 The Structure of Resumptive Restrictive Relatives In the previous chapters, we have established the existence of two strategies for the derivation of resumptive restrictive relatives: The first strategy involves movement of the external head from the relativized site to a clause peripheral position, and is given the representation in (la). The second strategy involves no movement: the external head is base-generated in its surface position and a null pronoun resumes it in the relativization site, as indicated in (lb).1 11 will continue to assume that the WRE is not itself the resumptive element. Chapter 5 provides the arguments to that effect. Note that in an analysis like (lb) nothing prohibits movement from taking place inside the relative. That is, the element occupying the base position, pro, may undergo movement to a clause initial position. This is in fact what Browning (1987) argues is the syntax of restrictive relatives generated by null operator movement in English. What (lb) crucially encodes is the fact that EH and pro are not part of the same chain. 190 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (1) a. Movement strategy E H i... WRE-EHi b. Base-generation E H i... WRE-pro, Restrictive relatives that are derived by movement display the properties in (2), those derived by base-generation display the properties in (3), as seen in chapters 2 and 3. (2) Properties of restrictive relatives generated xvith movement a. Relativized site can be a gap b. Island sensitivity c. Reconstruction effects for scope, bound variable anaphora, and principle C (3) Properties of restrictive relatives generated xvithout movement a. Relativized site cannot be a gap b. No island sensitivity c. No reconstruction effects for scope, bound variable anaphora, or Principle C 191 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In this chapter I provide a structure for resumptive restrictive relatives in LA. I argue that the two types of derivations: the head raising analysis and the base- generation analysis are independently required to distinguish two types of relative constructions in LA, which have been termed definite relatives and indefinite relatives in Aoun and Choueiri 1996. Definite relatives can be generated via (head-)raising, in which case they have the properties in (2). Indefinite relatives cannot involve movement and have only the properties in (3). The availability of a base-generation analysis can help explain cases of mismatch between the antecedent and the resumptive element in the relativized site, since in such an analysis the two elements do not form a chain. Finally, I discuss the relevance of the proposed structure in solving a problem for the raising analysis. This problem is posed by Lebeaux's generalization about the argument/adjunct asymmetry in triggering Principle C effects under reconstruction. 4.1 Background discussion While I will not provide an exhaustive summary of the literature on the syntax of restrictive relatives, I would like to discuss first some relevant aspects of the syntax of those constructions, which have been at the heart of the debate in the literature. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A survey of the research points to two important aspects in the syntax of restrictive relatives: (i) The structural configuration of the noun phrase containing the relative, i.e. whether the relative clause is adjoined to the norm phrase containing its antecedent or whether it is a complement (of the external determiner or the antecedent norm); and (ii) the relation between the head norm and the relativization site. In other words, is the head noun moved from inside the relative or is it matched with an operator that is generated inside tire relative construction? The two aspects outlined above are logically (and in fact) independent from each other. That is, some proponents of the complementation analysis have argued for a head raising derivation for restrictive relatives, as is the case in Kayne 1994 or Bianchi 1999. Kayne's structure for restrictive relatives is given in (4). In Kayne's analysis, the relative clause CP is generated as a complement to the external determiner (ED) and the antecedent noun phrase (EH) raises to the specifier position inside the relative clause. Thus, the external determiner and the (4) DP 193 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. external head of the relative clause do not form a constituent on this analysis. Other proponents of the complementation analysis have defended a base-generation analysis of relatives, as is the case for Ouhalla (1999), Platczack (2000), and Schmitt (2000). For both Ouhalla and Platczack, for instance, the relative clause is generated as a sister to N; the head noun where the antecedent of the relative clause is base generated.2 Similarly, advocates of the adjunction analysis have argued for deriving relative clauses via head-raising. This is the analysis defended in Vergnaud 1985. An analysis that argues for base-generation of the external head in a structure where the relative clause is in an adjunct position corresponds to the analysis proposed in Chomsky 1977. There, it is argued, based on the similarities between relative clause 2 Schmitt's 2000 proposal is different from that of Ouhalla 1999 and Platczack 2000. In her analysis, the head noun is generated in [Spec, AgrP] in a structure such as the one in (ib), which corresponds to the relative construction in (ia): (i) a. The books that Bill wrote b. [dp the [AgrP [NumP books]i [A gr- Agr [cp OPj [c that [ip Bill wrote tj ] That is, the definite determiner takes a relative clause as complement. The head of the relative, which is a NumP, is base-generated in the specifier projection that is part of the extended projection of the relative CP, and therefore does not enter in a relation with the definite determiner. The head of the relative is coindexed with a null operator which undergoes movement inside the relative CP. 194 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. constructions and other wh-constructions with respect to locality constraints, as well as the presence of wh-words associated to gaps, that restrictive relatives are derived by operator movement. The relation between the antecendent and the operator is established via predication: the relative clause is predicated of the external head and hence a matching relation is established between the two.3 4.2 The representation of raising resumptive relatives in Lebanese Arabic Studies in the syntax of resumptive relatives have mainly focused on the relation between the resumptive element inside the relative clause and the external head of the relative construction.4 For languages that make extensive use of the resumptive strategy like the various Arabic dialects, Modem Hebrew, or Irish, it was argued that a resumptive element inside a relative is generated independently of the external head, which in turn is base-generated in its surface position. The anaphoric relation between the external head and the resumptive pronoun was obtained via predication (see Chomsky 1977, Borer 1983, among others), or via A-binding (see 3 Browning 1987 develops the idea that the mechanism that allows the antecedent to be related to the operator sitting in the specifier position of the relative CP, is an agreement chain which obtains between the antecedent, the operator and its trace inside the relative clause. This agreement chain is what Browning refers to as complex predication. 4 With the noted exception of Ouhalla 1999. See also footnote 13. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. McCloskey 1990). Alternatively, Demirdache (1991) suggests that resumptive pronouns are in-situ operators that move at LA, and in that sense, they are akin to relative pronouns in languages like English.5 All the above-mentioned analyses were developed within the standard analysis of relativization in which a relative clause, the modifier of a noun, is structurally represented as an adjunct to the noun phrase containing the modified norm, the main idea being that modification is encoded in the representation as adjunction. I have so far shown that the relation between the external head of a resumptive restrictive relative in LA and the weak resumptive element inside the relative clause is not always one of (A-)binding and that resumptive restrictive relatives can be generated by raising the external head of the relative clause from the position of the WRE. This analysis is consistent with Kayne's 1994 proposal further developed in Bianchi 1999 that resumptive relatives have a doubling structure as a source. In 5 Demirdache's analysis tries to account for the observation that resumptive pronouns are in complementary distribution with wh-operators. While this observation seems to be true for resumptive restrictive relatives, it is not for wh-questions in Lebanese Arabic (see Aoun and Choueiri 2000) or free relatives in Hebrew (see Borer 1983). Furthermore, the absence of wh-operators in restrictive relatives in Hebrew as well as in Lebanese Arabic seems to be independent of the presence of a resumptive element, since the same generalization holds whether the restrictive relatives in those languages involve resumptive pronouns or gaps. 196 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. what follows I examine the structural configuration of the norm phrase containing the resumptive restrictive relative.6 4.2.1 The external determiner hypothesis In a theory where restrictive relative clauses are considered adjuncts of NPs (Chomsky 1977), it is a natural assumption that tire determiner introducing the relative construction be external to the relative clause (5). However, this assumption is not clearly necessary in a theory where relative clauses are derived by movement of the external head. In a sentence like I read the book that John bought, in which EH moves from inside the relative clause into its surface position, it is not clear whether the ED the is moved along with the noun book from inside the relative, as indicated in (6), or whether it is base-generated in D°, there.7 6 Chapter 5 examines in detail the relation between the resumptive pronoun and its antecedent. 7 In fact, Kayne (1994) assumes that a derivation where the determiner is moved along with the relativized element exists for relative clauses headed by the indefinite determiner a (ia) and constructions headed by the definite determiner occurring with very, as in (ib): (i) a. a man just walked in who we knew in high school b. The very man just walked in who we knew in high school His argument is based on the analysis he gives for extraposition. Briefly, relative clause extraposition is in fact relative clause stranding and involves the movement of the EH (see (4) in text), leaving the 197 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (5) NP NP RC the hook (6) DP CP DP that the book 4.2.1.1 Restrictive vs. non-restrictive relatives As has been often discussed in the literature, restrictive relatives differ from non- restrictive relatives in that, in the former, the clause falls within the restriction of the determiner introducing the relative construction (i.e. the external determiner). Thus, observe the following paradigm, discussed in Vergnaud 1985: relative clause behind. In a structure where the external determiner is generated separated from the EH (see (8) in the text), this movement is prohibited, as it would involve two terms that do not form a constituent. However, in the case of (ia) and (ib), the determiner is part of the relativized DP and hence can be extracted out of the relative clause construction (see Buring and Hartman 1997 for arguments against such an analysis). 198 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (7) a. Mary knows few boys who enjoy knitting. b. Mary knows few boys, who enjoy knitting. c. Mary knows few boys. The restrictive relative in (7a) differs from the appositive relative in (7b) in its entailments. Only the appositive relative entails the proposition expressed in (7c). It was suggested that this difference between appositive relatives and restrictive relatives is encoded in the structural representation of those constructions: restrictive relatives are adjoined to the noun phrase they modify, in a position lower than that of appositive relatives. In the DP analysis of noun phrases, argued for in Abney 1987, restrictive relatives are adjoined to NP, as illustrated in (8), appositive relatives would be adjoined to DP (9). (8) DP D the NP NP RC book that John bought 199 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (9) DP RC DP the book that John bought Under an analysis such as the one in (8), the restrictive relative is clearly c-commanded by the external determiner in the relative construction, and as a result is interpreted in the scope of this determiner.8 That is not the case for the appositive relative in (9). 8 Independent evidence for the structure given for restrictive relatives comes from the licensing of polarity items by the universal quantifier every, as discussed in Bianchi 1999. The generalization concerning those facts is that every licenses polarity items, only when they occur within its restriction. Thus we can observe the contrast in (i): (i) a. *Every student who attended Gall's lecture had ever read anything about phrenology, b. Every student who ever read anything about phrenology attended Gall's lecture. Under the assumption that the restriction of a quantifier is its c-command domain, the relative clause in (ib) is in the c-command domain of the universal quantifier every, which is what the representation in (9) encodes. 200 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.2.1.2 Amount relatives Another argument showing that the determiner introducing the relative construction is external to the relative clause comes from Carlson's 1977 discussion of amount relatives. The argument is built to show that, in amount relatives, the external determiner and the external head of the relative do not form a constituent. Carlson (1977) tries to account for a contrast he observes between two types of relatives, illustrated in (10a) and (10b), respectively. (10) a.* Some man there was on the life-raft died, b. Every man there was on the life-raft died. The type of relative exemplified in (10b) is what Carlson has termed amount relatives.9 These constructions occur with a specific range of determiners (class I), including the definite article the, the universal quantifier every, all, any, non-deictic that, and xvhat. They do not occur with (class II) determiners such as cardinal quantifiers, demonstratives, possessives, most, or each. The basic property differentiating between class I determiners and class II determiners has been noted by both Carlson (1977) and Vergnaud (1985) and was termed separability by the 9 See the discussion in chapters 2 and 3 pertaining to those constructions in Lebanese Arabic. 201 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. latter. Class I determiners can be separated from the noun they modifiy by an expression of amount (11); Class II modifiers cannot (12).1 0 (11) Class I determiners (12) Class II determiners a. The 40 men a* Ten many people b. Every few insects b* Few several incidents c. All fifty states c* Some eight moms Irrespective of how these constructions are to be analyzed, what they show is th at, in the context of relativization from the subject position of existential constructions, for instance, the external determiner is precisely one that cannot occur in the relativization site. In order to appreciate this generalization, let us consider the following sentences from Lebanese Arabic. (13) a. ?axatt 1-katub Hi keen fii Ta-t-taawle took.ls the-books that was.3sm in-it on-the-table 'I took the books that there was on the table.' 1 0 This difference in distribution can be taken as evidence that determiners of class I and those of class 1 1 do not occupy the same position inside the DP (see chapter 5 for further discussion). 202 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. ?axatt kail warde keen fii b-3 -3 nayne took.ls every flower was in-it in-the-garden 'I took every flower that there was in the garden.' The relative constructions in (13) are introduced by the definite article I- 'the' and the universal quantifier kail 'every' respectively. It is clear, from the examples below, that those same elements will trigger a definiteness effect if they were to introduce the subjects of existential constructions:1 1 (14) a* keen fii 1-katub Ta-t-taawle was.3sm in-it the-books on-the-table 'There were the books on the table.' b* keen fii kail warde b-3 -3 nayne was in-it every flower in-the-garden 'There was every flower in the garden.' 1 1 It comes as no surprise that (14a) is acceptable under the list reading, a fact that is irrelevant for the discussion at hand. 203 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This discussion clearly indicates that the determiner that introduces the relative constructions in (13) is external to the relative clause. This is what I have been referring to as the External Determiner (ED) in the relative construction.1 2 1 2 A similar argument is given in Bianchi 1999 and deals with scope assignment under reconstruction. Bianchi observes that the distributive reading is available in a sentence like (i): (i) Ho telefonato ai due pazienti che ogni medico vistera domain. Italian I phoned the two patients that every doctor will examine tomorrow. As Bianchi points out, this judgment is reversed in a sentence like (ii), where the DP in the scope of the quantifier is a definite noun phrase: (ii) Ogni medico visitera i due pazienti Italian every doctor will visit the two patients In (ii), the definite object must denote a set with exactly two members and hence the distributive reading in (ii) is impossible. By the same reasoning, the reconstructed DP in (i) could not include the external definite determiner. One of Kayne's 1994 arguments to the same effect was the contrast between (iiia) and (iiib) below, which involve double genitives: (iii) a* The friend of John's b. The friend of John's that you told me about 204 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.2.2 The complementation analysis Carlson's analysis of amount relatives suggests that, at least in those constructions, the relative clause is selected by the external determiner. The selectional restrictions that hold between the external determiner and other types of relative constructions, like the restrictive relatives, have been pointed out in the literature dealing with the topic (see Smith 1969 and Vergnaud 1985, among others). Kayne's 1994 proposal that relative clauses are generated as sisters of the external determiner encodes this idea in the structural configuration of the relative construction. Several pieces of evidence are given to corroborate the idea that there is indeed a dependency between the external determiner and the relative clause. The following contrast illustrates one such argument. The unacceptability of (iiia) was taken to indicate that in (iiib), the EH and the ED could not form a constituent, or else (iiia) would be acceptable, contrary to fact. Barker 1998 argues convincingly that those examples do not constitute an argument for the ED hypothesis. The author observes that double genitives such as the DP in (iiia) display a semantic effect, which he calls anti-uniqueness. Briefly, the unacceptability of (iiia) is due to the anti-uniqueness requirement imposed on the first of two DPs in a double genitive construction, which puts it at odds with the uniqueness implication of the definite article. Further restriction on the noun phrase combining with the definite article improves the sentence, as illustrated in (iiib). However, as Barker observes, if one uses nouns that give rise to uniqueness implications, then the restriction on it will not improve the status of the sentence. Thus using mother instead of friend in (iii), will yield unacceptable results, as seen below. (iv) a* The mother of John's b* The mother of John's that you told me about 205 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (15) a.* The Paris b. The Paris that I love The contrast between the unacceptability of (15a) and the acceptability of (15b) with the definite determiner can be accounted for as follows: Since the name Paris cannot be associated with the definite article, as the unacceptability of (15a) shows, the acceptability of (15b) indicates that it is the relative clause that licenses the presence of the definite determiner in (15b). Additional data showing selectional restrictions holding between the external determiner and the relative clause it introduces can be gleaned from constructions which involve the relativization of abstract norms like kind (16), way (17) (see Schmitt (2000)): (16) a. She is that kind of person. b. She is the kind of person *(that is always helpful). (17) a. He did it in that way. b. He did it in a way *(that annoyed me). The sentences in (16b) and (17b) clearly show the dependency between the presence of the determiner introducing the external head and the presence of the relative 206 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. clause. As we have observed earlier (see section 4.2.1.2), in the case of amount relatives, the dependency between the external determiner and the presence of the relative clause is expressed as a restriction on the possible range of determiners that can introduce those relative clauses. If selectional properties are to be expressed in structural terms as a sisterhood relation, a Kaynean structure like the one in (6) (repeated below as (18)) encodes that at no cost. (18) DP CP DP that the book 207 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.2.3 DP relatives Considering the extent of the arguments favoring the external determiner hypothesis, we will henceforth assume it to be true. The sisterhood relation holding between the ED (D° in (18)) and the relative CP in English restrictive relatives, I will argue, holds between the ED and another nominal projection in LA resumptive restrictive relatives. This nominal projection is NumP. Relative clauses in LA are nominal, i.e. they are clauses headed by an element that is determiner-like in that it bears features of definiteness and agreement with the relativized head.1 3 131 am borrowing the term DP relative from Ouhalla 1999. The author makes a distinction between CP relatives and DP relatives to account for cross-linguistic differences in the surface position of relative clauses. CP relatives, as the ones found in English and Hebrew occupy the N complement position at S-structure (ia), whereas DP relatives occupy the subject position (genitive) (ib). (i) a. CP relatives [dp D [ ip I [np N [CP relative] ] ] ] b. DP relatives [dp D [ ip [DP relative] I [ n p N ] ] ] At LF, CP relatives raise to the subject position and we end up with a uniform LF representation for both types of restrictive relatives. 208 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Relative clauses in LA are either introduced by an invariable morpheme specific to relative constructions,1 4 i.e. yalli, or they are not introduced by any morpheme. Restrictive relatives with a definite external head are always introduced by yalli (19a) and restrictive relatives with an indefinite external head are not introduced by any morpheme (19b): (19) a. 1-kteeb *(yalli) jtarayt-o mbeerih daa? the-book *(that) bought.ls-it yesterday lost.3sm 'The book that I bought yesterday is lost.' b. ?aasasit Laila walad (*yalli) dayyaT kteeb-o mbeeriti punished.3sf Laila boy (*that) lost.3sm book-his yesterday 'Laila punished a boy that lost his book yesterday.' 1 4 Sentential complements in LA are introduced by ?anno: (i) a. xabbarat-na Laila ?anno 1-mmaslin m?adrbiin told.3sf-us Laila that the-actors on-strike.p 'Laila told us that the actors are on strike.' b. xabariyyit ?anno 1-mmaslin m?adrbiin fee3 a?at-na story that the-actors on-strike.p surprised.3sf-us 'The story that the actors are on strike surprised us.' 209 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.2.3.1 Morphosyntactic properties of yalli Not only does the morpheme yalli occur in relatives introduced by a definite article, it is itself definite, as the following discussion indicates.1 5 In LA, nouns and their adjectival modifiers agree in definiteness (20). (20) a. 1-kteeb l-3diid wssil Tal-maktabe the-book the-new arrivedat-the-bookstore 'The new book arrived at the bookstore.' b. bakyit Laila Tan kteeb 3 diid talked.3sf Laila about book new 'Laila talked about a new book.' 1 5 The counterpart of yalli in Standard Arabic (?a)lladi occurs also in definite relatives but not in indefinite relatives. It is morphologically definite: it is introduced by the definite article al-. (i) qaraVtu 1-kitaaba *(lla5i) Jtaraytu-hu 1-baariha read.lS the-book that bought.IS-it yesterday 'I read the book that I bought yesterday.' (ii) qara?tu kitaaban (*lla5i) Jtaraytu-hu 1-baaritia read.lS book that bought.IS-it yesterday 'I read a book that 1 bought yesterday.' 210 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In (20a-b), the adjectives agree in definiteness with the nouns they modify: In (20a), the adjective l-jdiid 'the new' occurs with the definite article (a)l- since the noun it modifies, i.e. l-kteeb 'the book', is definite. On the other hand, the adjective jdiid. 'new' in (20b) does not occur with the definite article, since the noun it modifies, i.e. kteeb 'book', is indefinite. Furthermore, a sentence where the norm and its modifier do not show agreement in definiteness would be ungrammatical: (21) a* l-kteeb 3 diid wasil Tal-maktabe the-book new arrived at-the-bookstore b.* hokyit Laila Tan kteeb l-3 diid talked.3sf Laila about book the-new Consider now the following sentences involving relative clauses: (22) a. l-kteeb *(yalli) talabtii wasil Tal-maktabe the-book *(that) ordered.2sf-it arrived.3sm at-the-bookstore 'The book that you ordered arrived at the bookstore.' b. ?riina kteeb (*yalli) hkyit Tann-o Laila read.lp book (*that) talked.3sf about-it Laila 'We read a book that Laila talked about.' 211 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The contrast between (22a) and (22b) indicates that yalli is definite: yalli can only occur when the relativized DP is introduced by the definite article. By comparison with the paradigms in (20-21), we can assume yalli to be [+definite]. This would mean that the relative clause yalli talabtii 'that you ordered' in (22a) is definite, thus matching the relativized NP l-kteeb 'the book'. In (22b), ungrammaticality results from the clash between the definiteness of the relative clause in the presence of yalli and the indefiniteness of the relativized NP. Additional evidence that yalli is [+definite] comes from relative constructions headed by the quantifier kdl 'all/every', kdl 'all/every' in LA, as in many other dialects of Arabic, is ambiguous between the 'each/every' reading and the 'all' reading: walad boy l-kteeb the-book 1 -wleed the-children wleed children 212 (23) a. kali every b. kali all (24) a. kali all b* kail every Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. When it occurs with a singular noun phrase, kd l 'all/ every' is ambiguous between the distributive each and the collective all: if the noun phrase following it is indefinite (23a), it is interpreted as distributive each, if the noun phrase is definite (23b), on the other hand, the interpretation is that of a collective all. When the NP modified by kd l 'all/every' is in the plural, however, the only reading is the collective reading, and more importantly for our purposes, the NP following it must be definite, as indicated by the ungrammaticality of (24b). Returning to relative constructions headed by kdl 'all/every', consider the following sentences: (25) a. kail yalli baddun yatidaro 1-film fallo all that want.3p watch.3p the-movie left.3p 'All those who want to watch the movie left.' b. kail 1-wleed fallo all the-children left.3p 'All the children left.' The (syntactic) plurality of the NP modified by kdl 'all/every' is indicated by the plural agreement on the verb. Thus, we can conclude that the relative clause 213 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. introduced by yalli in (25a) must be definite otherwise, one would expect (25a) to pattern with (24b). This is contrary to fact.1 6 1 6 Phonological evidence can be adduced to show the similarity between yalli and the definite artilce in LA. Both elements undergo reduction and assimilation to a following coronal consonant (see Haddad and Kenstowicz 1980). The definite article I- undergoes total assimilation with a following coronal consonant (i): (i) a. s-samak 'the fish', z-zeet 'the oil', l-taccwle 'the table', d-daraj'the stairs', n-nuur 'the light', f-fams 'the sun', b. l-beeb 'the door', l-fayy 'the shade', l-kirse 'the chair', l-xaatim 'the ring', l-f)abr 'the ink', l-?alam 'the pen'. This assimilation process does not take place between an [1 ] and a following coronal sound morpheme-internally: (ii) a. kalse 'a sock' *kasse b. balde 'a village' *badde and it does not take place between an [1 ] and a following coronal across morphemes. (iii) a. 3 abal 'mountain' b. 3abal-na 'our mountain' *3 aban-na In definite relatives, yalli can optionally be reduced to 1 -: 214 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (iv) a. s-sabe yalli/1-tiimil I-kirse the-boy that/ the-carried.3sm the-chair 'The boy that carried the chair' b. 1-bant yalli/l-Vatat-ne l-?alam the-girl that/the-gave.3sf-me the-pen 'The girl that gave me the pen' c. 1-bant yalli/l-Fariidbihabb-a the-girl that/ the-Farid like.3sm-her 'The girl that Farid likes' In such cases, yalli may assimilate to the following coronal consonant: (v) a. s-sabe yalli/l-/d-darab 1-bant the-boy that/the-hit.3sm the-girl 'The boy that hit the girl' b. 1-bant yalli/1-/n-nisy it kteeb-a the-girl that/the-forgot.3sf book-her 'The girl that forgot her book' c. 1-bant yalli/l-/s-sam iir bihabb-a the-girl that/the-Samir like.3sm-her 'The girl that Samir likes' However, it is important to note, as Haddad and Kenstowicz do, that the difference between yalli and the definite article I- is that the latter assimilates obligatorily to the following coronal consonant whereas the latter dos so only optionally. 215 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.2.3.2 (p-features In addition to being [+definite], yalli also bears cp-features (Aoun and Choueiri 1996). First, it is important to note that yalli is not always preceded by a relativized NP, and may occur in what has been called in the literature, a headless relative: (26) a. ttasalna bi-yallirati njuufun bukra contacted.lp in-that fut. see.lp tomorrow 'We contacted those that we will see tomorrow.' b. Jsfna yalli zfelto la?anno ma rati na'Tzim-un saw.lp that upset.2p because neg. fut. invite.lp-them 'We saw those that you were upset because we will not invite them.' As can be seen from the translations in (26), when yalli occurs in a headless relative it may have the distribution of a definite noun phrase. I will assume that in such cases yalli heads the DP argument of the preposition bi 'in' in (26a) and the verb Jeef 'see' in (26b). The alternative would be to suggest that the argument of the preposition in (26a) and that of the verb in (26b) is a null pronominal. However, this alternative is untenable. Note that in the sentences in (26) an agreement clitic is absent from the preposition or the verb. In LA, null pronominals are generally identified locally within the domain of the head that selects them. Thus, we have the following paradigm: 216 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (27) a. Jtarit zeina katub-*(un) bought.3sf Zeina books-3p 'Zeina bought their books.' b. Jafnee-*(ha) saw.lp-3sf 'We saw her.' c. ttasalna fiy-*(un) contacted.lp in-3p 'We contacted them.' In (27a-c), a pronominal affix appears on the lexical head that selects the null pronominal argument and it cannot be omitted for the relevant readings to obtain. Thus, the absence of that affix on the preposition or the verb in (26a-b) indicates that the argument of those lexical items is not a null pronominal, but the noun phrase headed by yalli itself. In that sense, the sentences in (26) pattern with the following sentences which involve the demonstrative pronoun: (28) a. Jafna hayde saw.lp this 'We saw this one.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. ttasalna bi-haydole contacted.lp in-those 'We contacted those/ Further evidence that yalli bears (p-features is that it incorporates with the bound pronominal morpheme h(a)~ 'this/that': (29) a. Jaft halli rah ysallih-le s-siyyaara saw.ls this.that fut. repare.3sm-me the-car 'I saw the one that will repare my car/ b. Jaft z-zalame halli rah ysallih-le s-siyyaara saw.ls the-man this.that fut. repare.3sm-me the-car 'I saw the man that will repare my car/ This pronominal morpheme h(a)- may not occur on its own without being followed by an element bearing cp-features (see Aoun and Choueiri 2000): (30) * Jaft ha- saw.ls this 'I saw this/ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In LA, third person strong pronouns are formed with the morpheme h(a)~, as indicated in Table 2 (repeated here in (31)).1 7 (31) Table 2. Third Person Strong Pronouns in LA Masculine Feminine Singular h-uwwe h-iyye 'he' 'she' Plural h-0rme 'they' The strong pronouns in (31) can be factored into two elements: one of them carries the person feature namely h(a)-, and the other carries the number and gender features (see Aoun and Choueiri 2000). Lacking number and gender specifications, the third person morpheme h(a)- has an incomplete phonological matrix and merges with a noun phrase to acquire those features (32) (see Benmamoun 2000). 1 7 This is also the case for demonstrative pronouns that occur as free morphemes: (i) Table 3. Independent Demonstrative Pronouns in LA Masculine Feminine Singular hayd-a/-aak hayd-e/-iik 'this/that' 'this/that' Plural haydo -e/-iik 'these/those' In Aoun and Choueiri 2000, the demonstrative pronouns are factored into: (i) a third person morpheme ha-, (ii) a discontinuous deictic morpheme d_(k), and (iii) an agreement morpheme. 219 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (32) Jaft ha-*(z-zalame) saw.ls this-*(the-man) 'I saw this man/ Going back to (29), the acceptability of the sentences with ha+yalli indicates that yalli carries the relevant number and gender features that complete the feature matrix of the third person morpheme h ( a 1 9 1 8 The sentences in (29) provide further evidence that yalli is definite. It can be noted that the morpheme h(a)- always selects a definite noun phrase. Thus observe the contrast between (ia) and (ib) below: (i) a ha-l-walad this-the-child 'This child' b* ha-walad this-child 1 9 Cross-linguistic data from Standard Arabic provide motivation for the assumption that the counterpart of yalli, namely ?alladi agrees with its antecedent, not only in (p-features, but in case as well: (i) a. ra?aytu 1-waladayni llaQayni tuhibbuhuma 1-muYallima saw.IS the-boy.Dual.Acc that.3F.Dual.Acc like.3SF.them(dual) the-teacher 'I saw the two boys that the teacher likes.' 220 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. What I am proposing here is that yalli is a complex agreement morpheme and is generated in the Num position of a relative construction, as illustrated in (33) .2 0 (33) [dp ED [NumP [n um yMi] ] ] b. 3aa?at 1-fataataani llataani tufiibbuhuma 1-muVaIlima came.3SF the-girl.Dual.Nom that.3DF.Nom like.3SF.them(duaI) the-teacher 'The two girls that the teachers likes came.' As can be observed in (i) above, the Standard Arabic counterpart of yalli is inflected for person, gender, and number. In addition, the sentences in (i) illustrate that the relative clause complementizer 7alladi displays overt agreement in case with the relativized DP. 2 0 Although it is clear that Lebanese Arabic yalli is different from the English complementizer that, it cannot be equated with relative pronouns/determiners. Typically, relative pronouns like English wh- words who or which or even German determiners occur in constructions other than relative constructions. In English wh-words are used to form questions and in German, relative determiners occur as resumptive elements in left dislocation, as illustrated in (i). yalli in Lebanese Arabic is only used in relative constructions. (i) a. der Mann, der in seinen Biiro arbeitet the man who in his office works 'The man who works in his office.' b. den Hans, den habe ich gestem gesehen the Hans, him have I yesterday seen 'As for Hans, I saw him yesterday.' 221 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The selectional restriction between the external determiner and yalli in (33) is overtly reflected through agreement: the higher determiner ED, selects another head compatible with its (in) definiteness specification. When the ED is definite, yalli is selected. It is not uncommon to find such an agreement relation between two nominal heads in Semitic. We have already observed in (22) that nouns and adjectives agree in definiteness in Lebanese Arabic. In construct state nominals, one can find another instance of this definiteness sharing (34). (34) * 1 3 1 6 6 1 1 1 1 1 t-tleemiz *(l-)?aalye grade the-students *(the-)high 'The high grade of the students.' Accounting for the properties of construct state nominals has long been the subject2 1 of discussion in the literature (see Aoun 1978, Mohammad 1988, Ouhalla 1990, Hazout 1990, Ritter 1991, Fassi Fehri 1993, Siloni 1994, Borer 1996, and Benmamoun 2000, among others) and there is no consensus on how to represent those constructions. However, it is clear that in the construct state nominal illustrated in (34), the definiteness specification of the whole DP seems to be related to the definiteness of the second member of the construct. Thus, as can be noted from the 2 1 See chapter 5 for a reinterpretation of those facts in different terms and for further evidence that the construct state nominal in (34) is indeed definite. 222 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. example in (34), although the first member of the construct does not overtly bear the morphological definite marker 1 -, it must agree with a definite adjective, indicating that it is indeed syntactically definite. Similarly, in definite relatives, the definiteness of yalli dictates the definiteness of the EH. The structure given in (35) (see Ritter 1991) allows that feature sharing to take place after head movement of tleemiz 'students' (DP2) into the D hosting Taleemit 'grade' in DPi. (35) DPi NumPi Taleemit Numi ttleemiz The second member (DP2) of the construct state nominal moves (by XP movement) to a position dominated by DPi, from which its head noun can raise and adjoin to D for definiteness sharing. Interestingly, Haddad and Kenstowicz (1980) mention a similar case in the domain of relativization. Indeed the EH of a definite relative together with yalli can form a construct state nominal, as indicated in the following example: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (36) (*t-)taawilt lli Jtarayt-a wslit (*the-)table that bought.ls-her arrived.3sf 'The table that I bought arrived/ In such cases, the external head cannot take the definite article and appears in the construct state, as indicated by the fact the its feminine ending -t, which is not pronounced w hen the noun appears in the free state (37), has to be overtly present in (36). (37) *(t-)taawle lli jtarayt-a wslit *(the-)table that bought.ls-her arrived.3sf 'The table that I bought arrived.' The structure that can be given for the relative clause containing the construct state nominal in (36) is parallel to that in (35): (38) DPi NumP EH- yalli ttleemiz 224 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The construct state nominal we see in (36) is obtained by subsequent raising of EH and yalli to D. The minimal difference between the tree configuration in (38) and the one in (35) is that the Num head in relative clauses takes a clausal complement, whereas it dominates a noun phrase in construct state nominals. 4.2.4 The categorial nature of the moving element An analysis that assumes that the external determiner in a relative construction does not form a constituent with the external head in raising relatives raises questions as to the categorial nature of the moving element. Kayne 1994 suggests that in f/wf-relatives, the category of the raised element could probably be NP. Such an analysis accounts not only for the ED hypothesis but also for the fact that in raising relatives, the relativized site does not trigger definiteness effects in the context of relativization from existential constructions, and allows distributive readings, which are not allowed by definite DPs. However, a few arguments militate against such an analysis, and argue for a refinement of Kayne's structure along the lines suggested by Bianchi (1999). I will discuss two arguments, one discussed in Borsley 1997, and one based on LA relative constructions. As Borsley 1997 points out, raising an NP in f/znf-relatives is in contradiction with the fact that the relativized site in those constructions cannot generally be occupied 225 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. by bare NPs. Thus what Kayne is proposing in a structure like (39) for the relative construction the picture that Bill liked, is a structure whose base IP configuration is ungrammatical because it could not be generated (39b). (39) a. [dp the [c p [np picture ] [that [ip Bill liked ] ] ] ] b.* Bill liked picture The construction in (39b) is ruled out on the assumption that bare NPs could not be arguments (Stowell 1989, Longobardi 1994). Evidence that the relativized site cannot correspond to an NP comes also from LA relatives. It is observed (see chapter 2) that relativization from direct object positions always gives rise to resumptive relatives. In fact, this would not have been possible if the relativized element did not clearly correspond to a DP. Thus, non-referential objects like piano 'piano' in the expression bdlTab piano 'I play piano' cannot be relativized (40) P- 2 2 This raises questions as to the presence of resumptive elements in the relativization of idiomatic NP chunks, which are also non-referential. However, although referentiality will not distinguish idiomatic NP chunks from the non-referential NPs in (40), it must be noted that the former but not the latter can be quantified, indicating that although they are non-referential, idiomatic NP chunks like those in (ia) have more structure than non-referential NPs like those in (ib). In other words bare plurals, but not idiomatic NP chunks are bare NPs: 226 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (40) * 1-piano lli bWab-o the-piano that play.ls-it Bianchi 1999 suggests that the relativized element in f/znf-relatives is indeed a DP but with a null determiner. Thus the configuration in (39a) is more correctly represented as in (41): (41) [dp the [cp [ dp D° picture ] [that [ip Bill liked ] ] ] ] If the relativized external head in f/wf-relatives is indeed a DP with a null determiner that incorporates into ED, then we have an account for the absence of pied-piping with f/zrtf-relatives. The null determiner in the relativized head in the sentences in (42) is licensed by the ED through incorporation (Baker 1988, Pesetsky 1995). (i) a. ?axatt kam Yatta took.ls several/few nap 'I took a few naps.' b.* lTabt kam piano played.ls severa./few piano Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (42) a* The [author [recent books by f] ] you should read t b * The man with (that) you're sure to have a good time. (43) a. Which author did you read [recent books by t ] b. W ho did you leave with? The unacceptability of the sentences in (42) cannot be attributed to an ECP-type violation, under the assumption that the prepositions by and with cannot govern the trace of their complement. Such an account would not extend to the sentences in (43) where preposition stranding is acceptable. The unacceptability of the sentences in (42) could then be attributed to the licensing requirement on empty determiners: They need to incorporate into the ED. This is clearly not possible in (42) where the null D is embedded within a PP or a complex noun phrase. We thus eliminate the pied-piping option in f/uif-relatives. The difference thus reduces to the fact that the determiner head of the relativized noun phrase is only visible to the computation if it is not embedded within a PP or a DP. 4.2.5 The (in-) definiteness of the relativized noun phrase We thus identify the moving element in i/nZh-relatives as a DP with an empty determiner. The representation of a construction like (44a) is given in (44b). 228 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (44) a. 1-beet lli bihabb-o Sami the-house.sm that like.3sm-it Sami 'The house that Sami likes' b. [dp 1 - [N u m P [dp D° beet] [N u m H i [ bihabb-o Sami ] ] ] ] There are indications that the empty D° introducing the relativized DP in (44) is indefinite. This observation is suggested by the same facts that argue for the external determiner hypothesis. The relativized subject of existential constructions, as well as the relativized idiomatic NP chunks cannot bear the definite article for the relevant readings to be available. The contrasts in (45-46) illustrate this point: (45) a. 1-katub yalli keen fii Ta-t-taawle saaro Ta-r-raff the-books that was.3sm fii on-the-table become.3p on-the-shelf 'The books that there were on the table are now on the shelf.' b* keen fii 1-katub Ta-t-taawle i i was.3sm fii the-books on-the-table 'There were the books on the table.' (46) a. 1-Yatta yalli ?axadt-a baTd 1-rada feedit-ne the-nap that took.ls-3sf after the-lunch helped.3sf-ls 'The nap that I took after lunch helped me.' 229 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b.* ?axadt 1-ratta baTd 1-rada i i took.ls the-nap after the-lunch 'I took the nap after lunch/ Recall as well that the availability of distributive readings in relatives introduced by the definite article (in Italian as well as Lebanese Arabic) contrasts with the unavailability of such readings when simple definite NPs occur in the relativized position:2 3 -2 4 2 3 See Chapter 3 for the relevant examples from Lebanese Arabic. 2 4 Bianchi 1999 suggest that this indefinite requirement is imposed by the principle of Full Interpretation (FI). In a configuration like (i): (i) [dp the [C p [D» N] ] ] The determiner the needs to bind a variable inside its complement NP (Higginbotham 1985). If that is also the case for empty D°, then we face a situation where there is only one variable to be bound, the one inside the relativized NP, with two determiners to bind it. Thus, one of the determiners will have no variable to bind, which violates FI. The approach advocated by Heim (1982) to analyze indefinites as variables provides a way out: if the relativized D° is indefinite, it will provide the variable required by the ED and FI is respected. 230 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (47) a. Ogni medico visitera i due pazienti. *distributive reading 'Each doctor will examine the two patients.' b. Ho telefonato ai due pazienti che ogni medico 0 K distributive reading 'I phoned the two patients that every doctor visitera domani will visit tomorrow.' However, the claim that the relativized DP is non-definite runs counter the observation that, in many languages, relative pronouns are definite.2 5 Clearly, as Bianchi points out, Italian il quale, and French leqnel incorporate the definite article and are morphologically definite. In English, the «>/i-pronouns that can appear in relatives are the ones that can be interpreted as definite (Kuroda 1968). Thus, which and who can be relative pronouns, but not what.2 6 2 5 This objection can be made assuming that restrictive relatives involving relative pronouns and that- raising relatives involve the same type of moving element. This is not necessarily the case. 2 6 Kuroda 1968 argues that which and who can refer back to an entity previously introduced in the discourse (familiarity condition on definites), but not what: (i) a. Barry and Joseph are here. Which one/w ho do you want to see? b. You have a choice between Plurals and Events and Generalized Binding. Which one/*what do you want to read? 231 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. But it can be argued that what is morphologically definite may have the distribution of indefinites and may be interpreted as such. A case in point is presented in Ouhalla 1999.2 7 Ouhalla (1999) observes that definite relatives are sometimes interpreted like indefinite DPs rather than definite DPs. (48) a. ?riit-3lla kteeb read.ls-her.dat book 'I read a book of hers.' 'I read a book to her.' b. ?riit-3lla l-kteeb read.ls-her.dat the-book 'I read a book to her.' c. ?riit-alla l-kteeb lli read.ls-her.dat the-book that b-l-3 ariide in-the-newspaper 'I read the book of hers that they talked about in the newspaper.' 'I read to her the book that they talked about in the newspaper.' 2 7 Ouhalla's analysis is based on Moroccan Arabic. However, the same facts obtain in Lebanese Arabic and I will substitute Lebanese Arabic data for the Moroccan paradigm discussed in Ouhalla 1999. 232 hakyo Tann-o talked.3p about-it Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A sentence such as the one in (48a) where the direct object is indefinite is ambiguous between two readings, as illustrated in the English translation: one where the dative pronoun is interpreted as author of the book (author reading), and one where the dative pronoun is interpreted as a goal, i.e. the listener (benefactive readingj.2 8 As can be noted, also from the translation provided in (48b), a sentence with a simple definite object is unambiguous, having only the benefactive reading. Interestingly, the sentence in which the definite direct object is modified by a definite relative is ambiguous between the two readings. The paradigm thus draws a parallel between indefinite noun phrases and definite relatives: They allow the same range of interpretations. In that sense, definite relatives are different from definite noun phrases. This is the reason why Ouhalla suggests treating the definite determiner that introduces the relative construction in (48c) as an expletive determiner in the sense of Vergnaud and Zubizarreta 1992.2 9 2 8 Under the benefactive reading, the author of the book can be independently specified. Thus, a sentence like (i) is unambiguous: (i) ?riito-lla kteeb la-Edgar Allan Poe read.ls-her.dat. book to-Edgar Allan Poe T read a book of Edgar Allan Poe to her.' 2 9 A similar argument is made independently by Bianchi (1999), based on Hungarian definiteness agreement. The verb in Hungarian takes two forms depending on the definiteness of its direct object DP: 233 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (i) a. Akart egy konyvet/*a konyvet. He wanted.indef a book/* the book b. Akarta a konyvet/* egy konyvet. He wanted.def the book/*a book Restrictive relatives in Hungarian are obligatorily introduced by relative determiners that are formed by adding the definite article a to an interrogative determiner. Unlike in declaratives, the verb inside restrictive relatives must appear with the indefinite conjugation, as the contrast between (ii) and (iii) shows. (ii) a. egy konyv amit akart t a book w hich he wanted.indef b. a konyv amit akart t the book which he wanted.indef (iii) a* eg)' konyv amit akarta t a book w hich he wanted.def b* a konyv amit akarta t the book which he wanted.def Briefly, relativization of morphologically definite pronouns in Hungarian triggers indefinite morphology on the verb; otherwise, the results are unacceptable. Bianchi 1999 analyzes the definiteness of the relative pronoun/determiner as being an expletive feature, without syntactic or semantic consequences. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. However, there is evidence that the ED in definite relatives is not to be treated as an expletive determiner: for one, Ouhalla's 1999 analysis does not really account for the paradigm in (48), for which a semantic analysis is preferable, and second, it can be shown that definite relatives do indeed have the distribution and interpretation of definite noun phrases. It is important to note that Ouhalla's account of the paradigm in (48) assumes that the definiteness specification of the noun phrase in direct object position affects directly the availability of the author reading. It is not clear however that this is indeed the case. In fact, the author reading is again available with direct objects that are introduced by demonstratives, and those DPs are clearly morphologically definite (49). (49) ?riit3-lla ha-l-kteeb read.ls-her.dat this-the-book 'I read this book to her/ 'I read this book of hers/ I would like to suggest that the paradigm in (48-49) is to be accounted for using Barker's (1998) analysis of double genitives in English, by appealing to the anti uniqueness requirement imposed on the first DP member of a double genitive 235 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. construction.3 0 Only the author reading implies that the dative pronoun is an argument of the noun phrase containing the direct object. Assume that for the author reading to obtain in the constructions in (48-49), the direct object must obey the anti-uniqueness requirement. That is, the direct object kteeb 'book', in each of the sentences in (48-49) cannot carry an implication of uniqueness. The unacceptability of the sentence in (48b) under the author reading is then due to the incompatibility of that requirement with the uniqueness implication of definite articles. As in the case of double genitives, a restrictive relative modifying the direct object improves the status of the sentence.3 1 The availability of the author reading in (48a) and (49) is due to the fact that indefinites and demonstratives do not imply uniqueness. This analysis extends to cases where the noun phrase interpreted as author is not raised. These are cases where the author noun phrase is a full lexical DP:3 2 3 0 Thus the account for (48-49) would be the same as the one for the English paradigm in (i): (i) a. a book of hers b.* the book of hers c. that book of hers 3 1 See Barker 1998 for a discussion of how restrictive modification improves the status of definites in double genitives. 3 2 (50b) is acceptable under the irrelevant reading where George Bassil is the buyer of the painting in question. 236 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (50) a. b9*Tt rasm la-George Bassil b-l-mazeed l-'ialane sold.ls painting of-George Bassil in-the-auction the-public 'I sold a painting by George Bassil at the public auction/ b.* baTt r-rasm la-George Bassil b-l-mazeed l-'ialane sold.ls the-painting of-George Bassil in-the-auction the-public 'I sold the painting by George Bassil at the public auction/ c. boTt r-rasm l-3 diid la-George Bassil b-l-mazeed sold.ls the-painting the-new of-George Bassil in-the-auction l-'ialane the-public 'I sold the new painting by George Bassil at the public auction/ d. ba*it ha-r-rasm la-George Bassil b-l-mazeed l-'ialane sold.ls this-the-painting of-George Bassil in-the-auction the-public 'I sold this painting by George Bassil at the public auction/ The paradigm in (50) parallels the data in (48-49): When the object noun phrase is introduced by a simple definite article the sentence is unacceptable, thus contrasting with the case where an indefinite article introduces the object noun phrase. The presence of a demonstrative pronoun (50d) or restrictive adjectival modification (50c) improves the sentence. This is what Barker's 1998 analysis of double genitive 237 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. constructions predicts and I adopt it to account for the interpretive effects we observe with definite NPs and definite relatives in (48-50). One interesting feature of this account -which I have only informally presented here- is that it relies on the assumption that in definite relatives, the external definite determiner is indeed definite. This is consistent with the fact that definite relatives do display the behavior of definite norm phrases in contexts that require definiteness, as we will see shortly. Definite relatives can occur in partitive constructions, where the second of the two DP members must be definite (see also Schmitt 2000): (51) a. waladeen mn *(l-)wleed child.dual of *(the-)children T w o of the children/ b. waladeen mn 1-wleed lli keeno hon Tabukra child.dual of the-children that were here morning 'Two of the children that were here this morning/ As (51a) shows, the second DP member of a partitive construction cannot be indefinite. However, a definite relative can occur as the second DP member in a 238 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. partitive construction, which indicates that the definite relative is indeed a definite noun phrase. Furthermore, definite relatives can occur as the restriction of the quantifier kail 'all', which requires a definite DP in that position (when the DP is plural) (52) ,3 3 (52) a. kail *(l-)katub/ ha-l-katub/ katub 1-wleed all *(the-)books/ this-the-books/ books the-children 'All the books/all those books/all the students' books' b. kail 1-katub lli Jtarayt-un raaho b-l-harii?a all the-books that bought.ls-them went.3p in-the-fire 'All the books that I bought burned in the fire.' 3 3 When the DP in the restriction of the quantifier is singular, then ksll 'every/all' is ambiguous between two interpretations: it is interpreted as every if the noun phrase that follows is indefinite (ia), and as all if the following noun phrase is definite (ib). kali walad every child.sm kail 1-kteeb all the-book.sm 239 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The quantifier kail 'all' in LA can only introduce a definite noun phrase if the latter is plural. If the definite relative in (52b) weren't definite, the acceptability of (52b) comes as a surprise in light of the unacceptability of (52a) without the definite article. We thus have evidence that definite relatives have the same distribution as definite DPs and are introduced by a 'true' definite determiner. This definite determiner is what licenses the null determiner contained within the relativized noun phrase in raising relatives. The relativized DP has a distribution similar to that of an indefinite noun phrase, as can be seen from the fact that it doesn't trigger definiteness effects and yields distributive readings. It can be further argued that only definite determiners can license another D° head. This is evidenced in construct state nominals, illustrated in (53), for instance: (53) byuut kail mmasle houses every actress 'The houses of every actress' A much discussed property of those constructions is known as definiteness spreading. The phenomenon refers to what seems to look like the spreading of the definiteness feature from the second member of the construct onto the first one: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (54) a. byuut 1-mmasle *(l-)wees‘ Tiin houses the-actress *(the-)big.p 'The big houses of the actress' b. byuut mmasle (*l-)wees< Tiin houses actress (*the-)big.p 'An actress's big houses' Thus the construct state nominal in (54a) is interpreted as definite and must agree in definiteness with an adjective that follows it. Similarly, the construct state nominal in (54b) is indefinite and must agree with a following adjective in definiteness as well. However, it is not clear that what we observe in (54) is indeed definiteness spreading, for the nominal in (54b) is ambiguous between two interpretations: one where it means 'an actress's houses', with a non-referential interpretation for actress, and one where it means 'a specific actress's houses' (Woisetschlaeger 1983). Only in the former interpretation can the construct state be indefinite. Thus observe the following contrasts: (55) a. kail *(s-)suwar all *(the-) pictures 'All the pictures.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. kail suwar m'Tallme (mn ha-l-m?allmeet) all pictures teacher (of this-the-teachers) 'All the pictures of one of those teachers.' c. kail suwar 1-mTallme I all pictures the-teacher 'All the teacher's pictures' As pointed out earlier, the quantifier kail 'every/all' is ambiguous between the meaning of 'all' and the meaning of 'every'. When it is followed by a plural noun phrase, this NP must be definite, as indicated by the unacceptability of (55a) without the definite marker. If the construct state nominal in (55b) were to be interpreted as an indefinite, the acceptability of (55b) comes as a surprise. In fact (55b) patterns with (55c) where the second member of the construct state occurs with an overt definite marker. Therefore, the construct state nominal in (55b) must be definite under the interpretation where mTcillme 'teacher' refers to a specific teacher. Thus, the phenomenon of adjective agreement in definiteness (54) cannot be taken as the sole test for the definiteness of the construct state nominal.3 4 If we 3 4 For the moment, I can offer no explanations for the unacceptability of the definite article on the adjective in (54b). However, observe the unacceptability of an indefinite adjective in (i), where we would expect one, under the definiteness spreading analysis. 242 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. take the difference between non-specific/non-referential indefinite noun phrases and referential/ specific noun phrases to be one between a DP and an NP, then the generalization is that only a definite construct state nominal can license a second member as a full DP. In other words, only a definite determiner can license another D°. 4.2.6 The trigger for movement So far we have established that the moving element in raising restrictive relatives is a noun phrase headed by a null determiner, which is interpreted as an indefinite.3 5 It remains to be seen why this movement takes place. I will not provide a definite answer to the question, however, I will argue that the type of movement involved in resumptive restrictive relatives is not triggered by EPP features (see Chomsky 1998, 2000). This is in fact the first observation one can make about restrictive relatives in LA: movement to [Spec, yalli] in LA, is not triggered by any strong features on that element. If it were true that yalli is associated with an EPP feature, then we would (i) *hayda beet m'Tallme mn ha-I-mTallmeet weesi*T this house teacher.fs of this-the-teacher.fp big 'This is the big house of one of the teachers.' 3 5 See also Chapter 5 for a more detailed discussion of null 'determiner' licensing. 243 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. not expect to see it heading relative clauses without antecedents, as the one illustrated below: (56) ftarayte yalli fiabbayt-a bought.2sf yalli liked.ls-her 'You bought the one I liked/ Thus yalli does not require the presence of an overt DP in its specifier position for the sentence to be acceptable.3 6 I suggest that we represent the relative constructions such as the one in (56), as in (57), where yalli moves to occupy the D position of the relative clause. (57) DPi NumP yalli 3 6 As we have observed earlier in this chapter the 'headless' relative clause in (56) does not have a null pronoun as antecedent. 244 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Therefore, it is not yalli that triggers the movement in resumptive restrictive relatives. Movement is then triggered by the need to license the null determiner inside the relativized norm phrase. The licensing relation between the external determiner of the relative construction and the null determiner of the relativized head, which requires the relativized head to move from its base position, is evident in the relativization of norm phrases that can function as bare NP adjuncts (Larson 1985). Consider the following contrast: (58) a* ?addayna 1-layle Hi wsalto b-l-mataar spent.lp the-night that arrived.2p in-the-airport 'We spent the night that you arrived at the airport.' b. ?addayna 1-layle lli wsalto fiy-a b-l-mataar spent.lp the-night that arrived.2p in-her in-the-airport 'We spent the night that you arrived at the airport.' What the contrast between (58a) and (58b) indicates is that, if the whole noun phrase containing the relative clause is an argument of the matrix verb, then there must be a resumptive element in the relativization site. The reverse impHcation is also true. Thus, if a resumptive element occurs inside the relative clause, then the noun phrase containing the relative clause must be an 245 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. argument of the matrix verb. This is evidenced by the contrast between (59a) and (59b-c). (59) a. fallayna laylt Hi wsalto left.lp night that arrived.2p 'We left the night that you arrived/ b.* wsalna laylt lli ?addaytuw-a b-l-mataar arrived.lp night that spent.2p-it in-the-airport 'We arrived the night that you spent at the airport.' c.* fallayna laylt lli wsalto fiy-a left.lp night that arrived.2p in-it 'We left the night you arrived.' The relative clause in (59a-c) occurs as an adjunct to the matrix verb and thus no resumptive pronoun can be present inside the relative. If resumptive elements signal the presence of a full DP in the relativized argument position (as will be argued in Chapter 5), then, the generalization that accounts for the data in (58-59) is as follows: There is a dependency between the presence of an external determiner in a resumptive restrictive relative and the presence of another D in the relativized site. The presence of an overt external determiner in the relative construction 246 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. implies the presence of a (null) determiner inside the relative clause itself.3 7 This dependency is what underlies the licensing relation between the external determiner and the null D head inside the relativized NP, and therefore movement in raising relatives. For sentences like (59a) where there is no overt external determiner and where a weak resumptive element is also absent, two possible derivations are available, one which involves movement of the adjunct noun phrase to its surface position, and one which involves base-generating the adjunct in that position. Assuming the bare NP adjunct does not project a DP layer, in the absence of a null determiner to license in the relativized DP, the trigger for movement is missing, hence the adjunct relative construction in (59a) would be base-generated with the head laylit 'night' in its surface position, adjacent to yalli.3 6 The assumption that bare NP adjuncts do not project a DP layer runs counter the fact that they can sometimes appear with a definite article (60a), or even with a demonstrative (60b): 3 7 This dependency is not a chain dependency since the external determiner is not an occurrence (Chomsky 1998) of the relativized NP D°: they do not have the same values for their definiteness feature, for instance. 3 8 The resulting construct state nominal would be due to a PF merger (see Benmamoun 2000) of the adjunct noun phrase with yalli. 247 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (60) a. fallayna s-see'i'a Tajra left.lp the-hour ten 'W e left at ten/ b. waasliin ha-l-layle arriving.lp this-the-night 'We are arriving tonight.' The alternative analysis, which involves movement, would consider that bare NP adjuncts can poject a DP layer. The presence of a null D° inside the relativized NP would trigger its movement to the surface position for licensing purposes. As in the case of argument relatives, the resulting construct state nominal is derived by movement of both the relativized adjunct and yalli to the ED position.3 9 Two types of evidence favor the movement analysis for adjunct relatives. 3 9 The unacceptability of (i), which differs minimally from (59a) in having the ED overtly realized as a definite determiner, could be due to the incompatibility of the non-referentiality of the adjunct noun phrase with the semantics of the definite article. (i) a* fallayna s-seeTa lli wsslto left.lp the-hour that arrived.2p 'We left when you arrived.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. First, recall that those constructions are sensitive to islands and, in that respect, they parallel constructions which involve wh-questions (see Chapter 2). An illustrative example is provided below: (61) * wsalna Ta-l-hafle seeTt lli keeno 1 -bolisiyye Tam arrived.lp to-the-party hour that were.3p the-policemen Asp. yil?ato zzalame lli darab Maria 0 catching.3p the-man that hit.3sm Maria 0 'We arrived to the party at the time that the policemen were catching the man that hit Maria.' The sentence in (61) parallels the one in (62), which involves extraction of a wh- question word form inside a relative clause island: As can be seen by comparing (i) to the sentences in (60), the noun seeTa 'hour' in (i) is not interpreted 'referentially' to mean 'hour' but non-referentially to refer to time in general. 249 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (62) * ?eemta keeno 1-bolisiyye Tam yil?ato zzalame when were.3p the-policemen Asp. catching.3p the-man lli darab Maria 0 t that hit.3sm Maria 0 'When were the policement catching the man that hit Maria?' Thus, relativization of time adjuncts as well as wh-extraction of a time adverbial are sensitive to (relative clause) islands. Adopting the base-generation analysis for adjunct relatives will result in having two different analyses for the similar paradigms illustrating island sensitivity: in the case of wh-questions, the unacceptability of (62), for instance, is the result of illicit movement; in the case of (61), some interpretive mechanism must be invoked that has properties similar to those of A-movement chains with respect to island sensitivity. Apart from being undesirable, such an account would face problems with data that involves reconstruction, like the following: 250 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (63) rafi nibTat xabar la-l-?ahl seeTt lli bitwasslo fut. send.lp news to-the-parents hour that accompany.2p kail walad Ta-bayt-o every child to-house-his 'We will send notice to the parents when you accompany every child to his house.' (63) is ambiguous between a reading where there will be one notice sent to all the parents, at one time, after all the children are accompanied to their respective houses, and a reading where different notices are sent at different times, depending on the time each child is driven home. It is the availability of the latter reading that is of interest here. This is the distributive reading that results from the quantifier phrase kail walad 'every child' taking scope over the head of the relative seeTt 'hour'. Under the assumption that this reading results from reconstructing the head of the adjunct relative to a position inside the relative, which is c-commanded by the quantifier phrase, a base-generation account of the relativization of adjunct NPs would face problems. Clearly, this result is expected under the analysis where the bare NP adjunct reaches its surface position by movement. 251 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.3 Structure of base-generated resumptive restrictive relatives Thus, raising relatives in Lebanese Arabic have the representation in (64): (64) DPi ED NumP DP yalli Movement of the relativized DP is triggered by the need to license its null D° head. This is done by incorporation of this element into the external determiner of the relative clause construction (ED), yalli, the element that introduces definite relatives as well as headless relatives in Lebanese Arabic, heads NumP and agrees with ED in definiteness and cp-features. We have seen however that not all resumptive restrictive relatives are generated by movement. Specifically, those constructions that violate islands (65), and which do not exhibit reconstruction effects (see Chapter 3), are base-generated. 252 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (65) raddayt 1-film yalli rati to min duun ma tjufu-u returned.ls the-film that left.2p without Comp see.2p-3sm 'I returned the film that you left without watching it/ The relativized site inside the relative clause is occupied by a null pronominal (66) and the relativized DP is base-generated in its surface position. (66) 1-film -yalli... tjufu-u-pro The null pronominal in Lebanese Arabic is licensed locally by the agreement morpheme that shows up on the head that selects it. In (65) this head is the verb y fu u f see'. The relation between the null pronoun inside the relative clause and the external head is an interpretive one: the Chomsky/Browning predication analysis is enough to establish this relation. The null pronorm inside the relative clause provides the open position for the relative clause to be interpreted as a predicate which applies to the external head. The strongest evidence for the availability of a base-generation derivation for restrictive relatives comes from a class of relatives called Indefinite Relatives (Aoun 253 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and Choueiri 1996). Recall that those are relative constructions that are introduced by an indefinite determiner or a numeral:4 0 (67) ?riina kteeb/ tleet kstub 3eebit-o/ -tin Nada read.2p book/ three books brought.3sf-it/ -them Nada Ta-s-saff i t to-the-class 'We read a book/three books that Nada brought to class.' In a raising analysis, such as the one argued for in Kayne 1994 or Bianchi 1999, indefinite relatives differ from definite relatives in that, although they also involve raising, the indefinite article or the numeral forms a constituent with the relativized noun phrase and undergoes raising with it. According to this analysis, we would expect to observe in those constructions reconstruction effects of the type discussed in Chapter 3 for definite relatives. However, this is not the case: We observe no distributive readings in (68a-b), for instance. 4 0 They are also relative clauses that do not have the equivalent of yalli. 254 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (68) a. kawyit Nada JarJaf ?8lto ?8nno leezim yhittu-u ironed.3sf Nada tablecloth said.2p that should put.3p-it ?a kail taawle b-l-3 nayne on every table in-the-garden 'Nada ironed a tablecloth that you said they should put on every table in the garden.' b. maTTuule yha??i?o ma’ T tleet barraas talab possible question.3p with three guards asked.3sm r-ra?is ?snno ywa?fuw-un ?8deem kail bineeye the-president that stand.3p-them in-front every building 'It is possible that they question three guards that the president asked that they make them stand in front of every building.' If the relative constructions in (68a-b) were generated by raising the relativized head (along with its weak determiner) from a relative clause internal position, the fact that the distributive reading is absent in those constructions comes as a surprise. Take (68a), for instance, although the context of the utterance favors the distributive reading where, for each table, there is one tablecloth, that reading is still unavailable when the relativized head farfaf' tablecloth' is indefinite. i 255 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A similar observation can be made regarding the absence of bound variable anaphora readings in indefinite relatives. Thus, the pronoun contained within the relativized noun phrase in (69) cannot be bound by the negative quantifier inside the relative clause: (69) * ballajo yzahhro Ja3 areet mn Ja3 areet-[o]j Nada bta'Trif started.3p bloom.3p trees of trees-[his]i Nada know.3sf ?anno [wala 3nayneete]i byahmal-un that no gardner neglect.3sm-them 'Some trees of his that Nada knows that no gardner neglects started to bloom/ Interestingly as well, indefinites and numerals belong to the class of determiners that cannot introduce amount relatives (Class II), as evidenced by the unacceptability of (70), and indefinite relativized heads necessarily have a de re/'referential' reading, as illustrated in (71a-b). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (70) * tiatteet (tleet) katub keen fii Ta-t-taawle Ta-r-raff put.ls (three) books was fii on-the-table on-the-shelf 'I put (three) books that there was on the table on the shelf/ (71) a. 1-mhaasib rati y?eebil kam tabbax ?arrarit the-accountant will meet.3sm few cook decided.3sf Nada tjaxral-un b-l-matTam Nada hire.3sf-them in-the-restaurant 'The accountant will meet a few cooks that Nada decided to hire for the restaurant/ b. Dani rati ylee?e mara Tam bifattij Talay-a Dani will find.3sm woman Asp look.3sm on-her mn s-sobti i i from the-moming 'Dani will find a woman he has been looking for since the morning/ Both sentences in (71) are unambiguous. As in Hebrew restrictive relatives involving resumption, the relative construction in (71b) has only one interpretation where there is a specific woman, say his sister, which he has been looking for since Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the morning and that he will find.4 1 In (71a), only the referential reading, but not the amount reading, is available: that is, (71a) only has the interpretation where the accountant will meet the actual chefs that Nada opted to hire for the restaurant. The evidence presented here conspires to show that indefinite relatives are radically different in not allowing movement effects, in the same contexts where definite relatives do. Any analysis that takes movement to be the only strategy for the generation of restrictive relatives faces the burden of accounting for such a difference. In the analysis I am defending here, the relativized head in an indefinite relative is base-generated in its surface position and is related to a null pronoun inside the relative clause. As such, indefinite relatives have the interpretive properties of definite relatives involving islands. This is exacdy what the paradigm (68-71) shows. The question can be raised now as to the structural representation of the relativized noun phrase in indefinite relatives. More specifically, is there in those constructions a null D° to license? If so, what licenses it? Based on the discussion so far, one can say that in a structure like (64), the ED head cannot be occupied by an indefinite 4 1 The reading that is missing in (71b) is one where Dani will find any woman who fits the description that he has in mind; say one who wears the same hat size as Jodie Foster. 258 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. article, since those cannot license null D° heads in Lebanese Arabic.4 2 Thus, a structure like (72) is an illicit one. The facts suggest that in indefinite relatives, the relativized head is not in [Spec, NumP], but in Num itself4 3 This explains the absence of yalli in indefinite relatives: yalli and the indefinite relativized head are generated in the same position and thus are in complementary distribution (73). 4 2 In fact, I will argue in Chapter 5 that an indefinite article or a numeral cannot occupy the highest functional projection in a noun phrase. Recall also that the indefinite article in Lebanese Arabic is phonetically null. 4 2 and then it may raise to the position of the external determiner, which heads DPi in (72). (72) * DPi NumP 0 kteeb 259 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (73) DPi kteeb NumP An alternative analysis would be to say that the indefinite relativized head is generated in [Spec, NumP], as in (72), and that the Num head is null. Just as adjectives in Lebanese Arabic agree with the noun they modifiy in definiteness, the null Num head agrees with the indefiniteness of the relativized DP in that case. For this analysis to go through, one would have to make the further assumption that indefinite relatives do not involve any null determiner heads. Thus, an indefinite relative would have the representation in (74). (74) NumP NumP kteeb Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Again, the data in Lebanese Arabic militate against this analysis. Recall that yalli can introduce a headless relative, however, its null equivalent cannot, and a construction like the one in (75) is unacceptable as a headless relative, in the absence of yalli.*4 (75) a * 3 eebit-o Nada brought.3sf-it/him Nada 'What Nada brought' 'The one that Nada brought' b. yalli 3 eebit-o that brought.3sf-it/him 'What Nada brought' 'The one that Nada brought' Under the assumption that there is a null Num head that dominates the relative clause in (73), the inability of the construction in (75a) to function as a DP comes as a surprise. This fact can be easily explained under the assumption that the indefinite relativized head occupies the head of NumP: In the absence of yalli the construction in (75a) corresponds to a clausal element and not to a nominal element. 4 4 (75a) is, however, acceptable as a full sentence. 261 Nada Nada Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.4 The categorial nature of the resumptive relative clause in Lebanese Arabic The attentive reader may have noted the absence of labels on relative clauses in the representations provided so far. There is good reason to believe that the projection dominated by NumP in (73) in Lebanese Arabic relatives is not a CP, but a lower Top(ic)P projection. The evidence comes from a contrast that can be observed in LA between focus movement constructions and clitic-left dislocation. The generalization is that, while it is acceptable to have a clitic-left dislocated noun phrase inside a restrictive relative, focus movement is prohibited in those contexts. This is illustrated in (76-77).4 5 (76) a* z-zalame lli ZEINA Tarrafto Tale-e the-man that Zeina introduced.2p on-him 'The man that ZEINA you introduced to him / b* zalame ZEINA ‘ Tarrafto Tale-e man Zeina introduced.2p on-him 'A man that ZEINA you introduced to him / 451 use small caps to indicate that the fronted element in focus movement constructions is interpreted contrastively. This element is also contrastively stressed. 262 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (77) a. z-zalame lli Zeina Tarraftuw-a Tale-e the-man that Zeina introduced.2p-her on-him 'The man that Zeina, you introduced her to him / b. zalame Zeina Tarraftuw-a Tale-e man Zeina introduced.2p-her on-him 'A man that Zeina, you introduced her to him / The unacceptability of the sentences in (76) results from the fact that a noun phrase undergoes focus movement inside a relative clause: (76a) illustrates the fact for definite relatives, (76b), for indefinite relatives. The sentences in (77), which involve a clitic-left dislocated noun phrase instead of the focused norm phrase, are acceptable, whether clitic-left dislocation takes place inside a definite relative, as in (77a), or an indefinite relative, as in (77b). Interestingly, focus movement patterns with wh-question word movement, as can be noted from the unacceptability of the sentences in (78).4 6 4 6 It is important to note in this respect that a wh-in-situ is acceptable in both definite relatives (ia) and indefinite relatives (ib): (i) a. hayda z-zalame lli Tarrafto miin this the-man that introduced.2p who 'This is the man that you introduced who to him.' 263 Tale-e on-him Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (78) a.* z-zalame lli miin ‘ Tarrafto 'Tale-e the-man that who introduced.2p on-him 'The man that you introduced who to him / b* zalame miin ‘ Tarrafto ‘Tale-e man who introduced.2p on-him 'A man that you introduced who to him/ In simples sentences, focus movement, like wh-movement, can co-occur with clitic- left dislocation, and in such cases, the moved focused phrase or the fronted wh- phrase can precede the clitic-left dislocated phrase (see Aoun and Benmamoun 1998): (79) a. ‘ Ta-KARIM Zeina ‘ Tarraftuw-a on-Karim Zeina introduced.2p-her 'To K a r im Zeina you introduced her/ b. ‘ Ta-miin Zeina ‘ Tarraftuw-a on-whom Zeina introduced.2p-her 'To whom Zeina did you introduce her?' b. hayda zalame 'Tarrafto miin Tale-e this man introduced.2p who on-him 'This is a man that you introduced who to him/ 264 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Another important fact to note in this respect is that focus movement does not cooccur with wh-movement in simple sentences, which indicates that focused noun phrases and fronted wh-question words are in complementary distribution: they both move to the same position, namely [Spec, CP]. (80) a.* Ta-KARIM miin Tarrafto on-Karim who introduced.2p 'To Karim who did you introduced?.' b.* Ta-miin ZEINA Tarrafto on-whom Zeina introduced.2p 'To whom ZEINA did you introduced?' The paradigm in (79) provides evidence that CP dominates a projection hosting clitic-left dislocated noun phrases. Let's call this projection Top(ic)P. Thus, we can conclude that in resumptive restrictive relatives in LA, the highest projection in the relative clause, the one dominated by the Num head that hosts yalli in definite relatives, is a topic phrase, which is dominated by CP in simple clauses.4 7 4 7 The conclusion that the relative clause in Lebanese Arabic is a topic phrase provides a possible explanation for why LA has no wh-relatives: if wh-words move to [Spec, CP], the absence of such a projection in restrictive relatives in LA accounts for the absence of wh-restrictive relatives in the 265 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.5 A problem for the raising analysis Lebeaux's generalization about the argument/adjunct asymmetry in triggering Principle C effects under reconstruction presents a problem for the raising analysis of relative clauses and particularly to an approach in the spirit of Kayne 1994, which assumes that the relative clause is a complement of the ED (or any other nominal projection). The contrast in (81) illustrates Lebeaux's generalization: (81) a* Which claim that Johni was asleep was hei willing to discuss? b. Which claim that John; made was he; willing to discuss? As we have seen earlier (see Chapter 3), Chomsky 1993 accounts for the generalization using two assumptions: (i) late adjunction and (ii) Preference Principle for reconstruction. Basically, this account makes crucial use of the claim (originally defended by Lebeaux (1988) himself) that adjuncts can be inserted late in the derivation. It is clear how such an account presents problems for the complementation structure of restrictive relatives. This structure differs from that of norm complements in one aspect only: norm complements are generated as sisters of a language. See Bianchi 1999 for a discussion on language variation with respect to the categorial nature of the relative clause. 266 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. noun head, whereas relative clauses are generated as sisters of a functional head in the nominal projection. Being a complement, the relative clause could not be inserted late in the derivation and thus we lose the account of the contrast in (81).4 8 Another problem posed by Lebeaux's generalization was pointed out to me independently by Sportiche (1999) and Jamal Ouhalla (p.c.). This is a problem that any raising analysis of relative clauses faces, regardless of die structural configuration of those constructions. If the relativized head is generated inside the relative clause in a raising construction like (81b), then the relative clause must have been inserted low in the structure before the wh-phrase containing the relative undergoes movement. Thus, a raising analysis seems to force early insertion of relative clauses into the structure. Nunes 2001 presents a 'minimalist' analysis of the contrast in (81), which is also compatible with a head-raising analysis of relative clauses. Taking the concept of Move as Copy and Merge literally, Nunes develops the notion of Sideward Movement, "where the computational system copies a given constituent a of a syntactic object K and merges a with a syntactic object L, which has been independendy assembled and is unconnected to K." (304-305) 4 8 One way of tackling the problem is to question the generalization itself, which is exactly what Bianchi 1999 does. 267 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. According to Nunes, the sentence in (81b) can be given a Sideward movement account: The computational system forms the relative CP through raising and adjunction of which claim (82a). Then a copy of which claim is merged with the verb discuss. After further computations, the derivation reaches the step in (82b). Merging the two structures in (82a) and (82b), we obtain the structure in (83). This structure allows coreference between the name John inside the relative clause and the subject pronoun he. Two chains can be formed between the higher occurrence of the noun phrase which claim and each of the lower copies based on the c-command relation that holds between them. Further computational applications at PF yield the relevant output. (82) a. [c p i [which claim]k [c p i that John made [which claim]k ]] b. [cp2 was + Q [he willing to discuss [which claim]k ]] (8 3 ) [cp 2 [cp i [which claim]k [cpi that John made [which claim]k ]] [ c was + Q [he willing to discuss [which claim]k ]]] The contrast between (81a) and (81b) lies for Nunes in the fact that the sentencial complement is the sister of the head norm so that the structure that is formed by merging the clause that John was asleep with which claim is significantly different from the one in (82a): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (84) a. [which* [claimk [that John was asleep]]] b. [cp2 was + Q [he willing to discuss [which* claimk ] ]] When the structure in (84a) is merged with the structure in (84b), as in (85), no chains can be formed between the different occurrences of xvhich and claim because there is no c-command between them, and the derivation crashes, as it cannot proceed to PF. (85) [ [which* [claimk [that John was asleep]]] [era was + Q [he willing to discuss [which* claimk ] ]] ] While Nunes' analysis resolves the problem pointed out by Sportiche and Ouhalla, I believe it makes false predictions with respect to the data. That is, we expect to find in cases similar to (81b), certain effects of movement, like reconstruction effects into the relative clause. This is not the case however in Lebanese Arabic. In cases where coreference is allowed, the external head seems to be unambiguously referential: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (86) kam tabbax ?arrarit tjaYYal-un Nada b-l-matTam few cook decided.3sf hire-them Nada in-the-restaurant badda tTuud t?eebil want.3sf come.3sf meet.3sf 'How many cooks that Nada decided to hire for the restaurant does she want to meet again?' In (86), the relativized head is unambiguously referential. Distributive readings under reconstruction are also absent, as indicated by the unacceptability of interpreting the universal quantifier phrase in (87) as having scope over the relativized head: (87) * distributive ?ayya Ja3ra badda Zeina koll 3nayneete yiTtaT-a which tree want.3sf Zeina every gardener cut.3sm-it rah teexud < Ta-l-maJtal fut. take.3sf to-the-nursery 'Which tree that Zeina wants every gardener to cut will she take to the nursery?' 270 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Thus it seems that for the coreference reading to obtain, the external head of the relative must be interpreted outside the sope of elements that occur inside the relative clause. This is what we expect if a base-generation account is available for the generation of resumptive restrictive relatives. However, the structure provided in (73) falls short of a full account for the Lebeaux problem: in this structure the relative clause is a complement to the Num head. None of the facts discusses in this chapter make this configuration necessary. In fact, we could assume that in a base generated relative, the relative clause is adjoined to the external head. Thus the account for the argument/ adjunct asymmetry observed in (81) remains the same as Chomsky's 1993 account. 4.6 The final picture In this chapter, I have argued that resumptive restrictive relatives in LA are of two types: those that can be generated by movement and those that cannot be so generated. Each type is identified by the range of determiners that can introduce the external head. These two types require two types of structures, a complementation structure and an adjunction structure. Further investigation is needed to see if the two structures map onto the two generation strategies. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 5 Movement and the Nature of Resumptives 1 1 1 the previous chapters, we have seen that the presence of resumption in Lebanese Arabic restrictive relatives does not distinguish constructions derived by movement from those that are not so derived. We have identified several properties that, taken together, form solid evidence for the hypothesis that resumption in those constructions may indeed involve movement. In this chapter, I examine how the nature of the resumptive pronominal element interacts with the availability of movement in restrictive relatives in particular, and in other resumptive constructions, in general.1 I will argue that the weak resumptive element that generally appears in restrictive relatives in LA and other resumptive constructions is in fact an agreement marker resulting from the movement of the object through the specifier position of vP. I will also examine the parallelisms between resumptive constructions and clitic-doubling constructions, and the interpretive properties of resumptive constructions. 11 will mainly be concerned with resumptives in direct object positions. I leave the examination of how the analysis might extend to obliques and genitives for future research. 272 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.1 Hebrew free relatives In Hebrew free relatives, as in Lebanese Arabic restrictive relatives, resumption may cooccur with movement. Borer (1984) argues that Hebrew free relatives are derived by movement of the wh-phrase to the complementizer projection.2 The sentences in (1) form a minimal pair with the only difference between them being the presence of a strong pronoun ?oto 'it/him ' inside the relative in (lb) and a clitic -av 'it/him 7 in (la). (1) a. kaniti ?et ma Je hexlatet ?alav bought-I ACC what that decided-you on-it 'I bought what you decided on/ b* kaniti ?et ma je ra?it ?oto bought-I ACC what that saw-you it 'I bought what you saw/ 2 Borer (1984) follows Groos and van Riemsdijk 1979 in assuming that free relatives are headed by a null pronominal element; thus, a typical surface representation for free relatives would be as in (i) [1 am substituting DP for NP, CP for S, and IP for S]: (i) [dp e [cp wfc-phrasej je [jp ... t,... ] ] ] 273 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Borer provides the representation below as the underlying structure for the sentence in (la) (see footnote 1): (2) kaniti ?et [d p e [c p [e [ip hexlatet [pp ?al-av ma ]]]] bought-I ACC that decided-you about-it what The structure under PP in (2) is a 'doubling' structure:3 the pronominal clitic -av 'it' is analyzed as a bound morpheme attached to the preposition. Crucially, it does not occupy the argument position inside the prepositional phrase, which is instead occupied by the wh-pronoun ma 'what'. Such a representation as the one given in (2) is not available in the case of direct object pronouns like ?oto in (lb). Those pronouns are not clitics in Hebrew and are analyzed as occupying the argument position of the selecting verb.4 3 1 have used quotation marks around doubling because Modem Hebrew does not make free use of surface doubling constructions, where a clitic co-occurs with a DP, as represented in (2). According to Shlonsky 1997 Hebrew doubling constructions are found only in nominals or in PPs, but not in VPs. The analysis argued for in this chapter aims partly at accounting for this generalization. 4 Evidence for movement in free relatives in Hebrew comes from the fact that these constructions are sensitive to islands, as illustrated below: 274 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. As Shlonsky (1997) points out, ?et pronouns in Hebrew display characteristic properties of strong pronouns (see Cardinaletti and Starke 1994): they can be left dislocated (3a) and coordinated with noun phrases (3b). (3) a. ?oto, lo ra?inu ?etmol him NEG saw.lp yesterday 'Him, we didn't see yesterday.' b. lo ra?inu ?et Dani ve- ?oto NEG saw.lp ACC Dani and him 'We didn't see Dani and him.' This is in sharp contrast with clitics, which can neither be dislocated nor coordinated with DPs. Shlonsky analyzes Hebrew ?et pronouns as being bi- morphemic, involving the particle ?et and a null pronoun which is identified by the agreement morpheme that shows up on the particle. The paradigm of Hebrew accusative pronouns is given in table (4). (i) * ze ma fe pagafti ?et ha-?ish fe hexlit ?alav this what that met-I ACC the-man that decided on-it 'This is what I met the man who decided on.' 275 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (4) Table 4. Hebrew accusative pronouns Singular Plural 1 ?ot-i ?otanu 2m ?ot-xa ?otxem 2f ?ot-ax ?ot-xen 3m ?ot-o ?ot-am 3f ?ot-a ?ot-an Thus, as in Cardinaletti and Starke 1994, the distinction between clitics in Hebrew and ?et pronouns is reduced to a difference between heads and XPs. To complete the picture and further show that free pronouns in Hebrew are indeed argument XPs, Shlonsky (1997), citing Siloni 1994, argues that ?et pronouns are licensed in case positions. This conclusion is based on the contrast between free pronoun and full DP licensing with derived nominals. (5) a. kri?at ha-kacin ?et Dan le-seder hifti'Ta ?et J?ar calling the-officer ACC Dan to-order surprised.3sf ACC other ha-xayalim the-soldiers 'The officer's calling Dan to order surprised the other soldiers.' 276 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b * kri?at ha-kacin ?oto la-seder hifti'Ta ?et J?ar calling the-officer him to-order surprised.3sf ACC other ha-xayalim the-soldiers 'The officer's calling him to order surprised the other soldiers.' The unacceptability of (5b) is taken as an indication that (accusative) ?et pronouns need to check their case feature in AgrOP. Since this licensing position is absent in derived nominals, the pronoun in (5b) is not case licensed and the sentence is unacceptable. The discussion of the Hebrew facts regarding free relatives highlights the interaction between the nature of the resumptive pronominal and movement in resumptive constructions: Only when the sentence initial operator can be generated in a 'doubling' structure with the resumptive element is movement possible. The conclusion then seems to be that when movement is unambiguously available in the generation of an A-construction, the availability of resumption in that construction is correlated with that of a 'doubling' structure. My goal in the remainder of this chapter is to account for this generalization and examine its consequences. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.2 Italian clitic left dislocation revisited The study of the properties of reconstruction in Italian clitic left dislocation has led Cecchetto and Chierchia (1998) (C&C) to a similar conclusion. The reader may recall that what those authors were examining was the contrast between NP CLLD and PP CLLD. I provide a summary of the relevant properties of those two constructions in the table in (6). (6) Table 5. Clitic Left Dislocation in Italian NP CLLD PP CLLD ■ Obligatory clitic ■ Optional clitic ■ Availability of scope reconstruction ■ No scope reconstruction ■ Availability of doubling structure ■ Absence of doubling structure As the table in (6) clearly shows, the movement properties of NP CLLD, such as the availability of scope reconstruction, are linked to the availability of clitic doubling for NPs.5 Conversely, the non-availability of movement for PP CLLD, as evidenced by the absence of scope reconstruction, for instance, is linked to the absence of clitic doubling with PPs, as the following contrast from the Trentino dialect shows: 5 Italian, unlike Modem Hebrew, has surface clitic doubing constructions in VPs. 278 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (7) a. El *(me) vol mi He *(me) wants me 'He wants me.' b. (*Ghe) vago volintiera a Roma (*There) (I) go with pleasure to Rome 'I go with pleasure to Rome.' The analysis provided to account for the facts summarized in (6) is that while NP CLLD involves movement of the NP to a dislocated position in the sentence, a CLLDed PP is generated in the left periphery of the minimal clause containing the corresponding clitic and then fronted to its sentence initial position.6 The idea that emerges then is that a doubling structure, where the clitic is generated together with the CLLD NP (Big DP in (8b)) is the underlying structure for a sentence like (8a). 6 While I have adopted the basic insight that PP CLLD involves the base generation of the dislocated PP in the left periphery of the minimal clause which contains the corresponding clitic, with subsequent movement of the PP to a sentence initial position, I have argued against a chain binding approach to account for the properties of PP CLLD listed in (6) (see chapter 3). 279 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (8) a. Maria, Leo la incontra Maria, Leo her met 'Maria, Leo met her/ b. [TopicP ••• [ip Leo incontra [agrop • • • [vp ... [big dp Maria la] ...]]]] c. [TopicpM aria [ipLeo la incontra [agrop [big d p tMaria tia] [vp ... tBicD P •••]]]] As the surface representation of the sentence in (8c) indicates, NP CLLD involves movement of the doubling structure or Big DP to [Spec, AgrO], followed by subsequent movement of the clitic to attach to the verb and the fronting of the CLLDed NP to [Spec, TopicP].7 Such a doubling/Big DP structure is not available for PPs in Italian and therefore a representation parallel to (8b) cannot be posited for PP CLLD.8 It is important to 7 The derivation sketched through the representations in (8) is very similar to the one Boeckx (2001) assumes for resumptive constructions, but it is argued for in a different framework. Such a derivation raises die question as to how case is assigned to the name Maria and/or the doubling clitic: Is Big DP assigned the case of the direct object? How does this case 'percolate' to the relevant argument? 8 C&C further argue that movement is a last resort strategy in CLLD constructions. Their discussion is based on the behavior of dative PPs. It is well known that dative PPs have an intermediate status between DPs and adjunct PPs. On one hand, they can be clitic doubled, and on the other, the clitic related to a CLLD dative PP is always optional. One might expect then that there will be two derivations possible for constructions involving CLLD dative PPs. However, it is observed that CLLD 280 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of dative PPs patterns with that of adjunct PPs: a CLLD dative PP cannot reconstruct for scope or binding, as the sentences in (i-ii) illustrate. (i) A qualche professore, Leo (gli) ha assegnato ogni studente 3V but *V3 To some professor Leo to him has assigned every student 'To one professor, Leo assigned every student.' (ii) *A1 professore che 1,'aveva messo in difficolta, (gli) ho gia presentato [ogni studentejj To the professor that him had put in difficulties (I) to him have already introduced every student If movement is indeed a last resort strategy, then the parallel behavior of dative PP CLLD and adjunct PP CLLD is accounted for. The question can then be raised as to why NP CLLD must be generated by movement and the authors argue that it is because the NP must get structural case. PPs on the other hand do not need (structural) case. Based on evidence from left dislocation (Cinque 1990 and Aoun and Benmamoun 1998), I reject this idea. NPs can be generated in sentence initial position, as left dislocated elements, and be related to a pronominal element within the sentence, which might occur inside an island. Such sentences are perfectly acceptable: (i) Karim, miin baddo yaTrif leej [afiatu-u min 1-madrase Karim who want.3sm know.3sm why expelled.3p-3sm from the-school 'Karim, who wants to know why they expelled him from school?' A sentence like (i) could not have been derived by movement from the position of the resumptive element. The question arises as to how Karim is case licensed but here, with an acceptable result. All one could say is that, in those contexts, the absence of structural case assignment does not lead to a crashing derivation. 281 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. note that the presence or absence of a surface clitic doubling structure in Italian cannot be taken as direct evidence for the availability of movement or its non availability. We have seen that in Hebrew, a language that does not allow surface clitic doubling structures with noun phrases, can still allow movement in a resumptive free relative. The explanation of the facts however should still account for the correlation between the availability of clitic doubling constructions and the availability of movement in resumptive CLLD constructions in Italian. 5.3 Two analyses of clitic doubling The question now arises as to what licenses this Big DP or the 'doubling' structure that gives rise to resumptive constructions, especially in those languages like Modern Hebrew that do not realize those structures overtly. That is, what is the reason why PPs in Italian cannot have a doubling structure as in (8b) and why are Hebrew Pet pronouns excluded from such a construction as well? Borer 1984 suggests that the strong/clitic distinction is relevant. Only clitic pronouns can cooccur with a noun phrase inside the constituent termed BigDP in (8b). The difference is then reduced to the fact that strong pronouns are structurally more complex than clitics. This explanation does not however carry over to the case of PP CLLD where the resumptive element is indeed a clitic with the same structural properties as argument clitics in Italian. The discussion in Shlonsky 1997 regarding 282 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the nature of ?et pronouns and in C&C regarding the contrast between PP CLLD and NP CLLD suggests an answer in the agreement properties of those elements. In this section, I argue that in order for a 'doubling' structure to underlie movement in resumptive constructions, two conditions must be met: (i) the resumptive element must not occupy the argument position in the doubling structure (i.e. it must not get case from v) and (ii) the doubled XP is a noun phrase that agrees with the verb. That is, movement of direct objects in resumptive constructions is contingent upon movement to a vP peripheral position and hence on agreement with the verb. This will account, on one hand, for the difference between strong pronouns and clitics in their ability to act as resumptive elements in constructions generated by movement, and on the other, for the difference between DPs and PPs in being licensed inside a 'doubling' structure. Bianchi (1999) suggests that Belletti's 1994 structure of the BigDP accounts for doubling constructions as well as resumption. According to Belletti 1994, clitics are generated in the highest functional projection (D°) of the doubled/resumed DP, as illustrated in (9). (9) BigDP PP Do clitic DP doubled element 283 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Clitics are further endowed with a strong case feature that needs to be eliminated by Spell-Out and which is responsible for their movement to the AgrO projection {case absorption). The functional projection occupied by the clitic takes a complex constituent as complement, a PP. The P head selected by D° is responsible for assigning case to the doubled DP, making sure that this DP satisfies the case filter. Under certain, not very well understood, circumstances the preposition can be deleted (or null), thus allowing D° to select a 'bare' noun phrase. Such would be, for instance, the case of NP CLLD in Italian. The structure proposed in Belletti 1994 proves to have several advantages. It readily accounts for the presence of the "dummy" case marker in doubling constructions like (10a) in Romanian and (10b) in Spanish (Bianchi's 1999 (106a) and (106b) respectively). (10) a. Romanian Fiecare professor ii va examina pe every teacher themcL will examine P 'Every teacher will examine ten students.' b. Spanish Lo vimos a Juan himcL saw.lp P Juan 'We saw Juan.' 284 zece elevi ten students Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In (lOa-b) the constituent glossed as P occupies the P head in (9). Furthermore, it explains how a doubled element can satisfy the case filter and be assigned case when the doubling clitic is a case bearing element as well. Belletti's structure also accommodates cases where the doubled element is larger than a 'bare' NP, which is clearly the case in (11), a clitic-left dislocated construction from Lebanese Arabic: (11) kail 1-wleed jafnee-hun Tala ja 3 ra all the-children saw.lp-them on tree 'All the children, we saw them on a tree.' It is clear that the CLLDed constituent in (11) is a noun phrase, which involves a structure larger than an NP: the quantifier kali 'all' selects a definite plural noun phrase. This fact is not easily accommodated in analyses that assume the doubling structure to involve a simple DP headed by the clitic, as illustrated below: (12) DP doubled element 285 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. It is with such facts in mind that proponents of the analyses that assimilate resumptive constructions to doubling constructions have proposed the BigDP structure (see Belletti 1995, Uriagereka 1995, Checchetto and Chierchia 1998).9 A problem with Belletti's proposal however is posed by PP CLLD in Italian. Recall that C&C's analysis of those constructions as involving base-generation of the PP separated from its corresponding clitic rests on tire observation that PPs do not enter doubling constructions. However, the structure in (9) suggests that doubling constructions with PPs ought to be readily available and cannot be ruled out without additional stipulations. Sportiche (1992) proposes to analyze (Romance) clitics as complex agreement morphemes that head their own projection, which are called clitic voices; these are functional projections dominating the VP. Their heads (the clitics) enter into an 9 Other analyses are available in the literature which assimilate resumptive constructions to doubling constructions without providing an explicit structure to accommodate this similarity (see Kayne 1994, Dobrovie-Sorin 1990, Zubizarreta 1996, among others). Such analyses are anticipated in Kayne 1975, where it is suggested that in doubling constructions the clitic and its double form a constituent upon first merge. 286 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. agreement relation, a Spec-Head relation, with an XP, i.e. the doubled noun phrase, as dictated by the clitic criterion in (13) :10 (13) Clitic Criterion AtLF (i) A clitic must be in a Spec-Head relation with a [+F] XP (ii) A [+F] XP must be in a Spec-Head relation with a clitic In the context of accusative clitics [+F] corresponds to specificity. The reflex of the Spec-Head relation is understood to be expressed by case, number, gender and person matching between the clitic and the doubled XP. Thus under this analysis, in a sentence like (1 0b) (repeated below), 10 In the framework I am adopting here, i.e. that of Chomsky 1998,2000, Spec-Head relations are limited to the elimination of EPP features and have no special theoretical status. Checking is eliminated in favor of Agree, and checking domain in favor of local domain. However the clitic criterion can easily be recast in terms of Agree, as in (i): (i) Clitic Criterion (Minimalist Version) a. A clitic must agree with a [+F] XP. b. A [+F] XP must agree with a clitic. I will continue to use the term Spec-Head relation, keeping in mind that it has only a descriptive value. 287 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (10) b. Lo vimos a Juan himcL saw .lp P Juan 'We saw Juan.' the clitic lo would be generated as the head of a clitic projection CIP dominating the VP. The clitic then moves along with the verb to the sentence initial position (14a). The double a Juan raises at LF in order to satisfy the clitic criterion (14b). (14) a. [ip lo-vimos [q p tj0 ... [vp tv [d p a Juan] ] ] ] b. [ip lo-vimos [q p [dp a Juan] t/<,... [v p tv tD p ] ] ] In CLLD cases it is assumed that the clitic criterion is satisfied overtly and hence necessarily at LF. Under Sportiche's analysis then, the clitic lo in (10b) is not the argument of the verb vimos 'we saw'. The argument position is in fact occupied by the double a Juan. The overt movement of the double into the CLP explains the facts about participle agreement one can observe in languages like French. The generalization that accounts for the distribution of participle agreement in French (and other Romance languages) is given by Sportiche (1992) as follows (15), and is illustrated in (16): 288 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (15) Sportiche's generalization for participle agreement A participle may (or must, depending on the language) agree with its accusative direct object when it precedes the participle. Agreement is excluded when it follows the participle. (16) a. Jean a peint(*e) la porte Jean painted the door. b. La porte que Jean a peint(e) t the door that Jean painted c. Quelle maison as-tu construit(e) t ? Which house have you built? Participle agreement is taken to be the reflex of a Spec-Head configuration that obtains between the participle and the direct object, as a result of the movement of the double into AgrOP, the projection where the double is assigned case by the participle, on its way to C1P (or even CP). The absence of agreement morphology in (16a) is due to the fact that movement into AgrOP didn't take place overtly.11 1 1 Separating the projection where participle agreement is checked and the one where the clitic criterion is checked implies that the two types of agreement are not necessarily linked. This is exactly what we observe in cases where overt wh-movement of the direct object does not yield participle agreement, as illustrated in (16b-c) in the text. Sportiche (1992) suggests that AgrOP may be skipped in 289 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Under a Sportiche-type analysis, an explanataion for the unavailability of doubling structures with PPs is readily available. PPs do not enter into agreement relations.12 PPs do not carry agreement features like 9 -features or specificity. Thus PP fronting does not trigger participle agreement in French. (16c) contrasts with (17) where the option of having participle agreement is not available. (17) De quelles etageres as-tu pris(*es) les livres? 'From which shelves have you taken the books?' We have contrasted two analyses of doubling structures: one that assumes the clitic to be the head of the argument being doubled and selecting a PP complement (Belletti 1994); the other, assumes the clitic to be a complex agreement morpheme that enters an agreement relation with the doubled argument (Sportiche 1992). While both analyses fare well with the distinction between clitics and strong pronouns as resumptives, they differ in their predictions with respect to PP those cases. Within minimalist assumptions, this option is not available. I will return to this issue shortly. 12 Pied-piping cases present an obvious problem for this generalization. (16) in the text involves a case where a PP is pied-piped to the sentence initial position in order to check the strong [+wh] feature of CP. Thus whereas a [+wh] feature can percolate up and lead to the whole PP standing in for the wh- word that it contains, 9 -features do not seem to do so. The conditions under which pied-piping is licensed are beyond the scope of the analysis conducted here. 290 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. dislocation. Whereas a Sportiche-type analysis readily accounts for the absence of movement with PP dislocation due to the absence of agreement features on those phrases, a Belletti-type analysis does not. Furthermore, it can be noted that in a Sportiche-type analysis there is no mechanism for case absorption: the doubled NP is indeed the argument of the verb, which readily accounts for the presence of an agreement relation between it and the verb, an agreement relation that can be realized overtly under certain conditions. In Belletti's 1994 analysis, the doubled NP is not licensed by the verb, as it is assigned case by the dummy marker that introduces it. The agreement relation that can be established between the doubled element and the verb in those cases needs further stipulations to be accounted for. Finally, clitics under a Sportiche-type analysis are taken to be complex agreement morphemes and not the equivalent of pronouns: they do not carry case features. The predicate clitic le in French (18a), ethical dative clitics (18b), inherent clitics (18c), as well as adjunct clitics are all clear examples of non-cased clitics and a Sportiche-style analysis allows the existence of such clitics in principle, since clitics are not themselves arguments. (18) a. Louis estfou/Louis l'est French Louis is crazy/ Louis it-is b. Je t'acheterais un cadeau a Pierre I tell ya, I would buy Pierre a present 291 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. Pierre en a bave Pierre of-it drooled (=Peter suffered) d. Pierre y a fait ses etudes Pierre there has done his studies 5.4 W eak resumptive elements in Lebanese Arabic as agreement markers Similar questions can be raised regarding resumptive constructions in LA: What is the nature of the weak resumptive element in LA? What relation does it entertain with its antecedent in movement contexts? And what licenses the 'doubling' structure that underlies movement in resumption. I will argue that weak resumptive elements that appear on verbs in LA are agreement markers reflecting cp-features of the direct object with which the verb agrees. The argument will be based on four main observations listed in (19): (19) a. Accusative weak pronouns in LA are not incorporated pronouns. b. Accusative weak pronouns in LA cooccur with full lexical DPs. c. Accusative weak pronouns in LA are only associated with arguments. d. Accusative weak pronouns in LA, unlike Romance accusative clitics, do not carry single interpretation (i.e. [+specific]). 292 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. First, I would like to argue against treating weak pronominal elements as incorporated pronouns (see Benmamoun 2000 and Shlonsky 1987).1 3 Evidence that WREs in LA are not incorporated pronouns can be found internal to LA. Analyses that advocate pronoun incorporation (see McCloskey and Hale 1984, Fassi Fehri 1993) aim mainly at accounting for the complementary distribution one can observe between these pronominal affixes and full lexical DPs. While the generalization seems to apply in LA when one looks at a paradigm limited to cases of direct object norm phrases (2 0 ), it cannot extend to contexts where the weak pronominal element appears with subject DPs. (20) a. zarnee-ha (*Laila) mbeerih visited.lp-her (*Laila) mbeerih 'We visited Laila yesterday.' b. zarna *(Laila) mbeerih visited.lp *(Laila) mbeerih 'We visited Laila yesterday.' 13 For the opposite view, see Fassi Fehri 1993. As Benmamoun (2000) notes, the argument that weak pronominal elements in LA are not incorporated pronouns does not contradict the hypothesis that those pronominal elements historically derive from pronouns. 293 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In fact weak pronominal elements cooccur with subject DPs when the predicate is non-verbal. Thus, observe the following examples: (2 1 ) a. 1-wleed wayn(-un)? the-children where(-them) 'Where are the children?' b. 1-baneet kiif(-un)? the-girls how (-them) 'How are the girls?' c. Kariim badd*(-o) 1-kteeb Karim want.3sm the-book 'Karim wants the book.' d. Laila fiy*(-a) trxiib Laila in-her leave.3sf 'Laila can leave.' In all the sentences above, the weak pronominal element that shows up on the predicative head agrees with its subject DP. In (21a) and (21c) the predicate is 294 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. nominal. Badd, which is translated as 'want', does not carry verbal morphology in LA. In (21b), the predicate is adjectival and in (21d) the predicate is prepositional. 14 It is important to note that in all the above cases the doubled DP is not introduced by the dative preposition la 'to', which usually introduces doubled DPs, as illustrated in (2 2 ): (22) a. mjiit maT-a *(la)-Zeina walked.ls with-her *(to)-Zeina 'I walked with Zeina.' b. Nada Jeefat-o *(la)-Karim Nada saw.3sf-him *(to)-Karim 'Nada saw Karim.' 14 Additional instances can be found in the use of prepositions like ?9nd 'at' or roaf'with' to express possession, as illustrated in (i): (i) a. Zeina ma< T *(-a) kteeb teerix Zeina with-her book history 'Zeina has a history book with her.' b. Zeina Tand*(-a) kteeb teerix Zeina at-her book history 'Zeina has a history book.' 295 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In (21), we have then cases of DP subjects cooccuring with a weak pronominal element, giving us clear indication that those pronominal elements are not pronorms that absorb case. In fact, they are agreement markers, relating predicates to their subjects. What the sentences in (21) further show is that the distinction between what is called subject-verb agreement and object agreement is not the correct one to characterize the difference between the two sentences like (23a) and (23b): (23) a. Karim badd-o 1-kteeb Karim want-him the-book 'Karim wants the book/ b. Zeina fall-it Zeina left.3sf 'Zeina left.' In both cases the DP with which the predicate agrees is the subject. However, the morphological reflex of this agreement is different in both cases. The distinction in the morphology of agreement is in fact related to the categorical nature of the predicate, as we observed earlier: whereas the verb fall 'leave' in (23b) is [+V] and can therefore inflect for tense, the verb baddo 'want' in (23a) is [-V] and cannot inflect for tense. This is true for all the other heads that carry what we have called object 296 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. agreement, i.e. nouns, quantifiers, prepositions, etc. That is, only heads that can carry tense morphology can also carry what we have called subject agreement. The past tense of the verb baddo 'want', for example, is expressed periphrastically using the verb keen 'be', as in (24), -whereas fall 'left' in (23b) inflects for tense. (24) Kariim keen badd-o 1-kteeb Karim was.3sm want-3sm the-book 'Karim wanted the book.' Interestingly, the copula itself carries subject agreement morphology, as other verbs do: (25) a. Zeina keen-it b-l-beet Zeina was-3sf in-the-house 'Zeina was at home.' b. 1-wleed keen-o b-l-beet the-children were-3p in-the-house 'The children were at home.' We can conclude that the so-called subject agreement is more precisely agreement between a DP and a tense head; object agreement, on the other hand, is the reflex of 297 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. agreement between a DP and a head different from tense.15- 16 Thus, the weak resumptive elements in LA cannot be synchronically analyzed as incorporated pronouns and are better understood as agreement markers themselves. 15 Keeping in mind this conclusion, I will still refer to agreement between T and DP as subject agreement and that between DP and a head different from T as object agreement. 16 In Benmamoun 2000, it is argued that in the imperfective, verbs in Arabic do not raise to T and that T in the imperfective does not carry a [+V] feature to be eliminated by verb raising. This conclusion is at odds with the generalization that subject agreement morphology is the reflex of agreement between a T head and a DP in the following way: If subject agreement is realized on T, and a verb in the imprefective does not raise to T, then we expect that subject agreement in those cases will be realized on heads other than the verb, which is never the case. (i) a. 1-wleed rah yfill-o the-children fut. leave-3p 'The children will leave.' b. Lina Tam te-drus Lina Asp. 3sf-study 'Lina is studying.' A suggestion can be made to resolve this potential mismatch between syntax and morphology: The agreement morpheme is realized on the verb in the morphological component of the grammar. That is, after Spell-Out, terminal elements (i.e. heads) are rearranged in such a way as to attach the subject agreement morpheme on the category [+V]. For recent suggestions that head movement could be done post-Spell Out in the PF side of the grammar, see Chomsky 1998,2000. Whether head movement takes place in narrow syntax or in PF, one can contrast Lebanese Arabic with other Arabic dialects and 298 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. conclude that even in the imperfective, the verb in LA raises to T. An indication that this might be the case comes from the fact that in LA, unlike in the dialects of Arabic examined by Benmamoun, the verb in the imperfect can merge with the negative - / morpheme (ib) and easily appear before its subject (ii). In that sense, the imperfective verb in LA behaves like the perfective verb: (i) a. ma Takalto-J Lebanese Arabic neg ate.ls-J 'I didn't eat.' b. ma bteekul-j neg 3sf.eat-j 'She doesn't eat.' c. mij byiktib Egyptian Arabic neg write.3sm 'He doesn't eat.' (ii) a. byeekul Sami teffeefi Lebanese Arabic eat.3sm Sami apple 'Sami eats apples.' b. ?akal Sami teffeefi ate.3sm Sami apple 'Sami ate apples.' 299 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. As agreement markers, object agreement markers are also distinct from Romance- type clitics. An observation made by Shlonsky (1997) is that there are distinctions to be made between Romance clitics and Semitic weak pronominal elements: most importantly, while Romance clitics seem to attach on the highest verbal position in a tree, as seen in (26), (26) a. Je le vois danser. b. Je l'ai mange. Semitic weak pronominal elements attach to the closest c-commanding head, be it a lexical head or a functional one (27). (27) a. ?aalo ?onn-a ?8 3 it said.3p that-3sf came.3sf 'They said that she came/ b. ?axad kteeb-o took.3sm book-3sm 'He took his book/ If merger with the negative morpheme as well as the VSO order are obtained via head movement, then we can account for the observation that subject agreement is realized on the verb in the imperfective as well as the perfective in Lebanese Arabic. 300 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c. ?aasas8 t-o i i punished.3sf-3sm 'She punished him.' Furthermore, while Romance clitics can undergo movement as XPs, Semitic clitics are affixal in nature. Consider the following paradigm (Sportiche's (1992) (18a-b) and (2 0 b) respectively): (28) a. Jean a vu une/la/m a/cette photo de qui? Jean saw a / the picture of whom? b. ... dontj Jean a vu [une/ la/ *ma/ *cette photo tj ] ... of whom Jean saw a/ the/ *my/ *this picture c. Jean enj a vu [une/la/*ma/*cette photo tj ]. Jean of-him saw a/the/*m y/*this picture. The contrast between (28a) and (28b) illustrates the following generalization about extraction out of direct object DPs in French: (29) (=Sportiche's (1992) (17)) An XP may be extracted out of a DP in a given structure iff XP may otherwise appear as the possessor of this DP. 301 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This interpretation is blocked in the presence of an alternative possessor, i.e. ma 'm y', or a demonstrative, i.e. cette 'this', in (28b). (28c) shows that the same generalization holds of clitic extraction, indicating that clitic extraction is akin to XP movement in those cases. The paradigm in (28) has no equivalent in Lebanese, where clitics do not undergo movement at all.17 1 7 The equivalent of (28c) in LA displays ungrammaticality only in one case: (i) Laila xazza?it-le suura/ ??s-suura/ suurt-e/ ha-s-suura Laila tore.3sf-me.dat picture/ ??the-picture/ picture-ls/ this-the -picture 'Laila tore up for/of me a picture/the picture/my picture/this picture.' The dative pronominal element attached to the verb can be related (as theme, object) to the DP that follows, in all but the case where the DP is a simple definite. The dative pronominal there can only have a 'possessor' reading, a reading that is available in all the other cases. The paradigm in (i) is reminiscent of the one discussed in chapter 4 with respect to the issue of the definiteness specification of definite relatives. The object/theme reading seems to be incompatible with the presence of the definite article. Interestingly, if we replace the pronominal dative by a full lexical noun phrase, the sentences become unacceptable under the possessor reading in all but the case where the following DP is indefinite: (ii) Laila xazza?it la-Sami suura/ *s-suura/ *suurt-o/ *ha-s-suura Laila tore.3sf to-Sami picture/ *the-picture/ *picture-3sm/ *this-the-picture 'Laila tore up for Sami a picture/the picture/his picture/this picture.' 302 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Whereas weak resumptive elements in LA do not pattern like Romance clitics, they parallel the distribution of participle agreement in French (and the other Romance languages). They share the following properties: (i) LA weak pronominal elements appear when the postverbal object position is empty, (ii) they are optional if the object appears in a position preceding the verb, as the result of A-movement, for example, and (iii) they are associated with DPs, elements that carry 9 -features. These properties are illustrated in (30), (31), and (32) respectively. (30) a. ?ayya beeb dahanat-(o) Laila? which door painted.3sf-it Laila 'Which door did Laila paint?' b. Quelle porte Laila a-t-elle peint(e)? which door Laila has-she painted 'Which door has Laila painted?' (31) a. Zeina lammarat-a (*l-bineeye) Zeina built.3sf-it (*the-building) 'Zeina built it.' In cases where the dative possessor is a full noun phrase, the sentences display definiteness effects (ii). Thus, even if movement is involved in (i) under the possessor reading (Landau 1999), this will not be XP movement as the contrast between (i) and (ii) indicates. 303 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. Marie l'a construit(e) (*la maison) Marie it-has built (the house) 'Marie has built it.' (32) a. man ?ayya raff ?axattu-(*u) l-katub? from which shelf took.2p-3sm the-books 'From which shelf have you taken the books?' b. De quelles etageres as-tu pris(*es) les livres? 'From which shelves have you taken the books?' The optionality of the presence of these weak elements in (30) and their complementary distribution with full lexical DPs in (31) cannot be taken as evidence against those elements being analyzed as agreement morphemes. In French, where direct object participle agreement is clearly distinguished from direct object clitics (they can occur together in (31b)), this agreement morpheme is also optional, as evidenced by the acceptability of (30b) without the agreement morpheme. 5.4.1 Subject and object agreement morphology We have shown so far that weak resumptive elements in LA are to be analyzed as object agreement markers: affix-like elements that realize an agreement relation between a DP and a head different from T. In the cases I am focusing on here, this 304 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. head is the verb. Two characteristics of this type of agreement that need to be accounted for are: (i) that it is optional, and (ii) it occurs when the postverbal object position is empty. Thus, in LA object agreement contrasts with subject agreement in that the latter is generally obligatory. Shlonsky 1997 also observes that subject agreement morphology attaches to the verbal root before object agreement morphology. Based on this ordering of the different agreement morphemes with respect to the verbal head Shlonsky argues that subject agreement is lexical, that is, present on the verb upon its insertion into the numeration. Object agreement on the other hand, is affixal in nature and is the result of the movement of a DP into the projection that hosts the agreement affix. The distinction between subject agreement and object agreement can be reduced to a difference between derivational and inflectional morphology, as discussed in Benmamoun 2000. Examining the distribution of subject-verb agreement morphology in Standard Arabic, Benmamoun observes that this kind of morphology serves to distinguish verbs from nouns. As (34) indicates, the generalization is also true of LA. (33) a. yu-saaTid b. mu-saaTid Standard Arabic 3-help N-help 'He helps' 'Helper' 305 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (34) a. darb-it b. darb-e Lebanese Arabic hit.3sf hit-N 'She hit' 'A hit' In fact, there is a complementary distribution between person agreement morphology on the root and the nominalizer in (33), so that only the presence of the person morpheme can be said to identify the root as a verbal one. By analogy with Standard Arabic, in LA, person morphology can also be said to identify roots as being verbal ones. In (34a) the morpheme that stands for the entire subject agreement complex could also stand for the [+V] feature on the verb: In the perfective, the person morpheme, as well as number and gender morphology are fused together into one affix (Halle and Marantz 1993). In the imperfective, the person morpheme, the morpheme in complementary distribution with the nominalizer, is realized separately as a prefix, and gender and number morphemes are fused together and realized as a suffix: (35) a. t- seeTd- e b. m- seeTd- e In that respect, it is interesting to note that in LA, as in all Arabic dialects, there are no infinitives, i.e. verbal stems without person morphology. We thus have an 306 2 p- help- fs N- help- fs 'You (f.) help.' 'helper (fs.)' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. explanation for the obligatoriness of person agreement (and in the perfective subject agreement) morphology in Arabic, and its close connection to the verbal root in those languages. However, something remains to be said about the form of the number agreement morpheme, and generally the subject agreement morpheme. Namely, why is agreement obligatory for subjects in LA, but not for objects? (36) a. ?ayya fialiib 1-wleed Tam y-jrab-*(o) (u)? which milk the-children Asp. 3-drink-(p)(3sm) 'Which milk are the children drinking (it)?' In other words, why must the verb in (36b) show subject agreement but not object agreement? The aspect of Benmamoun's (2000) analysis discussed so far accounts for the obligatory presence of person morphology on the root and for its direct relation to the characterization of roots as verbal ones, but it does not make any claims as to the shape of number agreement. An answer is suggested by the different specifications of verbs and tense heads with respect to the EPP feature: whereas T obligatorily has an EPP feature, this feature is only optional on v, being determined by the presence of another EPP feature in the sentence (Chomsky 1998, see also chapter 1). If we take overt agreement morphology in LA to be the reflex of the elimination of EPP features, then agreement between subject and verb in 307 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. number is the reflex of the elimination of T's EPP feature; object agreement, on the other hand, is the reflex of the elimination of v's EPP feature. To appreciate how this analysis works in LA, consider the sentence in (37a) and its representation in (37b). (37) a. 1-wleed ?akal-o 1-gato the-children ate-3p the-cake 'The children ate the cake/ b. [tpSU T ... [v p SU v [V p VOB] ] The obligatory presence of an EPP feature on T requires movement of a DP into [Spec, T]; the subject l-wleed 'the children' generated in [Spec, v], being active due to its unchecked uninterpretable case feature moves to eliminate the EPP feature on T and have its case assigned. The movement results in the appearance of subject agreement. The affix shows up on the verb, which has moved to T in Arabic (see Shlonsky 1997, Benmamoun 2000, Aoun et al 1994, and for an alternative see footnote 16). Interestingly, in the VSO order in Lebanese Arabic, the verb shows subject agreement as well: 308 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (38) ?akal-o 1-wleed 1-gato ate.3p the-children the-cake 'The children ate the cake/ Following Aoun et al. 1994,1 assume that in such a case, the subject has moved from its base position in [Spec, v] to eliminate the EPP feature on T and that the verb has undergone movement past the tense head to a higher functional projection ( see also Ouhalla 1994). Object agreement results from the same mechanism as subject agreement, with the difference being that the EPP feature in v is only optionally available (Chomsky 1998). In a sentence like (37) or (38), where the object remains in situ, an agreement relation can be established with the verb (v), which results in case assignment. The object, which is in the local domain of v, need not move and no agreement morphology appears on the verb. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In the context of A-movement to a sentence peripheral position, the phase impenetrability condition requires there to be an EPP feature in v. The sentences below illustrate cases of wh-movement (39a), CLLD (39b), and relativization (39c).18 (39) a. ?ayya Taleeme baddkun kali walad ya*Traf-a which grade want.2p every child know.3sm-3sf 'Which grade do you want every child to know?' b. Taleemit fahs-o baddkun kail walad ya*Traf-a grade exam-his want.2p every child know.3sm-3sf 'The grade of his exam, you want every boy to know it.' c. 1-laleeme lli baddkun kail walad yaTraf-a the-grade that want.2p every boy know.3sm-3sf tolTit I came-out.3sf 'The grade that you want every boy to know it is out.' The avaHability of a distributive reading in the sentences in (39) can be taken as evidence that those sentences are generated by movement. If irrelevant details are 18 It is important to note here that the availability of the distributive reading in (39) cannot be the result of QR. In fact the universal quantifier phrase in all the sentences in (39) cannot take inverse scope over the subject of baddo 'want'. 310 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. omitted, the A-constructions above can have a similar representation, given in (40).” (40) [ O B w h / O B c l l d /Relativized OB... [v pOB [vp SU v [V OB] ] ] Due to the phase impenetrability condition, OB will not be able to reach its designated sentence peripheral position unless it moves to [Spec, vP] first. If it remains inside vP when the derivation has reached the next (strong) phase, it will not be accessible for the computation. OB moves to [Spec, vP] then in order to check the EPP feature in vP, thereby deleting the uninterpretable cp-features on V and its own case feature, and thus triggering object agreement. At the next phase level, i.e. CP, the object is thus accessible for movement and can be extracted out of the vP in order to reach its final landing site. The importance of the analysis adopted here is that it establishes a direct link between the presence of object agreement morphology and movement, which 19 In fact, CLLD moves noun phrases to an IP peripheral position below the complementizer. CLLDed elements can occur below the complementizer in LA: (i) lirfit Nada Torino Sami Jeefat-o Laila knew.3sf Nada that Sami saw.3sf-him Laila 'Nada knew that Laila saw Sami.' 311 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. immediately captures the generalization that agreement is present only when the postverbal object is empty. What about cases where the object is a null argument (pro)? Our analysis leads us to the conclusion that the null argument pro moves out of vP in order to trigger agreement morphology. Evidence that favors this analysis comes from the observation that (null) pronouns are interpreted higher than their lexical noun phrase counterparts. Consider the contrast below discussed in Aoun and Benmamoun 1998 (attributed by the authors to an observation by Maria Luisa Zubizarreta): (41) a. ?9 mm Karims bithibb-os mother Karim love.3sf-3sm 'Karim's mother loves him/ b.* bithibb-os ?amm Karims love.3sf-3sm mother Karim 'Karim's mother loves him.' In (41a) but not in (41b) coreference is possible between the name Karim and the null object. The unacceptability of (41b) is a Principle C violation. The name Karim is c- commanded by the null pronoun object and coreference is thus impossible. This indicates that the null object has moved out of the verb phrase in order to 312 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. c-command the post-verbal subject. Since the latter triggers agreement on the verb, it must be at least as high as TP, and therefore the null object must be in a projection dominating TP and the agreeing subject. The analysis that overt object agreement markers are the reflex of the elimination of EPP features on the verb carries over to subject agreement in Standard Arabic. Interestingly, the generalization about the distribution of subject agreement in Standard Arabic is the same as the one given for object agreement in LA: (42) Distribution of subject agreement in SA (Benmamoun 2000) The number suffix is obligatory when the postverbal subject position is phonetically null. This generalization is illustrated in (43) .20 20 The total absence of object agreement in LA differs, on the surface, from partial agreement of verbs with their postverbal subjects in Standard Arabic. We have seen that in the latter case person agreement is always present on the verb. It was suggested, following Benmamoun 2000, that this difference may pertain to the difference between derivational and inflectional morphology. That is, whereas person agreement for subject is of the domain of derivational morphology, object agreement is (entirely) of the domain of inflectional morphology. 313 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (43) a. 1-fatayaatu wasal-*(na) ?ila l-'haflati Standard Arabic the-girls arrived.3fp to the-party 'The girls arrived to the party/ b. wasalat-(*na ) 1-fatayaatu ?ila l-'haflati arrived.3fs(*3fp) the-girls to the-party 'The girls arrived to the party.' c. al-fatayaatu llawati wasal-*(na) ?ila l-'haflati the-girls that.fp arrived.3fp to the-party 'The girls that arrived to the party' (43b) indicates that an in situ subject, contrary to a preverbal subject (43a), does not trigger the presence of number agreement morphology on the verb in Standard Arabic. The case in (43c) parallels cases of object relativization in LA, where object agreement is obligatory as well. The paradigm of subject agreement in Standard (i) a. wasal-at 1-fatayaatu ?ila l-'haflati Standard Arabic arrived.3fs the-girls to the-party ‘The girls arrived to the party.' b. wasil baneet fa-l-hafle Lebanese Arabic I arrived.3sm girls to-the-party 'Girls arrived to the party.' For further discussion on this issue and how it relates to gender agreement, see Benmamoun 2000. 314 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Arabic can be accounted for if we assume that the EPP feature on T in Standard Arabic is optional: only when the subject undergoes movement to T is agreement morphology overtly realized on the verb.21 This is exactly what the paradigm in (43) suggests. As is the case with the optional presence of an EPP feature on v in LA, the optional presence of the EPP feature on T in SA is dictated by further movement of the subject to a sentence peripheral position; in (43c) this is illustrated through relativization.22'23 21 For a different analysis of the distribution of number agreement in Standard Arabic, see Benmamoun 2000. 22 There are differences between Standard Arabic and Lebanese Arabic with respect to object agreement. Standard Arabic weak pronominal elements are in complementary distribution with full lexical noun phrases and do not appear in A-constructions generated with movement (Fassi Fehri 1978, Demirdache 1991): (i) a* man ra?ayta-hu? who saw.2sm-him 'Who did you see?' b* ?ayya ra3 ulin ra?ayta-hu? which.ACC man saw.2sm-him 'Which man did you see him?' This indicates that in this language, weak pronominal elements are not agreement markers, but incorporated pronouns. Thus we can find them as resumptive elements in island contexts: 315 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (ii) ?ayyu ra3ulin ra?ayta 1-fataata llati darabat-hu? which.NOM man saw.2sm the-girl that hit.3sf-him 'Which man did you see the girl that hit him?' 2 3 As a consequence of this analysis, the preverbal subject in (43a) cannot occur in [Spec, T] in SA. It must have undergone further movement (dislocation) to the left. In that context, it is interesting to note the fact that preverbal subjects seem to be in complementary distribution with elements that generally appear in sentence initial position, such as m/i-elements (i-ii), a Relativized Minimality effect (Rizzi 1990): (i) a. maTa man takallama l-muTallimuna? with who talked.3sm the-teachers 'With whom did the teachers talk?' b. ?ayya ra3 ulin ra?a I-Tummaalu? which.ACC man saw.3sm the-workers 'Which man did the workers see?' (ii) a* mala man 1-muTaIlimuna takallamuu? with who the-teachers talked.3sp 'With whom did the teachers talk?' b* ?ayya ra3ulin 1-Tummaalu ra?aw? which.ACC man the-workers saw.3p 'Which man did the workers see?' A paradigm similar to the one (i) and (ii) can be constructed with clitic left dislocated objects. See Fassi Fehri 1993 and Shlonsky 1997 for discussions on subject positions in Standard Arabic. 316 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Summarizing the results of this section, the difference between the distribution of subject agreement and object agreement in LA is attributed to the obligatory presence of the EPP feature on T and its optional presence on v. Thus we observe that subjects generally agree with the verbs in LA, but objects do not have to. We have seen that in Standard Arabic, a language where the EPP feature on T is optional, subject agreement has the same distribution as object agreement in LA.2 4 5.4.2 Canonical chains and resumptive chains One problem raised by my analysis is the following: since the presence of WRE, analyzed here as agreement markers, is reduced to the presence of an EPP feature on vP, we predict that all cases of overt movement will yield agreement on the verb. This is contrary to fact. There are cases of overt wh-movement that clearly do not involve object agreement. Thus (44a) minimally contrasts with (44b):2 5 (44) a. ?ayya muhaame baddkun tPTzmo? which attorney.ms want.2p invite 'Which attorney do you want to invite?' 2 4 One difference remains to be accounted for: when the post-verbal subject position is null, number agreement is obligatory, but object agreement is optional. This is an issue 1 turn to immediately. 2 5 If the reader recalls from chapter 1, the contrast in (44) cannot be attributed to D-Iinking as which- phrases are always D-linked. 317 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. ?ayya mutiaame baddkun tiTzmu-u? which attorney .ms want.2p invite-3sm 'Which attorney do you want to invite?' In the theory adopted here, which assumes the phase impenetrability condition to regulate movement, one cannot say that in (44a) the specifier of vP can be optionally skipped on the way to CP. In fact, the only way an object can leave the vP is by moving to a peripheral position in that vP first. So how are we to account for the contrast in (44)? I would like to suggest that the DPs that induce agreement are the ones that involve a (nominal) CP, this being the highest functional projection associated with a nominal element (see Cardinaletti and Starke 1994).2 6 The CP carries the referential index of the noun phrase and realizes features that are related to discourse properties, such as [+specific]. IP, the complement of the nominal CP, realizes agreement features. C is often spelled out as a dummy marker. The one we see on agreeing DPs is usually identical to the dative preposition, as illustrated in (46) for the doubled DP in (45): 2 6 This CP is termed as such by analogy with the clausal CP: it is what makes the noun phrase or the IP become a "complement of". 318 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (45) ‘ iazamu-u la-l-talmiiz invited.3p-3sm to-the-student 'They invited the student/ (46) talmiiz student The evidence that the dative marker is nominal and different from other prepositions in the language is that datives can be doubled, triggering agreement on the verb (47a), whereas prepositional phrases cannot (47b). (47) a. fikitii-lo ?assa la-Sami told.2sf-3sm.dat. story to-Sami 'I told Sami a story.' b* Jalt-o 1-Yabra Tan r-raff took-away.ls-3sm the-dust.sf off the-shelf.sm 'I took the dust off the shelf.' 319 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Note that in the tree representation in (46) the definite article is the realization of I, the head where agreement features are projected (see Ouhalla 1999). This is a welcome result in LA where there is clear evidence that what is called the definite article is an agreeing feature as well.2 7 Thus, its presence does not always carry the semantics of definiteness and some nominal phrases can be interpreted as definite even when the definite article does not appear. Recall that in Semitic languages, agreement within norm phrases between nouns and adjectives involves definiteness agreement as well: (48) a. 1-kteeb *(l-)?ahmar the-book the-red 'The red book' b. kteeb ?aftmar book red 'A red book' Thus the definite article /- is required on an adjective agreeing with a definite noun and is absent when the norm is indefinite. It appears on adjectives to match the 2 7 IP is equivalent to NumP is more traditional analyses. The structure of noun phrases then parallels more closely that given for definite relatives where yalli is also generated in the Number Phrase of its corresponding DP -the projection where agreement relations are checked (see chapter 4). 320 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. definiteness feature on the corresponding nouns without adding any interpretive distinction on the resulting construction. On the other hand, there are noun phrases that lack any overt manifestation of the definite article that are necessarily interpreted as definite. Recall the discussion of construct state nominals in Chapter 4. (49) illustrates such a construction. (49) suwar kali mTallme pictures every teacher 'The pictures of every teahcer' Even in the absence of an overt definite marker on the 'head' norm suwar 'pictures' in (49), the construct state nominal is interpreted as definite, as can be seen from the English translation. Evidence that this noun phrase is indeed definite comes from the fact that it can occur as the second member in a partitive construction, where indefinite NPs are prohibited: (50) a. suura mn suwar kali mlallme picture from pictures every teacher 'One of the teacher's pictures' 321 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. suura mn *(s-)suwar picture from *(the-)pictures 'One of the pictures' Returning to the structure of agreeing noun phrases, when the highest functional projection, i.e. CP, is realized, the DP is strong in the sense of Cardinaletti and Starke 1994. In lexical noun phrases, the CP projection may be occupied by the dummy marker la 'to', but this projection may be null in some strong DPs. In that case, the null C head can only be licensed by strong determiners like every, which, demonstratives or the definite marker, but not by weak determiners like some, a, txoo (see also Chapter 4). Thus a CP noun phrase can have one of the two representations below: (51) CP to strong/ weak determiner (52) CP strong determiner 322 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. However, a representation such as the one in (53) is illicit. (53) * CP 0 i / \ NP weak determiner I suggest that weak determiners occupy an even lower projection in the tree configuration, as illustrated in (54), then the weak determiner is able to license the empty I projection, but not both the I head and the null C head. If licensing of a head by another head is achieved through incorporation of heads (Baker 1988 and Pesetsky 1995), the desired result can be obtained by resorting to the properties of head movement. In order for the weak determiner in (54) to license the null C head, it has to incorporate into that head. However, I cannot be skipped, as dictated by the Head Movement Constraint (Travis 1984, Baker 1988).2 8 If the weak determiner 2 8 Baker's 1988 version of the Head Movement Constraint is stated as in (i): (i) Head Movement Constraint (HMC) An X° may only move into the Y° that properly governs it. In the framework adopted here, the notion of government has no theoretical status. However, what is important to retain is that the HMC prevents movement of a head from skipping an intervening head, where the notion of intervention is understood in terms of c-command. 323 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. first incorporates into I, subsequent head movement to C will not be possible: the empty I head cannot pied-pipe the weak determiner to C,2 9 and excorporation is not possible either. We thus rule out the configuration in (54), which underlies the one in (53). (54) CP XP weak determiner NP That weak determiners do not occupy the same projection as strong determiners can be easily demonstrated, since the latter are not in complementary distribution with the former. In fact weak determiners can occur between strong determiners like every, which, the definite determiner, and the head norm, as illustrated below: (55) a. l-?araba*r wleed the-four children 'The four children' 2 9 This constraint can be understood as part of the general prohibition on phonetically empty elements (including traces) to pied pipe phonetically non-null material (see Chomsky 1998). 324 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. kali kam walad every few child 'Every few children' c. ?ayya fleet wleed which three children 'Which three children' The facts in (55) give plausibility to a structure where weak determiner and strong determiner do not occupy the same projection in the noun phrase, as illustrated in (54) (see also Szabolcsi 1994). If only norm phrases that instantiate the CP projection trigger agreement on the verb and if the CP projection can be licensed only by a dummy marker or a strong determiner, then we can account for the class of noun phrases that trigger object agreement and extend the analysis to account for some special cases where subject agreement in LA is lacking.3 0 3 0 The presence of fii in (ib) can be interpreted as an instance of CP licensing. The only way for an indefinite noun phrase to be licensed in preverbal position in LA, is if it is introduced in nominal structures by the preposition-like element fii, as illustrated by the unacceptability of (iib) without fii. 325 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In fact, only noun phrases introduced by strong determiners (56) can trigger agreement on the verb in LA. The relevant examples are provided in (57). (56) Determiners tlmt license object agreement a. kail 'every/ all' b. 1 - 'the' c. ?ayya 'which d. wala 'no' e. ha-/hay da 'this' (i) a. ?eebalt tleemiz m et.ls students 'I met some students.' b. *(fii) tleemiz ?eebalt-un fii students met.ls-3p 'I met some students.' Thus,_/n realizes the highest functional projection in an indefinite noun phrase, which allows the latter to move outside the vP thereby triggering agreement on the verb. For evidence that fii is to be analyzed as a nominal element see Mohammad (n.d.). 326 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (57) a. ?ayya wleed Tazamtuw-un which children invited.2p-3p 'Which children did you invite?' b. kail walad/1-wleed Tazamtu-u/ -un every/all child/ the-children invited.2p-3sm/ -3p 'Every boy/All the children you invited.' c. wala walad Tazamtu-u no child invited.2p-3p 'No child, you invited.' d. (ha-/hayda) 1-walad Tazamtu-u (this/this.sm) the-child invited.2p-3sm '(This) the child, you invited.' We have seen in chapter 4 that those determiners in (56) are the ones that license the null functional head inside a moved relativized noun phrase, which also triggers agreement in LA restrictive relatives. Weak determiners like the ones in (58), which I take to be generated lower than IP, do not trigger agreement on the verb, as illustrated in (59). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (58) Determiners tlrnt do not license object agreement a. kam 'few' b. indefinite c. numerals (59) a.* kam walad Tazamtuw-un few child invited.2p-3p 'Few children, you invited.' b.* walad/wleed Tazamtu-u/ -un child/ children invited.2p-3sm/ -3p 'A boy/boys you invited.' c.* TJriin walad Tazamtu-un twenty child invited.2p-3p 'Twenty children, you invited.' In my analysis here, the norm phrases introduced by the determiners in (56) are ambiguous between a full CP structure, in which case they trigger overt agreement on the verb, and an IP structure, in which they trigger no agreement. The optionality of overt agreement in cases like (44) (repeated below) reduces to the possibility of analyzing the noun phrase either as a full CP or as an IP. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (44) a. ?ayya rmihaame baddkun ti'i'zmo? which attorney.ms want.2p invite 'Which attorney do you want to invite?' b. ?ayya muhaame baddkun tiTzmu-u? which attorney.ms want.2p invite-3sm 'Which attorney do you want to invite?' The obligatoriness of agreement with noun phrases introduced by the dummy markers la- and fii (60) is due to the fact that those noun phrases are unambiguously CPs.3 1 (60) a. Jsft-*(o) la-ha-l-walad saw.ls-*(3sm) to-this-the-boy 'I saw this boy.' b. fii walad Jaft-*(o) fii boy saw.ls-*(3sm) 'A boy, I saw.' 311 will provide a more detailed account of clitic doubling constructions in LA shortly (see section 5.4.3). 329 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. We have observed that subject agreement is obligatory in LA due to the obligatory presence of an EPP feature on T. However, certain cases can be found where the subject does not agree with the verb; in those cases, the subject appears in postverbal position. (61) a. wosil xams treneet Ta-l-mfiatta arrived.3sm five trains to-the-station 'Five trains arrived to the station/ b. 1am yillab wleed la-t-tarii? Asp. play.3sm children on-the-street 'Children are playing on the street.' Generally, in cases like the ones illustrated in (61), the subject is either a bare noun phrase or a noun phrase introduced by a weak determiner. The verb is always in the third person singular. Having assumed that the tense head in LA is always generated with an EPP feature, we are led to the conclusion that a null expletive pronoun, which carries (default) third person singular features, is inserted to satisfy the EPP feature of T. The EPP feature of T being eliminated, the motivation for the movement of the subject disappears. An agreement relation (Agree, excluding Merge) is established between T and the subject, which results in case assignment.3 2 3 2 A null expletive also appears in sentences like (i): 330 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (i) a. keen 1-wleed baddun yilTabo was.3sm the-children want.3p play.3p 'It was the case that the children want to play.' b. keen daruure ?anno nfll l was.3sm important that Ieave.lp 'It was important that we leave.' The subject in the sentences in (i) is a sentential one and therefore cannot eliminate the EPP feature of T. Sentential subjects in LA therefore always appear following their predicates, as the unacceptability of the sentences in (ii) indicates, and the verbal copula element always carries third person singular morphology. (ii) a* 1-wleed baddun yilTabokeen the-children want.3p play.3p was.3sm 'It was the case that the children want to play.' b.* daruure ?onno nfll keen l important that Ieave.lp was.3sm 'It was important that we leave.' In contexts like those illustrated in (i), we must assume that the EPP feature of T is eliminated by merging a null expletive, in order to explain the acceptability of those sentences. For further discussion on the issue of expletive subject pronouns in Semitic, the reader is referred to Fassi Fehri 1993, Mohammad 1988, and Shlonsky 1997. 331 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. What is important to note here is that even when the in situ subject in (61) precedes the verb, in a wh-question for instance, no agreement morphology shows up on the verb: (62) kam treen wosil < Ta-l-mtiatta? few train arrived.3sm to-the-station 'How many trains arrived at the station?' Evidence that the noun phrases in (61) and (62) do not instantiate a CP projection is also available. These noun phrases are what Li 1998 calls quantity denoting norm phrases, to distinguish them from quantified noun phrases. Quantity denoting noun phrases are analyzed as lacking the highest functional projection of argument noun phrases.3 3 Thus they do not enter into scopal relations with quantified noun phrases or binding relation with other noun phrases. These generalizations are illustrated below: 3 3 Li 1998 analyzes these quantity denoting noun phrases as instantiating the Number Phrase, which in my terms would be an IP. In fact, in my analysis, quantity denoting noun phrases instantiate a projection below the NumP/IP level. 332 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (63) a* 7 8 1 1 1 yilTab wleed maT baTdun Asp. play.3sm children with each-other 'Children are playing with each other.' b. Tam yilTab xams wleed bi-taabteen *5>2 Asp. play.3sm five children with-ball.dual 'Five children are playing with two balls.' The indefinite noun phrase subject in (63a) cannot bind an anaphor inside a prepositional phrase and the quantity denoting norm phrase subject in (63b) does not interact in scope with the quantified noun phrase inside the prepositional phrase. More specifically, the sentence in (63b) can only have a meaning that corresponds to a situation where there are five children, two balls, and the children are playing with the balls. The distributive reading that corresponds to a situation where there are five children each of them playing with two balls is not available. When the subject is introduced by the particle fii, for instance, the bound anaphora reading and the distributive reading become possible: (64) a. fii wleed Tam yilTab-o maT baTdun fii children Asp. play.3p with each-other 'Children are playing with each other.' 333 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. fii xams wleed * 1 3 1 1 1 yilTab-o bi-taabteen fii five children Asp. play.3p with-ball.dual 'Five children are playing with two balls.' The particle fii licenses the CP projection in weak noun phrases and thus agreement morphology becomes available as well as the relevant readings for the sentences in (64). Another consequence of the assumption that weak determiners cannot license a null C in a noun phrase is that it enables us to account for the contrast between relative clauses and wh-questions involving how many phrases in allowing resumption. Thus whereas a resumptive element cannot occur related to a how many phrase in wh-questions (65a), this option is readily available in restrictive relatives (65b). (65) a* kam walad badda t?aases-un 1-m‘ Tallme few child want.3sf punish.3sf-3p the-teacher 'How many children does the teacher want to punish?' b. 1-kam walad Hi badda t?aasas-un 1-m‘ Tallme i i the-few child that want.3sf punish.3sf-them the-teacher 'The many children that the teacher wants to punish' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Whereas the resumptive agreement marker is perfectly acceptable inside the relative clause in (65b), it is not allowed inside a question (65a): In the presence of the weak determiner kam 'few', the null C in (65a) is not licensed.3 4 The analysis provided in chapter 4 predicts exactly what we see in sentences involving restrictive relatives; in the case of raising relatives, the null C in the nominal structure is licensed by the external determiner, which can only be a strong determiner in those constructions. 5.4.3 Clitic doubling in Lebanese Arabic Given the analysis argued for so far, we are led to give clitic doubling constructions such as those illustrated in (66) an analysis involving movement of the doubled noun phrase from the argument position inside the vP. 3 4 CLLD constructions pattern with wh-questions in that respect: (i) a* kam walad badda t?aasas-un 1-m’ i'allme few child want.3sf punish.3sf-3p the-teacher 'Few children, the teacher wants to punish.' b. fii kam walad badda t?aasas-un l-m’ i'allme fii few child want.3sf punish.3sf-3p the-teacher 'Few children, the teacher wants to punish them.' 335 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (66) a. Jaft-a la-ha-l-bant saw.ls-3sf to-this-the-girl 'I saw this girl/ b. Ttiit-un la-l-wleed l-katub I gave.ls-3p to-the-children the-books 'I gave the children the books.' The derivation of sentences like the ones in (66) would proceed as follows: (i) a first step involves movement of the doubled noun phrase to a vP peripheral position and (ii) subsequent movement across the doubled element derives the final word order. In the sentences in (66), the movement of the verb across the doubled norm phrase derives the final word order. In my account of clitic doubling in Lebanese Arabic, I build on Aoun 1999. Aoun (1999) analyzes doubled element as elements that serve to build non-canonical predication structures (Williams 1980). The author assumes that the doubled element is the subject of a complex predicate, which corresponds to the phrase to which this doubled element attaches (67). (67) doubled elements [p re d ic a te • • • V argument position,] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The argument position inside that phrase provides the predicate variable for the doubled element to bind. This variable/argument position in Aoun's analysis is occupied by a null pronoun. The surface word orders observed in (66) are obtained by extracting certain elements from the predicate past the doubled element. In some cases, the whole predicate can be moved around the doubled element.3 5 What I would like to argue here is that there is evidence to believe that doubled norm phrases in LA reach their predicate peripheral position via movement. That is, doubled elements can be generated in the argument position and the configuration in (67) is arrived at after the doubled element is extracted from vP. The evidence comes from the discussion of clitic doubling in double object constructions. (68) a. ‘ Ttiit Karim l-laTbe gave.ls Karim the-toy 'I gave Karim the toy/ 3 5 For a full discussion on the categorial nature of the phrase to which the doubled element attaches and how the surface word order is generated in clitic doubled constructions, the reader is referred to Aoun 1999. 337 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. *Ttiit-o la-Karim 1-bTbe I gave.ls-3sm to-Karim the-toy 'I gave Karim the toy.' c.* 7tiit-o yeeha la-Karim la-l-bTbe gave.ls-3sm 3sf to-Karim to-the-toy 'I gave Karim the toy.' d* ‘ Ttiit-a Karim la-l-bTbe I gave.ls-3sf Karim to-the-toy 'I gave Karim the toy.' Interestingly, in double object constructions, such as the one illustrated in (68), only one of the objects can be doubled, not both, as evidenced by the unacceptability of (68c). It is not that the succession of two 'dative' noun phrases is unacceptable: the sentence in (69) is acceptable. (69) Ttiit-a la-l-bTbe la-Karim gave.ls-3sf to-the-toy to-Karim 'I gave the toy to Karim.' (69), as opposed to (68c) involves a clitic doubled object inside a dative construction. By comparing (68c), (68d), and (69), we conclude that in double object constructions, 338 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. it is only the first of the two objects that can be doubled, the one that corresponds to the higher structural position (IO) in the tree configuration in (70). (70) vP IO V' V DO If the doubled noun phrase reaches is vP peripheral position via movement, this generalization can be easily explained as a minimality effect: in (70), the IO intervenes with the movement of the DO to the vP peripheral position. This fact cannot be easily accommodated in an analysis that assumes that the doubled noun phrase is base generated outside the vP and binds a pronoun in the argument position, as the relation between an antecedent and a pronoun is generally thought to be unbounded. Summarizing the analysis so far, we have argued that weak resumptive elements in LA are agreement markers, which appear on verbs when an object instantiating the full nominal projection is extracted from the vP. Thus in order to induce agreement 339 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. on the verb, two conditions must be met: (i) the noun phrase must have a full CP structure and (ii) v must have an EPP feature. We have seen that noun phrases that cannot license the highest functional projection do not induce agreement on the verb.3 6 Under the assumption that the EPP feature on v is contingent upon the presence of another EPP feature in the sentence, we account for the generalization that object agreement occurs when the postverbal position is empty. As agreement markers, weak resumptive elements will not appear when PPs are moved, since the latter do not have cp-features. 5.5 Further Consequences 5.5.1 Strong pronouns in resumptive constructions It has been observed that strong pronouns cannot participate in resumptive constructions that involve movement (Borer 1983, Bianchi 1999). We now have an explanation for the absence of movement in contexts where strong pronouns are used. Strong pronouns are full DPs that must occupy the variable position in a resumptive construction. That is they compete for the same position as the agreeing 3 6 For an interesting confirmation of this generalization in the context of nominals, see Engelhardt 2000, where it is argued that only "structurally prominent noun phrases can enter into agreement relations with a head" (p. 54). 340 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. noun phrase that undergoes movement. In a construction where both a strong pronoun and a full noun phrase occur in the same position upon first merge, only one of those elements will be able to move: once the EPP feature of the v is checked, no movement out of the vP can take place and the derivation crashes whether it is the strong pronoun that remains in the base position or whether it is the norm phrase. The remaining element will have its case feature unchecked and the derivation crashes on the assumption that case features are not interpretable at the interfaces. The same generalization, which we have seen applies in Hebrew (see section 5.1), also holds in Lebanese Arabic: (71) * ?ayya muttahame ?aalo ?anno kail bolisi which suspect said.3p that every policeman la?at hiyye b-l-3 8 rm l-majhuud? arrested.3sm she in-the-offense the-witnessed 'W hich suspect did they say that every policeman arrested her red- handed?' Thus, in LA as well, resumptive constructions involving movement and a strong pronoun in the base position are ungrammatical. It is not the case that strong pronouns cannot be resumptives because they cannot be interpreted as variables 341 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (see Bianchi 1999). In fact they can be interpreted as variables in contexts of bound variable anaphora, as observed for Lebanese Arabic in Aoun and Choueiri 2000 and Aoun etal. 2001 (72)37 (72) ?ayya muttahame ?aalit ?onno kail bolisi which suspect said.3psf that every policeman la?at-a hiyye b-l-3 8 rm l-majhuud? arrested.3sm-3sf she in-the-offense the-witnessed 'Which suspect said that every policeman arrested her red-handed?' The contrast between (71) and (72) can also be observed in English, a language that allows resumption only marginally (see Ross 1967, Kroch 1981, Chao and Sells 1983, and Prince 1990, among others): 3 7 Strong pronouns must be doubled by an agreement marker when they occur in subject or object positions. The same generalization holds of French strong pronouns, as argued by Kayne (2000). The sentence in (i) provides an example where the strong pronoun occurs in subject position: (i) wala muttahame xabbarit 1-muhaame ?anno hiyye rah tahrub no suspect.fs told.3sf the-attomey that she fut. flee.3sf 'No suspect told the attorney that she will flee.' 342 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (73) a* Every linguist that I hate him b. Every linguist thinks that I hate him. It was such a contrast that led Chao and Sells (1983) to make the distinction between intrusive pronouns and resumptive pronouns. Resumptive pronouns are those that can be operator-bound, i.e. bound by an operator from an A-position. English is a language that has no resumptive pronouns: its pronouns can only be anaphorically bound, that is bound from an A-position. Thus, the pronoun in (73b) is taken to be bound from the A-position occupied by the quantifier phrase prior to QR. Therefore, pronouns in English, which may appear inside islands in A-constructions (74), are in fact what the authors call intrusive pronouns: (74) I just saw a girl who Long John's claim that she was a Venusian made all the headlines. Intrusive pronouns are E-type pronouns (Evans 1980), which cannot be related to true quantifiers, which explains the unacceptability of (73a) where the pronoun cannot be given an E-type reading. The analysis to which this chapter leads is that the contrast between anaphorically bound pronouns and operator-bound pronouns can be attributed to the categorial nature of the pronominal element itself. Following recent proposals by Hornstein 343 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (2000) and Kayne (2001), if we assume that binding is the result of movement, the relation between an operator and a pronoun interpreted as a bound variable falls under that category. The contrast found in English and LA, has a direct account under the analysis argued for here. In (73b), as in (72), the operator and the pronoun can be generated together as a single constituent (Kayne 2001). The sentence is acceptable because each of the pronoun and the DP that binds it is associated with a different case position. They are both case licensed and the sentences are acceptable.3 8 In (73a), as in (71), this is not the case: both the noun phrase and the pronoun are associated with one case position and one of them will end up with an unchecked case feature, which results in the unacceptability of those sentences. In bound anaphora contexts, two case positions are available to accommodate two noun phrases; in resumptive contexts, only one case position is available. Therefore, strong pronouns in LA behave like English pronouns. The dual behavior of resumptive constructions with respect to movement in LA is due to the contrast between strong pronouns and agreement markers. Only the former need case licensing. We now have a direct account for the absence of movement in Hebrew free relatives involving ?et pronouns, following Borer's 1983 description of those facts. ?et pronouns are norm phrases that must be case licensed (Siloni 1994), and 3 8 The reader is referred to Kayne 2001 for a more detailed implementation of this proposal. 344 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. like other noun phrases, they must occur in the variable position in a resumptive construction involving movement.3 9 Irish resumptive constructions involving inflected prepositions fall under this analysis as well. (75) an fear aN raibh me ag caint leis the man ptc was I speaking with-3ps 'The man that I was speaking with' In fact, Irish resembles SA in that its inflectional morphology signals the presence of a null pronoun (see McCloskey and Hale 1984): agreement morphemes do not cooccur with lexical DPs. Thus if we take weak resumptive elements in Irish to 3 9 Shlonsky 1997 argues that Pet pronouns in Hebrew are ambiguous between weak and strong. Thus they can have either of the two representations in (i): (i) a. [cp [ip ... [np ] ] ] b. [ip ... [np ] ] The representation in (ib) corresponds to the representation Ritter (1995) gives for third person pronouns in Modem Hebrew. Ritter argues that the highest functional projection in pronouns hosts the person feature and that third person pronouns lack a specification for person. In any case, Pet pronouns in Hebrew are XPs that require case licensing. 345 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. signal the presence of a pronoun in the resumptive site, we can explain why these constructions in Irish can only be base-generated.4 0 The noun phrase that appears in the sentence peripheral A-position could not have been generated along with the resumptive element and thus could not have started in the variable position, which is in fact occupied by a pronoun. Certain Lebanese Arabic resumptive constructions seem to present a problem for the analysis argued for so far. Aoun et al. 2001 show that strong pronouns (and epithets as well) have a limited distribution as resumptive elements inside A- constructions generated by movement. Thus, strong pronouns related to subject positions can mark the locus of reconstruction sites. The examples below illustrate bound variable anaphora readings in clitic left dislocation and relativization, respectively. 40 Object resumptives in Irish are strong pronouns; they can be dislocated to the end of a sentence (Sells 1987): (i) a. Chonaic se i i mBaile Atha Cliath areir saw he her in Dublin last-night 'He saw her last night.' b. Chonaic se i mBaile Atha Cliath areir i Therefore, it comes as no surprise that A-constructions that involve resumptives in Irish are not generated by movement. This is the conclusion that McCIoskey (1990,2000) arrives at independently. 346 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (76) a. t3lmiiz-[a]i 1-ksleen ma baddna nxabbir [wala m( i'allme]i student-her the-bad neg. want.lp tell.lp no teacher.fs ?anno huwwe zaTbar b-l-fatts I that he cheated.3sm in-the-exam 'Her bad student we do not w ant to tell any teacher that he cheated on the exam/ b. t3lmiiz-[a]i 1-ksleen yalli ma baddna nxabbir [wala mTallmeji ?anno huwwe zaTbar b-l-fahs teacher.fs that he cheated.3sm in-the-exam harab ran-away.3sm 'The bad student of hers that we do not want to tell any teacher that he cheated on the exam ran away.' In both sentences in (76), the pronoun contained within the noun phrase talmiiza l- ksleen 'her bad student' can be interpreted as a variable bound by the negative quantifier wala mTallme 'no teacher' inside the sentence. This is an indication that the CLLDed norm phrase in (76a) as well as the EH of the definite relative in (76b) can reconstruct to a sentence internal position. This is also taken as evidence that the CLLDed norm phrase and the relativized element reach their respective surface 347 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. positions via movement and that the strong pronoun (SP) inside the sentence doubles those elements in their base position, as indicated in the representation in (77) (irrelevant details omitted): (77) [ c ll d n p /R e la tiv iz e d np . . . pronouni... ] j ... Q P ; ... SP - [ n p . . . pronoim i. .. ] j Aoun et al. provide an analysis for constructions like the one in (77), which is consistent with the analysis provided here: in sentences such as (76a) and (76b), the strong pronoun is not in the argument position. That position is occupied instead by the relativized NP or the CLLDed NP before movement. The strong pronoun is in fact in the position occupied by the strong pronouns in sentences like (78-79). (78) and (79) differ in that (78) is what has been called a verbless sentence in the Semitic literature.4 1 41 Verbless sentences can involve either a prepositional predicate or a nominal predicate, as illustrated in (78). The presence of a strong pronoun between the subject and the predicate in those sentences has been analyzed by many Semiticists as the realization of a tense head (see Doron 1982, Fassi Fehri 1993, Benmamoun 2000, and references cited therein). This is what has been called the pronoun copula to differentiate it from strong pronouns that occur in subject positions. In LA, the presence of the pronoun is not obligatory, as the following sentence, which minimally contrasts with (78b) clearly indicates: 348 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (78) a. huwwe Sami b-l-beet/ mmasil he Sami in-the-house/ actor.ms 'Sami is in the house/ an actor/ b. Sami huwwe b-l-beet/ mmasil Sami he in-the-house/ actor .ms 'Sami is in the house/an actor.' c.* huwwe Sami huwwe b-l-beet/ mmasil he Sami he in-the-house/ actor (79) a. huwwe Sami he Sami 'Sami ate the apple.' b. Sami huwwe Sami he 'Sami ate the apple.' c.* huwwe Sami huwwe Sami ?akal t-taffefia ate.3sm the-apple ?akal t-tsffeha ate.3sm the-apple huwwe ?akal t-taffeha he ate.3sm the-apple (i) Sami huwwe b-l-beet/ mmasil Sami he in-the-house/ actor.ms 'Sami is in the house/an actor.' 349 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The strong pronouns in (78a) and (79a) occupy a sentence initial position and cooccur with the subject of the main predicate. This position is associated with focus marking. That is, what precedes the pronoun is contrastively focused, otherwise, it is the whole sentence that is focused (wide focus). This is the case in (78a) and (79a). Thus, (78a-79a), but not (78b-79b), can be the felicitous answer to the question Vslhat happened? The fact that there can only be one huwwe in the clause, as indicated by the unacceptability of (78c-79c), indicates that the strong pronoun is associated with a specific projection. Let us call it FP. FP then dominates the highest subject position in the clause.4 2 42 Further evidence that focus marking is associated with a specific projection comes from the fact that the focused noun phrase that precedes huwwe in (79b), for instance, is in complementary distribution with other focused noun phrases in the sentence, as illustrated by the unacceptability of the sentences in (i) below (focused elements are marked with small caps): (i) a * Sam i Sami huwwe t-teffeha ?akal he th e-apple ate.3sm 'Sami ate the apple.' b* t-taffetha Sami huwwe ?akal he ate.3sm the- apple Sami 'Sami ate the apple.' c* Sami t-taffeha huwwe ?akal Sami the-apple he ate.3sm 'Sami ate the apple.' 350 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The association between the presence of the strong pronoun and focus in the clause provides an explanation for the generalization noted in Aoun et al 2001, that quantificational elements cannot be resumed by strong pronouns in the context of movement ((80) below is equivalent to their (12)): (80) * kali muttahame Trafto ?anno hiyye nhabasit each suspectfs knew.2p that she imprisoned.3sf 'Each suspect, you knew that she was imprisoned.' The sentence in (80) contrasts with the one in (81) where the universally quantified noun phrase is replaced by a non-quantificational norm phrase: (81) ha-l-muttahame Trafto ?anno hiyye nhabasit this-the-suspect.fs knew.2p that she imprisoned.3sf 'Each suspect, you knew that she was imprisoned.' In fact truly quantificational elements are banned from the focus position in (78) and (79) as well: (82) a.* kail walad b-z-zeeyid huwwe b-l-beet EACH CHILD IN-THE-ADDITION he in-the-house 'Each additional child is at home.' 351 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b * ?aktar mn tleet wleed hanne ?akalo taffefia MORE FROM THREE CHILDREN they ate.3p apple 'More than three children ate an apple/ This fact was observed in Heycock and Kroch 1999 concerning the focus position of pseudoclefts. The relevant contrasts are given in (83) (Heycock and Kroch's (72)): (83) a. What she bought was two books on linguistics. b.?*What she bought was fewer than three books on linguistics. c.??What she bought was every book on linguistics. We thus have an explanation for the contrast between (80) and (81). In addition we can provide an explanation for the unacceptability of sentences like (84), where the moved element is not a quantificational phrase. Thus (84) below minimally contrasts with (81): (84) * ha-l-muttahame Trafto ?anno habaso hiyye this-the-suspect.fs knew.2p that imprisoned.3p she 'Each suspect, you knew that they imprisoned her.' 352 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. If the strong pronoun heads FP, the projection that hosts focused elements and which dominates the highest subject position in the sentence, then a sentence like (84) cannot be generated.4 3 5.5.2 Note on the interpretation of resumptive constructions An analysis that assumes that weak resumptive elements in LA are the result of the elimination of v's EPP feature by a full noun phrase differs from one that assumes that they are clitics heading their own projection. We have seen that in the case of French, accusative clitics were associated with the projection that checks the [+specific] feature, which explains that noun phrases doubled by accusative clitics in French must be [+specific] (Sportiche 1992). This is not the case for noun phrases that are doubled by object agreement markers in LA. That is, object agreement markers do not impose a specific interpretation on the agreeing noun phrase, 43 It comes as no surprise that a sentence like (i) is acceptable: (i) ha-l-multahame Trafto ?anno fiabasu-a hiyye this-the-suspect.fs knew.2pthat imprisoned.3p-3sf she 'Each suspect, you knew that they imprisoned her.' (i) involves a base generated left dislocated construction where the argument position inside the sentences is occupied by the strong pronoun, as indicated by the fact that this pronoun is doubled by an agreement morpheme on the verb. I will briefly discuss resumption in island contexts in section 5.6. 353 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. although they impose conditions on the structure of those agreeing noun phrases, which necessarily affects their interpretation. This assumption allows us to explain why the interpretation of an agreeing noun phrase in clitic doubled constructions can be different from the interpretation of an agreeing noun phrase in a relative construction. To appreciate the differences in interpretations, let us examine a few examples where these differences yield differences in acceptability judgements: (85) a* ?axad-a la-l-ratta Sami took.3sm-3sf to-the-nap.fs Sami 'Sami took the nap.' b. ?arrarit Nada tjamrol-un la-kam tabbax.# Bas ma decided.3sf Nada hire.3sf-3p to-few cook. # But neg. b?yit-un ba?d found.3sf-3p yet 'N ada decided to hire few cooks. But she didn't find them yet.' (86) a. l-ratta lli ?axad-a Sami feedit-o the-nap that took.3sm-3sf Sami helped.3sf-3sm 'The nap that Sami took helped him.' 354 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. 1-kam tabbax lli ?arrarit Nada tjamral-un ma the-few cook that decided.3sf Nada hire.3sf-3p neg. la?yit-un baTd found.3sf-3p yet 'The few cooks that Nada decided to hire, she hasn't found yet.' Whereas an idiomatic reading is preserved in the context of relativization, as illustrated in (86a), this reading is not preserved in the context of clitic doubling, as the unacceptability of (85a) shows. The agreeing noun phrase la-l-ratta 'to the nap' cannot be interpreted at all and the resulting sentence has no meaning. Similarly, whereas the sentence in (86b) allows the amount reading, this reading is absent in the context of clitic doubling (85b). The scenario that can be provided to distinguish the amount reading from a 'referential' reading is the following: (87) Nada has decided to hire 7 cooks for the new restaurant she is opening. After one day of interviews, she decides on 4 of the candidates and plans to complete the hiring process on another day. In the scenario presented in (87), the amount reading corresponds to the number of cooks Nada has in mind to hire, i.e. 7. The referential reading corresponds to the number of actual candidates that she decided to hire, i.e. 4. 355 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Only the amount reading is compatible with the discourse continuation in (85b). Since that reading is absent, the discourse as given in (85b) is not felicitous. In relative contexts where the amount readings are preserved in the presence of object agreement markers, a sentence like (86b) is perfectly acceptable. If we follow Sportiche 1992 in assuming that movement in clitic doubling constructions involving direct objects is triggered by a specificity feature, we can explain the paradigm in (85). Both the idiomatic reading of idiomatic NP chunks and the amount reading of few NP are incompatible with specificity. Relativization, as we have seen in chapter 4, involves movement of a noun phrase headed by a null determiner, and therefore the readings that are available for relativized noun phrases are not available for clitic doubled noun phrases. What relativized norm phrases and clitic doubled noun phrases have in common is that they involve the highest nominal projection CP. Only a full noun phrase can host the [+specific] feature, but not all noun phrases that involve the highest nominal projection are ^specific].4 4 44 A specific noun phrase is necessarily referential. Under the standard assumption (Abney 1987, Szabolsci 1987, Stowell 1989, Longobardi 1994) that the referential index of a nominal expression is supplied by the highest functional nominal projection, then specific noun phrases will necessarily involve the CP layer. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The same generalization can be found to apply in the case of preverbal subjects in LA, which we have argued must involve the highest nominal projection as well. Preverbal subject in LA need not be specific: they can be interpreted generically when they involve kind/mass terms (88a), or existentially when they involve fii (88b). (88) a. Ha3ar bizahhir b-r-rabii? the-tree bloom.3sm in-the-spring 'Trees bloom in the spring/ b. fii wleed ‘ Tam yilTabo b-l-ha?le fii children Asp. play.3p in-the-field 'There are children playing in the field.' In this section I have pointed out the variability of interpretations that can be associated with constructions involving resumptive elements. I have clearly demonstrated that the presence of such elements is not associated with specific readings. This is the result of analyzing weak resumptive elements as agreement markers. 357 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.6 Concluding remarks In this chapter I have argued that the relation between resumption and movement is partly regulated by the categorial nature of the resumptive element itself: Strong pronouns, which are full noun phrases, cannot occur in resumptive constructions that involve movement.4 5 They compete with the moving element for the variable position. Clitics and agreement markers, on the other hand, which are not case licensed, can cooccur freely with moving noun phrases in resumptive constructions. This analysis provides a straightforward account for the different behavior of strong pronorms in resumptive constructions and bound variable anaphora contexts, an account that does not rely on an assumption that resumptive pronouns are a different breed of pronouns. I have also analyzed weak resumptive elements in LA as object agreement markers that are the result of the elimination of a non tense head's EPP feature via agreement with a nominal C.4 6 This analysis allows us to explain the distribution of agreement markers in resumptive constructions and the range of interpretations 45 Based on their parallel distribution with epithet noun phrases, Aoun and Choueiri 2000 argue that strong pronouns have the same structure as demonstrative noun phrases. 46 The question can be raised as to why overt agreement needs to be mediated by nominal C, especially since it is IP that holds the 9 -features. I have no answer for this question at the moment. 358 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. associated with them. Furthermore, it allows us to unify the account of the distribution of overt agreement in LA. A final note on resumption and last resort: In this chapter, I did not touch upon the relation between movement and base generation in resumptive constructions. Most of the cases observed and discussed were clear-cut cases where movement was either required or prohibited. However, two cases seem to argue for base generation as a last resort strategy in resumptive questions and clitic-left dislocated constructions. Thus, consider the following contrasts: the sentences in (89a-90a) illustrate constructions where the resumptive element occurs in non-island contexts, where movement is possible; the sentences in (89b-90b) illustrate constructions where the resumptive element occurs in an island context, where movement is prohibited. (89) a.* kail muttahame Trafto ?anno hiyye rihabasit every suspect.fs knew.2p that she imprisoned.3sf 'Every suspect, you knew that she was imprisoned.' 359 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. kail muttahame zaTlo l-?ahl lamina hiyye every suspect.fs upset.3p the-parents when she nhabasit imprisoned.3sf 'Every suspect, the parents were upset when she was imprisoned.' (90) a.* kam walad feefu-un b-l-ha?le few child saw.3p-3p in-the-field 'How many children did they see in the field?' b. kam walad zaTlit 1-m‘ Tallme lamma Jeefat-un few child upset.3sm the-teacher.fs when saw.3sf-3p b-l-tia?le in-the-field 'How many children was the teacher upset when she saw them in the field?' As observed in Aoun et al. 2001, the contrast in acceptability between (89a) and (89b) argues for movement as the primary strategy (true resumption as last resort) for generating resumptive constructions. This carries over to the contrast between (90a) and (90b) which involve weak resumptive elements. That is, everything else being equal, a representation like (91a) is preferred over a representation like (91b). 360 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (91) a. M ovem ent strategy Antecedents... WRE/SP-Antecedent; b. Base generation strategy Antecedent;... WRE-pro,/SP; In light of the acceptability of sentences like (89b) and (90b), the question can be raised as to the unavailability of a base generation strategy in (89a) and (90a). To account for the absence of that strategy in those sentences, we can say that when movement is possible, then the sentence must involve movement. In other words, relating an A-element to its variable via movement is the preferred strategy, relating them via binding is a last resort strategy.4 7 What is interesting to note in that respect is that the limited availability of the binding strategy is accompanied by a restriction on the range of interpretations available for the A-element in constructions like (89b) and (90b). We have already observed that amount readings as well as de dicto readings are not available when the resumptive element occurs inside an island (see chapter 3). Furthermore, antecedents of resumptive elements occurring inside islands cannot be truly quantificational (92). 4 7 For a technical implementation of this proposal, see Aoun et al. 2001, where it is argued that structures like (91b) involve more steps in the derivation than movement structures (91a). 361 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (92) a.* ?aktar mnmuttahame zaTlo l-?ahl lamma hiyye more of suspectfs upset.3p the-parents when she nhabasit imprisoned.3sf 'More than one suspect the parents were upset when she was imprisoned.' b* kam walad zyeede za'Tlit 1-mTallme lamma few child addition upset.3sm the-teacher.fs when Jeefot-un b-l-tia?le saw.3sf-3p in-the-field 'How many additional children was the teacher upset when she saw them in the field?' Sentences such as the ones in (92) are interpreted as topicalized constructions: the A-element is a topic and it is related to a pronoun inside the sentence, which has an E-type interpretation. Evidence that the antecedent in base-generated constructions like (92) must be topicalized comes from Standard Arabic, where topics are assigned default nominative case independent of the position of the pronoun inside the sentence. Thus, in Standard Arabic, a wh-phrase related to a resumptive element inside the sentence must be nominative as the contrast below indicates: 362 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (93) a. ?ayya ra3ulin ra?ayta-(*hu)? which.ACC man.GEN saw.2ms-(*3sm) 'Which man did you see?' b. ?ayyu rajulin ra?ayta-*(hu)? which.NOM man.GEN saw.2ms-*(3sm) 'Which man did you see?' c. ?ayyu ra3ulin ra?ayta 1-fataata llati darabat-hu which.NOM man.GEN saw.2ms the-girl that hit.3sf-3sm 'Which man did you see the girl that hit him?' Resumptive pronouns that are bound by topics, such as in (89b-90b) and (93) are what Chao and Sells 1983 terms intrusive pronouns. This analysis thus explains the restriction on the range of A-antecedents in base generated constructions. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. References Abney, S. P. 1987. The English noun phrase in its sentential aspects. Doctoral dissertation, MIT, Cambridge, Mass. Aoun, Joseph, and Elabbas Benmamoun. 1998. Minimality, Reconstruction, and PF Movement. Linguistic Inquiry 29:569-592. Aoun, Joseph, and Lina Choueiri. 1997. Resumption and last resort. Ms., University of Southern California, Los Angeles. Aoun, Joseph, and Lina Choueiri. 2000. Epithets. Natural Language and Linguistic Theory 18:1-39. Aoun, Joseph, and Yen-Hui Audrey Li. 1990. Minimal disjointness. Linguistics 28: 189-203. Aoun, Joseph, and Yen-Hui Audrey Li. 1993. Syntax of Scope. Cambridge, Mass.: MIT Press. Aoun, Joseph, Elabbas Benmamoun, and Dominique Sportiche. 1994. Agreement and conjunction in some varieties of Arabic. Linguistic Inquiry 25:195-220. Aoun, Joseph, Lina Choueiri, and Norbert Hornstein. 1999. Resumption, movement, and derivational economy. Ms., University of Southern California, Los Angeles and University of Maryland, Baltimore. Aoun, Joseph. 1978. Structure interne du groupe nominal en Arabe: l'?idafa. Analyses Tlieories 2:1-40. Aoun, Joseph. 1986. Generalized Binding. Dordrecht: Foris. Aoun, Joseph. 1999. Clitic-doubled arguments. In Beyond Principles and Parameters, ed. K. Johnson and I. Roberts, 12-42. Dordrecht: Kluwer. Ausin, Adolfo. 2000. Where does idiom interpretation apply? In Proceedings of the North East Linguistic Society, ed. M. Hirotani, A. Coetzee, N. Hall, and J.-Y. Kim, volume 30,43-54. Baker, Mark. 1988. Incorporation: A theory of grammatical function clmnging. Chicago: Chicago University Press. 364 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Barker, Chris. 1998. Partitives, double genitives, and anti-uniqueness. Natural Language and Linguistic Theory 16: 679-717. Barss, Andrew. 1986. Chains and anaphoric dependence: On reconstruction and its implications. Doctoral Dissertation, MIT, Cambridge. Belletti, Adriana. 1994. Case Checking and clitic placement: Three issues in (Italian/Romance) clitics. In Geneva Generative Papers 1.2,101-118. University of Geneva, Geneva. Benmamoun, Elabbas. 2000. The feature structure of functional categories: A comparative study of Arabic dialects. New York: Oxford University Press. Bhatt, Rajesh. 2000. Adjectival modifiers and the raising analysis of relative clauses. In Proceedings of the North East Linguistic Society, ed. M. Hirotani, A. Coetzee, N. Hall, and J.-Y. Kim, volume 30,55-67. Bianchi, Valentina. 1999. Consequences of antisymmetry: Headed relative clauses. Berlin: M outon de Gruyter. Bianchi, Valentina. 2000. Some issues in the syntax of relative determiners. In The syntax of relative clauses, ed. Artemis Alexiadou, Paul Law, Andre Meinunger, Chris Wilder, 53-82. Amsterdam: John Benjamins. Bianchi, Valentina. 2000. The raising analysis of relative clauses: A reply to Borsley. Linguistic Inquiry 31:123-140. Boeckx, Cedric. 2001. Resumption and derivation. Paper presented at the Workshop on Bound Pronouns, University of Southern California, Los Angeles. Borer, Hagit. 1984. Restrictive relatives in Modern Hebrew. Natural Language and Linguistic Theory 2: 219-260. Borer, Hagit. 1996. The construct in review. In Studies in Afroasiatic grammar, ed. J. Lecarme, J. Lowenstamm, and U. Shlonsky, 30-61. The Hague: Holland Academic Graphics. Borsley, Robert D. 1997. Relative clauses and the theory of phrase structure. Linguistic Inquiry 28: 629-647. Brame, Michael K. 1968. A new analysis of the relative clause: Evidence for an interpretive theory. Ms., MIT, Cambridge. 365 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Browning, Marguerite. 1987. Null operator constructions. Doctoral Dissertation, MIT, Cambridge, Mass. Buring, Daniel, and Katharina Hartman. 1997. Doing the right thing. The Linguistic Review 14:1-42. Cardinaletti, Anna, and Michal Starke. 1994. The typology of structural deficiency: On the three grammatical classes. Ms., University of Venice, University of Geneva, and Max Planck Institute, Berlin. Carlson, Greg N. 1977. Amount relatives. Language 58: 520-542. Cecchetto, Carlo, and Gennaro Chierchia. 1999. Reconstruction in dislocation constructions and the Syntax/Semantics interface. In Proceedings of the Seventeenth West Coast Conference on Formal Linguistics, ed. Kimary Shahin, Susan Blake, and Eun-Sook Kim, 132-146. Stanford: CSLI publications. Chao, Wynn, and Peter Sells. 1983. On the interpretation of resumptive pronouns. In Proceedings ofNELS 13,47-61. GLSA, University of Massachusetts, Amherst. Chomsky, Noam. 1977. On Wh-movement. In Formal syntax, ed. Adrian Akmajian, Peter W. Culicover, and Thomas Wasow, 71-132. New York: Academic Press. Chomsky, Noam. 1982. Some concepts and consequences of the theory of government and binding. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1993. A minimalist program for linguistic inquiry. In The view from Building 20: Essays in linguistics in honro of Sylvain Bromberger, ed. Kenneth Hale and Samuel Jay Keyser, 1-52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass: MIT Press. Chomsky, Noam. 1998. Minimalist inquiries: The framework. In Step by step: Essays on minimalist syntax in honor of Howard Lasnik, ed. Roger Martin, David Michaels, and Juan Uriagereka, 89-155. Cambridge, Mass.: MIT Press Chomsky, Noam. 2000. Derivation by phase. In Ken Hale: A life in language, ed. Michael Kenstowicz. Cambridge: MIT Press. Cinque, Guglielmo. 1982. On the theory of relative clauses and markedness. The Linguistic Review 1: 247-296. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Cinque, Guglielmo. 1990. Types of A-dependencies. Cambridge, Mass.: MIT Press. Cole, Peter. 1987. The structure of internally headed relative clauses. Natural Language and Linguistic Theory 5:277-302. Darrow, James. 2000. Reconstruction into resumptive pronoun positions. Ms., University of California at Santa Cruz, Santa Cruz. Demirdache, Hamida. 1991. Resumptive chains in restrictive relatives, appositives and dislocation structures. Doctoral dissertation, MIT, Cambridge, Mass. Dobrovie-Sorin, Carmen. 1990. Clitic-Doubling, lA7z-movement, and quantification in Romanian. Linguistic Inquiry 21: 351-397. Doron, Edit. 1982. On the syntax and semantics of resumptive pronouns. Texas Linguistic Forum 19,1-48. Department of Linguistics, University of Texas, Austin. Doron, Edit. 1983. Verbless Predicates in Hebrew. Doctoral dissertation, University of Texas at Austin, Austin, Texas. Dubinsky, Stanley, and Robert Hamilton. 1998. Epithets as antilogophoric pronouns. Linguistic Inquiry 29:685-693. Dwivedi, Veena. 1994. Syntactic dependencies and relative clauses in Hindi. Doctoral dissertation, University of Massachusetts, Amherst. Eid, Mushira. 1982. On the communicative function of subject pronouns in Arabic. Journal of Linguistics 19: 287-303. Elomari, Abdesslam. 1996. Operator movement and resumptive pronouns: The case of Moroccan Arabic. Ms., University of Southern California, Los Angeles. Emonds, Joseph. 1979. Appositive relatives have no properties. Linguistic Inquiry 10: 211-243. Engdahl, Elisabet. 1986. Constituent Questions. Dordrecht: Reidel. Engelhardt, Miriam. 2000. The projection of argument-taking nominals. Natural Language and Linguistic Theory 18:41-88. Evans, Gareth. 1980. Pronouns. Linguistic Inquiry 11:337-362. 367 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Fassi Fehri, Abdelkader. 1978. Relatives restrictives en Arabe. Ms., Faculte des Lettres, Universite de Rabat, Rabat. Fassi Fehri, Abdelkader. 1993. Issues in the Structure of Arabic Clauses and Words. Dordrecht: Kluwer. Fox, Danny. 1994. Relative clauses and resumptive pronouns in Hebrew: An optimality theoretic approach. Ms., MIT, Cambridge, Mass. Fox, Danny. 1995. Economy and Scope. Natural Language Semantics 3: 283-341. Fox, Danny. 1999. Reconstruction, Binding Theory, and the interpretation of chains. Linguistic Inquiry 30:157-196. Georgopoulos, Carol. 1984. Resumptive pronouns, syntactic binding, and levels of representation in Belauan. In Proceedings of the fourteenth annual meeting of the North-Eastern Linguistics Society. GLSA, University of Massachusetts, Amherst. Georgopoulos, Carol. 1985. The Syntax of variable Binding in Palauan. Doctoral Dissertation, University of California at San Diego, La Jolla. Grosu, Alexander, and Fred Landman. 1998. Strange relatives of the third kind. Natural Language Semantics 6:125-170. Grosu, Alexander. 1992. The structural diversity of restrictive and amount relatives. Ms., Tel Aviv University, Tel Aviv. Haddad, Ghassan, and Michael Kenstowicz. 1980. A note on the parallelism between the definite article and the relative clause marker in Arabic. In Studies in the Linguistic Sciences 10.2,41-48. University of Illinois, Urbana-Champaign. Haik, Isabelle. 1984. Indirect Binding. Linguistic Inquiry 15:185-223. Halle, Morris, and Alec Marantz. 1993. Distributed Morphology and the pieces of inflection. In The viexvfrom building 20: Essays in linguistics in honor ofSylvain Bromberger, ed. Kenneth Hale and Samuel Jay Keyser, 91-116. Cambridge, Mass.: MIT Press. Hallman, Peter. 2000. The structure of agreement failure in Lebanese Arabic. In WCCFL19 Proceedings, ed. by Billerey and Lillehaugen, 178-190. Somerville, MA: Cascadilla Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Harbert, Wayne. 1982. On the nature of the matching parameter. The Linguistic Review 2:237-284. Hazout, Ilan. 1990. Verbal nouns: Theta-theoretical studies in Hebrew and Arabic. Doctoral Dissertation, University of Massachusetts, Amherst. Heim, Irena. 1982. The semantics of definite and indefinite norm phrases. Doctoral Dissertation, University of Massachusetts, Amherst. Heim, Irena. 1987. Where does the definiteness restriction apply? Evidence from the definiteness of variables. In The Representation of (In)-definiteness, ed. by Eric Reuland and Alice ter Meulen, 21-42. Cambridge, Mass.: MIT Press. Heycock, Caroline. 1995. Asymmetries in reconstruction. Linguistic Inquiry 26:547- 570. Heycock, Caroline, and Anthony Krock. 1999. Pseudocleft connectedness: Implications for the LF interface level. Linguistic Inquiry 30: 365-398. Higginbotham, James. 1983. Logical form, binding, and nominals. Linguistic Inquiry 14: 395-420. Higginbotham, James. 1985. On semantics. Linguistic Inquiry 16: 547-593. Higginbotham, James. 1992. Anaphoric Reference and Common Reference. Ms. MIT, Cambridge, Mass. Hornstein, Norbert. 1995. Logical Form: From GB to Minimalism. Oxford: Blackwell. Homstein, Norbert. 2000. Move! A minimalist theory of construal. Malden, Mass.: Blackwell. Iatridou, Sabine. 1991. Clitics and island effects, Ms., MIT, Cambridge, Mass. Iatridou, Sabine. 1991. Topics in conditionals. Doctoral dissertation, MIT, Cambridge, Mass. Kayne, Richard. 1975. French syntax: The transformational cycle. Cambridge, Mass.: MIT Press. Kayne, Richard S. 1994. The antisymmetry of syntax. Cambridge, Mass.: MIT Press. 369 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Kayne, Richard S. 1999. A note on clitic doubling in French. Ms., New York University, New York. Kayne, Richard S. 2001. Pronouns and their antecedents. Ms., New York University, New York. Keenan, Edward. 1985. Relative clauses. In Language typology and syntactic description, Volume 2, ed. T. Schopen, 141-170. Cambridge: Cambridge University Press. Kroch Anthony. 1981. On the role of resumptive pronorms in amnestying island constraint violations. In Papers from the 17th Regional Meeting of the Chicago Linguistic Society, 125-135. Chicago Linguistic Society, University of Chicago, Chicago, 11 1 . Kroch, Anthony. 1989. Amount quantification, referentiality, and long wh- movement. Ms., University of Pennsylvania, Philadelphia. Kuroda, S.-Y. 1968. English relativization and certain related problems. Language 44: 244-266. Landau, Idan. 1999. Projection and bounding in possessor raising. In Proceedings of the 17th West Coast Conference on Formal Linguistics, ed. Kimary Shahine, Susan Blake, and Eun-Sook Kim, 405-420. Stanford: CSLI. Larson, Richard K. 1985. Bare-NP Adverbs. Linguistic Inquiry 16: 595-621. Larson, Richard. 1987. Missing prepositions and the analysis of English free relative clauses. Linguistic Inquiry 18:239-266. Lebeaux, David. 1992. Relative clauses, licensing, and the nature of the derivation. In Perspectives on Phrase Structure: Heads and Licensing, ed. Susan Rothstein and Margaret Speas, 209-239. Syntax and Semantics 25, San Diego: Academic Press. Li, Yen-Hui Audrey. 1998. Argument determiner phrases and number phrases. Linguistic Inquiry 29: 693-703. Li, Yen-Hui Audrey. 2001. Relativization: Head-initial relative constructions. Ms., University of Southern California, Los Angeles. Longobardi, Giuseppe. 1987. Extraction from NP and the proper notion of head government. In The syntax of noun phrases, ed. Alessandra Giorgi and Giuseppe Longobardi, 57-112. Cambridge: Cambridge University Press. 370 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N- movement in syntax and Logical Form. Linguistic Inquiry 25: 609-665. Mahajan, Anoop. 2000. Relative asymmetries in Hindi correlatives. In The syntax of relative clauses, ed. Artemis Alexiadou, Paul Law, Andre Meinunger, Chris Wilder, 201-229. Amsterdam: John Benjamins. May, Robert. 1985. Logical Form: Its structure and derivation. Cambridge, Mass.: MIT Press. McCawley, J. D. 1981. The syntax and semantics of English relative clauses. Lingua 53: 99-149. McCloskey, James, and Kenneth Hale. 1984. On the syntax of person-number inflection in Modem Irish. Natural Language and Linguistic Tlteory 1:487-533. McCloskey, James. 1990. Resumptive pronouns, A-binding and levels of representation in Irish. In Syntax of the modem Celtic languages, ed. Randall Hendrick, 199-248. Syntax and Semantics 23, San Diego: Academic Press. McCloskey, James. 2000. Resumption, successive cyclicity, and Locality of operations. Ms., University of California at Santa Cruz, Santa Cruz. Mohammad, Mohammad. 1988. On the parallelism between IP and DP. In Proceedings of the seventh West Coast conference on formal linguistics, 241-254. Stanford: CSLI. Mohammad, Mohammad. 1989. The Sentence structure of Arabic. Doctoral Dissertation, University of Southern California, Los Angeles. Mohammad, Mohammad, n.d. The syntax of indefinite subjects in equative sentences in Palestinian Arabic. Ms., University of Florida, Gainesville. [Available from Semitic Archives website http: / / www.usc.edu/ dept/ LAS/ linguistics/ Semitic] Montalbetti, Mario. 1984. After binding: On the interpretation of pronouns. Doctoral Dissertation, MIT, Cambridge, Mass. Munn, Alan. 1994. A minimalist account of reconstruction asymmetries. In Proceedings of the 24th meeting of the North East Linguistic Society, 397-410. University of Massachusetts, Amherst. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Nunes, Jairo. 1995. The copy theory of movement and linearization of chains in the Minimalist Program. Doctoral dissertation, University of Maryland, College Park. Ouhalla, Jamal. 1990. The syntax of head movement: A study of Berber. Doctoral Dissertation, University College, London. Ouhalla, Jamal. 1991. Functional and Parametric Variations. London: Routledge. Ouhalla, Jamal. 1994. Verb movement and word order in Arabic. In Verb movement, ed. David Lightfoot and Norbert Homstein, 41-72. Cambridge: Cambridge University Press. Ouhalla, Jamal. 1999. Predication and expletive determiners in definite relatives. Ms., Queen Mary-London University, London. Ouhalla, Jamal. 2001. Parasitic gaps and resumptive pronouns. In Parasitic gaps, ed. Peter Culicover and Paul Postal, 147-179. Cambridge, Mass.: MIT Press. Partee, Barbara Hall. 1972. Opacity, coreference, and pronouns. In Semantics of natural language, ed. Harman and Davidson, 415-441. Dordrecht: Reidel. Pesetsky, David. 1987. W7i-in-situ: Movement and unselective binding. In The representation of (in)definiteness, ed. Eric Reuland and Alice ter Meulen. Cambridge, Mass.: MIT Press. Pesetsky, David. 1995. Zero syntax: Experiences and cascades. Cambridge, Mass.: MIT Press. Platzack, Christer. 2000. A Complement-of-N0 account of restrictive and non- restrictive relatives: The case of Swedish. In The syntax of relative clauses, ed. Artemis Alexiadou, Paul Law, Andre Meinunger, Chris Wilder, 265-308. Amsterdam: John Benjamins. Prince, Ellen. 1990. Syntax and discourse: A look at resumptive pronouns. In Proceedings of the sixteenth annual meeting of the Berkeley Linguistics Society, 482-97. BLS, University of California, Berkeley. Reinhart, Tanya. 1997. Wh-in-situ in the framework of the Minimalist Program. Natural Language Semantics 6: 29-56. Riemsdijk, Henk van and Edwin Williams. 1981. NP-structure. The Linguistic Review 1:171-217. 372 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Riemsdijk, Henk van. 1989. Swiss Relatives. In Sentential complementation and the lexicon, ed. D. Jaspers, W. Klooster, Y. Putseys and P. Seuren, 343-354. Dordrecht: Foris. Ritter, Elisabeth. 1991. Two functional categories in noun phrases: Evidence from Modem Hebrew. Syntax and Semantics 25:37-62. Ritter, Elizabeth. 1995. On the syntactic category of pronouns and agreement. Natural Language and Linguistic Theory 13:405-443. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press. Romero, Maribel. 1998. Problems for a semantic account of scope reconstruction. In Reconstruction: Proceedings of the 1997 Tubingen Workshop, ed. Graham Katz, Shin-Sook Kim, and Heike Winhart, 127-153. Universitat Stuttgart and Universitat Tubingen. Ross, J.-R. 1967. Constraints on variables in Syntax. Doctoral Dissertation, MIT, Cambridge, Mass. Rullmann, Hotze, and Sigrid Beck. 1998. Reconstruction and the interpretation of which-'phiases. In Reconstruction: Proceedings of the 1997 Tubingen Workshop, ed. Graham Katz, Shin-Sook Kim, and Heike Winhart, 223-256. Universitat Stuttgart and Universitat Tubingen. Safir, Kenneth. 1986. Relative clauses in a theory of binding and levels. Linguistic Inquiry 17:663-689. Safir, Kenneth. 1996. Derivation, representation, and resumption: The domain of Weak Crossover. Linguistic Inquiry 27: 313-339. Safir, Kenneth. 1999. Vehicle change and reconstruction in A-chains. Linguistic Inquiry 30:587-620. Sauer land, Uli. 1998. The meaning of chains. Doctoral dissertation, MIT, Cambridge, Mass. Schachter, Paul. 1973. Focus and relativization. Language 49:19-46. Schmitt, Cristina. 2000. Some consequences of the complement analysis for relative clauses, demonstratives and the wrong adjectives. In The syntax of relative clauses, ed. Artemis Alexiadou, Paul Law, Andre Meinunger, Chris Wilder, 309- 348. Amsterdam: John Benjamins. 373 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Schneider-Zioga, Patricia. 1997. A predication analysis of clitic pronouns in Greek. In Proceedings of the 28th meeting of the North East Linguistic Society. University of Massachusetts, Amherst. Sells, Peter. 1984. Syntax and semantics of resumptive pronouns. Doctoral dissertation, University of Massachussets, Amherst. Sells, Peter. 1987. Binding resumptive pronouns. Linguistics and Philosophy 10: 261- 298. Sharvit, Yael. 1996. The syntax and semantics of functional relative clauses. Doctoral dissertation, Rutgers University, New Brunswick. Sharvit, Yael. 1999a. Functional relative clauses. Linguistics and Philosophy 22: 447- 478. Sharvit, Yael. 1999b. Resumptive pronouns in relative clauses. Natural Language and Linguistic Theory 17:587-612. Shlonsky, Ur. 1992. Resumptive pronouns as a last resort. Linguistic Inquiry 23: 443-468. Shlonsky, Ur. 1996. Constituent questions in Palestinian Arabic. Ms., Universite de Geneve. Shlonsky, Ur. 1997. Clause structure and word order in Hebrexv and Arabic: An essay in comparative Semitic syntax. New York: Oxford University Press. Siloni, Tal. 1991. On participial relatives and complementizer D°: A case study in Hebrew and French. Natural Language and Linguistic Theory 13: 445-487. Siloni, Tal. 1994. Noun phrases and nominalizations. Doctoral Dissertation, University of Geneva, Geneva. Sportiche, Dominique. 1992. Clitic constructions. Ms., University of California, Los Angeles. Sportiche, Dominique. 1998. Partitions and atoms of clause structure. New York: Routledge. Sportiche, Dominique. 1999. Relative Clauses. Ms., University of California, Los Angeles. 374 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Stowell, Tim. 1989. Subjects, specifiers, and X-bar Theory. In Alternative conceptions of phrase structure, ed. Mark Baltin and Anthony Kroch, 232-262. Chicago: University of Chicago Press. Suher, Margarita. 1997. Clitic-doubling of strong pronouns in Spanish: An instance of object shift. In Grammatical Analyses in Basque and Romance Linguistics, ed. Jon Franco, Alazne Landa, and Juan Martin. Amsterdam: John Benjamins. Suner, Margarita. 1998. Resumptive restrictive relatives: a crosslinguistic perspective. Language 74: 335-364. Szabolcsi, Anna. 1994. The noun phrase. In The syntactic structure of Hungarian, ed. F.Kiefer and K. E. Kiss, 179-274. San Diego: Academic Press. Szabolcsi, Anna. 1997. Ways of scope taking. Dordrecht: Kluwer. Travis, Lisa. 1984. Parameters and effects of word order variation. Doctoral Dissertation, MIT, Cambridge, Massachusetts. Ueyama, Ayumi and Hajime Hoji. 2001. Resumption in Japanese. Paper presented at the Workshop on Bound Pronouns, University of Southern California, Los Angeles, May 2001. Uriagereka, Juan. 1995. Aspects of the syntax of clitic placement in Western Romance. Linguistic Inquiry 26: 79-123. Vergnaud, Jean-Roger and Maria-Luisa Zubizarreta. 1992. The definite determiner and the inalienable constructions in French and English. Linguistic Inquiry 23: 595-652. Vergnaud, Jean-Roger. 1985. Dependances et niveaux de representation en syntaxe. Amsterdam: John Benjamins. Wahba, Wafaa. 1984. WJz-constructions in Egyptian Arabic. Doctoral dissertation, University of Illinois at Urbana-Champaign, Urbana, Illinois. Wilder, Chris. 1994. Coordination, ATB, and ellipsis. Ms., Max Planck Institute, Berlin. Williams, Edwin. 1980. Predication. Linguistic Inquiry 11: 203-238. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Williamson, Janis S. 1987. An indefiniteness restriction for relative clauses in Lakhota. In The Representation of (In)definiteness, ed. by Eric J. Reuland and Alice G. B. ter Meulen, 168-190. Cambridge, Mass.: MIT Press. Woisetschlaeger, Eric. 1983. On the question of definiteness in 'an old man's book'. Linguistic Inquiry 14:137-154. Zubizarreta, Maria-Luisa. 1996. Word order, prosody, and focus. Cambridge, Mass.: MIT Press. Zwart, Jan-Wouter. 2000. A head raising analysis of relative clauses in Dutch. In The syntax of relative clauses, ed. Artemis Alexiadou, Paul Law, Andre Meinunger, Chris Wilder, 349-385. Amsterdam: John Benjamins. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Form and meaning: Negation and question in Chinese

PDF

A cut above: Fashion as meta -culture in early -twentieth -century Russia

PDF

A syntax-based statistical translation model

PDF

Generals and particulars in Thucydides

PDF

A culture of sociability: Popular speech in ancient Rome

PDF

At home in the world: Indian literature in the postcolonial academy

PDF

Asymmetry of scope taking in wh -questions

PDF

Johann Krieger and sacred vocal music in Zittau, 1682--1735

PDF

Ellipsis constructions in Chinese

PDF

Discourse functional units: A re-examination of discourse markers with particular reference to Spanish

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

Grammaticalization and the development of functional categories in Chinese

PDF

Children in transition: Popular children's magazines in late imperial and early Soviet Russia

PDF

Connectionist phonology

PDF

Binding the new together with the old: Fifteenth -century writers on the origins of the Polish state and people in the face of earlier tradition

PDF

A measurement of sodium cluster polarizabilities in an electric deflection experiment

PDF

John Locke on the resemblance theses and the primary -secondary quality distinction

PDF

Arthur Waley: Translator of Chinese poetry

PDF

A case grammar of the parker manuscript of the "Anglo-Saxon chronicle" from 734 to 891

PDF

In search of the sun: Yukio Mishima's body aesthetic

Asset Metadata

Creator Choueiri, Lina Georges (author)

Core Title Issues in the syntax of resumption: Restrictive relatives in Lebanese Arabic

Degree Doctor of Philosophy

Degree Program Linguistics

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

Tag language, linguistics,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor [illegible] (committee chair), [illegible] (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-214772

Unique identifier UC11338990

Identifier 3073760.pdf (filename),usctheses-c16-214772 (legacy record id)

Legacy Identifier 3073760.pdf

Dmrecord 214772

Document Type Dissertation

Rights Choueiri, Lina Georges

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