Verbal complementizers in Arabic

A class of Modern Standard Arabic complementizers known as ‘ʔinna and its sisters' demonstrate unique case and word order restrictions. While CPs in Arabic allow both Subject‐Verb (SV) and Verb‐Subject (VS) word order and their subjects show nominative morphology, CPs introduced by ʔinna ban a verb from directly following the complementizer. Preverbal subjects in ʔinna clauses show accusative case marking, while postverbal subjects show nominative morphology. Previous research explains these restrictions as default case or Multiple Case Assignment, both problematic for Case Theory as they violate the Activation Principle. This dissertation explains word order and case effects of ʔinna within the framework of Phase Theory and Feature Inheritance (FI). Morphological, historical, and usage evidence point out that ʔinna‐type complementizers have verbal properties similar to illocutionary verbs. Taking Case to be a reflection of phi features that T heads receive from higher heads (e.g. Complementizers) via Feature Inheritance, the nominative‐accusative alternation on preverbal subjects can be attributed to the selection of C heads: phi features on null complementizers and conditionals reflect as NOM, while phi features on Verbal Complementizers (VCs) reflect as ACC. VCs show similar Case behavior to the English Prepositional Complementizer for. They differ in distribution; while for only introduces a subordinate clause, and takes infinitival TP complements, VCs introduce a matrix clause and require finite TP complements, lending stronger support to Feature Inheritance theory than English for. Nominative postverbal subjects in ʔinna clauses are explained as an effect of antiagreement at Spell‐Out. Postverbal subjects and the Case probe on T are PF local, allowing for impoverished case agreement. Preverbal subjects and the Case licenser belong to different Phonological Phrases. To satisfy the Recoverability Condition, full case agreement is required between T and the subject, resulting in accusative morphology on the subject. Finally, the requirement that ʔinna‐clauses have an intervener between ʔinna and the verb is explained by associating the full phi features of ʔinna with the EPP. As the phi set is inherited from ʔinna to T, the EPP property is satisfied by the preverbal subject or by adverbial intervening between ʔinna and the verb.

VERBAL COMPLEMENTIZERS IN ARABIC by Hossam Eldin Ibrahim Ahmed A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Linguistics The University of Utah December 2015 Copyright © Hossam Eldin Ibrahim Ahmed 2015 All Rights Reserved T h e U n i v e r s i t y o f U t a h G r a d u a t e S c h o o l STATEMENT OF DISSERTATION APPROVAL The dissertation of Hossam Eldin Ibrahim Ahmed has been approved by the following supervisory committee members: Edward J. Rubin , Chair June 3, 2015 Date Approved Patricia Hanna , Member June 3, 2015 Date Approved Aniko Csirmaz , Member June 3, 2015 Date Approved Howard Lasnik , Member June 3, 2015 Date Approved Kleanthes K. Grohmann , Member June 3, 2015 Date Approved and by Edward J. Rubin , Chair of the Department of Linguistics and by David B. Kieda, Dean of The Graduate School. ABSTRACT A class of Modern Standard Arabic complementizers known as ‘ʔinna and its sisters' demonstrate unique case and word order restrictions. While CPs in Arabic allow both Subject‐Verb (SV) and Verb‐Subject (VS) word order and their subjects show nominative morphology, CPs introduced by ʔinna ban a verb from directly following the complementizer. Preverbal subjects in ʔinna clauses show accusative case marking, while postverbal subjects show nominative morphology. Previous research explains these restrictions as default case or Multiple Case Assignment, both problematic for Case Theory as they violate the Activation Principle. This dissertation explains word order and case effects of ʔinna within the framework of Phase Theory and Feature Inheritance (FI). Morphological, historical, and usage evidence point out that ʔinna‐type complementizers have verbal properties similar to illocutionary verbs. Taking Case to be a reflection of phi features that T heads receive from higher heads (e.g. Complementizers) via Feature Inheritance, the nominative‐accusative alternation on preverbal subjects can be attributed to the selection of C heads: phi features on null complementizers and conditionals reflect as NOM, while phi features on Verbal Complementizers (VCs) reflect as ACC. VCs show similar Case behavior to the English Prepositional Complementizer for. They differ in distribution; while for only introduces a subordinate clause, and takes infinitival TP complements, VCs introduce a matrix clause iv and require finite TP complements, lending stronger support to Feature Inheritance theory than English for. Nominative postverbal subjects in ʔinna clauses are explained as an effect of antiagreement at Spell‐Out. Postverbal subjects and the Case probe on T are PF local, allowing for impoverished case agreement. Preverbal subjects and the Case licenser belong to different Phonological Phrases. To satisfy the Recoverability Condition, full case agreement is required between T and the subject, resulting in accusative morphology on the subject. Finally, the requirement that ʔinna‐clauses have an intervener between ʔinna and the verb is explained by associating the full phi features of ʔinna with the EPP. As the phi set is inherited from ʔinna to T, the EPP property is satisfied by the preverbal subject or by adverbial intervening between ʔinna and the verb. To Laila and Kareem. TABLE OF CONTENTS ABSTRACT ............................................................................................................................ iii LIST OF ABBREVIATIONS ................................................................................................... viii ACKNOWLEDGEMENTS ........................................................................................................ x 1. INTRODUCTION ........................................................................................................... 1 1.1. Theoretical Framework ................................................................................................... 3 1.2. Chapter Overview ............................................................................................................ 5 2. RESEARCH PROBLEM ................................................................................................. 10 2.1. Word Order, Case, and Agreement Under Null C.......................................................... 10 2.1.1. Arabic Word Order ................................................................................................ 11 2.1.2. Case Marking in Arabic .......................................................................................... 25 2.1.3. Agreement in Arabic .............................................................................................. 28 2.1.4. Case Marking on Fronted DPs ............................................................................... 32 2.2. Arabic Conditional Phrases ............................................................................................ 35 2.3. The ʔinna Clause ............................................................................................................ 41 2.3.1. ʔinna and Its Sisters ............................................................................................... 42 2.3.2. ʔinna and Its Sisters: Distribution .......................................................................... 46 2.3.3. ʔinna and Its Sisters: Word Order ......................................................................... 49 2.3.4. ʔinna and Its Sisters: Case Marking ....................................................................... 53 2.3.5. ʔinna and Its Sisters: Summary.............................................................................. 56 2.4. Accusative Subjects After ʔinna: The Research Problem .............................................. 57 3. REVIEW OF LITERATURE ............................................................................................ 62 3.1. Default Case Analysis ..................................................................................................... 62 3.1.1. Mohammad (1988) ................................................................................................ 63 3.1.2. Mohammad (2000) ................................................................................................ 70 3.1.3. Ouhalla (1994) ....................................................................................................... 76 3.1.4. Soltan (2006) ......................................................................................................... 82 3.1.5. Summary of Default Case Accounts in ʔinna Clauses ............................................ 90 3.1.6. Default Case According to Schütze (1997, 2001b) ................................................ 91 3.2. Multiple Case Assignment ............................................................................................. 97 3.2.1. Multiple Structural Case ........................................................................................ 98 3.2.2. Multiple Morphological Case .............................................................................. 102 vii 3.3. (NOM) Case Is the Realization of Another Feature ..................................................... 104 3.3.1. Mood Features as a Case Licenser ...................................................................... 105 3.3.2. Case as a T Feature (Pesetsky and Torrego 2001; Pesetsky and Torrego 2004) . 109 3.4. Summary ...................................................................................................................... 112 4. VERBAL COMPLEMENTIZERS ................................................................................... 114 4.1. Introduction ................................................................................................................. 114 4.2. Verbal Properties of ʔinna ........................................................................................... 115 4.2.1. Morphological Evidence ...................................................................................... 117 4.2.2. Evidence from Classical Arabic ............................................................................ 126 4.2.3. Other Complementizers ...................................................................................... 130 4.3. Illocutionary Properties of ʔinna ................................................................................. 132 4.3.1. Complementary Distribution with Illocutionary Verbs ....................................... 133 4.3.2. Interaction with Adjuncts .................................................................................... 135 4.3.3. Other Complementizers ...................................................................................... 137 4.4. ʔinna a Verbal Complementizer .................................................................................. 139 4.5. ʔinna Not a Verb .......................................................................................................... 142 5. CASE FROM C ........................................................................................................... 148 5.1. Case Valuation ............................................................................................................. 148 5.1.1. Accusative Subject Under Canonical ʔinna+SV Word Order ............................... 150 5.1.2. Case with Adjuncts+VS Word Order .................................................................... 153 5.1.3. Expletive Pronouns After ʔinna ........................................................................... 159 5.1.4. Summary .............................................................................................................. 164 5.2. Problems Avoided ....................................................................................................... 165 5.2.1. Default Case Avoided .......................................................................................... 165 5.2.2. Multiple Case Avoided ......................................................................................... 170 6. CONCLUSION ........................................................................................................... 171 6.1. Summary ...................................................................................................................... 171 6.2. Theoretical Implications .............................................................................................. 172 6.2.1. Anticase and Antiagreement ............................................................................... 173 6.3. Areas for Further Research .......................................................................................... 174 REFERENCES .................................................................................................................... 178 LIST OF ABBREVIATIONS 1S first person singular 1p first person plural 2fs second person feminine singular 2ms second person masculine singular 3ms third person masculine singular 3mp third person masculine plural 3fs third person feminine singular 3fp third person feminine plural ABS absolutive ACC accusative CA Classical Arabic CI Conceptual‐Interpretive interface DAT dative DCA Default Case Analysis EA Egyptian Arabic f feminine FI Feature Inheritance fut future ix GEN genitive IND indicative LSF Long Split Focalization MCA Multiple Case Assignment MCC Multiple Case Checking MSA Modern Standard Arabic NOM nominative PC Prepositional Complementizer SM Sensorimotor Interface SO Syntactic Object SUBJ subjunctives SVAA Subject‐Verb Agreement Asymmetry VC Verbal Complementizer ACKNOWLEDGEMENTS They say it takes a village to raise a child. Many people call their dissertations their baby, and now I know why. This work would not have been possible without the help and support of so many people that it seems my role in it was to contribute the imperfections in spite of their best efforts. Ed Rubin has helped me grow with the patience of saints. He always knew exactly when to give me space to explore, and what kind of direction would help me find the way forward. His continuous encouragement and belief in me helped me navigate academic as well as personal circumstances. Often times he would explain once and then again in different ways what I stubbornly could not see immediately. Thanks, Ed, for being such a great mentor, and for letting me claim the privilege of walking into your office without making appointments. Aniko Csirmaz raised the bar of what kind of professor I would like to be. A leader by example, Aniko showed me how to strive for and achieve excellence. I learned from her how to balance genuine care and firm guidance to my students. As an academic, Aniko helped me develop so many research skills some of them she might not even remember she did. Many of the quality parts of this dissertation are the product of good research habits I acquired through Aniko's consistent guidance. I started my first day at the University of Utah in Patricia Hanna's class as an MA student, and it was the hardest class I had ever taken. After two courses with her, and our occasional conversations, xi she changed me from a person who makes fun of people who talk about dead renaissance philosophers, to a person who knows why colleagues from different linguistic approaches and traditions have difficulties talking to each other, and can talk to colleagues from different traditions in linguistics without facing such difficulties, to a person who tells his colleagues that whatever they do is grounded in philosophy, and that they'd better know what that ground is. Professor Hann, I owe this aspect of my intellectual maturity to you. I remember when I asked Howard Lasnik to serve on my committee, I was worried he might not have the time for a student from a department so far away. He was worried I would take up all of his time, and I told him I was happy with whatever time he could give. To my pleasant surprise, I could see how he would take the time not only to offer general guidelines, but also to actually annotate my comps papers, and give feedback that took more time than I had expected of him. From thousands of miles away, Kleanthes Grohmann would send me follow‐up emails that not only pointed me to interesting literature that he somehow knew I overlooked, but also gave me assurance that I was going in the right direction. I also appreciate him putting up with the time difference. Both Kleanthes and Howard were flexible and reassuring as I shifted my dissertation topic, and made me feel safe in approaching them, in the past and future, as I rely on their guidance for my academic growth. I am truly blessed with my family and friends who have given me such strong and unconditional support. My parents and siblings have always been there for me, and I knew that if ever I fell they would catch me no matter what. I could go anywhere and be anyone, and enjoy their love and support. Words cannot describe the support Dalia xii has given to me on the long road towards this dissertation. You had to put up with all the frustrations of graduate school, and gained none of the benefits. No one else would have tolerated what we have gone through over the past years, but you have stepped up and kept us a great family with the kind of strength and love that only you could give. Rachel Hayes‐Harb and Rania Abou Shaeshaa changed the way I view knowledge. Rachel helped me see that it is not enough to acquire knowledge, or to find something new. She made me realize the ethical responsibility of making this knowledge available and valuable to the community. Rania showed me how sharing knowledge can build and help a community, and how such an act of sharing is an intuitive and spontaneous part of humanity. My colleagues in the graduate program at the University of Utah have been such a wonderful network of support. I will not risk forgetting a name, so I will mention none. You know who you are. Thank you guys for being such a great community where we shared the ups and the downs. Thank you for all the times when I showed up (and Dalia, and the kids), and thank you for forgiving me when I couldn't be there. Hamza Benzina single‐handedly made the proposal defense for this dissertation possible. I am proud to have been your mentor, and honored that you have made me your friend. I have received enormous support personally and professionally from Fernando Rubio. His mentorship helped me become the well‐rounded professional‐academic‐teacher that I am today. From best practices in teaching, to best practices in research, to best practices in professional conduct, I hit the jackpot having him as my mentor and friend. It is amazing how quickly Maghiel van Crevel, Petra Sijpesteijn and Ahmad Al‐ xiii Jallad made it their personal duty to support me in completing this thesis. Maghiel's encouragement, support, and advice made possible the timely completion of this dissertation. Without Petra's support in practical matters, and encouragement and assurance in times of panic, this dissertation would not have been completed. From Day one, Ahmad has been a great ‘buddy' as he likes to call me. He made himself available for chats about areas only peripheral to his specialization, and I can only hope to be able to pay forward all the professional support and backing I have received from him, since there is no way I can pay him back all the awesomeness. To name the rest of the community that made this dissertation possible will exceed the page limit imposed by the University, so I have to stop here. For those who made this work possible, just because your name is not in these lines does not mean I have forgotten how you helped me get here. Salt Lake City, June 18, 2015 Dedication: To Laila and Kareem - you will be surprised how little daddy knows 1. INTRODUCTION This dissertation is primarily concerned with the behavior of a class of complementizers in Modern Standard Arabic (MSA) known as ‘ʔinna and its sisters' in traditional grammar. Specifically, it investigates case alternation and word order restrictions associated with these complementizers. Although MSA allows both subject‐verb‐ object (SV(O)) and verb‐subject‐object (VS(O)) word orders after null complementizers (as in 1 and 2), clauses introduced by ʔinna do not allow the latter, as can be seen in the ungrammaticality of 4. Subjects of ʔinna‐clauses show accusative, rather than nominative, morphology (3). (1) Zayd‐un qaabala saalim‐an Zayd‐ACC met Saalim‐ACC ‘Zayd met Saalim.' (2) qaabala zayd‐un saalim‐an met Zayd‐NOM Saalim‐ACC ‘Zayd met Saalim.' (3) ʔinna zayd‐an qaabala saalim‐an indeed Zayd‐ACC met Saalim‐ACC ‘Zayd met Saalim.' (4) *ʔinna qaabala zayd‐un saalim‐an indeed met Zayd‐NOM Saalim‐ACC ‘Zayd met Saalim.' The case behavior of subjects of ʔinna‐clauses is unexpected. If we assume that T is the locus of Case licensing, by the time the complementizer is merged into the 2 structure, the subject DP will have already received nominative, and is unavailable for further valuation. If we assume that C is the source of Case licensing through Feature Inheritance (FI) or through agreement with T, accusative subjects and the ungrammaticality of postverbal subjects are unexpected because TP complements of null C and ʔinna are otherwise identical. This dissertation resolves this Case mismatch through proposing that ʔinna and its sisters in Arabic are a class of Verbal Complementizers (VCs), complementizers that share morphological and lexical features with verbs, while still belonging to category C. Taking Case valuation in subject position to originate at C (Chomsky 2007a; Chomsky 2008), the nominative‐accusative variation can be explained by [±verbal] feature on C heads. [+verbal] features in VCs transfer phi features and accusative Case to T heads, which value Case features on the subject as accusative. [‐verbal] features on other Arabic complementizers (null C and conditionals) contribute phi features that are interpreted as nominative. This dissertation contributes by examining a phenomenon that has been taken for granted in Arabic generative syntax as a result of Case assignment under a ‘default mechanism,' and provides further evidence to a discussion in syntax as a whole about the nature of Case, how Case is related to phi features and agreement, and syntax‐morphology interface. The proposal of this thesis further supports the (Minimalist) goal that linguistic features must be interpretable at LF. This introduction is structured as follows. In section 1.1, I lay out the general theoretical framework. In section 1.2, I provide summaries of each chapter in the 3 dissertation. 1.1. Theoretical Framework In this section, I lay out the theoretical assumptions that make the backdrop of the arguments in this dissertation. This dissertation assumes the general framework of the Minimalist Program (Chomsky 2008; Chomsky 2007a).1 Derivation begins with a selection of a Lexical Array (LA) from the Lexicon (Lex) into the Workspace, where syntactic computations take place. Structure is built via operation Merge: two Syntactic Objects (SOs) α and β are taken from LA, and are merged to create a new Syntactic Object {γ,{α, β}}. Lexical Items (LIs) are SOs, and the output of operation Merge is also an SO. Merge is feature‐driven. A Syntactic Object that carries an unvalued feature, called a probe, will search the workspace for an SO that carries a matching valued feature, a goal. When a goal is found, the unvalued feature on the probe is valued via operation Agree, the goal is rendered inactive (unavailable for further agreement with other probes). If the probe contains an Edge Feature (EF), Merge takes place. If the goal is contained in the probe, Merge copies the goal and merges with the copy, resulting in Internal Merge (IM; Move). Chomsky (2007) argues that only phase heads are probes. For the purposes of this dissertation, this approach includes phi features and Case on C. To derive subjects in the specifier of T (spec,T), FI takes place whereby phi features are transmitted from C to T, T becomes a probe by virtue of inheritance, locates a goal in 1 My analysis assumes Chomsky (2008) and previous research it is based on, unless otherwise stated. 4 spec,vP where relevant, and moves it to spec,T under the right circumstances. FI (Chomsky 2008) captures the correlation between phi features on T heads and their ability to be selected by a complementizer. In English matrix clauses, for example, infinitival T heads cannot be selected by the null complementizer (but rather by an ECM verb or the complementizer for in subordinate clauses), and they do not have any phi features (do not show subject agreement). Prior to Phase Theory, this generalization has been captured in terms of selection: null C selects phi‐complete (finite) T, and V (and for) selects phi‐defective (infinite) T.2 Richards (2007) notes that FI has two advantages over selection. First, it explains the correlation between Phi and C. It is C that carries phi features and Tense features, if it is not merged into the structure, T does not inherit such features. Second, FI allows for a more uniform characterization of phase heads, as v carries phi features as well. Richards argues that FI is conceptually necessary to accurately account for the timing of feature valuation and transfer, and the Phase Impenetrability Condition (PIC). Epstein and Seely (2002) show that feature valuation and transfer to the interface must take place simultaneously, otherwise the interpretable/uninterpretable distinction will be lost. But if transfer takes place simultaneously with feature valuation, PIC will no longer be tenable, as chain heads will also be shipped to the interfaces and will no longer be available as ‘escape hatches' for further computation. Richards argues that this problem disappears if phi features are not T features. This way, T can merge to the structure, and inheritance, valuation, and transfer can all take place at the phase level when C is merged to the structure. 2 In Chapter 5, I show that Arabic has a third type of infinitivals which has only a partial phi set. 5 Chomsky notes that it is sometimes felt intuitively that FI is countercyclic, but such perceived countercyclicity is no more so than it is for Case valuation in‐situ. However, Feature valuation is certainly different from structure building via Merge.3 He also notes (Chomsky 2008:144) that although the device of inheritance is in narrow violation of NTC, it still satisfies SMT because of the role FI plays at the C‐I interface. Of course, there is the possibility that there is another device that has the same effect as FI without NTC violation, hence better satisfying SMT. For the purposes of this dissertation, I assume FI as the best available device that satisfies SMT so far. At a certain point of the derivation, operation Spell‐Out takes place. Spell‐Out sends the domain of the phase (the complement) to the interfaces (PF to the sensorimotor (SM) interface, and LF to the Conceptual‐Interpretive (CI) interface). vP and CPs (and possibly DPs and PPs) are phases. Completed Phases are available for further computations subject to PIC. The domain of the phase is inaccessible: it is invisible to higher probes outside the phase. A probe can only search the edge of a phase; its head and specifier. 1.2. Chapter Overview Chapter 2 lays out the research problem. ʔinna ‘that/indeed' is a member of a set of Arabic complementizers that may not be followed by a verb‐first TP. The 3 If for nothing else, FI and Case‐in‐situ differ in how they can be proven to violate the No‐Tampering Condition/Extension Condition. If structure can be only extended at the root, it is easy to see how inserting a specifier of a none‐phase head after the merge of a higher phase head entails breaking an already existing c‐command relation (in whatever set‐theoretic representation we assume), and re‐create that relation after inserting the specifier. Case valuation in‐situ does not have the same shortcoming. The valuation of unvalued Case on a target DP via Agree does not entail breaking any previously established relations. 6 preverbal subject of the clause is in the accusative, in contrast to its nominative morphology in all other contexts. Assuming that morphological case mirrors syntactic Case, the accusative case on the subject poses a licensing problem: by the time the complementizer is merged, the subject has already moved to spec,T, valued its Case feature, and is inactive and invisible for the Case probe on C. Chapter 2 lays out the larger context of word order, Case/case, and agreement facts in Arabic. The aim is to give the reader an overview of the contexts and restrictions where ʔinna‐clauses occur, and contrast the properties of ʔinna‐clauses with those headed by other complementizers. Arabic allows a variety of word orders: subject‐first, verb‐first, pro‐drop, and nonovert copular clauses, with restrictions related to definiteness of the subject and the position of the object. SV agreement interacts with word order: SV word order shows full agreement (gender and number), while VS word order shows only gender agreement. Arabic is a nominative‐accusative language. The subject carries nominative case and the object accusative, regardless of word order. The only exception is when the sentence is introduced by an ʔinna‐type complementizer, where the preverbal subject is in the accusative. In Chapter 3, I review previous accounts related to the issue. Under Default Case Analysis (DCA) (Mohammad 2000; Mohammad 1988; Soltan 2006; Ouhalla 1994), nominative subjects in spec,T have their Case valued by a default procedure, in the absence of a Case probe. In the presence of a Case licenser (such as ʔinna), subjects in spec,T are available to receive accusative from them. Multiple Case Assignment (MCA) (Bejar and Massam 1999) allows structural nominative Case to be assigned/valued to 7 the subject in spec,T, then again the subject is available for the accusative probe when the complementizer ʔinna is merged into the structure. A third approach is to analyze Case as a reflection of another feature. Leung (2011) proposes that Case is a reflection of the [mood] feature of complementizers. Under this analysis, the nominative‐accusative variation depends on the mood of the CP as determined by the complementizer. Pesetsky and Torrego (2001; 2004) propose that Case is an uninterpretable T feature on D. Under that approach, complementizers such as English that or Arabic ʔinna are, in fact, of category T and have moved to C via head movement. Variation in Case/case can then be taken to reflect different types of T heads. In Chapter 4, I show that ʔinna‐type complementizers are Verbal Complementizers (VCs). Unlike other complementizers (null C and conditionals), VCs share some morphological and lexical features with verbs or verblike lexical items. ʔinna‐type complementizers also share semantic and pragmatic features with verbs indicating assertion, hope, and regret. These morphological and semantic characteristics are not shared with other complementizers. Verbal complementizers in Arabic, like the prepositional complementizer for in English, license Case for subjects and value Case for those subjects with the same case value that their (original) lexical category (preposition or verb) would value to its complement (accusative in both cases). Chapter 5, examines the implications of considering ʔinna‐type complementizers Verbal Complementizers, and explains case and word order restrictions in the ʔinna‐clause. Case alternation across different complementizer types are expressed in terms of the presence or lack of the verbal property of the complementizer at hand. Null C and 8 conditionals lack the verbal property and value Case on their goal subjects as NOM. ʔinna‐type complementizers are VCs and value Case on their goals as ACC. Word order restrictions are accounted for in terms of EPP: ʔinna requires a TP complement that has EPP, while other complementizers accept TP complements with or without EPP. The EPP feature in the ʔinna complement TP can be satisfied by the subject in spec,TP, resulting in SV word order. It can also be satisfied by an adjunct AP or PP to the T head, allowing instances of postverbal subjects only in the presence of preverbal adjuncts. Case asymmetry after ʔinna, where postverbal subjects are nominative and preverbal subjects are accusative is accounted for as a Spell‐Out condition. When the subject and the Case licenser are in the same phonological phrase, i.e. in postverbal subjects, ACC is weakened allowing nominative to appear in the morphology. When the subject is in a different phonological phrase, i.e. with preverbal subjects, ACC is enforced by the need for recoverability, weakening the effect of the verbal property of ʔinna and surfacing as the nominative morpheme. Chapter 6 is the conclusion. It discusses theoretical implications of the proposed analysis in Chapter 5. The proposed analysis weakens DCA and MCA approaches. The anticase account for postverbal ʔinna subject is consistent with antiagreement account as a Spell‐Out morphological effect, and that account supports the account that Case is a reflect of phi features. The anticase account is argued to be different from DCA, in that DPs are always licensed in the narrow syntax, and the morphological realization of Case is systematically accounted for as a Spell‐Out operation. Finally, given that phi features and Case go hand‐in‐hand, the implications of 9 implementing the proposed algorithm in terms of FI versus Agreement are discussed, with pros and cons for each account briefly discussed, and are left for future research. 2. RESEARCH PROBLEM Clauses introduced by a certain class of Arabic complementizers, known in traditional grammar as ʔinna and its sisters, display unique word order and case behavior. The set of ʔinna‐type complementizers consists of ʔinna (that, indeed), ʔanna (that), liʔanna (because), kaʔanna (as if), lakinna (but), layta (if only), laʕalla (hopefully). To help see such unique behavior, I compare ʔinna clauses with CPs headed by null complementizers and those headed by conditionals. I begin by giving an overview of word order, Case and agreement in MSA clauses headed by null C (section 2.1). I then provide an overview of CPs headed by conditionals in section 2.2. In section 2.3, I describe in detail word order and case facts of ʔinna clauses (clauses introduced by ʔinna or one of its ‘sisters'). In section 2.4, I describe in general terms the problem with accusative subjects in ʔinna clauses. 2.1. Word Order, Case, and Agreement under Null C In this section, I first describe possible word order variation under null complementizers, then I describe agreement and case facts for each possible word order. This overview of word order is relevant as it highlights the restrictions imposed by ʔinna. Facts about agreement are relevant as much of the literature which touches on the case/Case behavior with ʔinna do so within the context of what Soltan (2006) refers to as the Subject‐Verb Agreement Asymmetry (SVAA). 11 I assume that a sentence with no overt complementizers is a CP with null C, based on the coordination evidence in 5. (5) Zaydun yuHibbu mona wa layta mona tuHibbu zaydan (ʔayDan) Zayd loves Mona and if only Mona loves Zayd (too) ‘Zayd loves Mona and hopefully/if only Mona loves Zayd too.' The two phrases before and after the conjunction wa (and) are of the same type. The phrase after wa contains a complementizer (layta/laʕalla - if only/hopefully), indicating that the postconjunction phrase is a CP. This in turn makes the preconjunction phrase, which lacks an overt complementizer, a CP headed by a null C head. 2.1.1. Arabic Word Order In this section, I first describe word order in sentences with intransitive, transitive, and ditransitive verbs. I will then move to describing word order in sentences with no overt copula. Following Mohammad (2000), I show that in sentences with a subject, a verb, and an object, out of the six logically possible word orders, four are syntactically well‐formed. Sentences where both the subject and the object are preverbal are unacceptable. I aim by this description to familiarize the reader with an overview of possible word orders that can be contrasted with the more restricted set of available word orders under ʔinna‐type complementizers. In terms of word order, Arabic allows verb‐first, verb‐second, and pro‐drop sentences, as in 6 - 8. DPs ending in /a/ in Arabic do not show case marking for phonological reasons, they are intentionally selected for the examples in this section, as a more detailed discussion of case will follow. 12 (6) Nadia waSalat Nadia arrived ‘Nadia arrived.' (7) waSalat Nadia arrived Nadia ‘Nadia arrived.' (8) waSalat arrived (3FS) ‘She arrived' Transitive verbs demonstrate the word orders in 9 - 11, where the object follows the subject. (9) Nadia qaabalat Nuura Nadia met Nora ‘Nadia met Nora.' * ‘Nora met Nadia.' (10) qaabalat Nadia Nuura met Nadia Nora ‘Nadia met Nora.' * ‘Nora met Nadia.' (11) qaabaluu Nadia met.3mp Nadia ‘They met Nadia.' In double object constructions, the indirect object precedes the direct object (12). The indirect object can also occur as part of PP (13). The word order in 12 is strict, as can be seen in the ungrammaticality of alternative interpretations. The subject‐direct object order in 13 is strict, but the two objects have a free word order in dative shift, as can be seen in the grammaticality of 14. 13 (12) manaHat Nadia Nuura hadaaya gave Nadia Nora presents ‘Nadia gave Nora presents.' * ‘Nora gave Nadia presents.' * ‘Nadia gave presents Nora' (13) manaHat Nadia hadaaya li‐Nuura gave Nadia presents to‐Nora ‘Nadia gave presents to Nora.' * ‘Nora gave presents to Nadia' (14) manaHat Nadia li‐Nuura hadaaya gave Nadia to‐Nora presents ‘Nadia gave presents to Nora.' * ‘Nora gave presents to Nadia' The examples above show that in the absence of overt case marking, Arabic allows preverbal and postverbal subjects, and objects maintain a postverbal, postsubject position. Sentences where argument DPs show case marking may have some more flexibility as they relax the relative order requirement between postverbal subjects and objects. Overt case marking allows for more flexibility in word order. When case marking is not obscured by phonological processes, four of the six logically possible word orders of the subject, verb, and object are possible in Arabic, as the data in 15 (Mohammad 2000:3) shows. Mohammad notes that sentences where both the subject and the object are preverbal are ‘marginal' (p.2), and marks them with question marks (15d, 15f). He does not specify in what way they are marginal. I interpret this statement to mean that these sentences (15d and 15f) are ungrammatical under normal (unmarked) contexts, without heavy discourse, presupposition, PF preposing/postposing operations, or splitting the sentence into two phonological phrases. Such interpretation is consistent 14 with my own intuitions. (15) a. qaabala zayd‐un ʕamr‐an met.3MS zayd‐NOM Amr‐ACC ‘Zayd met Amr.' b. qaabala ʕamr‐an zayd‐un met.3MS Amr‐ACC zayd‐NOM ‘Zayd met Amr.' c. zayd‐un qaabala ʕamr‐an zayd‐NOM met.3MS Amr‐ACC ‘Zayd met Amr.' d. ?zayd‐un ʕamr‐an qaabala zayd‐NOM Amr‐ACC met.3MS ‘Zayd met Amr.' e. ʕamr‐an qaabala zayd‐un Amr‐ACC met.3MS zayd‐NOM ‘Zayd met Amr.' f. ?ʕamr‐an zayd‐un qaabala Amr‐ACC zayd‐NOM met.3MS ‘Zayd met Amr.' (Mohammad, 2000: 3. his (3)) Arabic exhibits another restriction on sentence‐initial DPs: they cannot be indefinite. 16 is grammatical with a definite subject, al‐waziir (the minister). The subject in 17 and 18 is indefinite; the indefinite subject waziir may not be sentence‐initial. (16) al‐waziir‐u fii lijtimaaʕ‐i the‐minister‐NOM in the.meeting‐DAT ‘The minister is at the meeting'. (17) fii lijtimaaʕ‐i waziir‐un. in the.meeting‐GEN minister‐NOM ‘A minister is at the meeting'. 15 (18) *waziir‐un fii lijtimaaʕ‐i minister‐NOM in the.meeting‐DAT ‘A minister is at the meeting'. Sentences like 18 may be grammatical if the indefinite subject has a certain degree of specificity.4 I do not attempt to explain this restriction here, but rather take it as a given. This restriction on indefinites is especially important as it is used to show that ʔinna assignment of ACC is nonlocal and can be long‐distance. Copular sentences in past and future tense, as well as negative copular sentences show copulas overtly, as can be seen in 19 - 24. The sentences in 19 and 20 are past, 21 and 22 future, and 23 and 24 are negated present copular sentences. (19) kaanat hind‐un ustaathat‐an was Hind‐NOM professor‐ACC ‘Hind was a professor.' (20) hind‐un Kaanat ustaathat‐an Hind was professor (F.NOM) (3FS) (F‐ACC) ‘Hind was a professor.' (21) Sa‐takuun hind‐un ustaathat‐an will‐be Hind professor (3FS) (f‐NOM) (F‐ACC) ‘Hind will be a professor.' (22) hind‐un satakuun ustaathat‐an Hind will be professor (F‐NOM) (3FS) (F‐ACC) ‘Hind will be a professor.' 4 cf. Brustad (2000, p. 21) for a discussion on definiteness in Spoken Arabic and Mohammad (2000:9) for definite subjects in MSA. 16 (23) laysat hind‐un ustaathat‐an is not Hind professor (3FS) (F‐NOM) (F‐ACC) ‘Hind is not a professor.' (24) hind‐un laysat ustaathat‐an Hind is not professor (F‐NOM) (3FS) (f‐ACC) ‘Hind is not a professor.' The examples in 19 - 24 show that the past tense copula (kaan), the future tense copula (sayakuun), and negative copula (lays) have the same word order distribution as transitive verbs; they allow VS and SV word orders. The present declarative counterpart (25) of these sentences does not have an overt copula. (25) hind‐un ustaathat‐un Hind‐NOM professor‐NOM ‘Hind is a professor.' This type of sentence consists of a subject DP followed by a predicate DP (25), AP (26), or PP (27). (26) Al‐ustaathatu jamiilatun / taʕbaanatun/miSriyyatun professor beautiful / tired/Egyptian ‘The professor (F) is beautiful (F) / tired (F) / Egyptian (F).' (27) al‐astaathithatu fii l‐maktabi the‐professors in the‐office ‘The professors are at the office.' Word order in null‐copular sentences is highly rigid. Any change in the word order in 25 and 26 is marked, perhaps as high rhetoric or focus, which I ignore. In sentences with PP predicates (like 27) the subject is obligatorily sentence‐initial, unless it is indefinite where the ban on sentence‐initial indefinites kicks in, and the indefinite 17 subject occurs after the predicate PP. This data will be relevant in our discussion of whether equative sentences are null‐copular sentences, or whether they are indeed verbless, which in turn bear on the discussion of (default) case assignment and ʔinna. 2.1.1.1. Null‐Copula or Verbless? Because there is no overt copula in sentences like 25 -27, there has been controversy on whether there is a copula at all. The issue of whether sentences like 25 are null‐copular (Bakir, 1980; Fassi Fehri, 1993, 2012, among others) or are ‘verbless' (merely a projection, lexical or functional, that does not subcategorize for a VP) (Benmamoun 2000 and references cited there; Aoun et al. 2010) is still largely an open discussion. I will outline very briefly the major arguments for and against a null‐copula analysis. The account proposed in this dissertation is concerned with a higher domain than that of the copula (i.e. that of C‐T), which makes it compatible with both null‐copula and verbless analyses. I adopt the null‐copula analysis, merely for ease of exposition. Proponents of the verbless sentence analysis claim that sentences like 26 have a structure like in 28, where T lacks a [V] feature, hence does not require a copula. (28) To support this claim, they contrast verbless sentences with their past and future counterparts (19 - 24) where the predicate shows accusative morphology, unlike verbless sentences (25), where the predicate is in the nominative. If there were a covert 18 copula, it is hard to explain why the present copula fails to assign the same case as the past and future counterparts.5 Additional evidence comes from negation in Moroccan Arabic (Benmamoun 2000). In Moroccan Arabic,6 negation of copular sentences is achieved via the circumfix ma‐shi on the copula (29 and 30). (29) ma‐kan‐shi Omar f‐d‐dar NEG‐be.past.3MS‐NEG Omar in‐the‐house ‘Omar was not in the house.' (30) Omar ma‐kan‐shi f‐d.dar Omar NEG‐be.past.3MS‐NEG in‐the‐house ‘Omar was not in the house.' A null‐copular analysis would predict that the negative present tense counterpart of these sentences will allow mashi in the same positions as their past counterpart, contrary to facts evidenced in 31 and 32, where the sentence in 31 is ungrammatical. (31) *ma‐shi Omar f‐d‐dar NEG Omar in‐the‐house ‘Omar was not in the house.' (32) Omar ma‐shi f‐d‐dar Omar NEG‐be.past.3MS‐NEG in‐the‐house ‘Omar was not in the house.' The same point has been made by Aoun and colleagues (2010). They further support a verbless analysis with data showing that a verbless sentence fails to allow a modal like yqdar ‘can, may' indicating permission. Bahloul (2008) makes a similar point 5 This question is related to a larger issue of whether arguments need case, or predicates as well (by virtue of being DPs). In all cases, if we assume that null copulas do not move out of their base position to T, they will be local to the complement, allowing anticase feature suppression to take place at Spell‐Out (see anticase account in Chapter 5). 6 Benmamoun notes that the argument is originally proposed for Egyptian Arabic by Jelinek (1981), but it is also applicable to Moroccan Arabic. 19 with qad ‘may' indicating possibility. According to those authors, if there were a null copula, these modals should be allowed in the sentences. Indeed, sentences with modals are only grammatical with the overt present tense verb yakoon ‘be'. Proponents of the null‐copular analysis, on the other hand, offer an alternative analysis. Fassi Fehri (2012) deals with null copulas as a (lack of) response to agreement requirements. Comparing copular sentences in Arabic, English, and French, he argues that as English and French do not have overt T morphology in present tense, the overt realization of the copula in sentences like 33 and 34 is motivated by agreement with the subject. In Arabic, the counterpart (35) does not require an overt copula because the agreement requirement is not enough to force an overt realization of the copula. (33) The man *(is) at the house. (34) L'homme *(est) a la maison. (35) Ar‐rajulu (*yakuunu) fii almanzili Empirical evidence seems to support a null‐copular analysis as well. When morphological content additional to agreement is present, the present copula yakuun appears. Take, for example, the auxiliary qad (may), which requires an infinitive. In contrasting 36 with 37 and 38, and building on Fassi Fehri's analysis, it can be argued that the presence of qad invokes (infinitival) T morphology on the copula, which, together with agreement, are enough to force an overt realization of takuun. (36) hind‐un ustaathat‐un Hind‐NOM professor‐NOM ‘Hind is a professor.' 20 (37) hind‐un qad *(takuuna) ustaathat‐an Hind‐NOM may be3FS professor‐ACC ‘Hind may be a professor.' (38) qad *(takuuna) hind‐un ustaathat‐an may be3FS Hind‐NOM professor‐ACC ‘Hind may be a professor.' The evidence from negation in Moroccan Arabic is not necessarily applicable to MSA. Indeed, the (lack of) parallelism Benmamoun draws in the sentences 29 - 32 does stand in MSA. In MSA, the present tense negation of the so‐called verbless sentence is achieved by means of the negative copula laysa. The sentences in 23 and 24, repeated here as 39 and 40, show that MSA negated sentences allow the negative copula laysa in the same positions that are available for the overt past tense copula. These sentences suggest that in the present tense, the copula with negation is overt, but the copula without negation is covert. (39) laysat hind‐un ustaathat‐an is not Hind professor (3FS) (F‐NOM) (F‐ACC) ‘Hind is not a professor.' (40) hind‐un laysat ustaathatan Hind is not professor (F‐NOM) (3s) (F‐ACC) ‘Hind is not a professor.' Fassi Fehri (1993:87-88)7 makes a similar point with laysa. He uses agreement morphology (the third person feminine singular -t of laysat) as evidence that not only is there a copula in the present tense, but that it also projects an AgrP. The other piece of 7 There are a few typos and a relevant example is misnumbered in this part of the book; they may be a bit confusing at first. The reader is encouraged to approach that reference cautiously. 21 evidence Fassi Fehri uses to support a null copula analysis is how the so‐called verbless sentence interacts with tense. Sentences that contain a verb in the present restrict the use of adverbs of time such as l‐ʔaana (now) or l‐ʔamsi (yesterday), as can be seen in the contrast between 41a and 41b. (41) a. yaaʔkulu r‐rajul‐u l‐ʔaan‐a eats the‐man‐NOM the‐now‐ACC ‘The man eats now.' b. *yaaʔkulu r‐rajul‐u ʔamsi eats the‐man‐NOM yesterday ‘The man eats yesterday.' (Fassi Fehri 1993:87, his (3)) The use of a temporal adverb is restricted by the tense of the verb. The sentence in 41b is ungrammatical because the temporal adverb yesterday is incompatible with the present tense verb morphology, unlike 41a. Similarly, the sentence in 42b is ungrammatical, while it is grammatical in 42a. Fassi Fehri takes the ungrammaticality of the former to be due to incompatibility between the temporal adverb ʔamsi (yesterday) and the tense of the sentence. He further takes this as evidence of the existence of a TP in the sentence, hence a null copula. (42) a. r‐rajul‐u mariiD‐un l‐ʔaan‐a the‐man‐NOM sick‐NOM the‐now‐ACC ‘The man is sick now.' b. * r‐rajul‐u mariiD‐un ʔamsi the‐man‐NOM sick‐NOM yesterday (Fassi Fehri 1993:87, his (4)) It may be easier to explain the nominative case of the complement (rather than accusative) if we stipulate there is no covert copula in sentences like 36, allowing for nominative on the complement via some sort of case copying/agreement/percolation 22 from the subject, as proposed by Abu‐Joudeh (2013) or via a default case analysis where nominative is the morphological case that appears in the absence of a case assigner. However, it is much harder for a verbless‐sentence analysis to account for the structural Case licensing of the second DP (ustaathat) in 36. It is not possible to assume that predicative DPs in Arabic do not need Case licensing if they acquire their case in some other way. Both proposals - that case on the predicative DP is received via agreement with the subject, and that they do not need Case - are invalidated by the fact that in embedded verbless sentences like 43 the predicate DP ustaathatan (professor) shows different case marking from that of the subject. This means that case agreement with the subject does not take place, leaving the need to justify nominative morphology on the predicate as a reflect of syntactic licensing. (43) qulna ʔinna hind‐an ustaathat‐un say1PL that Hind‐ACC professor‐NOM ‘We said that Hind is a professor.' Furthermore, a verbless‐sentence analysis is faced with the burden of explaining why Arabic sentences in all tenses, negative and declarative, should have a verbal projection in their derivation, with the sole exception of present declarative sentence. A unified account that postulates a phonologically null copula is simpler, and preferable, to an account that postulates two different types of predication for largely similar constructions. In this section, I have briefly outlined arguments for and against a null‐copular analysis of subject‐predicate sentences. For detailed arguments supporting the existence of a verbal projection in Arabic, as well as its implications to word order and 23 agreement, the reader is referred to Mohammad (2000) and Mohammad (1989). 2.1.1.2. Word Order with Fronted DPs While Mohammad (2000:2) notes the marginality of SOV and OSV word orders, he indicates that such word orders are acceptable if there is a resumptive pronoun co‐referring with the object. Implementing this to the sentences in 15d and 15f, we get the sentences in 44. (44) a. Zayd‐un amr‐an qaabala‐hu Zayd‐NOM Amr‐ACC met3MS‐him ‘Zayd met Amr.' b. amr‐an Zayd‐un qaabala‐hu Amr‐ACC Zayd‐NOM met3MS‐him ‘Zayd met Amr.' Mohammad notes that 44a and 44b are the counterparts of in 15d and 15f, with the resumptive -hu marking the gap. He notes that this is an acceptable word order, more acceptable than the 15 counterparts. However, there is evidence that the correspondence that Mohammad notes between resumptive pronouns and the acceptability of otherwise ungrammatical sentences is independent from the position of the object. DP extraction from pretty much any island is permitted if it leaves behind a resumptive, as can be seen in 45. (45) a. Nuura kalbu‐*(ha) ʕaDDa Nadia Nora dog‐her bit Nadia ‘Nora's dog bit Nadia.' b. Nuura Zaarat bayt‐a‐*(ha) Nadia Nora visited house‐ACC‐her Nadia ‘Nadia visited Nora's house.' 24 c. amriika ʕaashat Nuura fii‐*(ha) America live Nora in‐it ‘Nora lived in America.' The sentences in 45 show that DPs can be extracted from a variety of positions, leaving a resumptive behind. 45 show that a DP (Nuura) can be extracted from a Construct State phrase (DP island) in the subject (45a) or object (45b) position, leaving behind a resumptive pronoun (‐ha). In 45c, the DP amriika is extracted from a PP, leaving the resumptive (fii‐ha). This data shows that the word orders in 15d and 15f are a type of A‐bar movement that involves the left periphery, perhaps focus or topicalization. I will relate some analyses on left‐dislocated/topicalized elements and show their relevance to the case/Case assignment in ʔinna clauses in the review of literature. To sum up, section 2.1.1 explores word order variation in MSA. It sets the testing ground for ʔinna by showing available word orders in MSA under null‐copula. We have seen in this section that out of possible variations (logically possible, or possible under repair techniques such as pauses or the use of resumptive pronouns), available word orders in Arabic can be reduced to four possibilities: VSO, VOS, SVO, and OVS. All other word orders are either ungrammatical or are derived from one of these word orders (e.g. via A' movement for purpose of topicalization or focus). Later sections will test the effect of ʔinna on permissible word order in Arabic. Further discussion focuses on those four basic word orders. 25 2.1.2. Case Marking in Arabic In the previous section, I have described possible word order variation in Arabic. This section describes case facts across word orders. I show that case marking is largely insensitive to word order. Arabic marks DPs with one of four cases: nominative, accusative, dative, or genitive. The dative case is marked on DPs that are complements of PPs, as can be seen in 46. Genitive marks the second DP in Construct State DPs. (46) ustaath.u l.Hisaab.i akala mangat‐an fii l.maktab.i professor the.math ate mango in the office (NOM) (GEN) (ACC) (DAT) ‘The math professor ate mango in the office.' In 46, Hisaab is marked with the dative marker -i as it is the second word in the ‘Idafa' (Construct state) construction. Maktab is marked with the homophonous dative case marker as it is the complement of the preposition fii. DPs in the dative or genitive do not interact with ʔinna, and I will not discuss them further. Accusative is marked on DPs that function as adverbials, as can be seen in the accusative marker -an on the nominals indicating manner or time in 47. (47) Al‐ustaathu akala mangatan SabaaHan/saʕiidan/musriʕan the‐professor ate mango morning/happy/fast (NOM) (ACC) (ACC) ‘The math professor ate mango in the morning/happily/quickly.' Complements of overt verbal elements (including overt copula) are in the accusative and subjects are nominative. In 48, zayd is marked with the nominative marker -un and ʕamr is marked with the accusative marker -an regardless of the word order. 26 (48) a. qaabala zayd‐un ʕamr‐an met3MS zayd‐NOM Amr‐ACC ‘Zayd met Amr.' b. qaabala ʕamr‐an zayd‐un met3MS Amr‐ACC zayd‐NOM ‘Zayd met Amr.' c. zayd‐un qaabala ʕamr‐an zayd‐NOM met3ms Amr‐ACC ‘Zayd met Amr.' d. ʕamr‐an qaabala zayd‐un Amr‐ACC met3MS zayd‐NOM ‘Zayd met Amr.' The same facts are true for copular sentences where the copula is overt. In 49 - 54, hind carries nominative as the subject and ustaathat carries accusative as the complement of the copula. (49) kaanat hind‐un ustaathat‐an was Hind‐NOM professor‐ACC ‘Hind was a professor.' (50) hind‐un Kaanat ustaath‐at‐an Hind‐NOM was3FS professor‐F‐ACC ‘Hind was a professor.' (51) satakuun hind‐un ustaath‐at‐an will be3FS Hind‐NOM professor‐F‐ACC ‘Hind will be a professor.' (52) hind‐un sa‐takuun ustaath‐at‐an Hind‐NOM will‐be3FS professor‐F‐ACC ‘Hind will be a professor.' (53) laysat hind‐un ustaath‐at‐an is.not3FS Hind‐NOM professor‐F‐ACC ‘Hind is not a professor.' 27 (54) hind‐un laysat ustaath.at‐an Hind‐NOM is not3FS professor‐F‐ACC ‘Hind is not a professor.' Both objects of ditransitive verbs carry accusative marking, as can be seen in 55, where both objects albint and alsayyarat show accusative markers. (55) manaHa al‐walad.u al‐bint.a al‐sayyarat.a gave3MS the‐boy‐NOM the‐girl‐ACC the‐car‐ACC ‘The boy gave the girl the car.' When the indirect object is preceded by a preposition, it receives DAT, as can be seen in 56 and 57. (56) manaHat al‐bint‐u as‐sayyarat‐a li‐l‐walad‐i gave3FS the‐girl‐NOM the‐car‐ACC to‐the‐boy‐DAT ‘The girl gave the car to the boy.' (57) al‐banaat‐u manaHna as‐sayyarat‐a li‐l‐walad‐i the‐girls‐NOM gave3FP the‐car‐ACC to‐the‐boy‐DAT ‘The girls gave the car to the boy.' In section 2.1, I have stated that null‐copular sentences consist of a subject DP followed by a predicate DP, AP, or PP. In sentences of the type DP‐DP or DP‐AP, both phrases carry nominative marking, as can be seen in 58 (repeated from 26 with more detailed morphological description), where both words in the sentence carry nominative markers (‐u and -un respectively, depending on definiteness). In terms of agreement, the two phrases agree in number and gender. In 58, both the subject and the predicate are feminine singular (as indicated by the feminine singular marker -at). In 59, both the subject and the predicate are masculine plural, as can be seen in the masculine plural marker -uun. 28 (58) al‐ustaath‐at‐u jamiil‐at‐un / taʕbaan‐at‐un/miSriyy‐at‐un the‐professor‐FEM‐NOM beautiful / tired/Egyptian‐FEM‐NOM ‘The professor is beautiful/ tired/Egyptian.' (59) al‐muhadis.uun taʕbaan.uun the‐engineers‐NOM tired‐NOM ‘The engineers are tired.' In sentences with PP complements, the DP complement to P shows DAT, and the subject shows NOM. The examples in 60 show the NOM marker on the definite subject (‐u) (a), indefinite subject (‐un) in sentence‐initial position (b and c). (60) a. al‐ustaath‐u fii l‐maktab‐i the‐professor‐NOM in the‐office‐DAT ‘The professor is at the office.' b. ustaath‐un fii l‐maktab‐i professor‐NOM in the‐office‐DAT ‘A (certain) professor is at the office.' c. fii l‐maktab.i ustaath‐un in the‐office‐DAT professor‐NOM ‘There is a professor at the office.' All in all, I have shown in this section that word order does not affect case marking. Comparing case data in null‐C sentences (and conditionals, which are discussed later) with their ʔinna‐clause counterparts will detail the research problem. In the next section, we will see that Agreement, on the other hand, shows sensitivity to word order. 2.1.3. Agreement in Arabic In this section, I give an overview of agreement in Arabic. I start by describing number and gender facts in Arabic, then I describe gender and number agreement as they relate to word order. Unlike case, which is largely insensitive to word order, 29 subject‐verb agreement is sensitive to word order. Verbs agree with preverbal subjects in number and gender, but if the subject is postverbal, the verb shows only gender agreement. Arabic has two grammatical genders: feminine and masculine, and is realized on nouns and adjectives. The feminine marker is the suffix -at, as in: (61) muhandis‐un - muhandis‐at‐un engineer‐NOM - engineer‐f‐NOM. For animates, grammatical gender generally reflects natural gender, such as: (62) qiT‐un - qiT‐at‐un. cat‐NOM - cat‐F‐NOM Inanimate objects also carry grammatical gender, as can be seen in the feminine marker agreement on the following adjective: (63) madrasat‐un kabiir‐at‐un - marsam‐un kabiir‐un school‐NOM large‐F‐NOM - atelier‐NOM large‐NOM ‘A large school' - ‘A large atelier.' Gender designation of inanimates is arbitrary; there is nothing intrinsic about school or atelier to give them their respective gender in Arabic. Gender assignment for inanimate nouns is also unrelated to the feminine marker -at. In 64, although daar does not carry a feminine marker, it is grammatically feminine and induces feminine agreement on the adjective kabiir when modifying it. (64) daar‐un kabiir‐*(at)‐un house‐NOM large‐f‐NOM ‘A large house.' Finally, inanimate plurals are treated as singular feminine as far as agreement is concerned. 65 and 66 show that the plural form of both madrasat and marsam takes 30 adjectives that show singular feminine morphology, which is not the case with the animate muhandisat in 67, where the adjective following the plural is feminine plural marker ‐aat. (65) madrasat‐un kabiir‐at‐un - madaris‐un kabiir.at‐un school‐NOM large‐F‐NOM - schools‐NOM large.f‐NOM ‘A large school' - ‘Large schools' (66) marsam‐un kabiir‐un - maraasim‐un kabiir‐at‐un atelier‐NOM large‐NOM - ateliers‐NOM large‐F‐NOM ‘A large atelier' - ‘large ateliers' (67) muhandis‐at‐un jamiil‐at‐un - muhandis‐at‐aan jamiil‐at‐ aan - muhandis‐aat‐un jamiil‐aat‐un engineer‐F‐NOM pretty‐F‐NOM - engineer‐F‐DUAL.NOM pretty‐ F‐DUAL.NOM - engineer‐FP‐NOM pretty‐FP‐NOM ‘A pretty female engineer' - ‘Two pretty female engineers' - ‘Pretty female engineers' Arabic marks singular, dual, and plural. Regular masculine duals and plurals are marked by ‐aan/‐ayn/‐ayn and -uun/‐iin/‐iin for nominative/accusative/dative respectively, as can be seen in 68, irregular (broken) plurals are built via root‐and‐pattern morphology (69). (68) muhandis muhandis‐aan/ayn muhandis‐uun/iin engineer engineer‐DUAL‐NOM/ACC engineer‐PL‐NOM/ACC (69) rajul‐un - rijaal‐un, kitaab‐un - kutub‐un man‐NOM - men‐NOM book‐NOM - books‐NOM Feminine dual markers are the same as masculine counterparts. Feminine plurals are marked by -aat, as can be seen in 70. (70) Muhandis‐at‐un muhandis‐at‐aan/ayn muhandis‐aat‐un/in engineer‐F‐NOM engineer‐F‐DUAL.NOM/DUAL.ACC engineer‐FP‐NOM/ACC 31 I now turn to describing how SV agreement in Arabic interacts with word order. The verb always shows person agreement. In SV word order, the verb also shows number and gender agreement. For example, in 71, the verb manaH agrees with the plural feminine preverbal albanaat, by the agreement marker -na, showing both gender and number. (71) al‐banaat‐u manaHna as‐sayyarat‐a li‐l‐walad‐i the‐girls‐NOM gave3FP the‐car‐ACC to‐the‐boy‐DAT ‘The girls gave the car to the boy.' VS word order, on the other hand, shows only partial agreement morphology on the verb, where only gender, but not number, appears on the verb. In 72, the verb manaH shows the feminine singular marker -at regardless of the subject being singular (albint) or plural (albanaat). (72) manaHat al‐bint‐u/al‐banaat‐u as‐sayyarat‐a li‐l‐walad‐i gave the‐girl‐NOM/the‐girls‐NOM the‐car‐ACC to‐the‐boy‐DAT (3FS) (fs) /(fp) (fs) (fs) ‘The girl/girls gave the car to the boy.' The same facts are true for masculine subjects: in 73, the verb shows third person masculine singular agreement with both singular (alwalad) and plural (alʔawlaad) subjects. (73) manaHa al‐walad‐u / al‐ʔawlaad‐u al‐bint‐a as‐sayyarat‐a gave the‐boy‐NOM/ the‐boys‐NOM the‐girl‐ACC the‐car‐ACC (3MS) (ms) / (mp) (fs) (fs) ‘The boy/boys gave the girl the car.' In 74, the verb in SV position shows singular masculine agreement -a with the singular subject in 74a and plural masculine agreement -uu with the plural subject in 74b. 32 (74) a. al‐walad‐u manaHa al‐bint‐a as‐sayyarat‐a the‐boy‐NOM gave the girl‐ACC the‐car‐ACC (MS) (3MS) (FS) (FS) ‘The boy gave the girl the car.' b. alʔawlaad‐u manaH‐uu al‐bint‐a as‐sayyarat‐a the boys‐NOM gave‐ACC the‐girl‐ACC the‐car‐ACC (MP) (3MP) (FS) (FS) ‘The boys gave the girl the car.' As for null‐copular sentences, the two phrases (subject and predicate) agree in number and gender. In 75, both the subject and the predicate are feminine singular (as indicated by the feminine singular marker -at). In 76, both the subject and the predicate are masculine plural, as can be seen in the masculine plural marker -uun. (75) al‐ustaathatu jamiilatun / taʕbaanatun/miSriyyatun the‐professor beautiful / tired/Egyptian (DEF‐prof‐FS‐NOM) (FS‐adj‐INDEF‐NOM)) ‘The professor is beautiful/ tired/Egyptian.' (76) al‐muhadisuun taʕbaanuun the‐engineers tired (DEF‐MP‐NOM) (INDEF‐MP‐NOM) ‘The engineers are tired.' 2.1.4. Case Marking on Fronted DPs In section 2.1.1.1, I noted Mohammad's (2000) restrictions on word order that arguments can be preposed leaving a resumptive pronoun, and that would obtain some word orders that are not otherwise permissible, especially in relation to the object occurring before the subject. This section describes case facts related to this type of fronting. Data related to this type of fronting will bear on our discussion in later sections on Case assignment, and on the structure and Case behavior in sentence‐initial DPs in 33 Arabic. According to Mohammad (2000:151-152), fronted (scrambled) objects maintain their case marking, as can be seen in 77 - 79, where ʔaxaahu is marked accusative in its canonical word order (77), scrambled across the subject (78), or fronted (79). The sentences in 77 - 79 do not involve a resumptive pronoun coreferring with the fronted element. (77) raʔaa l‐walad‐u ʔax‐aa‐hu saw3MS the‐boy‐NOM brother‐ACC‐his ‘The boy saw his brother.' (Mohammad 2000:151) (78) raʔaa ʔax.aa‐hu l‐walad‐u saw3MS brother‐ACC‐his the‐boy‐NOM ‘The boy saw his brother.' (Mohammad 2000:151) (79) ʔax‐aa‐hu raʔaa l‐walad.u brother‐ACC‐his saw3MS the‐boy‐NOM ‘The boy saw his brother.' (Mohammad 2000:152) It is possible for the fronted object to be marked nominative, according to Ryding (2005) and Badawi et al. (2003). They both refer to the construction as a ‘topic‐comment construction.' The examples in 80 and 81 are a case in point.8 (80) haathihi l‐furSat‐u na‐jidu‐ha fii l‐qaahirat‐i this def‐opportunity‐NOM we‐find‐it in Cairo‐DAT ‘We find this opportunity in Cairo.' (adapted from Ryding 2005:67, her (56)) 8 The similarity of the internal DP structure of the fronted object is incidental. The data in both references are corpus‐based, and it seems that this construction is very common. 34 (81) haathihi l‐mihnat‐u tuqaabilu‐ha ʕiddat‐u Suʕubaat‐in this def‐profession‐NOM meet‐it many‐NOM difficulties‐DAT ‘Many difficulties face this profession.' (Badawi et al. 2003:327) In the two sentences in 80 and 81, the fronted DP haathihi lfurSatu/hathihi lmhnatu ‘this opportunity/this profession' are marked with nominative (‐u), leaving a resumptive pronoun in the accusative (the clitic -ha) in the object position. We know that the sentence‐initial DP is not the subject of the verb because the verb shows first person subject agreement in 80, and not the expected third person if this opportunity were the subject of the sentence. The nominative case is not exclusively associated with fronted elements extracted from object position. Fronted DPs out of PPs, for example, also have nominative case marking, as in 82. (82) al‐ʕarab‐ui kaanat la‐humi ʕilaaqat‐un bi‐Hayaat‐i ʔisbaanyaa the‐arabs‐NOM was to‐them relation‐NOM with‐life‐DAT Spain ‘Arabs had a relation with the life of Spain.' The fronted DP alʕarabu, which carries the nominative marker -u, is extracted from its original position as complement to the preposition la, marked by the dative resumptive pronoun -hum. The data above suggests that the resumptive pronoun and case maintenance of the fronted DP are in complementary distribution. In Mohammad's data, the accusative element is not associated with a resumptive (i.e. a DP maintains its Case/case if it is not associated with a resumptive), and in Badawi and Ryding data, the nominative element is associated with a resumptive that shows a different case (i.e. if the resumptive keeps the case/Case in the extraction site, the fronted DP shows nominative). In fact, Bakir (1980) argues extensively that, indeed, this is the case, and uses this fact to argue that 35 there are two types of fronting: topicalization (for the nominative‐resumptive sentences) and left‐dislocation (for the case maintenance‐gap sentences). When stating the research problem in the next chapter, I will show that DPs that had already been licensed and valued for nominative as subjects unexpectedly show ACC morphology. When describing data that shows the details of this problem, (object) left‐dislocated DPs that maintain their accusative case will not shed much light on the problem, as they are sentence‐initial and are expected to show ACC anyway. This type of data, hence, will be ignored when describing the research problem. So far, we have seen a description of possible word order and case properties of Arabic sentences headed by null complementizers. Arabic allows SV(O) and VS(O) word orders. OVS is only allowed as part of A' fronting (focusing or left‐dislocation). Sentences without an overt verb have a null copula. Arabic is a nominative‐accusative language. Subjects are nominative, and objects are accusative, regardless of word order. Verbs agree with their subjects in number, gender, and person in SVO word order. In VS word order, only partial agreement (person and gender) obtains on the verb. 2.2. Arabic Conditional Phrases This section describes the first type of overt complementizers: conditionals. I will show here that CPs introduced by conditionals have the same word order, case, and agreement properties as those introduced by null complementizers. The next section, in contrast, will show that ʔinna‐type complementizers impose restrictions on word order and case which are not shared by conditionals or null complementizers. Arabic has four conditionals: 36 (83) ʔitha ʔin law lawla if (type I) if (type II) if (type III) if not (unless) I gloss the first three conditionals as if introducing different types of conditionals. The three types of conditionals differ in terms of the likelihood of the event in the conditional clause as the following examples show. ʔitha is used when the event indicated by the verb in the conditional clause is certain to happen: (84) ʔitha waSal‐ta l‐bayt‐a kallem‐ni if arrived‐you the‐home‐ACC talk.to‐me ‘If you arrive home call me.' ; ‘call me when you get home.' Phrases introduced by ʔin indicate that the event of the conditional clause is unlikely: (85) ʔin waSal‐ta mubakkir‐an kallem‐ni if arrived‐2MS early‐ACC talk.to‐me ‘If you arrived early call me.' law introduces an impossible, hypothetical, or counterfactual phrase: (86) law kun‐tu r‐raʔiis‐a la‐rafaʕ‐tu l‐ʔujuur‐a if was‐1S the‐president‐ACC then‐raised‐1S the‐salaries‐ACC ‘If I were the president, I would have raised salaries.' lawla, meaning ‘if not for' also indicates impossibility in a reverse manner: the second predication is impossible because the first is true, as can be seen in 87. Morphologically, it is made up of law and the negation particle laa. (87) lawla (faaza) Obama la‐faaza l‐jumhuuri‐uun if.not (won) Obama then‐won the‐rupublicans‐NOM ‘If it were not for Obama, the Republicans would have won.'; ‘If Obama had not won, the Republicans would have won.' I do not discuss in detail the use or differences between the three conditional particles. In terms of word order, conditionals seem to allow all four SV(O) word orders allowed under null complementizers, as can be seen in the sentences in 88 - 90. The 37 sentences in a show that VSO word order is possible for each of the conditionals. The sentences in b show examples of VOS, c SVO, and d show that OVS word order is possible for each of the conditional complementizers. (88) a. law qaabala zayd‐un ʕamr‐an la‐Htadana‐hu if met3MS zayd‐NOM Amr‐ACC then‐hug3MS‐him ‘If Zayd met Amr, he would hug him.' b. law qaabala ʕamr‐an zayd‐un la‐Htadana‐hu if met3MS Amr‐ACC zayd‐NOM then‐hug3MS‐him ‘If Zayd met Amr, he would hug him.' c. law zayd‐un qaabala ʕamr‐an la‐Htadana‐hu if zayd‐NOM met3MS Amr‐ACC then‐hug3MS‐him ‘If Zayd met Amr, he would hug him.' d. law ʕamr‐an qaabala zayd‐un la‐Htadana‐hu if Amr‐ACC met3MS zayd‐NOM then‐hug3MS‐him ‘If Zayd met Amr, he would hug him.' (89) a. ʔitha qaabala zayd‐un ʕamr‐an fa‐saya‐Htadana‐hu if met3MS zayd‐NOM Amr‐ACC then‐will‐hug3MS‐him ‘If Zayd meets Amr, he will hug him.' b. ʔitha qaabala ʕamr‐an zayd‐un fa‐saya‐Htadana‐hu if met3MS Amr‐ACC zayd‐NOM then‐will‐hug3MS‐him ‘If Zayd meets Amr, he will hug him.' c. ʔitha zayd‐un qaabala ʕamr‐an fa‐saya‐Htadana‐hu if zayd‐NOM met3MS Amr‐ACC then‐will‐hug3MS‐him ‘If Zayd meets Amr, he will hug him.' d. ʔitha ʕamr‐an qaabala zayd‐un fa‐saya‐Htadana‐hu if Amr‐ACC met3MS zayd‐NOM then‐will‐hug3MS‐him ‘If Zayd meets Amr, he will hug him.' (90) a. ʔin qaabala zayd‐un ʕamr‐an la‐Htadana‐hu if met3MS zayd‐NOM Amr‐ACC then‐hug3MS‐him ‘If Zayd met Amr, he would hug him.' 38 b. ʔin qaabala ʕamr‐an zayd‐un la‐Htadana‐hu if met3MS Amr‐ACC zayd‐NOM then‐hug3MS‐him ‘If Zayd met Amr, he would hug him.' c. ʔin zayd‐un qaabala ʕamr‐an la‐Htadana‐hu if zayd‐NOM met3MS Amr‐ACC then‐hug3MS‐him ‘If Zayd met Amr, he would hug him.' d. ʔin ʕamr‐an qaabala zayd‐un la‐Htadana‐hu if Amr‐ACC met3MS zayd‐NOM then‐hug3MS‐him ‘If Zayd met Amr, he would hug him.' In terms of tense,9 conditionals show a very strong tendency to select past complements, as is seen in the sentences 88 - 90. It is possible, albeit uncommon, that the conditional phrase be in the present tense. The main clause may or may not match the tense of the conditional clause. In 91, the conditional clause is in the past tense, and the main clause can be in the present or future. Users of MSA find conditional phrases in the present tense acceptable when they see them, but usually do not produce them because, I speculate, this structure is common in Classical Arabic (CA), Quranic and religious texts. Although all the above examples involve pronouns in the second part of the conditional, this choice is not mandated by grammar ‐ merely for convenience. The then‐clause does not need to have a pronoun (resumptive or otherwise). In 91, for example, the second clause (Republicans will rejoice) has no pronominal reference to the if statement. This fact is true for all conditional sentences. (91) law faaza romny la‐fariHa/sa‐yafraHu l‐jumhuuriyuun if won Romny then‐rejoice/will‐rejoice the‐republicans ‘If Romny wins, Republicans will rejoice.' 9 Although I refer to this restriction in terms of tense, the selectional restriction may well be that of a perfective aspect. 39 Whether conditionals allow null copulas, and what possible word orders available is inconsistent across conditionals. Recall that null‐copular sentences allow DP‐DP/ AP/PP word order, as well as PP‐DP provided that the DP is indefinite. Overall, conditionals introducing null‐copular sentences in MSA are found ungrammatical by myself and my language informants.10 However, some sentences are worse than others. The sentences in 92 - 95 show the best (least objectionable) sentences I could elicit. Sentences in a are with law, in b with itha, and c with ʔin. The number of question marks (indicating ‘marginality') is based on how varied the responses I received from informants, and how strongly they (dis)agree with a certain sentence. (92) DP‐PP a. ? law zayd‐un fii d‐ddaar‐i la‐qaabala‐ni if Zayd‐NOM in the‐house‐DAT then‐met‐me ‘If Zayd were at home, he would have met me.' b. ? sa‐ʔu‐qaabilu‐hu itha huwwa fi d.daar.i will‐1s‐meet‐him if he in the‐house‐DAT ‘I will meet him if he is at the house.' c. ?? uqaabilu.hu ʔin huwwa fi daari‐hi meet1S‐him if he in house‐his ‘I would meet him if he were at his hous.' (93) PP‐DP a. law fi d‐daar‐i rajul‐un la‐qaabala‐ni if in the‐house‐DAT man‐NOM then‐meet3MS‐me ‘If there were a man in the house, he would have met me. 10 I have also spent about 30 hours looking at concordances of conditionals in newspaper and modern literature corpora in ARABICorpus (www.arabicorpus.byu.edu), and could not find any instances of a conditional followed by a null‐copular complement. The concordance tool in the corpus does not allow for a search of the type ‘conditional followed by words that are nonverbal all the way to the following period.' So, I had to resort to other less efficient techniques. Just because I could not find any results does not mean there are not any, and even if the construction is indeed absent in the corpus, this does not necessarily mean it is ungrammatical. However, this result is enough to raise deep suspicions about null‐copular conditionals in MSA. 40 b. ? ʔitha fi d‐daar‐i rajul‐un fa‐sa‐yuqaabala‐ni if in the‐house‐DAT man‐NOM then‐fut‐meet3MS‐me ‘If there is a man in the house, he will meet me.' c. ??? ʔin fi d‐daar‐i rajul‐un la‐qaabala‐ni if in the‐house‐DAT man‐NOM then‐met‐me ‘If there were a man in the house, he would meet me.' (94) DP‐AP a. law r‐rajul‐u murattab‐un uHibbu‐hu akthar if the‐man‐NOM tidy‐NOM like1S‐him more ‘If a man is tidy I would like him (even) better.' b. ʔitha r‐rajul‐u murattab‐un uHibbu‐hu akthar if the‐man‐NOM tidy‐NOM like1S‐him more ‘If a man is tidy I would like him (even) better.' c. ?? ʔin r‐rajul‐u murattab‐un uHibbu‐hu akthar if the‐man‐NOM tidy‐NOM like1S‐him more ‘If a man is tidy, I would like him better.' (95) DP‐DP a. uqaabilu‐hu law huwwa axuu‐ka meet1S‐him if he brother.NOM‐your ‘I will meet him if he is your brother.' b. ʔaaxuthu‐ha itha hiyya hadiyyat‐un take1S‐it if it.fem gift‐NOM ‘I will take it if it is a gift.' c. ? ʔaaxuthu‐ha ?in hiyya hadiyyat‐un take1S‐it if it.fem gift‐NOM ‘I will take it if it is a gift.' The degraded status of null‐copular conditionals might follow from the overall tendency of MSA to prefer past tense in the conditional clause. The acceptability of the sentences in 92 - 95 might be due to interactions between MSA and dialects. For example, sentences with law are found less degraded than ones with itha. In Egyptian 41 Arabic, law may introduce a null‐copular sentence, while itha may do so only for a subsection of the population. The sentences in 95 are overall better than other sentences, but they all have pronominal subjects. Pronouns in Egyptian Arabic are indicated to have some copular properties, as noted by Eid (1983). In terms of case and agreement, conditional sentences do not display any different behavior than that of null‐C sentences, as can be seen in 88 - 95. Preverbal and postverbal subjects are nominative, and so are subjects and complements of null‐copular sentences. Objects and complements of overt copulas are accusative. Verbs agree with preverbal subjects in number and gender, and agree with postverbal subjects in gender only. 2.3. The ʔinna Clause Now that we have seen word order and case/Case properties of Arabic sentences, I move to the data that is at the core of this dissertation. This section will describe word order and case behavior in ʔinna clauses. As we will see in the next section (2.4), the case and word order properties unique to clauses introduced by this class of complementizers need to be explained. ʔinna and its sisters pose a restriction on word order in the clauses they head, and they induce accusative marking on the otherwise nominative subjects. In 2.3.1, I first give an inventory of ʔinna and its sisters and their distribution, then I move to describing permissible word orders in ʔinna clauses. This section will intentionally use many examples where DPs are not overtly marked for case, to keep the discussion focused on word order. The next section will describe case marking and agreement in 42 sentences introduced by ʔinna and its sisters. 2.3.1. ʔinna and Its Sisters In this section, I list the members of the ʔinna family and their meaning. ʔinna has six ‘sisters', listed in 96. Morphological analysis will show that three of the members are morphologically complex, consisting of one of the other sisters combined with a conjunction. (96) ʔanna liʔanna kaʔanna lakinna layta laʕalla ʔinna that because as if but if only hopefully indeed/that liʔanna and kaʔanna are morphologically related to ʔanna. They are composed of ʔanna preceded by ka‐ (‘like,' indicating similarity) and li‐ (‘because,' indicating reason). The two complementizers, respectively, mean ‘for/because that…' and ‘it is the same as that….' To illustrate, consider 97 and 98, where ka‐ is used to connect DPs and indicate similarity. kaʔanna is used in 99 and 100 to show similarity in a larger sense; states‐of‐affair introduced by complete sentences. (97) uHibu.hu ka‐ʔax‐i love1S.him like‐brother‐my ‘I love him like my brother. (98) zayd‐un qawwi‐un ka‐l‐ʔasad‐i Zayd‐NOM strong‐NOM like‐the‐lion‐DAT ‘Zayd is as strong as a lion.' (99) uHibu.hu kaʔanna‐hu ax‐i love1S‐him as‐if‐him brotherther‐my ‘I love him as if he were my brother.' (100) kaʔanna zayd‐an mariiD‐un as.if Zayd‐ACC sick‐NOM ‘It is as if Zayd is sick.' / ‘Zayd seems sick.' 43 Similarly, li‐ ‘to' is used to indicate cause or reason, as can be seen in 101 and 102. (101) saafara li‐yadrus‐a traveled‐3MS to‐study3MS‐SUBJ ‘He travelled to study.' (102) ʔishtara kitaab‐an li‐yaqraʔa‐hu bought3MS book‐ACC to‐read3MS‐it ‘He bought a book to read it.' Combining li‐ with ʔanna ‘that', liʔanna introduces a clause that is a cause/explanation of the main clause, as can be seen in 103 and 104. (103) saafara liʔanna‐hu mariiD‐un traveled3MS because‐he.ACC sick‐NOM ‘He travelled because he is sick.' (104) ʔishtara l‐kitaab‐a liʔanna‐hu yuHilbbu Shakespeare bought3MS the‐book‐ACC because‐he.ACC love3MS Shakespeare ‘He bought the book because he loves Shakespeare.' Similarly, lakinna is composed of the conjunction laakin (but) and ʔinna (indeed/that). The sentence in 105 shows that laakin can be used to conjoin two DPs (Zayd and Amr). (105) lam yusaafira zayd‐un laakin ʕamr‐un did‐not travel3MS Zayd‐NOM but Amr‐NOM ‘Zayd did not travel, but Amr.' As a complementizer, laakinna may only introduce a full clause, as can be seen in the ungrammaticality of 106, which is identical to 105 except in the use of laakinna rather than laakin. (106) *lam yusaafira zayd‐un laakinna ʕamr‐un/ ʕamr‐an did‐not travel3MS Zayd‐NOM but Amr‐NOM/Amr‐ACC ‘Zayd did not travel, but Amr.' laakin can be used, nonetheless, to introduce a complete clause, as in 107, where 44 laakinna indicates a contrast between the two clauses before and after it. (107) saafara zayd‐un laakinna ʕamr‐an kaana mariid‐an travelled3MS Zayd‐NOM but Amr‐ACC was3MS sick‐ACC ‘Zayd travelled, but Amr was sick.' Knowing the morphological composition of kaʔanna, laakinna, and liʔanna will help us reduce the number of complementizers that need to be investigated in detail in subsequent sections. We have seen that morphologically they are related to ʔinna and ʔanna. The rest of this section will show that all sisters of ʔinna share distribution and case/Case properties. After we have established that these three complementizers conform morphologically and syntactically to the same behavior of ʔinna and ʔanna, detailed discussions will ignore them for convenience. The ʔinna family, then, will be considered to consist of four members only (ʔinna, ʔanna, layta, laʕalla), with the implication that the other three members are only morphological derivatives of ʔinna and ʔanna. layta is used to express regret, as can be seen in 108, where layta expresses that the speaker regrets that the proposition of the complement to layta is not true; a wish that it were. (108) layta zayd‐an qaabala ʕamr‐an if only zayd‐ACC met3MS Amr‐ACC ‘If only Zayd had met Amr.' or ‘I wish Zayd had met Amr.' laʕalla is used to express the speaker's hope that the proposition in the complement is/will be true: (109) laʕalla l‐jaww‐a (sa)yataHassanu γdadan hopefully the‐weather‐ACC (fut.)improve3MS tomorrow ‘Hopefully the weather will improve tomorrow.' 45 All seven complementizers seem to express a belief or attitude of the speaker about the proposition of the sentence - that of assertion (ʔinna and ʔanna)11, hope (laʕalla), or regret (layta). This may explain why layta and laʕalla are part of this set, although they may not be considered complementizers in other languages. More importantly, they occur in complementary distribution with the other sisters, and they display the same case effects and word order requirements. 2.3.2. ʔinna and Its Sisters: Distribution ʔinna and its sisters can introduce independent clauses, as well as subordinate clauses, with the restrictions I outline in this section. Five of the seven complementizers may introduce an independent clause: ʔinna, kaʔanna, lakinna, layta, laʕalla, as can be seen in 110 - 117. 11 At matrix CPs, it is harder to see how ʔinna and ʔanna indicate assertion. After all, a sentence with a null complementizer can also be considered an assertion. The literature on ʔinna and ʔanna usually gloss them as ‘indeed' to show emphasis. It is also easier to see the difference in embedded clauses. Compare: i. na.ʕrifu ʔinna/*ʔin nadia fii l‐mabna we.know that/*if Nadia in the‐builiding ‘We know that Nadia is in the building.' ii. laa na‐ʔrifu *ʔinna/ʔin kaanat nadia fii l‐mabna not we‐know *that/if were Nadia in the‐builiding ‘We do not know if Nadia is in the building.' When the subordinate clause is factual, as indicated by the matrix verb ‘know,' ʔinna is used. When it is doubtful, as indicated by ‘we do not know,' ʔin is used, in a restriction comparable to the English that/whether (if). Comparing ii to iii where the verb know is negated, but is in the past: iii. lam na‐ʕrifu ʔinna/ʔin kaanat nadia fii l‐mabna not.past we‐know that/if were Nadia in the‐builiding ‘We did not know if Nadia was in the building.' The sentence is acceptable with both subordinators. ʔinna indicates that Nadia indeed was in the building, but we did not know that fact, while ʔin indicates that the truth of the subordinate proposition is unknown. 46 (110) zayd‐un qaabala ʕamr‐an zayd‐NOM met3MS Amr‐ACC ‘Zayd met Amr.' (111) ʔinna zayd‐an qaabala ʕamr‐an indeed zayd‐ACC met3MS Amr‐ACC ‘Indeed Zayd met Amr.' (112) kaʔanna zayd‐an qaabala ʕamr‐an it.is.as.if zayd‐ACC met3MS Amr‐ACC ‘It seems as if Zayd met Amr.' (113) layta zayd‐an qaabala ʕamr‐an if.only zayd‐ACC met3MS Amr‐ACC ‘If only Zayd had met Amr.' (114) laʕalla zayd‐an qaabala ʕamr‐an hopefully zayd‐ACC met3MS Amr‐ACC ‘Zayd hopefully met Amr.' (115) lakinna zayd‐an qaabala ʕamr‐an but zayd‐ACC met3MS Amr‐ACC ‘But Zayd met Amr.' (116) *ʔanna zayd‐an qaabala ʕamr‐an that zayd‐ACC met3MS Amr‐ACC ‘Zayd met Amr.' (117) *liʔanna zayd‐an qaabala ʕamr‐an because zayd‐ACC met3MS Amr‐ACC ‘Because Zayd met Amr.' lakinna may need some contextual support. It is used to indicate (the speaker's belief) that whatever comes after it contradicts already stated or implied assumptions in the conversation. In the presence of such context, lakinna can introduce an independent clause, as is evident by the grammaticality of 115. ʔanna and liʔanna may not introduce 47 an independent clause, as can be seen by the ungrammaticality of 116 and 117. Two of the complementizers do not introduce a subordinate clause: layta and laʕalla. They can only introduce matrix clauses. Verbs that indicate hope or regret in Arabic do not select a CP complement introduced by these complementizers, as can be seen in the ungrammaticality of 118 and 119. (118) *nadimtu/ tamanayytu layta zayd‐an qaabala ʕamr‐an regretted1S/hoped1S if.only zayd‐ACC met3MS Amr‐ACC ‘I wished/regretted if only Zayd had met Amr.' (119) *tamanayytu laʕalla zayd‐an qaabala ʕamr‐an hoped1S hopefully zayd‐ACC met3MS Amr‐ACC ‘I hoped Zayd hopefully met Amr.' The verbs expressing hope and regret above take infinitival TP complements instead, as can be seen in 120. The verb tamanayytu (I hope) has an infinite TP complement introduced by the infinitival ʔan (to). (120) tamanayytu ʔan yuqaabala zayd‐un ʕamr‐an hoped1S to meet zayd‐NOM Amr‐ACC ‘I hoped for Zayd to meet Amr.' The ungrammaticality of 118 and 119 may not be surprising, given that the meaning of hope or regret is already encoded in the complementizer, making the verbs tamanna (hope) or nadim (regret) selecting one of the two complementizers superfluous. liʔanna introduces only subordinate clauses. The verb in the matrix clause does not have to be a ‘subordinating' verb, a verb that takes a CP complement. Any verb can be used in the matrix verb position in 121. Recall from the last section that laakinna is morphologically the conjunction laakin (but) and ʔinna (indeed). The conjunction but does not require any specific kind of verb in a higher position, which can explain why 48 laakinna can occur so freely in subordinate clauses, and why it can only introduce subordinate clauses. The conjunction part of liʔanna needs a higher clause to conjoin to. (121) Saafara/naama/taʔaxxara liʔanna‐hu mariiD‐un traveled3MS/slept3MS/was.late3MS because‐he sick‐NOM ‘He travelled/slept/ran late because he is sick.' kaʔanna can introduce a subordinate clause, interchangeably with ʔanna, as can be seen in the grammaticality of both 122 and 123. kaʔanna is used to indicate that the speaker thinks the subordinate clause is not true (I am not sick, I feel as if I were), while ʔanna is used to indicate that the speaker thinks the subordinate clause is true (I feel I am sick). (122) ʔashʕuru ʔann‐i mariiD‐un feel1S that‐I sick‐NOM ‘I feel that I am sick.' (123) ʔashʕuru kaʔann‐i mariiD‐un feel1S as.if‐I sick‐NOM ‘I feel as if I am sick.' ʔinna and ʔanna may introduce subordinate clauses. They occur in complementary distribution, depending on the selecting verb. qaal ‘say' selects ʔinna (124), while all other verbs of saying (such as iddaʕa - claim) select ʔanna (125). (124) qaala ʔinna/*ʔanna zayd‐an qaabala ʕamr‐an say3MS that zayd‐ACC met3MS Amr‐ACC ‘He said that Zayd met Amr.' (125) iddaʕa *ʔinna/ʔanna zayd‐an qaabala ʕamr‐an claim3MS that zayd‐ACC met3MS Amr‐ACC ‘He claimed that Zayd met Amr.' 49 2.3.3. ʔinna and Its Sisters: Word Order This section is concerned with what word orders are possible in the ʔinna‐clause; matrix and embedded CPs headed by ʔinna or any of its sisters. Regardless of the type of clause (matrix or embedded) and the actual complementizer, word order restrictions are the same for the ʔinna‐clause. For ease of exposition, I will mainly give data with ʔinna heading a main clause. Unless otherwise stated, the reader is free to assume that examples are also applicable to any of ʔinna's sisters, in their respective clauses. I will first show restrictions on permissible word orders in sentences that contain overt verbal elements; that a verb may not directly follow ʔinna. I then move to describing ʔinna interaction with word order in null‐copular sentences; that ʔinna does not impose any additional word order restrictions. Finally, I describe word order in sentences with topicalized DPs, showing that ʔinna may precede DPs extracted from any phrase. In terms of word order, ʔinna may not be followed by a verb. The VS sentence in 127 is ungrammatical, while its SV counterpart in 126 is grammatical. (126) ʔinna Nadia waSalat indeed Nadia arrived3FS ‘Indeed the girls arrived.' (127) *ʔinna waSal.at Nadia indeed arrived Nadia ‘Indeed Nadia arrived.' This restriction means that out of the four word orders that we consider acceptable in 15, only two word orders are acceptable complements to ʔinna: SV(O), as in 126 above and OVS, as in 128. Note that in 128 neither DP shows case marking, but single 50 masculine agreement on the verb and the feminine resumptive object pronoun ‐ha indicate that Hamza (a man's name) is the subject, and Nadia the object. (128) ʔinna Nadia qaabala‐ha Hamza indeed Nadia met3MS‐her Hamza ‘Indeed, Nadia, Hamza met her.' The other two acceptable word orders in 15, VSO and VOS, are only acceptable if the verb is separated from ʔinna by an adjunct or an expletive pronoun. In 129, ʔinna is followed by an adverbial, lyawm (today), and the verb‐first complement is acceptable. Without lyawm the sentence is ungrammatical. (129) qaala ʔinna *(lyawwm‐a) tattaDiHu Haqiiqat‐u l‐ʕayyinaat‐i said3MS that today‐ACC become‐clear3FS truth‐NOM the‐samples‐GEN ‘He said that today the truth about the samples comes out' The adjunct separating ʔinna and the verb can also be a PP. In 130, the complement to ʔinna is has VS word order, but the verb yaʔkulu (eat) is separated from ʔinna by the PP adverb of place fii l‐yabaan (in Japan), and the sentence is grammatical. The omission of the adjunct renders the sentence ill‐formed. (130) samiʕtu ʔanna *(fii l‐yabaan‐i) yaʔkulu r‐rijaal‐u ʔasamak‐an saamat‐an heard1S that in Japan‐DAT eat the‐men‐NOM fish‐ACC poisonous‐ACC ‘I heard that in Japan men eat poisonous fish.' The other case where the ʔinna clause may have a VS word order is when the complement is introduced by an expletive pronoun. Indeed, if ʔinna is followed by an expletive or an adverbial, the restrictions on word order are lifted, and all well‐formed word orders without ʔinna, which we have discussed in section 2.1, are available and 51 grammatical. The sentences in 131 show this fact. The sentences with the expletive (the clitic -hu; third person masculine singular) can be followed by SV (as in a), VS (b), all the already well‐formed null‐copular sentences c - e. (131) a. ʔdaaf‐at ʔinna‐hu Nadia saafar‐at added‐3FS that‐expl Nadia traveled‐3FS ‘She added that Nadia travelled.' b. ʔdaaf‐at ʔinna‐hu saafar‐at Nadia added‐3FS that‐expl travelled‐3FS Nadia ‘She added that Nadia travelled.' c. ʔdaaf.at ʔinna‐hu Nadia γaneyyat‐un added.3FS that‐expl Nadia rich‐NOM ‘She added that the Nadia is rich.' d. ʔdaaf‐at ʔinna‐hu Nadia fii l‐hind‐i added‐3FS that‐expl Nadia in the‐India‐DAT ‘She added that Nadia is in India.' e. ʔdaaf‐at ʔinna‐hu fii bayt‐i‐na rajul‐un added.3FS that‐expl in house‐DAT‐our man‐NOM ‘She added that a man is in our house.' Agreement patterns point to the direction that the pronoun indeed is a non‐referential; an expletive pronoun. The subject of the sentences is Nadia, feminine, while the pronoun is masculine (‐hu as opposed to feminine -ha). The use of the expletive in MSA is restricted to post‐ʔinna position, but can occur sentence‐initially in other varieties of Arabic such as the closely related CA, and Egyptian Arabic. I return to these facts in more detail in Chapter 4. ʔinna does not pose any additional restrictions on the word order of null‐copular sentences. 132 and 133 show that the acceptable word orders for null‐copular 52 sentences, discussed in section 2.1.1 examples 25 - 27, are also acceptable when introduced by ʔinna, and unacceptable word order without ʔinna remains unacceptable with the introduction of ʔinna. (132) ʔinna hind‐an ustaathat‐un indeed Hind‐ACC professor‐NOM ‘Indeed Hind is a professor.' (133) a. ʔinna fii l‐maktab‐i asaatithat‐an indeed in the‐office‐DAT professors‐ACC b. * ʔinna asaatithat‐an fii l‐maktab‐i indeed professors‐ACC in the‐office‐DAT ‘Indeed professors are at the office.' ʔinna may head a clause with fronted DPs of the type discussed in section 2.1.4 without any additional word order restrictions. This fact is true for fronted DPs extracted from the object position (134), or from other positions (135). (134) ʔinna Nadia qaabala‐ha Hamza indeed Nadia met‐her Hamza ‘Indeed, Nadia, Hamza met her.' (135) ʔinna Nadia ishtara bayt‐a‐ha Hamza indeed Nadia bought3MS home‐ACC‐POSS3FS Hamza ‘Nadia Hamza bought her house.' In 135, Nadia is extracted from a Construct State phrase. Note that the fronted element may not be at a higher position than ʔinna, as can be seen in the ungrammaticality of 136. (136) *Nadia ʔinna ishtara bayt‐a‐ha Hamza Nadia indeed bought3MS home‐ACC‐POSS3FS Hamza ‘Nadia Hamza bought her house.' To sum up, ʔinna may not be followed by a verb. It allows subject‐first sentences 53 with an overt verb, and sentences with null copula. It only allows verb‐first word orders only if ʔinna is separated from the verb by an adjunct AP or PP, or by an expletive pronoun. The next section shows case marking in the ʔinna clause under the various permissible word orders. 2.3.4. ʔinna and Its Sisters: Case Marking In this section, I describe case marking in clauses introduced by ʔinna, and the differences between them and that of clauses introduced by a null complementizer or conditionals. This description will show that for each of the permitted word orders described in the previous section, the following case marking obtains: 1. In ʔinna‐SV(O) word orders, S shows accusative marking. 2. In null‐copular sentences (ʔinna‐DPdef‐DPindef, ʔinna‐DPdef‐PP, ʔinna‐PP‐DPindef), the subject DP shows accusative marking, regardless of word order. 3. In fronted‐DP word orders, the fronted DP shows accusative marking, and the resumptive pronoun shows whatever case marking determined by the extraction site: nominative for the subject. 4. In ʔinna‐expl‐VS(O), the expletive shows accusative marking and the subject shows nominative marking. 5. In ʔinna‐adverbial‐VS(O), the adverbial shows accusative and subject nominative case marking. I now move to showing each case marking in each of the possible word orders of ʔinna phrases described in the last section. When ʔinna is followed by a DP, that DP are in the accusative case regardless of which word order follows the DP. In 137, the DP 54 albanaat (girls) shows accusative case marking, while in the absence of ʔinna, it is nominative. (137) ʔinna albanaat‐i waSalna indeed the girls‐ACC arrived ‘Indeed the girls arrived.' This is also true for null‐copular sentences. In 138, which is the ʔinna counterpart of 25, hind is in the accusative after ʔinna while it is nominative in the absence of ʔinna. Note that the effect of ʔinna is limited to the subject DP in null‐copular sentences. The complement ustaathat (professor) maintains its nominative case regardless of the presence of ʔinna. (138) ʔinna hind‐an ustaathat‐un indeed Hind‐ACC professor‐NOM ‘Indeed Hind is a professor.' At least as far as null‐copular sentences are concerned, the effect of ʔinna is structural, not linear. In sentences with indefinite subjects, like 139, where subjects do not appear clause initial, they are still in the accusative case, as can be seen through the accusative case of asaatithatan (professors). Accusative case marking after ʔinna is not determined by adjacency, but rather by structural distance from ʔinna. (139) ʔinna fii l‐maktab‐i asaatithat‐an indeed in the‐office‐DAT professors‐ACC ‘Indeed there are professors at the office.' When ʔinna is followed by an expletive, as in 140, the expletive receives accusative case, and the arguments in the rest of the sentence receive whatever case they would be assigned in the absence of ʔinna. Note the nominative case on the subjects and predicates in the examples, and dative after the preposition and in Construct State. They 55 demonstrate the same case marking behavior described in the case section. (140) a. ʔdaaf‐at ʔinna‐hu hind‐un saafar‐at added‐3FS that.expl(ACC) Hind‐NOM traveled‐3FS ‘She added that Hind travelled.' b. ʔdaaf‐at ʔinna‐hu saafar‐at hind‐un added‐3FS that‐expl(ACC) travelled‐3FS Hind‐NOM ‘She added that Nadia travelled.' c. ʔdaaf‐at ʔinna‐hu hind‐un γaneyyat‐un added‐3FS that‐expl(Acc) Hind‐NOM rich‐NOM ‘She added that the Nadia is rich.' d. ʔdaaf‐at ʔinna‐hu hind‐un fii l‐hind‐added‐ 3FS that‐expl(Acc) Hind‐NOM in the‐India‐DAT ‘She added that Nadia is in India.' e. ʔdaaf‐at ʔinna‐hu fii bayt‐i‐na rajul‐un added‐3FS that‐expl(Acc) in house‐DAT‐our man‐NOM ‘She added that a man is in our house.' At first glance, it may seem redundant to note that the expletive shows accusative case marking. However, given that we have seen in the last section that the expletive is not the subject of the main verb, and in this section we have seen that accusative case marking after ʔinna is not an adjacency effect, the fact that the expletive shows accusative after ʔinna may be noteworthy. Finally, we have seen in 129 that a VS word order is permitted after ʔinna if the verb is separated from ʔinna by an adverbial. As 129 shows, the postverbal subject Haqiiqat shows nominative case marking. If the adverb after ʔinna is followed by the subject, the subject is in the accusative, as in 141: (141) qaala ʔinna l‐yawwm‐a Haqiiqat‐a l‐ʕayyinaat‐i tattaDiH said3MS that today‐ACC truth‐ACC the‐samples‐DAT become.clear3FS ‘He said that today the truth about the samples comes out' 56 In 141, the adverbial12 is followed by a SV construction, and the preverbal subject Haqiiqat carries the accusative marker ‐a. 2.3.5. ʔinna and Its Sisters: Summary To sum up, section 2.3 describes word order and case behavior of ʔinna clauses. Compared to sentences headed by null complementizers or conditionals, only two of the four permissible word orders are available in ʔinna clauses. In terms of case, subjects of ʔinna clauses show accusative, as opposed to the nominative case morphology in sentences headed by conditionals or null complementizers. The ungrammaticality of VS ʔinna‐headed clauses is relaxed if an adjunct PP, AP, or an expletive intervenes between ʔinna and the verb. In terms of case marking, subjects in sentences with intervening PPs or APs are in the accusative. In sentences containing an expletive pronoun after ʔinna, the subject is in the nominative in both preverbal and postverbal positions. The discrepancy in case marking between the expletive pronoun and APs and PPs indicates that the accusative case marking on the expletive is structurally licensed, unlike the inherent accusative of APs. This section shows case and word order facts in ʔinna clauses as opposed to clauses introduced by null complementizers and conditionals. The next section shows why this data, especially the accusative case marking on subjects, is theoretically problematic, and states the research problem of the dissertation. 12 The accusative morphology on l‐yawwm.a (today) is independent from ʔinna. Non‐PP adverbials in Arabic show accusative morphology as a reflection of their inherent case. 57 2.4. Accusative Subjects After ʔinna: The Research Problem Case behavior of the accusative subjects in ʔinna clauses is unexpected, assuming that morphological case mirrors structural Case. In approaches that assume that T is the locus of Case licensing, by the time the complementizer is merged into the structure, the subject DP will have already received nominative, and is unavailable for further valuation. Under the assumption of this dissertation that C is the source of Case licensing through FI, it is unclear why ʔinna should license accusative rather than nominative subjects, or restrict complements in postverbal subjects as discussed earlier, especially since TP complements of null C and those of ʔinna are otherwise identical at least in SV word order. To give a concrete example, take the sentence in 142. (142) ʔinna zayd‐an qaabala ʕamr‐an indeed zayd‐ACC met3MS Amr‐ACC ‘Zayd met Amr indeed.' First, let us consider the sentence under older assumptions, that subject Case is valued under spec‐head agreement with T. At the step of the derivation where the TP is built, the structure of the sentence is as shown in 143, ignoring functional projections between VP and T (v, AspP, …). The verb qaabal moves to T under head movement. The probe on T (an unvalued D feature) is valued by agreeing with the active goal zayd, which is visible to the probe as it has unvalued Case. zayd moves to spec,TP to satisfy the EPP property on T. T values the unvalued Case on zayd with NOM. 58 (143) The activity requirement of a goal to be accessible for computation in a given derivation (Chomsky 2001:5 - 8; Chomsky 2005:19; Chomsky 2000:123; Chomsky 2008:142) effectively bans multiple case assignment. By the time the complementizer ʔinna is merged, the DPs in question will have already received nominative case, and are unavailable for the accusative. In the following step in the derivation (144), ʔinna merges into the structure. (144) Empirical data shows zayd with accusative morphology. This can only result from a 59 probe on C. However, zayd should not be visible for the probe, as it already has a valued Case feature from 143. The derivation should crash. Current assumptions for this dissertation can explain case behavior under ʔinna less problematically. Under FI, subject case/Case is accounted for in terms of C rather than T. The sentence in 145 would then have a derivation as follows: when T merges with vP in 145, both the verb and the subject are available targets. Only ʕamr ‘Amr' is in the domain of vP phase and is inaccessible under PIC. It is also inactive, as its ACC case has been valued by the light verb head. The verb moves to T under head movement. (145) [TP qaabalaq [vP zayd tq (case) ʕamr Case=ACC]] met Zayd Amr In the next step of the derivation 146, merge takes place (a) as well as FI (b) and valuation. C merges to the structure, marking the higher CP phase (a). In b, FI takes place, whereby phi and Case features are transferred to T (now also filled by the verb). The subject Zayd is active because its Case is unvalued, and is accessible in the specifier position of the lower vP phase. Phi and Case probes, now on T, search the domain of the CP phase and locate Zayd. Feature valuation takes place (c) where phi features on T are valued as third person singular, Case on zayd is valued as ACC. The EPP property of T is satisfied (d) by moving zayd to spec,T. (146) a. [CP ʔinnaACC, phi features [TP qaabalaq [vP zaydCase= tq (ACC) ʕamr Case=ACC]] b. [CP ʔinnaACC, phi features, EPP [TP qaabalaqACC, phi features, EPP [vP zaydCase= tq (ACC) ʕamr Case=ACC]] c. [CP ʔinnaACC, phi features, EPP [TP qaabalaACC, phi=3MS, EPP [vP zaydCase=ACC tq (ACC) ʕamr Case=ACC]] d. [CP ʔinnaACC, phi features, EPP [TP zaydCase=ACC qaabalaACC, phi=3MS, EPP [vP tzayd tq (ACC) ʕamr Case=ACC]] 60 Recall that all the operations described in (a - d) take place simultaneously. Under FI, the second step (146) can be represented by (d) only; I have opted to breaking down the operation in several lines for clarity of exposition. The final step is Spell‐ Out (147). At PF, relevant copies are pronounced or deleted, and the morphological component assigns morphemes to valued features. (147) ʔinna zayd‐an qaabala ʕamr‐an indeed Zayd‐NOM met Amr‐ACC ‘Indeed Zayd met Amr.' FI seems to avoid problems related to the activation principle. It still needs to show how the feature structure of ʔinna‐type complementizers account for Case/case and word order restrictions as opposed to those of null complementizers and conditionals. Specifically, why should Case licensed by ʔinna be realized as accusative in preverbal subjects, but nominative in postverbal subjects? FI still needs to show how the feature structure of ʔinna‐type complementizers account for Case/case and word order restrictions as opposed to those of null complementizers and conditionals. This question is compounded if we bear in mind that TP complements, at least in SV word order, are identical across the three types of complementizers as far as finiteness and agreement are concerned. To address this question, the rest of the dissertation examines different steps at the derivation where Case/case for subjects can be valued. I first discuss previous accounts in the literature involving case marking and ʔinna, whereby the grammar could allow for zayd in 142 to be visible to the Case probe on ʔinna for revaluation, in effect allowing MCA. Other accounts posit that the nominative morphology in non‐ʔinna subjects is ‘default case;' case assigned by some postsyntactic 61 operation in the absence of a structural Case assigner, and the probe on ʔinna, if present, would be able to value ACC with no problems. I will show that current complementizer‐based accounts for subject Case do not quite capture Arabic data, and propose a fuller complementizer‐based account. 3. REVIEW OF LITERATURE In Chapter 2, I indicate that the accusative case of the subject in ʔinna clauses is problematic. In this chapter, I review relevant literature related to subject Case assignment in Arabic, case assignment in ʔinna (and other) clauses, and MCA. Section 3.1 outlines literature that touches on ʔinna subjects, mostly in terms of referring to the nominative case in Arabic as default case, as well as one of the most commonly referred to concepts of default case, that of Schütze (1997, 2001). Section 3.2 reviews literature on MCA. Finally, section 3.3 reviews literature on Case assignment via Complementizer. Literature on the subject does not discuss the restriction on VS word orders, after ʔinna. This is probably because many authors are more interested in the agreement asymmetry associated with VS‐SV word order variation. I will touch on SVAA in the review of literature as it is relevant in the context of case/Case and the ʔinna clause. 3.1. Default Case Analysis Under DCA, the nominative case marking on the (subject) DP is not the result of any syntactic operation: it comes out of the syntax with no case marking,13 and is 13 Depending on the specific implementation of DCA, default case is given in the absence of morphological or abstract case. I will discuss specific implementations in their respective sections. 63 assigned default case (nominative) to satisfy a PF (morphology) requirement that every nominal in Arabic must carry case marking. When a syntactic Case assigner, such as ʔinna, is present, the DP in question gets its case (accusative) from its relation with that assigner, and there is no need for default case to be invoked. Accounts that involve ʔinna clauses and invoke default case can be seen in Mohammad (1988, 2000), Ouhalla (1994), and Soltan (2006). In this section, I summarize the relevant data and arguments made in these works. I will discuss in section 4.3 default case as argued for by Schütze (1997, 2001), as it is one of the best articulated accounts of DCA. The overall review of work involving DCA shows that DCA for Arabic is either assumed rather than substantiated, or is defined structurally where an ‘elsewhere condition' is not the best account. 3.1.1. Mohammad (1988) Mohammad argues that the nominative case is the default case for subjects in Arabic based on evidence that it cannot receive case under government from AGR. Working within the GB tradition, he notes (p. 225) that there are five properties attributed to pro‐drop languages (Chomsky 1981; Rizzi 1982): i. Missing subjects. ii. Absence of expletives.14 iii. Apparent absence of that‐trace effect. iv. Free subject‐verb inversion. 14 This property is not true for Arabic, as we have seen the expletive pronoun ‐hu in the last section. There is also another more commonly used expletive; hunaaka, equivalent to the English there. 64 v. Availability of Rule R (Chomsky 1981). Mohammad focuses his paper on items iv and v. Rule R is an optional rule that delivers the effects of affix‐hopping within the GB framework. This rule lowers AGR from Infl to V, which allows AGR to govern the postverbal subject, and assigns NOM to the subject with no need for the subject to move above the verb. Non‐pro‐drop languages do not make use of the optional Rule R, subjects cannot receive NOM, and SV word order variation is not allowed. This rule, according to Mohammad, is based on the assumption that AGR is the nominative case assigner (at least in Arabic), and he sets out to test that claim. All in all, Mohammad shows that there is lack of correspondence between agreement and case assignment in Arabic. He first shows that there are nominative non‐agreeing subjects (Mohammad's terminology for partial agreement), then shows that there are nominative nonsubjects, and finally that there are accusative subjects. To show that nominative subjects do not have to agree with the verb, Mohammad uses data from Classical and Palestinian Arabic to demonstrate SVAA, which is discussed in 2.1.1.1 and 2.1.2. Verbs must agree with preverbal subjects in both gender and number: 148b is ungrammatical as the verb shows only gender agreement, while its counterpart 148d which shows both gender and number agreement on the verb is grammatical. Postverbal subject, on the other hand, may only show gender agreement, as can be seen in the ungrammaticality of 148c where the verb shows number agreement, and the grammaticality of 148a where the verb shows only gender agreement. 65 (148) a. jaaʔat l‐banaat‐u came3FS the‐girls‐NOM ‘The girls came.' b. * l‐banaat‐u jaaʔat the‐girls‐NOM came3FS ‘The girls came.' c. *jiʔna l‐banaat‐u came3FP the‐girls‐NOM ‘The girls came.' d. l‐banaat‐u jiʔna the‐girls‐NOM came3FP ‘The girls came.' He proceeds to show that there are nominative NPs (DPs) that are not governed by AGR. Using data similar to our 75 and 76, repeated here as 149 and 150, he shows that both the subject and predicate ‘(if the predicate can receive Case, that is)' are nominative (p. 230). (149) al‐ustaath‐at‐u jamiil‐at‐un / taʕbaan‐at‐un/miSriyy‐at‐un the‐professor beautiful / tired/Egyptian (def‐prof‐FS‐NOM) (beautiful /tired /Egyptian ‐FS‐indefNOM) ‘The professor is beautiful/ tired/Egyptian.' (150) al‐muhadis.uun taʕbaan.uun the‐engineers tired (def.PM.NOM) (indef.PM.NOM) ‘The engineers are tired.' Mohammad dismisses nominative Case assignment under government by AGR by showing that there are data where exclusive Case assignment under government by AGR violates uniqueness, where there is a mismatch between the number of governors and the number of nominative NPs. According to Mohammad, ‘If nominative Case can 66 only be assigned under government by AGR, we must, then, conclude that AGR can assign nominative Case to two positions. This will constitute an unwarranted deviation from the assumption in GB (cf. Chomsky, 1981) that one and only one Case per Case assigner, and one and only one Case per assignee' (p. 230). He further shows that the nominative on the predicate DPs are not related to AGR indirectly through some kind of percolation from the subject. In sentences like 151, the embedded subject lwalada (the boy) must be in the accusative case, and the predicate must be in the nominative case/Case. This means that case marking on the subject and predicate are independent of each other. (151) a. qultu ʔinna l‐walad‐a mariid‐un said1S that the‐boy‐ACC sick‐NOM 'I said that the boy is sick.' b. *qultu ʔinna l‐walad‐a mariid‐an said1S that the‐boy‐ACC sick‐ACC 'I said that the boy is sick.' c. *qultu ʔinna l‐walad‐u mariid‐un said 1S that the‐boy‐NOM sick‐NOM. 'I said that the boy is sick.' So far, Mohammad shows that there are nominative subjects that do not agree with the verb, and that there are nominative nonsubjects. The data in 151 also show that there are subjects that are not nominative (lwalada in a). Having shown that there is no relation between AGR and nominative, Mohammad attempts to identify where nominatives in Arabic come from. He looks at left‐dislocated NPs, as in 152 and topicalized NPs, as in 153. In both sentences, the sentence‐initial nominative NP is outside the domain of AGR, and is associated with a 67 resumptive pronoun that shows accusative [c]ase. In both sentences the sentence‐initial NP would be ungrammatical to carry anything other than nominative (p. 234). (152) l‐walad‐u qultu ʔanna‐hu jaaʔa the‐boy‐NOM said1S that‐he ACC came3MS 'The boy, I said that he came.' (153) l‐walad‐u raʔaa‐hu ahmad‐u the‐boy‐NOM saw3MS‐him ACC Ahmed‐NOM 'The boy, Ahmed saw him.' He suggests that the nominative case, then, is not assigned by AGR, but it is a default case: Following Mohammad (1986), I suggest that nominative Case in Arabic is a default Case, i.e. it is a Case assigned to a +N category if this category is immediately dominated by a sentential category and it is not governed by a lexical Case assigning governor. This is confirmed by the following examples where the nominative Case assigned to TOPIC positions, as in the examples discussed previously, can be overridden: (154) [his(34)] qultu ʔanna l‐walad‐a ra?aa‐hu ar‐rajul‐u said 1S that the‐boy‐ACC saw3MS‐him the‐man‐ACC 'I said that the boy, the man saw him.' (155) [his (35)] inna l‐walad‐a qultu ʔanna‐hu jaa?a that the‐boy‐ACC said1S that‐he came3MS 'The boy, I did say that he came.' In both (34) and (35) there is a TOPIC in the accusative which would be in the nominative if it were not embedded under ʔanna/ʔinna. The ease with which the nominative Case can be overridden does indeed provide strong support to the claim that nominative Case in Arabic is a default Case. (Mohammad 1988:234-235) 3.1.1.1. Review of Mohammad (1988) Mohammad convincingly shows that AGR is not the nominative case assigner by showing that there can be a mismatch between the number of AGR heads in a sentence 68 and the number of nominative DPs in that sentence. His argument is based on a number of theoretical assumptions. First, he assumes that morphological case is a strict reflection of structural case; he consistently refers to upper‐case Case. He also commits to a strict one Case assigner assigning exactly one Case, and each NP will receive one and only one Case. These are standard assumptions within GB, and largely within MP ‐ whether it is checking, assignment, or valuation, there is exactly one assigner/valuator/checker for each DP. I will discuss these assumptions in later sections that discuss DCA and MCA. In terms of data, Mohammad's selection of data points is appropriate, and addresses the point in hand: that an AGR cannot be the nominative assignment. He states that his data are from CA. However, I think that he is using the terms CA and Standard Arabic interchangeably and the data are meant to demonstrate MSA facts. The sentences in themselves can be from MSA or CA. However, at least one of the grammaticality judgments does not apply to CA. He claims that (p. 234) fronted NPs must be in the nominative case, and nothing else (but c.f. the discussion of examples 152 and 153 above). This is, in fact, not true for CA, as can be seen in 156.15 (156) Wa kulla ʔinsaanin ʔalzam.naa.hu Taaʔira.hu fii ʕunuqi.hi and each human impose.we.him fate.his in neck.his (ACC) (GEN) (ACC) (GEN) ‘And [for] every person We have imposed his fate upon his neck' (Quran 17:13) In this sentence, the left‐dislocated DP kulla ʔinsaanin shows accusative case marking (the final -a on kulla), which is the same case for accusative resumptive pronoun 15 Thanks for Ahmed Al‐jallad (p.c.) for bringing the sentence to my attention. 69 associate -hu. If Mohammad's claim were true, the verse would show the fronted DP in the nominative case (kullu ?insaanin), contrary to fact. Mohammad's statement, however, is true for MSA. My own grammaticality judgment, as well as Bakir (1980) confirm the case distribution for nominative fronted DPs in MSA. As far as offering an alternative explanation to the nominative case, i.e. in terms of default case, this is where Mohammad's paper falls short. First of all, he commits to a government‐based account of Case, at least for accusatives. Providing two different accounts - default for nominative and government for accusative - lacks elegance. Mohammad's account of default case refers to Mohammad (1986), which is a conference presentation with no printed record of the line of argumentation thereof, so I limit my review to the statements at‐hand. Mohammad defines default case structurally as ‘a Case assigned to a +N category if this category is immediately dominated by a sentential category and it is not governed by a lexical Case assigning governor.' His definition is problematic in two ways. First, if he could structurally assign default case so precisely, his definition ceases to be that of a ‘default' case. It can simply be restated in terms of a Case assigner plus some locality requirement. His definition amounts to something like ‘nominative is assigned by a sentential head16 (perhaps an empty C), and there is another case assigner intervening between that head and the +N category,' which is the approach of later literature in Case Theory that I adopt for this dissertation. Also, if we look at his default case as a kind of ‘elsewhere condition,' it goes against his earlier argument that for each case, there has to be one and only Case 16 He does not specify what he means by a sentential category, so I will take it to be CP. 70 assigner, and nominal categories must be assigned one and only one Case. His definition of default case deprives it from its ‘defaultness,' and throws the ball back to MCA issues. Under feature‐based assumptions of MP, Mohammad's argument regarding NOM are partially valid. AgrP is abandoned in favor of operation Agree stemming from T. Taking Case as a feature of T, which would be the Minimalist equivalent of Mohammad's argument, his argument against Case on T would still stand. As we will see in Chapter 4, in the absence of FI (where Case comes from C, rather than T), NOM/ACC alternation cannot be accounted for through T alone. Mohammad's proposal of DCA, however, is inconsistent with the Minimalist requirement that all unvalued features (such as Case) must be valued by a higher probe, or the derivation crashes. To sum up, Mohammad argues against NOM being assigned under government by AGR. He uses the accusative subject of the ʔinna clause as evidence for considering NOM a default case in Arabic, defined as dominance by a sentential category with no case assigner sister. His argument against AGR‐NOM relationship is strong, but his argument for default case is not strong as it leads to contradictions within Case Theory, even the version that he assumes. 3.1.2. Mohammad (2000) Mohammad (2000, Chapter 2) continues the same line of argumentation, but within the Minimalist Program. The aim of that chapter is to identify how different word orders are derived in MSA. It is of concern to us as it discusses interaction between case and preverbal DP positions in Arabic, and it discusses some of the word order 71 restrictions after ʔinna, and tries to explain them. Mohammad assumes a version of the established Minimalist assumptions (Chomsky, 1986, 1995; Koopman and Sportiche, 1991; Pollock, 1989, and others). Namely he assumes the following projection hierarchy, together with the VP‐internal subject hypothesis: (157) CP>TP>AGRS>AGRO>VP17 Mohammad explains word order variation in Arabic as follows: the verb always moves to T. In verb‐first sentences, the subject (and I assume the object) remains in‐situ: in spec,VP. In verb‐second sentences, the subject raises to spec,TP (attributing the variation to the<