University of Southern California Dissertations and Theses

Silence in answers: a study of ellipsis in Hindi

(USC Thesis Other)

Silence in answers: a study of ellipsis in Hindi

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content SILENCE IN ANSWERS: A STUDY OF ELLIPSIS IN HINDI by Bhamati Dash A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (LINGUISTICS) August 2022 Copyright 2022 Bhamati Dash ii Acknowledgements When I was three and had skinned my knee, my father told me that the Dr. in front of his name did not mean that he could treat me and make me feel better. We had to go to a ‘real’ doctor he said. Ever since then I have been intrigued by the ‘not real’ kind. When I was seventeen and had decided to pursue higher education in Humanities, my father gave me a research idea for a Ph.D. degree in Linguistics. In a way this has been a long time coming and now that I have finally made it here, I have a lot of people who I owe my thanks to, who have aided me in this journey. First and foremost, I am deeply indebted to Prof. Andrew Simpson, my advisor here at USC without whose support, gentle guidance and patience this endeavor would not have been possible. I owe the existence of this research to his guidance as he is the one who introduced me to the topic and only under his tutelage did I learn that sometimes silence speaks a lot louder than words, even syntactically! I would like to express my deepest appreciation to my committee for their invaluable contribution and guidance throughout my Ph.D. years. This project would not have seen its completion without the unwavering support of Prof. Roumyana Pancheva, Prof. Audrey Li, Prof. Stefan Keine, Prof. Namkil Kim and Prof. Jeremy Goodman. Whether it be Audrey’s belief in my project, Roumi’s encouragement to expand my research purview, Stefan’s insightful questions that made me think deeper or Namkil’s and Jeremy’s ever-ready support as my external members, I benefited from it all. I must also thank Stefan for allowing me to be a part of his Hindi cyclic Agree research project. Having seen him approach syntax research up close in my formative years as a researcher gave me invaluable insight into the process. Many thanks to Elsi Kaiser who guided me through my experimental syntax research. Working with her on this quantitative project has been a momentous experience for this otherwise armchair syntactician. iii I would not even have applied to USC for Ph.D. had it not been for Prof. K.A. Jayaseelan and Prof. R. Amritavalli’s belief in my work and my abilities. I had already started my Ph.D. in EFLU, Hyderabad under R.Amritavalli when they encouraged me to apply abroad and I have not looked back since. I cannot even begin to express my thanks to them for the role they have played in my academic career. I’d also like to extend my gratitude to all my linguistics teachers from EFLU and USC over the years. A special mention to all my syntax teachers: K. A. Jayaseelan, R. Amritavalli, P. Madhavan, Anuradha Sudarshan at EFLU and Andrew Simpson, Roumyana Pancheva, Audrey Li, Maria Luisa Zubizarreta, Stefan Keine at USC. They made me fall in love with syntax and have ensured my continued dedication to the field. I very much appreciate all my other teachers from EFLU and USC as well. Though I was quite lost to syntax from the beginning, I have learned so much about so many different aspects of Linguistics from them. Thank you, Rahul Balusu, Hemanga Dutta, Indranil Dutta, Utpal Lahiri, Hemalatha Nagarajan, M. Hariprasad ,Shruti Sircar, K. G. Vijayakrishnan at EFLU and Louis Goldstein, Hajime Hoji, Canan Ipek, Khalil Iskarous, Karen Jesney, Toby Mintz, Deniz Rudin, Barry Schein, Rachel Walker, Mary Washburn at USC. I gratefully acknowledge the assistance of our program administrators at USC, Lisa Jo and Guillermo Ruiz. A special note of thanks to Guillermo who made the administrative aspect of Ph.D. so effortless for me that the only thing I was left to do was research. My life at the Linguistics Department at USC and in LA was made fulfilling by all my friends and my cohort here. I very much appreciate the support that I received from Ulli Steindl, Thomas Borer, Saurov Syed, Mythili Menon, Priyanka Biswas and Arunima Chaudhary in my initial years here. They kept me from being homesick during my first few years. I will never forget the camaraderie I found in my cohort. Ana Besserman, Hayeun Jang, Maury Lander-Portnoy, Narae iv Son and I discovered LA together and held each other’s hands through the first couple of years of Ph.D. Ph.D. was not easy and it was not supposed to be and I am very appreciative of the bonds that we forged through this trial of fire. I would also like to mention the syntax plus group at USC who through the years have provided me with unwavering support both during group meetings and at after-meeting drinks. I have lived a life here, a life different from the one I have lived before or probably will live again. My friends have been a huge support system and I am extremely grateful to each and every one of them. Thank you, Madhu, for the long discussion sessions on any and every topic under the sun from cognate words to vamps in tv serials, for commiserating together on the challenges of taking the road less taken (how many people actually study Linguistics!) and last but not the least for reading through my drafts and giving me valuable suggestions. Thank you, Betul, for making sure that I had faith in myself, for always encouraging me to do my best and for your indispensable guidance on the path to a more fulfilling life. Thank you, Haley, for all our syntax talks and for being the best travel companion to conferences. Thank you, Mallika, for our horror movie weekends, and for the many long drives that we took, they were the perfect de-stressors. Thank you, Kavita and Thank you, Nitin, for always being encouraging from afar and for always pushing me to be worthy of your friendship and care. Thank you, Hiteshi, for being the best roommate ever and for teaching me that there is a life beyond research and reading stories by my lonesome. Thank you, Puja, Swati, Satya and Satish for treating me like your family all these years and for feeding me homemade Indian food whenever I felt homesick. Thank you, Ananya, Priscilla and Athul for always listening to my frustrated rants and for dreaming of building a future together and changing the world from across different continents and different time-zones. And thank you all also for indulging me whenever I needed language data for my research. I got the friendship, v the support, the food and my data and I am incredibly grateful for this skewed equation! I would be remiss if I did not extend my gratitude to my therapist. When the last few years grew tough (sometimes Ph.D. in an obscure field in a foreign land amidst a global pandemic drives you to therapy!), the therapy sessions were my only way to make sense of everything and they helped me a great deal in surviving and conquering. Last but not the least, I owe to my family whatever accomplishment I can put to my name. My mother has always encouraged me to pursue my dreams even if said dreams took me 8000 miles away for years together. She has always been ready to extend whatever support she can, whether it be collecting language data from her workplace for my research or from agreeing to not indulge in thoughts of my marriage (important for her!) until I finish my studies. I would also like to thank my younger sister who has been my silent pillar all through out. She has been my closest confidante and has never hesitated in accommodating howsoever she could, to ensure that I have the support I need. I started my thanks-giving with my father and now I would like to end with him. He was my inspiration growing up. Though he did not live to see me pursue Linguistics, I like to believe that being the daughter of a Linguistics professor played a role in my love affair with the subject. I take this opportunity to dedicate this dissertation to my father. Thank you papa. vi Table of Contents Acknowledgements………………………………………………………………………..ii List of Figures …………………………………………………………………………....ix List of Abbreviations ......................................................................................................... xi Abstract……………………………………………………………………………..........xii Chapter 1: Introduction ....................................................................................................... 1 1.1 Structure of Polar Questions in Hindi ......................................................... 6 1.1.1 A note on kya ............................................................................................ 11 1.2 Fragment Answers to Polar Questions ...................................................... 13 1.3 Literature Review...................................................................................... 17 1.3.1 Merchant (2005)........................................................................................ 17 1.3.2 Holmberg (2015) ....................................................................................... 19 Chapter 2: Fragment Answers in Hindi: An Account of ‘Verb Stacking’ ........................ 22 2.1 Hindi Data and Observations .................................................................... 23 2.2 Potential Analyses ..................................................................................... 27 2.2.1 A Pro Drop analysis .................................................................................. 28 2.2.2 An Argument Ellipsis analysis.................................................................. 33 2.2.3 A Subject pro drop + VP Ellipsis analysis ................................................ 38 2.2.4 An IP Ellipsis analysis .............................................................................. 41 2.3 Proposed Analysis: a υP Ellipsis Approach .............................................. 47 2.3.1 Subject in spec, υP .................................................................................... 48 2.3.2 Verb stranding υP Ellipsis ........................................................................ 53 2.3.3 Proposal………………...…………..…………………………………….55 2.3.3.1 Merchant’s E feature ................................................................................. 56 2.3.3.2 Two PolPs in Hindi ................................................................................... 57 2.3.4 Account…………………………………………………………………..61 2.4 Conclusion ................................................................................................ 67 vii Chapter 3: Biased Questions and Answers in Hindi ......................................................... 70 3.1 Negative Polar Questions .......................................................................... 70 3.2 Data - Answer Patterns ............................................................................. 73 3.3 The nature of polarity particles in Hindi ................................................... 77 3.4 Prosodic Properties of the two answer systems ........................................ 82 3.5 Analysis of Biased Answers ..................................................................... 84 3.5.1 Holmberg (2015) – A Syntactic Analysis ................................................. 85 3.5.2 Roelofsen & Farkas (2015) – A Lexical Semantic Analysis .................... 87 3.5.3 Tang (2020) – A Cartographic Analysis ................................................... 87 3.5.4 Asher & Reese (2005) – A Pragmatics -Semantic Analysis ..................... 88 3.5.5 Proposed Analysis ..................................................................................... 89 3.5.5.1 Evidence from positive/neutral polarity questions ................................... 96 3.6 Semantics of Bias ...................................................................................... 98 3.7 Other Biased Questions........................................................................... 100 3.8 Conclusion .............................................................................................. 105 Chapter 4: Other Types of Question - Answers .............................................................. 107 4.1 Introduction to Polar Alternative Questions (PAQs) in Hindi ................ 110 4.1.1 Analysis of PAQs in Hindi ..................................................................... 114 4.1.2 Analysis of Answers to PAQs in Hindi .................................................. 120 4.2 Introduction to Alternative Choice Questions (ACQs) in Hindi ............ 124 4.2.1 Han & Romero (2004) ............................................................................ 127 4.2.2 Analysis of ACQs in Hindi ..................................................................... 128 4.2.3 Analysis of Answers to ACQs in Hindi .................................................. 139 4.2.4 Interim Conclusion: Section 4.1 and 4.2 ................................................. 141 4.3 Introduction Polar Questions with Narrow Focus (PFQs) in Hindi ........ 142 4.3.1 Holmberg’s Analysis of Questions and Answers with Narrow Focus ... 144 4.3.2 Analysis of PFQs in Hindi ...................................................................... 146 4.3.3 Analysis of answers to PFQs in Hindi .................................................... 151 4.4 Conclusion .............................................................................................. 158 viii Chapter 5: Conclusion..................................................................................................... 160 5.1 Verb stranding υP Ellipsis in Hindi: Manetta (2019) ............................. 161 5.2 IP Ellipsis in Hindi .................................................................................. 165 5.2.1 Analysis for IP Ellipsis in Hindi ............................................................. 170 5.2.2 Fragment Answers to polar Questions .................................................... 172 5.2.3 Conclusion: IP Ellipsis in Hindi ............................................................. 173 5.3 Doubling of Functional Projections ........................................................ 179 5.4 Crosslinguistic Implications.................................................................... 186 5.5 Conclusion .............................................................................................. 197 References………………………………………………………………………………198 ix List of Figures Chapter 1 Figure 1: Polar Questions in Hindi ............................................................................................... 11 Figure 2: Merchant (2005) ............................................................................................................ 18 Chapter 2 Figure 3: Argument Pro Drop Analysis ........................................................................................ 30 Figure 4: Argument Ellipsis Analysis ........................................................................................... 36 Figure 5: Subj pro drop + VP Ellipsis ........................................................................................... 38 Figure 6: Holmberg's PolP Ellipsis ............................................................................................... 43 Figure 7: Verb stacking as υP Ellipsis .......................................................................................... 63 Figure 8: Polarity particle as IP Ellipsis ....................................................................................... 65 Chapter 3 Figure 9: Possible Combinations of polarity features (Roelofsen & Farkas 2015) ...................... 80 Figure 10: PRAAT waveform and spectrogram for (8i) – (nahi) nahi paDhii thii ....................... 83 Figure 11: PRAAT waveform and spectogram for (11i) – nahi // PADHII thii ........................... 83 Figure 12: Polarity-based answer – (nahi), nahi paDhii thii ......................................................... 92 Figure 13: Truth-based answer – nahi // PADHII thii .................................................................. 93 x Chapter 4 Figure 14: Or not (υP) PAQs ...................................................................................................... 115 Figure 15: ‘or not’ PAQs ............................................................................................................ 118 Figure 16: ‘or not υP’ PAQs ....................................................................................................... 119 Figure 17: Answer to ‘or not’ PAQs .......................................................................................... 122 Figure 18:Answer to ‘or not υP’ PAQs ...................................................................................... 123 Figure 19: Account of y/n reading of ACQs in Hindi ................................................................ 129 Figure 20: Account of alt reading of ACQs in Hindi.................................................................. 139 Figure 21: Fragment answer for alt-reading of ACQs ................................................................ 141 Figure 22: PFQs in Hindi ............................................................................................................ 148 Figure 23: Fragment answer to FPQs ......................................................................................... 155 Chapter 5 Figure 24: IP Ellipsis in Hindi .................................................................................................... 171 Figure 25: IP Ellipsis in Welsh ................................................................................................... 188 Figure 26: IP Ellipsis in Russian ................................................................................................. 189 xi List of Abbreviations 1/2/3 - 1 st person/2 nd person/3 rd person ACC - accusative AUX - auxiliary DAT - dative ERG - ergative F - feminine FUT - future GEN - genitive HAB - habitual INF - infinitive M - masculine NEG - negation NOM - nominative OBL - oblique P - postposition PFV - perfective Pl - plural PRES - present PROG - progressive Prt - particle PST - past Q - question particle Qprt - Q particle Sg - singular SM - scope marker Voc - vocative particle xii Abstract Merchant (2005) and Holmberg’s (2015) research on fragment answers posits that answer forms have structures similar to the questions they are responses to, thereby arguing that the fragment answer forms are derived from the non-pronunciation/ellipsis of significant portions of a full clausal sentence. Answers to yes/no questions like (1) below are argued to have an underlying sentential structure which is only partially pronounced (non-pronounced material is represented with strike-through across words). Welsh, Irish, Finnish, Mandarin Chinese, Japanese and Russian are ‘verb echo’ languages which show a pattern wherein the structurally highest verbal element is repeated as an answer. Holmberg (2015) analyzes ‘verb echo’ patterns as a structure derived by IP Ellipsis with the highest verb having raised to a position higher than IP. Interestingly, in Hindi, the entire verbal sequence needs to be repeated in an answer as can be seen in (2). I term the pattern in Hindi as ‘verb stacking’. 1. Q: all mair aros can mair stay ‘Can Mair stay?’ A: gall Mair aros can Mair stay ‘Yes.’ verb echo -Welsh (Holmberg 2015) 2. Q: (kya) raaghav kitaab paDh paa rahaa thaa? Q raaghav book read able PROG.3MSg AUX.PST.3MSg ‘Was Raghav able to read the book?’ xiii A: haaN paDh paa rahaa thaa raaghav kitaab yes read able PROG.3Msg AUX.PST.3MSg raaghav book ‘Yes.’ verb stacking - Hindi The primary objective of this dissertation is to provide an account of the novel fragment answer patterns attested in Hindi. Any account of fragment answers in Hindi should ideally be able to account for obligatory verb stacking given that the verbal sequence does not form a complex head and does not raise higher than IP. We also need to account for the fact that IP ellipsis is independently attested in the language but as we do not see verb echo patterns, fragment answers cannot be derived by IP ellipsis. Yet another empirical observation that needs to be accounted for is that in Hindi the verb stacking pattern is the unmarked answer pattern whereas answering a yes/no question simply with a polarity particle is a marked response. I propose that ellipsis in fragments answers is triggered by a Merchant (2005) style Ellipsis feature hosted by the polarity head which causes the complement of the polarity head to be elided under semantic identity. The default polarity head in Hindi is merged IP-internally just above the ʋP, leading to the deletion of ʋP after the main verb has raised out of it, resulting in the verb stacking pattern. I show that this fragment answer pattern follows from an independently motivated ellipsis process of υ-stranding υPE (Bhatt & Dayal 2007, Manetta 2019) in Hindi. The proposed analysis assumes head movement of the verb outside of ʋP to the IP-internal polarity head. Bhatt & Dayal (2007) and Manetta (2019) among others have all argued for the main verbal element in Hindi to move out of ʋP, though there is not full agreement on the identity of the head to which the verbal element is assumed to move to. I propose that the verb moves to a polarity head projected immediately above the ʋP and below the aspect head, for the purpose of lexical exponence of the polarity feature. xiv Furthermore, I suggest that this head movement is syntactic as it feeds into the syntactic process of ellipsis and is not completely determined by morphological considerations. In addition to arguing for an IP-internal default polarity head I motivate a higher polarity phrase specified for contrastively focused polarity. Romero (2006) among others has argued for two different positions of verum focus which may syntactically correspond to two distinct polarity phrases, one being a default projection and the other involving contrastive focus. Most recent literature on polarity emphasis presents polarity in clauses by giving a distributed structure, a projection on the left periphery which encodes focused polarity along with a lower polarity projection (Breitbarth et al. 2013). The analysis proposed for Hindi here supports this generalization. In Hindi, the higher projection is only licensed when parallel constructions have opposing polarity making contrastive polarity explicit in the discourse. I argue that when merged, the higher polarity head can host an ellipsis feature and licenses ellipsis of its complement IP. I further propose that verbal ellipsis in Hindi is constrained and licensed by polarity heads. The relevance of polarity comes into play only when it is being questioned, as in polar questions, or opposed, as in contrastive polarity ellipsis structures. I postulate that the IP-internal polarity phrase licenses verb stranding ʋP ellipsis and the contrastive polarity phrase above IP licenses IP ellipsis. Given that the only possible verbal ellipsis options are ʋP ellipsis or IP ellipsis, I further provide support for a general account of ellipsis being constrained by the two phases: ʋP and CP and licensed by polarity (Bošković 2014, McCloskey 2017). Besides the two dedicated positions for polarity in Hindi, I further motivate yet another projection which could potentially host a polarity particle assuming a cartographic approach to performative projections in Hindi. While examining fragment answer patterns to negative biased polar questions, I argue, following Roelofsen & Farkas (2015) that polarity particles may be xv anaphoric to the antecedent questions and signify (dis)-agreement of the fragment answer with the highlighted alternative of the question. Based on the prosodic properties of polarity particles in fragment answer patterns, I argue that when they have such an anaphoric function, they occupy a higher position in a performative projection in the extended left periphery. The dissertation, consequently, argues for three different clausal representations for polarity of a sentence. The default polarity head carrying the sentential polarity of a clause is merged in an IP-internal position. When in discourse there is explicit contrastively focused polarity, sentential polarity occupies a higher position and the dedicated contrastive polarity head takes an IP complement. Finally, when a polarity particle is anaphoric to a previous sentence and references the polarity of the previous sentence, it occupies a position in the extended left periphery. Such a clausal representation is shown to be able to model fragment answers to all kinds of questions in Hindi. Additionally, I posit that typological variation in fragment answer patterns is determined by the position of polarity phrases in different languages. While an IP-internal default polarity head accounts for verb stacking ellipsis pattens in fragment answers, a higher polarity phrase in other languages results in the different ‘verb echo’ (Holmberg 2015) pattern exhibited in those languages (Irish (McCloskey 2017), Finnish (Holmberg 2015), Welsh (Holmberg 2015), Russian (Gribanova 2017), Mandarin (Wei 2016) and Japanese (Sato et al. 2018). Since polarity phrases may occur in different positions in the clausal structure in different languages, I propose that this variation in the physical location of a polarity phrase directly results in differences in the amount of sentential material that can be left unpronounced in an answer, producing fragments of differing types and sizes as semantically parallel answer forms. Thereby, in principle, all an answer-form requires is a valued polarity feature in a language and as soon as that requirement is fulfilled, xvi ellipsis of all subordinate material takes place. Hence the position of the polarity head in a clause determines the size of ellipsis and consequently the size of the answer fragments. A stronger generalization I posit is that cross-linguistically fragment answers are always derived by verbal ellipsis, even in languages like Hindi which attest to mechanisms like pro drop and argument ellipsis. These other mechanisms could independently account for some silent syntactic structure but not that of fragments as answers as can be consistently observed in Hindi fragment answer patterns. 1 Chapter 1: Introduction This thesis examines the hypothesis that the syntax of answers is to a large extent similar to the syntax of questions with respect to different kinds of questions in Hindi. Amongst the various analyses given for the structure of questions in the post-GB framework, it has been a popular assumption that the base structure of a question is similar to that of its declarative counterpart wherein the arguments and the adjuncts are merged in their base positions. This is easily observed in wh in-situ languages where the wh-word occupies the same position as its non-wh counterpart. In wh-movement languages, the wh-constituent has been argued to have moved from its base position for the purposes of interpretation as an interrogative. Even in polar questions, cross- linguistically, the base structure of the question is considered mostly parallel to its declarative counterpart. It is an established observation that different kinds of questions behave similarly, as has been argued in various accounts from Katz & Postal (1964) to Cheng (1991) to Holmberg (2015). It has further been hypothesized that the answers to questions have a similar syntactic structure to that of the questions. Given that questions are derived from their declarative counterparts, it seems an obvious extension that the answers to those questions will share certain similarity with the questions. Usually, a complete answer consists of a repetition of the proposition in the question structure with the answer to the questioned constituent. In a wh-constituent question, this entails that the answer consists of the full proposition with the wh-word replaced with an appropriate non-wh expression (3-4). The answer needs to contain the value of the wh- 2 word such that the proposition expressed would be true. And in case of polar questions, the answer consists of repeating the full proposition along with the polarity of the proposition (5-6). The answer to a polar question conveys the polarity of the proposition such that the proposition conveyed would be true. 3. Q: Who read the book? 4. A: Raghav read the book. 5. Q: Did Raghav read the book? 6. A: Yes, Raghav read the book. However, answers are not always overtly composed of the entire proposition. (7) and (8) are natural answer forms to questions in (3) and (5) respectively. 7. A: Raghav (did). 8. A: Yes, (he did). Such answers are called fragment answers. Fragments because they do not repeat the entire proposition but only the relevant expression that targets the questioned constituent. In the case of wh-constituent questions, fragment answers essentially need to constitute the answer to the wh- word and in the case of polar questions, the polarity of the proposition is an essential part for the fragment answer. How the essential answer components are codified and what else remains in the answer is subject to cross-linguistic variation. An assumption in the literature concerning the syntax of fragment answers is that, both for wh-constituent questions and polar questions, fragment answers are semantically propositional and have a full-fledged clausal syntactic structure (Hankamer 1979; Morgan 1973, 1989; Stanley 2000; Reich 2002, 2003; Merchant 2005; Holmberg 2015; Gribanova 2017). This assumption holds weight in the sense that it respects the form-meaning (syntax-semantics) mapping. In (7) and (8) even though only a fragment is overtly 3 realized, semantically they are equivalent to (4) and (6) respectively. Most of the clause is then argued to be phonetically null thereby giving only fragments as answers. Merchant (2005) gives an ellipsis analysis for the clausal structure of fragment answers in (7) to wh-constituent questions like in (3) by essentially arguing that the answer in (7) has been derived from its full-fledged form in (4). Holmberg (2015) extends a similar analysis to fragment answers to polar questions by arguing that answer forms like in (8) are fully clausal constituents (just like in (6)) having undergone ellipsis. Form-meaning (syntax-semantics) mapping has been a keystone of linguistic formalism which embodies the traditional belief that syntactic structure should be parallel to and representative of the meaning conveyed. And as mentioned, a clausal structure in syntax has been argued to convey a propositional interpretation. The objective of this dissertation is to study fragment answers to different types of questions in Hindi and explicate the details of how structures which overtly are lacking, convey propositional interpretations. Research into syntactically smaller structures like fragments which convey the full propositional interpretations has weighed in on the fundamental question of how grammatical structure is formalized in a way that leads to the meaning conveyed. I have argued for a PF deletion ellipsis approach à la Merchant (2005) to account for answer fragments in Hindi. It respects the syntax-semantics mapping as it proposes a structure where constituents can be covert without compromising the meaning. I support the argument that fragments have a full-fledged clausal structure, parts of which are silent, having undergone ellipsis. Furthermore, I show that languages differ in how much unpronounced structure there is, which I propose is determined by the structural distinctions in a given language. The underlying philosophy that my dissertation vouches for is that this ‘syntactic silence’ (defined as the non-pronunciation of words in clausal structure) is a linguistic expression governed by usual 4 syntactic constraints. The overarching question in my dissertation is what the structure of silence conveys about the clausal structure of Hindi which might not have been obvious looking at fully overt structures in isolation. Another research questions I address is how the clausal structure of Hindi is in comparison with that of other languages, which further accounts for the different patterns of ellipsis contributing to a typological overview. This dissertation has four main chapters. In chapter 2, I provide a detailed account of fragment answers to neutral polar questions in Hindi. Hindi generally has multiple phenomena that could potentially account for missing syntactic structure, I discuss all of them and eventually motivate a verb stranding υP ellipsis approach to account for the obligatory verb stacking fragment answer patterns attested in the language. Chapter 3 focuses on fragment answer patterns to negative biased polar questions. Having established in the previous chapter that fragment answers are derived by verb stranding υP ellipsis, I extend the analysis to answers to negative biased polar questions as well. Furthermore, this chapter focuses on a detailed discussion of the nature and properties of polarity particles which constitute an essential part of fragment answers to polar questions. Next in chapter 4, I examine slightly different kinds of question-answer pairs. Polar questions can also be formulated with an overt disjunction of the positive and negative alternatives of the proposition. The first part of the chapter is dedicated to the structure of such polar alternative questions and their answer forms. Next, I explore alternative choice questions which are formed by disjunction of two distinct propositions and finally I study polar questions which have a constituent with narrow focus. I motivate a verb stranding υP ellipsis account for fragment answers to all the three different kinds of question structures. Though in this dissertation, I have not delved into wh-constituent questions and their fragment answer forms, the derivation of such answer forms can be considered to be a straight-forward application of the same verb stranding υP ellipsis 5 account. Finally in Chapter 5, I conclude this dissertation by providing evidence for verb stranding υP ellipsis and IP ellipsis, independent of fragment answer forms. I then motivate why verb stranding υP ellipsis is the only viable approach to formulate answer forms to different questions in Hindi. While concluding, I explore the cross-linguistic implications of the attested fragment answer forms in Hindi and their argued for analysis. I analyze the parametric variation in fragment answer forms essentially to be a product of the difference in position of the polarity projection in a clausal structure. I further propose that ellipsis in general is licensed by polarity heads, Hindi having two of them. The positioning of the two polarity projections, I argue, result in υP and IP ellipsis thereby constraining ellipsis by the two phases: υP and CP. In the rest of this introduction, I briefly explore the structure of polar questions in Hindi in section 1. As mentioned earlier, the syntax of questions sheds light on the syntax of answers and vice versa. An established syntax of polar questions in Hindi would lead us to postulate an analysis for the syntax of fragment answers in Chapter 2. Before we delve into the fragment answer patterns attested in Hindi, I introduce the different kinds of fragment answer patterns attested as answer forms to neutral polar questions cross-linguistically in section 2. I also lay out the attested answer patterns in Hindi and motivate the need for research into the novel Hindi patterns that has consequences for the cross-linguistic analysis of fragment answers. In section 3, I give a brief overview of Merchant’s (2005) ellipsis account for fragment answers to wh-constituent questions and Holmberg’s (2015) ellipsis account for fragment answers to polar questions, both of which constitute the essential framework for the analysis proposed for Hindi fragment answers in this dissertation. 6 1.1 Structure of Polar Questions in Hindi Hindi is an SOV language. Hindi typically formulates polar questions with the same word-order as its declarative counterpart but with a rising intonation on the verbal sequence, i.e., the last sequence in the word order as can be seen in (9). Rising intonation has been associated with L/H- H% by Butt et al. (2017) and Biezma et al. (2017). 9. Q: raaghav-ne kitab paDhii?↑ Raghav-ERG book read.PFV.3FSg ‘Did Raghav read the book?’ 10. Q: kya raaghav-ne kitab paDhii? ↑ Q 1 Raghav-ERG book read.PFV.3FSg ‘Did Raghav read the book?’ An optional question particle kya is also attested in polar questions in Hindi. Bhatt & Dayal (2020) observe that kya when present in the polar question does not substitute the rising intonation. Distributionally, it has been observed that the question particle kya can occur in any position in Hindi. 11. Q: (kya:) anu=ne (kya:) uma=ko (kya:) kita:b (%kya:) PQP Anu-ERG PQP Uma=Acc PQP book.F PQP [di:]↑ (kya:)? give.Pfv.F PQP ‘Did Anu give a/the book to Uma?’ Bhatt & Dayal (2020: 1118) 1 I have glossed kya as Q in this dissertation signifying a question particle. Crucially, I make no commitments to kya being analyzed as Q-particle à la Hagstrom (1998) or Cable (2007). I employ Q as a glossing shorthand signifying a question particle. 7 The unmarked position for kya however is the clause initial position for a neutral polar question interpretation. The clause final position comes a close second. I will not discuss here the complete distributional properties of kya (Refer to Bhatt & Dayal 2020 for a detailed overview). Bhatt & Dayal (2020) and in some earlier research in Syed & Dash (2017) propose an interrogative ForceP which takes a CP complement in polar questions in Hindi. An operator in ForceP which has an interrogative force Q has been argued for polar questions in Bailey (2013) as well. Bhatt & Dayal associate ForceP[+Q] with rising intonation which is obligatory to form a polar question. In Syed & Dash (2017), we have argued that the interrogative operator in ForceP was also specified with an open polarity feature, [± Polarity]. In that paper we argued for a Q-particle style analysis (Hagstrom 1998, Kishimoto 2005, Cable 2007) wherein the operator in ForceP c-commands the question particle kya which has been merged as adjoined to a focus projection. The operator establishes an AGREE relationship with the question particle. The schematized analysis for polar questions by Syed & Dash is given in (12) while that by Bhatt & Dayal is given in (13). 12. [ForceP Op[+Q, ± Polarity] [FocP kya [FocP [ IP … ]]]] 13. [ForceP kya: [CP C 0 [+Q] [TP … ]]] Bhatt & Dayal argue against a Q-particle style treatment for the Hindi kya mainly due to restrictions on its embeddability and non-occurrence in non-polar questions. Polar questions with kya can only be embedded under rogative predicates like ask, wonder etc. but not responsive predicates like know. Further, in Hindi, kya is attested only in polar questions and alternative questions but not in wh-constituent questions. They draw a distinction between Hindi kya which they term as a polar question particle and Q-particles like Japanese -ka/-no which they allow to be an overt realization of all C[+Q] as it occurs in all interrogatives and has no embeddability 8 restrictions. They argue that kya is not on overt realization of the interrogative operator at C unlike Japanese -ka/-no, it is instead merged in the Force projection in the left periphery. In both the analyses (Syed & Dash 2017 and Bhatt & Dayal 2020), the interrogative operator present is distinct from kya and ensures a polar interrogative interpretation. It is not the focus of this dissertation to argue in favor of one analysis versus the other. For our purposes it will be sufficient to adopt that there is some formal mechanism in the C domain that interprets the proposition expressed as a polar question such that the alternatives generated as the answer set are {p, ¬p}. Following Hamblin (1958) questions are interpreted as the disjunctive set of alternatives which form the set of possible answers to the question. I further adopt Holmberg (2015) which proposes that a formalization of question-answer pairs would capture the following intuitions: 14. i. answers to questions, even the simple yes or no are full clauses. ii. questions and answers are related in an intrinsic way and answers should have a syntax derivable from the syntax of questions. iii. all kinds of questions: polar questions, alternative questions and wh-constituent questions should share a similar structure. He proposes that all questions essentially contain a free variable that he refers to as the question variable. He further posits that the question variable is a disjunctive set of alternatives. Thereby the question is a proposition with a free variable which is equivalent to a set of propositions, restricted by the range of the variable. He further posits that the answer assigns a value to the variable thereby picking out one of the alternative propositions to be true. This formalism captures all the insights mentioned in (14). Since all answers are one of the propositions mentioned in the alternative set, they are all full clauses with fragment answers having undergone some sort of ellipsis. Furthermore, the relationship between a question and an answer is transparent, the 9 question consists of a free variable whereas its answer provides a value to the variable with everything else being the same. Finally, the different kinds of questions are all argued to have a question variable, with the only difference being in the values that could be assigned to the variable. I follow Holmberg in the formalization of a polar question. Every polar question has a polarity head with the value [+Pol] or [-Pol] left open, a question variable restricted to two values. Holmberg posits that the polarity projection is merged as the highest head in IP, taking IP as its complement. In chapter 2, I present a variation to this analysis that I argue for Hindi but in principle, I do adopt that there is a polarity feature in every declarative sentence and a polarity variable in all polar questions hosted by a polarity head and which extends a polarity projection in the clausal structure. Holmberg further posits that the polarity variable moves to spec,CP position to attain sentential scope so that the denotation of the question would be equivalent to the alternative set of propositions, {p, ¬p} for neutral polar questions. He then argues for a Q-force feature which denotes ‘tell me the value of the variable in spec,CP such that p, expressed by PolP is true. So essentially, a question has three components: 15. i. a question variable - in polar questions the range is limited to two values {p, ¬p}, in alternative questions the range is overtly specified as a disjunctive set of two or more items and in wh-constituent questions the range is infinite in principle but can be restricted by lexical and pragmatic considerations. ii. movement of the question variable to the C domain to get sentential scope so that the question can be interpreted as a set of alternative propositions. iii. in direct questions a Q-force feature which takes a CP complement and is interpreted as a request to the addressee to assign a value to the question variable. 10 Holmberg’s account of polar questions is similar to other accounts which have an interrogative operator in the C domain (Bailey 2013, Syed & Dash 2017 and Bhatt & Dayal 2020 among others for Hindi polar questions) but a crucial addendum is the question variable that Holmberg proposes primarily to account for the similarity in the structure of questions and answers. I will adopt Holmberg’s account for this dissertation. A polar question can thereby be formally schematized as in (16) as per Holmberg (2015). 16. Q: Did Raghav read a book? Syntactic Representation = [ForceP Q [CP Did [± Pol] [CP [PolP Raghav [± Pol] [TP Raghav read a book]]]]] Polar questions in Hindi are primarily similar to that in English but with certain crucial differences. A major difference is that I motivate the polarity phrase in Hindi to be IP-internal. Arguments for this proposal are detailed in Chapter 2 of this dissertation. Furthermore, as we noticed Hindi optionally has a question particle kya in polar questions and does not attest subject auxiliary inversion. The schematization of a polar question in Hindi is in (17). 17. (kya) raaghav-ne kitab paDHii?↑ Syntactic Representation = [ForceP Q [CP [± Pol] [CP [ IP [PolP [± Pol] [υP raaghav-ne kitab paDHii]]]]] 11 1.1.1 A note on kya The account for polar questions that I adopt from Holmberg (2015) argues for question particles in different languages occurring with polar questions to be an overt instantiation of the question variable [± Pol] which undergoes overt or covert movement to the C-domain. Holmberg’s main motivation for this treatment is that the particles that he discusses (among others, rǔu in Thai following Yaisomanang 2012) are disjunctive particles in the language. Different analyses for their occurrence in polar questions entail treating polar questions as disjunction of two clauses: p and ¬p with the second clause having undergone ellipsis leaving just the disjunction in the clause final position (for more detail on this account see Jayaseelan 2008). I will not adopt a disjunction + ellipsis account for polar questions in Hindi, the primary reason being that the question particle in Hindi is not a disjunction marker but rather homonymous with a thematic wh word kya meaning ‘what’ and there is no evidence that the neutral polar questions in Hindi are bi-clausal structures Figure 1: Polar Questions in Hindi 12 having undergone ellipsis. Furthermore, as mentioned earlier there is an argument to be made for kya being merged in the force projection rather than in the C head as a question variable as explicated by Bhatt & Dayal (2020). They explain the embeddability restrictions on kya by positing that kya is not hosted by the C head but rather in the Force projection. Rogative predicates can embed ForcePs and hence can embed kya structures but responsive predicates can only embed a CP thereby ruling out kya structures in main clauses with responsive predicates. Though a detailed account of the nature and syntactic positioning of kya is not the primary focus of this work, I do maintain from previous work (Syed & Dash 2017) that kya is a focus sensitive operator that needs to be in a c-commanding configuration with a focused constituent. The particle takes a focused constituent as an argument to generate focused alternatives over. In Holmberg’s account kya would take as its argument the question variable generating an alternative set with different values for the variable which form the possible answer set to the question. For now, I would simply posit that kya is merged in the extended C domain and leave question as to the specific host of kya open. I do agree that any analysis of kya would need to account for the following properties: 18. i. kya is optional and only occurs (at least overtly) in polar and alternative questions. kya is not attested in wh-constituent questions in Hindi. ii. questions with an overt kya cannot be embedded under responsive predicates. iii. kya seems to demarcate the clauses into two halves 2 . To the left is the presupposed not-at-issue content and to the right of kya is the at-issue content, i.e., any constituent to the right of kya can be questioned 3 . If the at-issue constituent is not a clausal phrase but a constituent they will essentially bear focal stress. 2 Clause medial kya will be discussed in detail in Chapter 4, section 4.3. 3 This claim has been disputed to certain extent in the literature. In Syed & Dash (2017) we claim that the constituent to the immediate right of kya is its focused argument and hence the at-issue content. Bhatt & Dayal (2014) claim that anything to the right of kya can be questioned. Later in Bhatt & Dayal (2020) they extend the scope of kya to include 13 In this section I have given a brief overview of polar questions and have adopted Holmberg’s (2015) account of the same in Hindi. Holmberg’s account allows me to be faithful to the fundamental principle in this dissertation, that of the structure of answers to be parallel to the structure of questions. After having laid out the structure of polar questions, in the next section I introduce the different forms of fragment answers attested cross-linguistically to polar questions. Fragment Answers to Polar Questions As mentioned earlier, the specific way that pertinent information is codified in fragment answer forms and the structure of fragment answers in general is subject to a great degree of cross- linguistic variation as we can see in (8, 19-24), but Holmberg (2015) gives a similar ellipsis account for all types of answer fragments to polar questions. 19. Q: on-ko marja voinut kantaa tuon kiven yksin rannalta has-Q Marja could carry that stone from shore ‘Could Marja have carried that stone from the shore alone?’ 20. A: on (voinut (kantaa)) has could carry ‘Yes.’ - Finnish (Holmberg 2015) 21. Q: all mair aros can mair stay ‘Can Mair stay?’ the immediate left constituent and everything to its right. Biezma et al. (2017) maintain that though kya is a focus sensitive operator that in an unmarked structure can take as its focused argument anything to its right, prosody can overcome this restriction. With focal stress, any constituent to the left of kya can also be questioned as well. All accounts however agree that with neutral intonation only the constituent to its right comes under the natural scope of kya. The ability of other constituents being under the scope of kya is dependent on prosody. 14 22. A: gall (*aros) - ‘Verb Echo’ can stay ‘Yes.’ - Welsh (Holmberg 2015) 23. Q: ban co hoc hanh cham chi khong friend CO study diligently NEG ‘Did you work hard?’ 24. A: cham chi diligently ‘Yes.’ - Vietnamese (Simpson 2014) Answer fragments to polar questions differ across languages, and each language can potentially have more than one way of answering a polar question. As we see in (8), polar questions in English are usually answered with either just the polarity particle or with a polarity particle along with the subject in the pronominal form and the verb. Another way of answering a polar question is by simply repeating the finite verb. As can be seen in (20), a polar answer in Finnish is formed by repeating the structurally highest verb from the question. The other verbal elements can also be optionally repeated. Welsh in (22) is different from Finnish in the sense that only the highest verbal element can form an answer fragment and having multiple verbal elements is ungrammatical. Holmberg (2015) defines the pattern attested in Welsh and Finnish where the answer is the structurally highest verbal element, as a ‘verb echo’ pattern. Simpson (2014) observes an interesting way of answering polar questions with an adjunct. The fragment answer then simply constitutes of the narrowly focused adjunct as attested in Vietnamese in (24). Yet another pattern 15 of answering polar questions is illustrated below with the Hindi example in (25-26). The unmarked form of answering a polar question in Hindi is the polarity particle with the verbal sequence. 25. Q: (kya) raaghav-ne kitaab paDhii? Q Raghav-ERG book read.PFV.3Fsg ‘Did Raghav read the book?’ 26. A: haaN paDhii yes read ‘Yes.’ - Hindi Holmberg (2015) among others argues that the answers to polar questions have unpronounced structure, i.e., they are full-fledged clauses. Even just the polarity particles as answers have full clausal structures. A pattern hitherto unattested in Holmberg (2015) and in other works to the best of my knowledge is the obligatory presence of the entire verbal sequence as an answer. This pattern is attested in Hindi as can be seen in (28). Henceforth, I use the term ‘verb stacking’ for this pattern. The pattern noted in (28) is different from that noted in Finnish in (20) as in the Finnish example the stacking is optional whereas the answer in Hindi obligatorily has to contain all the verbal elements present in the question and that too in the same order. 27. Q: (kya) raaghav kitaab paDh paa rahaa thaa? Q raaghav book read able PROG.3Msg AUX.PST.3MSg ‘Was Raghav able to read the book?’ 28. A: (haaN) paDh paa rahaa thaa - ‘Verb Stacking’ yes read able PROG.3Msg AUX.PST.3MSg ‘Yes.’ 16 At first glance, it appears that only the arguments present in the question are missing from the fragmented answer in Hindi (26) and (28). There are different syntactic mechanisms that can potentially be employed to account for the missing structure like pro drop of arguments, arguments ellipsis, larger ellipsis of different constituent sizes. Hindi is a language which allows pro-drop of both the subject and the object and hence the answer forms could have a pro drop account, with no ellipsis involved. Hindi also allows for argument ellipsis, and that could also be used to account for the attested answer patterns. However, in this paper, I argue for a υP ellipsis account for (28), wherein the subject stays in spec,υP and the verb moves out of υP giving rise to verb stranding υP ellipsis. The ellipsis of the υP is triggered by an ellipsis feature à la Merchant (2005); E ans on a polarity head which in Hindi clausal structure takes υP as its complement. Primarily, in this thesis, I argue for an account of fragment answers in Hindi adopting the idea that the fragments are clausal in nature with unpronounced structure attained by ellipsis of specific constituents. I argue against a pro drop and an argument ellipsis approach though these are independently attested in the language. I propose a polarity phrase to be able to account for the attested answer patterns. Adopting from Laka (1990), Rizzi (1997) and Holmberg (2015) among others that there is a polarity head higher in the I or the C domain, I argue for a structure with two distinct positions for polarity. The polarity head thus can be merged either in the lower IP-internal position or in the higher position. It is merged with an ellipsis feature Eans which triggers the ellipsis of the complement of the head. When in the higher position, the polarity head takes IP as its complement, thereby resulting in IP ellipsis when the head is merged with the ellipsis feature. The polar polarity particle as an answer is then derived via IP ellipsis. ‘Verb stacking’ as seen in (27), I argue, is the result of υP ellipsis which is obtained when the polarity head with the ellipsis feature is merged in the lower IP-internal position leading to the deletion of the υP complement. The 17 analysis for fragment answers to neutral polar questions is detailed in Chapter 2. But before we delve into our analysis, in the next section I present a literature review of the two major work that form a cornerstone for my analysis. 1.3 Literature Review With regard to the parallel mapping of the clausal syntactic structure to propositional semantic interpretation in answers and their further parallelism with the structure of questions; the two seminal works have been Merchant (2005) and Holmberg (2015). Merchant (2005) explores a movement and ellipsis approach to account for fragment answers to wh-constituent questions. Merchant (2005) introduces the idea of an Ellipsis feature E which could be of different types depending on the size and other properties of the unpronounced structure. When hosted by a head it results in the ellipsis/non-pronunciation of the complement of the head. Holmberg (2015) makes a more radical claim by arguing that the polarity particle as an answer to a polar question has a propositional interpretation and maps on to a clausal syntactic structure, as do the other types of fragment answers to a polar question. The account proposed here for ‘verb stacking’ in Hindi polar fragment answers draws its influence from both these works, which are briefly described in the following sub-sections. 1.3.1 Merchant (2005) Merchant (2005) following Ginzburg and Sag (2000) argues that there is unpronounced structure in the fragment answers to wh-constituent questions like (29). 29. Q: Who did she see? A: John. 18 Merchant (2005) provides evidence from grammatical dependencies he calls ‘connectivity and anti-connectivity effects’ to argue for a clausal structure underlying what overtly looks like a DP constituent. There is evidence from case and agreement, scope and binding, licensing of polarity items, distribution of pronominal elements and other syntactic properties that the fragment has moved from its argument position to a higher position in the C domain and a certain part of the clause has been elided. The structure argued for (29) is given in (30). 30. Essentially, the fragment answer John has a clausal structure of [John [she saw t2]] where John has been moved to spec,FP and then TP/IP has been elided. Merchant (2005) unifies all types of ellipsis by positing an ellipsis feature E, here hosted by the functional projection F (which could be argued to be FocusP). This ellipsis feature can potentially be hosted by any head and licenses the head to elide its own complement. The function of the ellipsis feature is to make sure the complement though phonetically unpronounced is semantically interpreted. Ellipsis is only possible when the unpronounced structure has semantic recoverability. Sluicing in English is thereby accounted for by a sluicing ellipsis feature with the given specification: E s [uwh*, uQ*], merged with the C head. Whereas fragment answers to an English wh-constituent question is Figure 2: Merchant (2005) 19 triggered by another type of an ellipsis feature, Ef [uF*] merged on the Focus head in the C domain. The * on the ellipsis feature translates to an EPP feature wherein the head merged with the E feature has to have an occupied specifier resulting in the movement of the fragment from its base position while the E feature in itself triggers the deletion of the complement of the head. Hence the fragment answers are given a movement and ellipsis account to derive a constituent fragment from a clausal structure. 1.3.2 Holmberg (2015) Holmberg (2015) extends Merchant’s (2005) idea that fragment answers are full clausal structures to answers of polar questions. Essentially, he argues for a clausal analysis for (32) and (34). 31. Q: Is John coming? 32. A: Yes.  Yes [John is coming]. Yes, he is.  Yes [he is coming]. 33. Q: tul-i-vat-ko lapset kotiin? come-PST.3Pl.Q children home ‘Did the children come home?’ 34. A: tul-i-vat  tulivat [lapset <tulivat> kotiin] come-PST-3Pl ‘Yes.’ -Finnish (Holmberg 2015) One piece of evidence for this analysis comes from the tensed verb which suggests that the answer is not merely the verb fragment by itself, as it bears tense which is argued to be a clausal/propositional property. Arguments that the polarity particle in (32) has a clausal structure are not as clear but the intuition that it has a propositional interpretation bears on its syntactic analysis. 20 For Holmberg (2015) polar questions have a question variable [±Pol] with two possible values, i.e., the polarity of the proposition mentioned in the question is a variable, it could either be affirmative or negative. This variable basically generates a disjunctive set of propositions. The polarity variable, to get a sentential scope, moves to the C domain and is bound by the interrogative operator/Qforce which essentially ‘asks’ that the variable receive a value. The schematization of a polar question in English is repeated in (35) with answer structure schematization in (36). 35. Q: Did Raghav read a book? Syntactic Representation = [ForceP Q [CP Did [± Pol] [CP [PolP Raghav [± Pol] [TP Raghav read a book]]]]] 36. A: i. Yes, he did (read a book). Syntactic Representation = [ForceP Q [CP-FocP Yes [+Pol] [CP [PolP He [Pol’ did [+Pol] [TP He did read a book]]]]]] ii. No, he did not (read a book). Syntactic Representation = [ForceP Q [CP-FocP No [- Pol] [CP [PolP He [Pol’ did not [- Pol] [TP He did not read a book]]]]]] The answer then assigns a value to this variable which would ensure that the resulting proposition is true. This value is encoded in various ways, one of which is by the focused polarity particle yes/no in (32) or the highest verbal element in (34). Since the structure provided by the questions is used by the answer and once the variable in the question has been given a value by the focused feature in the answer, much of the parallel structure in the answer can go unpronounced. This leads to an ellipsis structure for the answer under identity. In (34), the value of the polarity variable is affirmative/true/[+]. This focused value is merged in the C domain and requires a lexical 21 exponence. The lexical exponence can either be provided by the overt polarity particle ‘yes’ like in (32) or the highest verb/aux element in the clausal spine moves up to the C domain to lexicalize the polarity value like in (34). In both cases, the IP is elided in the answer under identity with the question structure. The longer fragment in (32) ‘yes, he is.’ is derived by a VP ellipsis account. Although, it is unclear in Holmberg (2015) how a lexical exponence analysis would account for the longer fragment. An attempt is made to address this issue of lexical exponence with Hindi answers in the present work. In the next chapter, I provide a detailed synopsis of fragment answers to neutral polar questions in Hindi and argue for a verb stranding υP ellipsis approach to account for the same. Essentially, I modify Holmberg’s analysis to better fit the Hindi answer patterns thereby extending the typological overview for fragment answers to polar questions cross-linguistically. 22 Chapter 2: Fragment Answers in Hindi: An Account of ‘Verb Stacking’ In the previous chapter we observed how the Hindi fragment answer patters are different from the answer patterns so far attested in the literature. We crucially identified the obligatory verb stacking pattern which is different from the verb echo pattern described in Holmberg (2015). Any cross- linguistic account for fragment answers should be able to account for both these patterns as well as motivate the difference between them. An account of fragment answers in Hindi should ideally be able to account for obligatory verb stacking when the verbal sequence does not form a complex head and does not raise higher than IP. We also need to account for the fact that IP ellipsis is independently attested in the language but as we do not see verb echo patterns, fragment answers are not derived by IP ellipsis. Yet another empirical observation that needs to be accounted for is that in Hindi the verb stacking pattern is the unmarked answer pattern whereas answering a polar question simply with the polarity particle is a marked response. As mentioned in the previous chapter, I propose that ellipsis in fragments answers is triggered by a Merchant (2005) style Ellipsis feature hosted by the polarity head which causes the complement of the polarity head to be elided under semantic identity. The default polarity head in Hindi is merged IP-internally just above the ʋP leading to the deletion of ʋP after the main verb has raised out of it resulting in the verb stacking pattern. I show that this fragment answer pattern follows from an independently motivated ellipsis process of υ-stranding υPE (Bhatt & Dayal 2007, Manetta 2019) in Hindi. A higher polarity phrase (merged above IP) in other languages results in the different ‘verb echo’ (Holmberg 2015) pattern exhibited in those languages (Irish (McCloskey 2017), Finnish (Holmberg 2015), Welsh 23 (Holmberg 2015), Russian (Gribanova 2017), Mandarin (Wei 2016) and Japanese (Sato et al. 2018)). Evidence from question-answer pairs in typologically different languages is presented as support for the view that polarity phrases may actually occur in different positions within sentences in different languages. I propose that this variation in the physical location of a polarity phrase directly results in differences in the amount of sentential material that can be left unpronounced in an answer, producing fragments of differing types and sizes as semantically parallel answer forms. Thereby in principle all the answer needs is the valued polarity feature in a language and as soon as that requirement is fulfilled everything else gets elided. Hence the position of the polarity head in a clause determines the size of ellipsis and consequently the size of the answer fragments. This paper has the following sections: Hindi data and the empirical observations are presented in Section 1. Section 2 then details various competing analyses for Hindi, which potentially could account for the attested answer patterns and the reasons they are rejected. Section 3 charts out the proposal and the account for fragments answers to polar questions in Hindi in detail. Section 4 concludes. 2.1 Hindi Data and Observations Polar questions in Hindi can be answered in various ways as can be seen in the following paradigm. 1. Q: (kya) raaghav-ne kitaab paDhii? Q Raghav-ERG book read.PFV.3Fsg ‘Did Raghav read the book?’ 24 2. PosA: i. ?haaN Yes ii. haaN paDhii yes read.PFV.3FSg iii. haaN raaghav-ne kitaab paDhii yes raaghav-ERG book read.PFV.3FSg iv. #paDhii - ‘Verb echo’ read.PFV.3FSg v. #raaghav-ne kitaab paDhii raaghav-ERG book read.PFV.3FSg ‘Yes’. NegA: vi. ?nahii no vii. nahii paDhii no read.PFV.3FSg viii. nahii raaghav-ne kitaab nahi paDhii no raaghav-ERG book NEG read.PFV.3FSg ‘No.’ The unmarked answer form of a polar question is the polarity particle + verb form instantiated in (2ii, vii). Another way of answering is of course simply stating the polarity particle as in (2i, vi). These two forms are called fragments as the entire clausal structure is not present/pronounced in the answer. An answer with the entire clause repeated from the question is also possible as shown in (2iii, viii). The focus of this paper is to give an analysis of the fragment answers attested in Hindi. An attempt has been made to account for the acceptability of (2i, ii, vi, vii) as answers to a polar question and at the same time account for the unattested patterns like (2iv, v). The empirical 25 observation is that the answer has to have an overt polarity particle and that the unmarked form is the polarity particle + verb from. Expanding the generalizations on fragment answers, if the question contains more than one verbal element in the verbal sequence, then the fragment answer obligatorily has to contain all the verbal elements in the same order has given in the question, a pattern I refer to as ‘verb stacking’. 3. Q: (kya) raaghav kitaab paDh paa rahaa thaa? Q raaghav book read able PROG.3Msg AUX.PST.3MSg ‘Was Raghav able to read the book?’ 4. PosA: i. ?haaN yes ii. (haaN) paDh paa rahaa thaa - ‘Verb Stacking’ yes read able PROG.3Msg AUX.PST.3MSg iii. haaN raaghav kitab paDh paa rahaa thaa yes raaghav book read able PROG.3Msg AUX.PST.3MSg iv. #paa rahaa thaa able PROG.3Msg AUX.PST.3MSg v. #rahaa tha PROG.3Msg AUX.PST.3MSg vi. #padh read vii. #thaa - Verb echo (Holmberg 2015) AUX.PST.3MSg ‘Yes.’ 26 NegA: viii. ?nahii no ix. nahii paDh paa rahaa thaa no read able PROG.3Msg AUX.PST.3MSg x. nahii raaghav kitaab nahi paDh paa rahaa thaa no raaghav book NEG read able PROG.3Msg AUX.3Msg xi. #nahii paa rahaa thaa no able PROG.3Msg AUX.PST.3MSg xii. #nahii rahaa tha no PROG.3Msg AUX.PST.3MSg xiii. #nahii padh no read xiv. #nahii thaa no AUX.PST.3MSg ‘No.’ Hence, we observe that the unmarked form of an answer to a polar question in Hindi is the polarity particle + verbal sequence (2ii, 4ii). We attempt an analysis of this pattern of ‘verb stacking’ as an answer form. Notice that the ‘verb stacking’ is obligatory and complete. Any answer form which does not have the parallel, complete verbal sequence to the question is unattested as can be seen in (4iv-vii, xi-xiv). Use of just the polarity particle can also be a fragment answer (4i and viii). Interestingly, using only the highest verb/aux is not an acceptable fragment answer in Hindi (4vi, xiv) unlike in ‘verb echo’ languages like Finnish, Welsh, Mandarin etc. (Holmberg 2015). An interesting difference noted here is when we compare (2ii) and (4ii). The polarity particle in (4ii) 27 seems to be optional 4 though when it is unpronounced the answer is followed by a gestural nod. However, no amount of gestural nod can lead to the polarity particle in (2ii) being silent. This issue and the implications it has on the proposed account is discussed in in the section 5.3.1. The empirical observations attained from the data presented above in (2) and (4) is (i) a polarity particle has to be overtly realized in a fragment answer to polar questions in Hindi 5 , (ii) the entire verbal sequence has to be repeated in the answer and (iii) native speaker intuitions suggest the polarity particle + verbal sequence to be the unmarked form. In the upcoming sections, I present an analysis that can explain these empirical facts. Before the proposed analysis is presented, I discuss certain potential analyses which at first glance, could potentially account for the verb stacking pattern observed in (4ii, ix) but eventually fall short. 2.2 Potential Analyses When we compare the unmarked answer pattern attested in (4ii) to the polar question in (3) – repeated below in (5) and (6), we realize that the only elements missing from the answer are the argument DPs as the answer constitutes of just the verbal sequence. A number of syntactic mechanisms could account for these missing arguments. The four mechanisms considered in this section are pro drop, argument ellipsis, verb stranding VP ellipsis (for missing object DPs) and IP ellipsis. All the four phenomena mentioned here are independently attested in Hindi and hence could each potentially account for (6). 4 speaker variation has been observed with regards to this generalization. We would need certain experimental evidence to determine the robustness of this generalization. 5 If for now, we ignore speakers who can omit the polarity particle in cases exemplified in 4ii. 28 5. Q: (kya) raaghav kitaab paDh paa rahaa thaa? Q raaghav book read able PROG.3Msg AUX.PST.3MSg ‘Was Raghav able to read the book?’ 6. A: (haaN) paDh paa rahaa thaa - ‘Verb Stacking’ yes read able PROG.3Msg AUX.PST.3MSg ‘Yes’. In this section I review each of the four aforementioned phenomena and how they could give a potential analysis for (6). The conclusion that we reach is that none of these phenomena can satisfactorily account for the attested fragment answer patterns in Hindi and hence would need a different analysis than the ones mentioned in this section. 2.2.1 A Pro Drop analysis Under a pro drop analysis, (6) would have an account where the subject and the object DPs are pros – phonologically empty pronominal elements. Pro drop is independently attested in Hindi. Hindi could be argued to be a partial pro-drop language. Though there is rampant pro-drop, a 3 rd person DP with a specific referent cannot be replaced with a pro without a context providing an antecedent (comparing and contrasting Hindi with another Indo-Aryan language Marathi in Holmberg et al (2009)). 7. *(usne) aam khayaa he mango eat.PST.3MSg ‘He ate a mango.’ 29 However, it can be dropped if an antecedent (linguistic or non-linguistic) is available in the context as shown in (8). 8. Context: Raghav is looking for a book and Dharmesh knows that. 6 Q: [e] [e] milaa (kya)? find.PST.3MSg Q ‘Did you find it?’ A: [e] [e] DhUnD rahaa hun. [e] [e] pataa search PROG.1Msg AUX.PRES.2Msg know nahi kahaan rakh diyaa. Neg where keep give.PST.3MSg ‘I am looking for it. I don’t know where I kept it.’ [e] – null argument There doesn’t seem to be any person restrictions on the null pro or any (in)definiteness restriction. In (9) we see that a 3 rd person pronoun can be dropped in Hindi declarative sentences given there is an available antecedent in the context (unlike in Finnish – Holmberg et al (2009)) and in (10) we see that the null 3 rd person pronoun can have an arbitrary interpretation (unlike in Italian - Holmberg et al (2009)). 9. raaghav laaiibrarii me hai. [e] ghar der-se raaghav library in AUX.PRES.3Sg home late-P aayegaa come.FUT.3MSg ‘Raghav is in the library. He will come home late.’ 10. garmiyoN-me [e] jaldi uThte heiN summer-in soon wake-up.3MPl AUX.PRES.3MPl ‘One wakes up early in the summer.’ 6 Context inspired by a similar example for Thai in Holmberg (2015). 30 The empirical conclusion relevant for our purposes is that arguments can be phonologically null in Hindi in both the subject and object positions as long as there is a linguistic or contextual antecedent present. And this phenomenon of pro drop does not seem to be restricted by any person or interpretation specifications. Given these generalizations, (6) could have an analysis wherein the subject and the object DP are phonologically empty pros accounting for their non-overt appearance in the answer. It is possible as the question is always present to supply the linguistic antecedents for the missing arguments in the answer. The schematized structure is in (11). 11. Figure 3: Argument Pro Drop Analysis 31 Though this analysis could account for the answer pattern of questions like in (5), it fails to account for slightly more complicated polar questions. In (12) we have a polar question with an adverbial adjunct. The attested affirmative fragment answer in (13) does not have the adverbial adjunct overtly realized but the adverb is still interpreted in the answer. An account where only the DP arguments can be substituted by an empty pro cannot account for the inclusion of (specifically non-DP) adjuncts. We assume here that adjuncts, specifically non-DP adjuncts cannot be substituted by pros. The fact that adverb inclusion is obligatory when an adverb is present in a question is shown by the negative answer in (13) which exclusively negates the adverb. 12. Q: (kya) raaghav-ne acche-se kitaab paDhii Q Raghav-ERG well book read.PFV.3Fsg ‘Did Raghav read the book properly?’ 13. PosA: haaN paDhii Yes read.PFV.3Fsg ‘Yes (Raghav read the book well/carefully).’ # Yes (Raghav read the book). NegA: nahi jaldi-me paDhii no hurry-in read.PFV.3Fsg ‘No, he read the book in a hurry.’ Another piece of evidence against an argument pro drop analysis comes from missing quantificational arguments in the fragment answers. Phonologically empty pro cannot substitute quantificational DPs because an overtly realized pronoun cannot substitute for quantificational elements (example in 14), as quantificational elements lack a definite referent in the discourse. 32 14. Q: (kya) kisiii bachche-ne kitaab paDhii? Q some child-ERG book read.PFV.3FSg ‘Did some child read the book?’ 15. A: # 7 haaN us-ne kitaab paDhii yes him-ERG book read.PST.3FSg ‘Yes, he read that book.’ Consequently, if a question has quantificational DPs as arguments and those arguments are not overtly realized in the answer, the missing arguments cannot be due to pro drop. This is exactly what we see in (16) wherein the object is a quantificational DP. In the answer in (17) the object DP is not overtly realized but still interpreted in the answer, which suggests that the object DP is not a pro. 16. Q: (kya) raaghav-ne koi kitaab paDhii? Q raghav-ERG some book read.PFV.3FSg ‘Did Raghav read some book?’ 17. A: haaN paDhii yes read.PST.3FSg ‘Yes (Raghav read some book).’ Hence, we conclude that though Hindi independently attests rampant pro drop whenever an appropriate antecedent (contextual or linguistic) is present, all fragment answer patterns cannot be accounted for by this analysis. As the pro drop analysis cannot account for adjunct inclusion and 7 This will be an acceptable answer only when usne ‘him’ is interpreted deictically. 33 missing quantificational arguments in the answer, we reject this analysis. The next potential analysis that I consider is that of argument ellipsis. 2.2.2 An Argument Ellipsis analysis Certain East Asian languages (Oku 1998, Kim 1999, Takahashi 2006) attest the phenomenon of argument ellipsis wherein just the arguments of the verb are elided. The phenomenon of argument ellipsis is similar to pro drop in the sense that in both the cases the arguments are not overtly realized. It is primarily different from pro drop because argument ellipsis can account for sloppy interpretation of the argument whereas a pro drop account cannot. Simpson et. al (2013) have argued that Hindi also attests genuine argument ellipsis of its internal arguments. The evidence comes from empirical patterns which cannot be accounted for by the alternative analysis of verb stranding VP ellipsis. The availability of sloppy interpretations of null arguments has been taken to suggest that missing arguments are not pros but a part of an elided constituent (Otani and Whitman 1991, Sener and Takahashi 2009 among others), This is so, because pro is just the phonological null counterpart of an overtly realized pronoun and an overtly realized pronoun cannot have multiple referents hence preventing a sloppy interpretation. Hindi attests sloppy reading of objects which has been taken as potential evidence for verb stranding VP ellipsis like in (18). 18. amit apni premikaa-ko pyaar kartaa hai amiti selfi girlfriend-ACC love do.PRES.3MSg AUX.PRES.3 ‘Amit loves hisi/self’si girlfriend.’ 34 ravi [e] bhi pyaar kartaa hai ravij also love do.PRES.3MSg AUX.PRES.3 ‘Ravi also loves hisi/j girlfriend. ’ Strict ✓, Sloppy ✓ Simpson et al. (2013) However, verb stranding VPE has been argued to be possible only in cases where the verb in the antecedent and the ellipsis structure is identical (Goldberg 2005). Simpson et. al (2013) present evidence that sloppy interpretations of the missing object can be obtained even though the verb in the otherwise parallel structures is not the same, as shown in (19). 19. amit apni premikaa-ko pyaar kartaa hai amiti selfi girlfriend-ACC love do.PRES.3MSg AUX.PRES.3 ‘Amit loves hisi/self’si girlfriend.’ magar ravi [e] nafrat kartaa hai but ravij hate do.PRES.3MSg AUX.PRES.3 ‘But Ravi hates hisi/j girlfriend. ’ Strict ✓, Sloppy ✓ Simpson et al. (2013) Simpson et al (2013) present another piece of evidence for argument ellipsis. They argue that if the missing internal arguments could only be a result of verb stranding VPE, (20) cannot be accounted for by VPE. In (20) only the indirect object is missing but the direct object is overt. The way VPE works in Hindi is that the verb raises out to a VP external position prior to the ellipsis of the entire VP remnant. Both the direct and the indirect internal arguments should then be inside the elided constituent. However as shown, only the indirect object is missing in (20). Since there is no evidence that the direct object has moved out of the VP from its base position an ellipsis of 35 the VP would have resulted in the non-pronunciation of the direct object as well. Instead what is attested in Hindi in cases where there are two internal arguments, is that they can be silent independent of each other, i.e. one internal argument can be missing with the other being overt. This suggests that the ellipsis site is more restricted than originally assumed. This conclusion taken with the empirical observation that the independently missing internal argument can still have sloppy interpretation leads to a conclusion that the missing argument is not due to substitution by a pro but it is a more targeted ellipsis of the argument itself. 20. amit-ne apni premikaa-ko ek kitab dii amiti-ERG selfi girlfriend-ACC one book give.PFV.3FSg ‘Amit gave a book to hisi/self’si girlfriend.’ ravi-ne bhi [e] ek kitab dii ravij-ERG also one book give.PFV.3FSg ‘Ravi also gave a book (to hisi/j girlfriend). ’ Strict ✓, Sloppy ✓ Simpson et al. (2013) Hence, we see that these two facts can only be accounted for by an analysis wherein the missing internal arguments are a result of genuine ellipsis of the arguments and not of the entire VP. It being ellipsis, would account for the sloppy interpretation in (19) and (20) and it being ellipsis of the DP and not the entire VP would account for the fact that only one of the internal argument could be independently missing as in (20) and the verbs need not be identical as in (19). Given that argument ellipsis is independently attested in Hindi, the fragment answer in (6) could have an analysis as schematized in (21). Let us recall, that in the fragment answer in (6) the only elements missing are the arguments. 36 21. This analysis has an advantage over the pro drop analysis, as it can explain the missing argument DPs when they are quantificational in nature as in (14 and 16). There is no restriction on ellipsis of quantificational elements but there is a restriction on them being substituted by a pro. However, we reject this analysis on multiple accounts. First, Simpson et al (2013) argue that the subject in Hindi cannot undergo argument ellipsis. Their evidence comes from the fact that missing subjects never have a sloppy interpretation as can be seen in (22). Hence the missing subject in (6) cannot be due to the argument ellipsis of the subject. Figure 4: Argument Ellipsis Analysis 37 22. giita-ko lagtaa hai uske bete-ko puraskaar milegaa gita-DAT feel AUX.PRES.3 her son-DAT prize get.FUT.MSg ‘Gitai feels heri/k sone will win the prize.’ suniita-ko bhi lagtaa hai [e] milegaa sunita-DAT also feel AUX.PRES.3 get.FUT.MSg ‘Sunitaj also feels hei/*j/k will.’ Strict ✓, Sloppy  Simpson et al. (2013) Moreover, adverb inclusion as shown in (12 and 13) cannot be accounted for by simple argument ellipsis as schematized in (21) given that adverbs are not arguments and cannot undergo independent ellipsis like an internal argument. In (23ii) we see that only the adverb is missing whereas the direct object is present, in this case there is no adverb inclusion. We see that adjunct inclusion is only possible when the direct object is also missing (23i), thereby arguing for a verb stranding VP ellipsis to be the only possible account for adverb inclusion. 23. raam-ne chomsky-kaa nayaa lekh do baar paDhaa ram-ERG Chomsky-GEN new article two times read.PFV.3MSg ‘Ram read Chomsky’s new article twice.’ i. raaj-ne bhi paDhaa raj-ERG also read.PFV.3MSg ‘Raj also read (the article twice).’ ii. raaj-ne bhi vo lekh paDhaa raj-ERG also that article read.PFV.3MSg ‘Raj also read that article.’ Not communicated: twice. Simpson et al. (2013) 38 As seen in the previous two subsections, neither the pro drop nor the argument ellipsis account can explain adverb inclusion in fragment answers. However, verb stranding VP ellipsis could explain the attested pattern of adverb inclusion. In the next subsection we examine a VP ellipsis analysis. 2.2.3 A Subject pro drop + VP Ellipsis analysis This analysis is schematized in (24). The missing subject in (6) could be suggested to be a result of a pro in the subject position while the missing object results from verb stranding VP ellipsis in which the verb raises to some higher VP-external position prior to ellipsis of the VP. 24. Figure 5: Subj pro drop + VP Ellipsis 39 As stated earlier, the missing subject never has a sloppy interpretation and hence cannot be due to any sort of ellipsis. The missing object results in a sloppy reading and hence is a result of ellipsis. It is VP ellipsis and not argument ellipsis because the VP ellipsis can account for adverb inclusion whereas argument ellipsis cannot. This analysis is a distributed analysis in the sense that it has two independent syntactic operations applying together to derive the fragment answer pattern in (6). Verb stranding VP ellipsis is independently attested in Hindi (Bhatt & Dayal 2007, Simpson et. al 2013, Manetta, to appear) and so is subject pro drop as we have already discussed. This analysis could account for adverb inclusion as shown in (13) assuming that the adverb achche-se ‘well’ is a VP level adverb and hence falls inside the ellipsis site. It will also account for the inclusion of quantificational object DPs like in (17). However, it might seem that a verb stranding VP ellipsis analysis cannot explain the inclusion of quantificational subjects in the interpretation of the answer if they are not overtly realized as can be seen in (25-26). As already mentioned, a pro cannot substitute for quantificational DPs and here is where this account falls short. Thus, we see that the missing subject in a fragment answer cannot actually be due to pro drop just like a quantificational object. 25. Q: (kya) kisii bacce-ne kitaab paDhii? Q some child-ERG book read.PFV.3FSg ‘Did some child read the book?’ 26. A: haaN paDhii yes read.PST.3FSg ‘Yes.’ 40 Another issue with this analysis is adverb inclusion of higher adverbs. Adverbs which can take scope over the subject have to be higher than the VP level, they have to at the least attach at the υP level. Inclusion of such adverbs cannot be accounted for verb stranding VP ellipsis. In (27) we have the adverb phir-se ‘again’ which in the answer (28) is overtly not realized but the answer has a repetitive reading of ‘again’ which suggests that the subject is under the scope of the adverb (von Stechow 1996, Johnson 2004), thereby suggesting that the adverb is higher than the subject. The adverb inclusion in this case cannot be accounted for if the ellipsis site is lower than the position of the adverb. 27. Q: (kya) raaghav-ne phirse kitaab paDhii Q Raghav-ERG again book read.PFV.3Fsg ‘Did Raghav read the book again?’ 28. A: haaN paDhii yes read.PFV.3Fsg ‘Yes. (Raghav performed the action of reading the book again.)’ There is yet another wrinkle with a distributed analysis of the type mentioned here. Such an analysis has two independent phenomena occurring - subject pro drop and verb stranding VP ellipsis. There is no restriction whatsoever which would condition these two seemingly independent phenomena to always occur together, i.e., ideally subject pro drop should be able to apply without VPE and VPE should in turn be independent of subject pro drop. However, we never see answer patterns which exhibit such expected patterns. Examples are presented in (30). 41 29. Q: (kya) raaghav-ne kitaab paDhii? Q Raghav-ERG book read.PFV.3Fsg ‘Did Raghav read the book?’ 30. A: # haaN raaghav-ne paDhii - V stranding VPE yes Raghav-ERG read.PFV.3Fsg #haaN kitaab paDhii - Subject pro drop yes book read.PFV.3Fsg Intended interpretation: ‘Yes.’ It would be anticipated that fragment answers could produce patterns where one phenomenon applies but the other does not. The attested patterns, however, suggest that if indeed there are two different phenomena responsible for the patterns of fragment answers they have to apply together, which is an unmotivated stipulation and is unaccounted for. 2.2.4 An IP Ellipsis analysis Holmberg (2015) in his analysis for fragment answers argues for an IP Ellipsis (actually PolP ellipsis where Pol is the highest head in the IP) approach to account for ‘verb echo’ languages like Finnish and Welsh (examples from the introduction section repeated below). As can be seen in (32) and (34), the highest verb/aux can be a suitable answer fragment to a polar question. Finnish differs from Welsh in the way that a ‘verb echo’ pattern is optional in Finnish but obligatory in Welsh. The fragment answer in Finnish can have multiple verbal elements but in Welsh the answer has to be only the highest verbal element. 42 31. Q: on-ko marja voinut kantaa tuon kiven yksin rannalta has-Q Marja could carry that stone from shore ‘Could Marja have carried that stone from the shore alone?’ 32. A: on (voinut (kantaa)) has could carry ‘Yes.’ - Finnish (Holmberg 2015) 33. Q: all mair aros can mair stay ‘Can Mair stay?’ 34. A: gall (*aros) can stay ‘Yes.’ -Welsh (Holmberg 2015) Holmberg (2015) argues for a polarity phrase as the highest phrase in the IP domain, this polarity phrase hosts a [±Pol] variable in its head which moves up to the C domain to receive sentential scope. The variable when it receives a focused polarity value in the answer is either lexicalized by the polarity particle or the highest verbal element in the IP domain which has moved up to the C domain to provide lexical exponence to the valued polarity feature. The fragment answer is then accounted for by the ellipsis of the entire PolP. The analysis for (32) is schematized in (35). 43 35. In (35) the highest verbal element on ‘has’ initially moves to the Pol head and then to the C domain where it is used to lexicalize the focused polarity value [+Pol]. Then the entire PolP is elided under identity with the question thereby giving us just the highest verbal element as the fragment answer. As mentioned earlier, Hindi does not have ‘verb echo’ patterns. The attested fragment answer patterns are repeated below in (37). 36. Q: (kya) raaghav kitaab paDh paa rahaa thaa? Q raaghav book read able PROG.3Msg AUX.PST.3MSg ‘Was Raghav able to read the book?’ 37. A: i. haaN Yes ii. (haaN) paDh paa rahaa thaa - ‘Verb Stacking’ yes read able PROG.3Msg AUX.PST.3MSg Figure 6: Holmberg's PolP Ellipsis 44 iii. #paa rahaa thaa able PROG.3Msg AUX.PST.3MSg iv. #rahaa tha PROG.3Msg AUX.PST.3MSg v. #padh read vi. #thaa - Verb echo (Holmberg 2015) AUX.PST.3MSg ‘Yes.’ As per Holmberg (2015), (37i) would be derived by IP ellipsis with the polarity particle being merged in the C domain to provide lexical exponence to the valued polarity feature [+Pol]. However, (37ii-vi) cannot be explained with a similar analysis. Simply adopting Holmberg’s PolP ellipsis (2015) analysis in Hindi, we incur certain non-trivial issues. With this account there is no easy explanation for why the valued polarity feature cannot be lexicalized by the highest verbal element in the Hindi verbal spine (i.e., 37vi is not an acceptable answer form). Given that IP ellipsis is possible in general in Hindi 8 as sluicing is attested in Hindi (Bhattacharya & Simpson 2012) and in this case if the fragment answer in (37i) is assumed to have a clausal structure having undergone IP ellipsis, it is unclear as to why Hindi does not attest ‘verb echo’ patterns. Also, unlike Finnish, Hindi does not attest to the optional occurrence of more than one verbal element and thereby we cannot explain why (37iii-v) are unattested answer forms. There is no easy explanation for verb stacking being obligatory and it also being the unmarked answer pattern. I thereby conclude that the analysis discussed in Holmberg (2015) for Finnish and Welsh cannot be applied as it is for Hindi fragment answers. 8 refer to Chapter 5, Section 5.2 of this dissertation for a review of independently attested IP ellipsis. 45 One way to give an IP ellipsis account of obligatory verb stacking attested in Hindi is to argue that all the verbal elements have moved higher than IP, as a complex head. The verbal sequence would be hosted in the highest verbal head position (presumably I) and then move higher to the C domain to provide lexical exponence to the valued polarity feature. Then PolP/IP ellipsis effectively elides arguments and adjuncts and the verbal sequence has moved higher than the ellipsis site yielding the attested pattern (as in 37ii). However, there is no empirical evidence supporting the idea that all the verbal elements form a complex head in Hindi. In fact, the verbal elements can be broken up as shown in (38). There seems to be little motivation for all the verbal elements in the spine to move higher than the I head so that IP ellipsis would result in a verb stacking pattern. To say then that this complex head formation is obligatory in answers but not in usual declarative contexts would feel like a stipulation. 38. raaghav kitaab paDh paa rahaa thaa raaghav book read able PROG.3Msg AUX.PST.3MSg ?raaghav paDh kitaab paa rahaa thaa raaghav read book able PROG.3Msg AUX.PST.3MSg ?raaghav paDh paa rahaa kitaab thaa raaghav read able PROG.3Msg book AUX.PST.3MSg One motivation for an IP ellipsis account would have been that subject in Hindi obligatorily raises to spec,IP. If the subject in Hindi is higher up in spec,IP ellipsis of any constituent lower than IP will not be able to account for the missing subject in the fragment answers. We have already looked at arguments as to why the missing subject cannot be derived by any other mechanism like pro 46 drop or argument ellipsis. To recall, missing quantificational subjects cannot be due to pro drop and can only be accounted for by ellipsis and subjects in Hindi cannot undergo argument ellipsis. It is not unreasonable to assume that the subject is in spec,IP. In many languages, subjects move to spec,IP from spec,υP either for the purposes of agreement or to satisfy an EPP feature on the I head when Agree is possible in-situ. However, Bhatt (2005) extensively argues that the subject in Hindi need not move to spec,IP at least for the purposes of agreement. Moreover, there has been extensive debate on the (non)-existence of EPP features in OV languages (for a detailed review see Haider 2013). One piece of evidence comes from the fact that OV languages in general do not attest expletive subjects. A potential way to justify that would be to suggest that there is no EPP feature on the I head, so the spec,IP position can stay empty. Thus, we could argue that the subject does not need to raise to spec,IP as there is no EPP feature on I triggering this particular movement. Hindi itself is an OV language which does not attest any expletive subjects. Since subject in Hindi does not raise to spec,IP for agreement purposes and assuming there is no EPP feature on I, there seems to be no empirical observation or motivation to claim that subject in Hindi has to raise up to spec,IP. Hence, the argument that IP ellipsis is a potential analysis for fragment answers in Hindi to account for the missing subject in answers does not hold as there is no clear evidence that the subject in Hindi has to be at spec,IP and thereby only IP ellipsis could potentially account for the missing subject. Concluding this section, we have examined various possible syntactic mechanisms like pro drop, argument ellipsis, pro drop+VP ellipsis and IP ellipsis which could potentially explicate the missing elements in a fragment answer. And we have discussed how all the analyses detailed above fail to account for the attested answer patterns. The empirical conclusion from the attested patterns seems to require us to look for an analysis which would involve ellipsis of a larger constituent so 47 that it could explain inclusion of higher-level adverbs and (non)-quantificational arguments in both the subject and object position. The adopted analysis preferably should also not have to rely on all the verbal elements forming a complex head. Finally, it should be able to account for the unattested fragment answer patterns. The analysis that I argue for, and which fulfils all these criteria, is developed in the next section. 2.3 Proposed Analysis: a υP Ellipsis Approach In this section, I develop an analysis of Hindi fragment answers which attributes all the attested patterns to an operation of V raising υP ellipsis. I argue for the ellipsis size to be larger than VP, it is actually at the level of little υP. Ellipsis of υP instead of VP allows us to have the subject inside the ellipsis site and furthermore can also explain the missing higher-level adverbs which scope over the subject as they too can be under the scope of the targeted ellipsis constituent. The only way that we can argue that ellipsis of υP elides the subject as well is if the subject in Hindi can stay in spec,υP. Further, we need independent evidence that the verb in Hindi moves out of υP to a position higher than υ prior to ellipsis. Only then can we account for the fact that all the verbal elements are outside of the scope of the ellipsis site. Hence for this proposal it is relevant to independently motivate that the subject in Hindi can stay inside υP and the verb moves out of it. I give arguments for each in the next subsections before I present the proposal. 48 2.3.1 Subject in spec, υP McCloskey (1997), in his overview article on the properties and positions of subjects cross- linguistically, presents scopal evidence in favor for the subject potentially occupying a position lower than spec,IP by following Ladusaw (1992). Ladusaw (1992) argues that the relative scope of modals, negation and adverbs is straightforwardly determined by their surface position, as these elements cannot undergo QR. A subject which overtly occupies a higher position than such elements, but scopes lower would then provide evidence for a lower origin. Following Miyagawa (2001) and Miyagawa & Arikawa (2007) who present relevant paradigms for Japanese, we examine independent empirical evidence that the subject in Hindi indeed might stay in spec,υP. Miyagawa & Arikawa (2007) argue that Japanese has an EPP condition which can be fulfilled either by the subject or the object. When the object moves to spec, IP the subject essentially stays lower in spec,ʋP and that is clear in the OSV scrambled constructions as in those constructions the subject can’t take scope over negation which it otherwise can when the subject is in spec, IP and the object stays lower. In (39) we see that the subject with the universal quantifier zen’in ‘all’ overtly occupies a higher position than the negation and even scopes outside of negation. The ‘not>>all’ reading is not easily available in this case as reported in the paper. However, in (40) wherein the object has scrambled to a higher position than the subject, the ‘not>>all’ reading becomes readily available. Miyagawa & Arikawa (2007) take this as evidence for the subject to originate at a lower spec,υP position and then move to a higher position over negation to spec,IP to satisfy the EPP feature on I. When the object moves to spec,IP to satisfy the EPP feature like in 49 (40), the subject ends up staying in its base position under the scope of negation thereby making the ‘not>>all’ reading available. 39. zen’in tesuto-o uke-na-katta all-NOM test-ACC take- NEG -PAST ‘All did not take the test.’ *not>>all, all>>not 40. tesuto-oi zen’in-ga ti uke-na-katta test-ACCi all-NOM take- NEG -PAST ‘That test, all didn’t take.’ not>>all, all>>not Japanese – Miyagawa & Arikawa (2007) In (41) we examine a Hindi example to figure out the scopal interaction between the universal quantifier sab ‘all’ in the subject position and sentential negation. The only reading available in this case is the ‘not >> all’ reading signifying that the quantifier subject occupies a lower position than negation 9 . Furthermore, in Hindi subject NPIs can be licensed by negation as can be seen in (42). Given that negation has to c-command an NPI to be able to license it, it can be argued that negation scopes higher than the subject NPI. 41. sab-ne kitaabein nahii paDhiiN all-ERG book.Pl NEG read.PST.3FPl ‘Everybody didn’t read the books.’ not >> all 9 Note that certain quantifiers (‘only’) always scope higher than negation, irrespective of whether they are in the subject or the object position. The assumption here would be that somehow only quantifier always raises to a higher position. 1. sirf teen larko-ne kitaab nahi paDhii only three boys-ERG book NEG read.PFV.3FSg ‘Only three boys didn’t read the book.’ only>> NEG , * NEG >>only 2. larko-ne sirf teen kitaabeiN nahi paDhiiN boys.ERG only three books NEG read.PFV.3FPl ‘The boys didn’t read only three books.’ only>> NEG , * NEG >>only 50 42. ek-bhi larke-ne kitaab nahi paDhii one-even boy-ERG book NEG read.PFV.3FSg ‘Not even a single boy read the book.’ Bhatt & Homer (2014) take (42) to be evidence of negation being high up in the Hindi clause. However, there is no clear evidence that it has to be the case. In fact, the linear position of negation in Hindi is such that it has to be adjacent to the main verb (Mahajan 1988). Examples (43 -45) show that negation in the unmarked order is always to the left of the verb and in some marked cases where there is additional material to the right of the verb it can also be right of the verb. Essentially, it forms a complex head with the main verb as nothing can occur in between negation and the main verb. Even very low adverbs like ‘easily’, ‘quickly’ etc. cannot intervene between negation and the verb. Assuming that all the verbal elements do not form a complex head and raise to I, we can conclude that negation is attached lower in the clausal spine right above υP. 43. raaghav-ne kitaab nahi paDhii raaghav-ERG book NEG read.PFV.3FSg 44. raaghav-ne kitaab paDhii nahi thii raaghav-ERG book read.PFV.3FSg NEG AUX.PFV.3FSg 45. *raaghav-ne nahi kitaab paDhii (thii) raaghav-ERG NEG book read.PFV.3FSg AUX.PFV.3FSg Then, (41) and (42) would then suggest that the subject stays lower than negation in Hindi. This can be taken as evidence for the subject being able to stay in its lower base position, that of spec,υP. Additionally, there have been cross-linguistic arguments concerning interpretive differences in subjects when the linear order of the subject is alternated with other constituents like certain 51 adjuncts. These interpretive differences are associated with two different positions for the subjects. Diesing (1992) in her study of indefinites in German argues for two different positions of subjects as they are interpreted differently when they occur in different positions. She notes that when adverbs like ja doch ‘indeed’ are inserted the linear ordering of a bare plural subject with respect to the adverb results in different interpretations of the subject. In (46), when the subject is to the left of the adverb (i.e., higher than the adverb) the bare plural has a generic interpretation. For an existential reading of the bare plural, it is placed to the right of the adverb as can be seen in (47). The conclusion then is that subject in a lower position lends itself to an existential interpretation in German whereas in the high position it is interpreted generically. 46. …weil Kinder ja doch auf der Strasses spielen …because children indeed on the street play ‘…because children play on the street. (in general)’ 47. … weil ja doch Kinder auf der Strasses spielen … because indeed children on the street play ‘…because there are children playing on the street.’ German- Diesing (1992) Similar evidence can be used to argue for two different subject positions in Hindi. In (48), when the subject is to the left of the location adjunct, it has a specific interpretation and in (49) when subject is to the right of the adjunct, the subject can only have an existential interpretation. The conclusion for German carries over to Hindi to a certain extent in the sense that the existential interpretation of the subject is obtained only when the subject stays in its lower base position in spec,υP. By way of contrast the specific interpretation of the subject is obtained when the subject moves higher to the spec,IP position. 52 48. kitaabeiN mej-par rakhi heiN book.Pl table-on keep.PST AUX.PRES.3FPl ‘The books are kept on the table.’ 49. mej-par kitaabeiN rakhi heiN table-on book.Pl keep.PST AUX.PRES.3FPl ‘There are books on the table.’ Another piece of potential evidence that the subject in Hindi might stay in a lower υP-internal position comes from scope-marking wh constructions. Scope-marking kya in the multi-clausal wh constructions in Hindi has been argued to be immobile and mark the edge of the matrix υP (Manetta 2010). In (50) we see, the subject amit in the second embedded clause can potentially occur in between the scope marker and the verb thereby giving evidence for a lower position for the subject. Since the scope marker kya occurs at the edge of υP, amit has to occupy a position inside υP and that position can be argued to be the base position spec,υP. 50. raaghav-ne kya kahaa ki kya amit-ne sunaa ki kaun raghav-ERG what say that what amit-ERG hear that who ayaa come.PST.3MSg ‘Who did Raghav say Amit heard had come?’ The three major pieces of evidence presented above suggest that subjects in Hindi can stay in their position of base merge, i.e., in spec,υP. Having independently motivated this, we can then argue that a υP ellipsis analysis could account for the missing subject in the attested fragment answer patterns in Hindi. The subject in spec,υP would be inside the ellipsis site and ellipsis of υP can 53 thus explain the missing subject in answers. Since the missing subject is due to ellipsis and not due to pro drop it would also explain that quantificational subjects which cannot be substituted by pro can be missing as well. 2.3.2 Verb stranding υP Ellipsis In a υP ellipsis account for verb stacking in fragment answers, the verb has to move out of υ due to some head-attracting feature to explain the fact that the answer patterns have the entire verbal sequence present. Verb stranding verb phrase ellipsis is independently attested in Hindi as can be shown by (51) wherein the adjunct inclusion and availability of sloppy interpretation points to a verb phrase ellipsis account as opposed to a pro drop or an argument ellipsis account of the missing argument (Bhatt & Dayal (2007), Simpson et al (2013), Manetta (2019). 51. raam-ne apanaa nayaa lekh do baar paDhaa raam-ERG self new article two times read.PFV.3MSg ‘Rami read hisi/selfi new article twice.’ raaj-ne bhi paDhaa raaj-ERG also read.PFV.3MSg ‘Rajj also read (hisi/j article twice).’ Manetta (2019) has a detailed discussion of verb stranding verb phrase ellipsis in Hindi where she argues that the main verb forms a complex head with υ and raises out of the υP layer following which υP gets elided with its remaining content. One piece of evidence she presents is from verb phrase ellipsis cases with complex predicate constructions. In (52), assuming that the light verb of the complex predicate is base merged at υ and the main verb is merged at V, verb phrase ellipsis 54 attests only those structures in which the entire V+υ complex as moved out of the elided constituent. As we can see, in Hindi the light verb cannot be stranded eliding only the main verb. This suggests that the main verb and the light verb form a complex head and furthermore this complex head raises higher than little υ to escape ellipsis. The claim that (52ii) is an instance of verb stranding vP ellipsis and not pro drop or argument ellipsis can be proved by adjunct inclusion in (52ii). 52. i. kabir-ne us kitaab-ko pahli baar paR lii-yaa. kabir-ERG this book-ACC first time read take-PFV.M ‘Kabir managed to read this book for the first time.’ ii. Meena-ne bhi __ paR lii-yaa. Meena-ERG also read take-PFV.M ‘Meena also read (this book for the first time).’ iii. ?*Meena-ne bhi _ lii-yaa. Hindi – Manetta (2019) Manetta (2019) argues that the higher head, that the V+υ complex moves to is Asp head following a similar analysis for Russian by Gribanova (2017). The evidence for that comes from habitual aspect in Hindi, which occurs strictly adjacent to the verb, thereby suggesting that V+υ+Asp form a complex head. I present a different analysis which is developed later in the proposal. She furthermore argues that the ellipsis site for the verb phrase ellipsis is higher than VP and lower than AspP and considers it to be υP following Merchant 2013 (for English), McCloskey 1991 (for Irish), Gribanova 2017 (for Russian) among others. She argues the same for Hindi by showing that adverbs like the repetitive ‘again’ which attach higher than the VP level, to the υP can be missing and still be interpreted and that is only possible if the ellipsis site is bigger than VP. 55 Furthermore, in Hindi aspectual heads can be different for the antecedent and the elided constituent suggesting that the ellipsis site is lower than AspP. Hence, I assume following Manetta (2019) that Hindi verb ellipsis patterns can be characterized as verb stranding υP ellipsis. In this sub-section we have independently motivated the two phenomena: (i) subject in Hindi can stay in spec,υP and (ii) verb phrase ellipsis in Hindi is the ellipsis of υP wherein the V+υ complex raises out of the υP. Let us recall that we need these two phenomena for a υP ellipsis analysis of fragment answers essentially so that we can account for the missing subject, for inclusion of higher-level adverbs and to explain the obligatory verb stacking pattern attested in Hindi fragment answers. The proposals necessary for our account follow in the next sub-section. 2.3.3 Proposal An account of fragment answers in Hindi should ideally be able to account for obligatory verb stacking when the verbal sequence does not form a complex head and does not raise higher than IP. We also need to account for the fact that IP ellipsis is independently attested, and even employed in forming polarity particle fragment answers, even though it is difficult to motivate an IP ellipsis analysis for verb stacking patterns. Verb stacking as an answer pattern cannot be motivated by reasons pertaining to lexical exponence of the valued polarity feature as well. If all the focused polarity feature needs is a lexical exponence, it is unclear why the entire verbal sequence needs to be repeated in the answer, just the highest verbal element should have sufficed. Hence, we should also formulate an answer for why verb echo answer pattern is not attested in the language. Yet another empirical conclusion that needs to be accounted for is that in Hindi the verb stacking pattern is the unmarked answer pattern whereas answering a polar question simply with 56 the polarity particle is a marked response. The proposal laid out below attempts to address these issues. I propose that ellipsis in fragments answers is triggered by a Merchant (2005) style Ellipsis feature hosted by the polarity head which causes the complement of the Pol head to be elided under semantic identity. More importantly the polarity head has two distinct positions in the Hindi clausal spine where it can merge into the structure. The lower IP-internal Pol phrase triggers υP ellipsis resulting in verb stacking as a fragment answer pattern (37ii), and the higher phrase in the C domain where polarity can be merged triggers IP ellipsis thereby accounting for the polarity particles as fragment answers (37i). I present the details of the two proposals in the following sub- sections before detailing the account. 2.3.3.1 Merchant’s E feature I propose that ellipsis in fragment answers is triggered by an ellipsis feature Eans which is merged with a polarity head. As defined by Merchant (2005), this ellipsis feature results in the complement of the head it is hosted on, to remain unpronounced at PF. Hence, what we get is the ellipsis of the complement of the Pol head. This unpronounced complement can be interpreted at LF under semantic identity with its antecedent. The unpronounced structure in the fragment answer then has to have a semantically identical antecedent in the corresponding question. Allowing ellipsis to be triggered only when the E feature is merged with the polarity head constrains the ellipsis site in fragment answers. This way we can answer the question of why fragment answers in Hindi are formed only due to IP or υP ellipsis and not any other kind of constituent ellipsis. As I argue in the upcoming subsection, IP and υP are the only potential ellipsis sites in a fragment answer as Pol can only take these two phrases as its complement. Moreover, 57 now ellipsis is not simplistically determined by maximal identity and hence we can argue for an account wherein υP ellipsis is the unmarked ellipsis site even when IP ellipsis should in principal have preference on account of maximal identity. Finally, having an ellipsis feature trigger ellipsis would unify the syntax of ellipsis across different ellipsis structures. 2.3.3.2 Two PolPs in Hindi As argued, attested fragment answers in Hindi can potentially be formed by υP ellipsis or IP ellipsis given that all the other analyses considered fall short of explicating the data. Having assumed that ellipsis is trigged by the Eans feature merged with a polarity head, a claim could be made that there are two distinct positions for polarity head in Hindi. One available position would be right above υP, such that the polarity head takes υP as its complement, triggering υP ellipsis when the Eans feature is merged with the head. The second position is a higher position, arguably in the C domain which takes IP as its complement. I adopt from Laka (1990) the claim that every clause has a polarity head and projects a polarity phrase. Holmberg (2015) assumes the presence of a PolP in every finite clause and argues PolP to be the highest head in the IP domain for Finnish, English, Mandarin among other languages. He admits that this assumption is not universal and suggests PolP to be IP-internal in Thai. I propose that Hindi, has an IP-internal PolP, specifically PolP is the highest head in the υP domain in Hindi. I further propose there is a higher head in the C domain which can host a polarity feature as well, specifically when the polarity feature is contrastively focused. I call this head the Contrastive Polarity head which projects a Contrastive Polarity phrase in the C domain. The idea of a higher polarity head being a projection of focused polarity, finds support in Romero and Han (2002), Han and Romero (2006), Romero (2006) among others. The 58 mentioned research gives an account of Ladd’s p/¬p ambiguity and epistemic biases in negative biased questions by arguing for interpretive differences and scopal interactions based on two different positions for polarity. For instance, Romero (2006) presents a discussion of negative biased questions and the biases corresponding to those questions. She gives an analysis for a commonly observed fact in English that in (53) when the negation is lower in the question there is not necessarily an inherent bias for a particular response, whereas in (54) when the negation has been preposed/moved higher the bias is for the affirmative proposition. 53. Is John not coming? Bias: none in particular. 54. Isn’t John coming? Bias: John is coming. The claim made in these works is that there are two dedicated positions for the verum operator with the higher position being the one wherein the verum operator is focused. I assume the verum operator as defined by Höhle (1992) to be the semantic counterpart of the syntactic polarity head. If there are two distinct positions for the verum operator and if the verum operator is syntactically placed on a polarity head, then there ideally should be two distinct positions for a polarity head. Though languages might vary on the exact positions of these polarity heads, the common assumption would be that the higher head is associated with focused polarity whereas the lower position would be the base position for the polarity head. There are empirical motivations for the claim that there are two distinct positions for polarity in the Hindi clausal spine and that the higher position is associated with contrastively focused polarity, whereas the lower position is not specified for focus. One piece of evidence comes from the negative answer form for a simple polar question in Hindi. In (56) we see that the negative answer can have two negative elements. It could be argued that though this answer form 59 is not preferred due to an anti-OCP type effect of repeating the same constituent next to each other, it still provides evidence for two syntactic positions that could host polarity. This becomes further clear if the two elements are separated by another element like an adverb. In that case the two negative elements are perfectly acceptable in the answer fragment. 55. Q: (kya) raaghav-ne kitaab paDhii? Q Raghav-ERG book read.PFV.3Fsg ‘Did Raghav read the book?’ 56. NegA: (nahi), nahi paDhii no NEG read.PFV.3FSg ‘No.’ (nahi), bilkul nahi paDhii no at all NEG read.PFV.3FSg ‘No.’ Another piece of data, relevant to the claims made here comes from verb phrase ellipsis structures in Hindi. As noted earlier in detail, verb phrase ellipsis structures in Hindi are of the verb stranding kind as shown in (57). Usually, the verb obligatorily has to move out of the υ domain to a higher head and the elided constituent is the υP and not a higher or a lower phrase. 57. raaghav kitaab acche-se nahi paDh rahaa thaa, raaghav book well-P NEG read PROG.3MSg AUX.PST.3MSg ‘Raghav was not reading a book properly.’ rohit bhi [e] nahi *(paDh rahaa) tha rohit also NEG read PROG.3MSg AUX.PST.3MSg ‘Rohit was also not reading (a book properly).’ 60 However, in certain specific situations, Hindi does attest an IP level verbal ellipsis. Interestingly, this IP level verbal ellipsis is only licensed when the polarity is contrastively focused. An example is given in (58). The polarity in the antecedent is affirmative while in the elided structure it is negative, thereby rendering a contrastive polarity configuration which makes verbal ellipsis at a higher level possible. 58. raaghav kitaab acche-se paDh rahaa thaa, raaghav book well-P read PROG.3MSg AUX.PST.3MSg ‘Raghav was reading a book properly.’ magar rohit nahi [e] but rohit NEG ‘Rohit was not (reading the book properly).’ Comparing (57) and (58) we can conclude that the verbal sequence in Hindi can get elided only when the polarity in the ellipsis site is contrastively focused. In other terms, it could be argued that polarity unspecified for focus can trigger only υP ellipsis but contrastively focused polarity triggers IP ellipsis. Converging this empirical observation with the assumption that it is actually the E ans feature on the polarity head which triggers the ellipsis of its complement, we infer that there are two distinct positions for polarity, a PolP at the edge of the υP domain taking ʋP as its complement and a PolfocP in the C domain which takes the IP as its complement and the ellipsis feature Eans can merge at either of these positions. Having motivated the two proposals needed for a comprehensive account of fragment answers in Hindi, in this next sub-section I lay out the details of the proposed analysis. 61 2.3.4 Account Let us recall, the major objective here is to provide an account of obligatory verb stacking fragment answer pattern attested in Hindi. This pattern is difficult to be accounted for by an IP ellipsis approach even when IP ellipsis is independently attested in Hindi and also employed in deriving fragment answer patterns. A lexical exponence approach like the one proposed by Holmberg (2015) for verb echo patterns in Finnish, Welsh, Mandarin etc. also seems to be lacking in its ability to justify the entire verbal sequence to be obligatorily present in fragment answers. The verb stacking fragment answer pattern to be accounted for is repeated in (60). I propose that the fragment answer in (60) is obtained by υP ellipsis wherein the subject stays in spec,υP and the verb is moved out of the υP. 59. Q: (kya) raaghav kitaab paDh paa rahaa thaa? Q raaghav book read able PROG.3MSg AUX.PST.3MSg ‘Was Raghav able to read the book?’ 60. A: (haaN) paDh paa rahaa thaa yes read able PROG.3MSg AUX.PST.3MSg ‘Yes.’ I adopt Holmberg’s (2015) idea that the polarity head in the question is a variable [±Pol] which is bound by an interrogative operator in the C domain. The function of the interrogative operator binding the variable is to give the variable a sentential scope and to look for a value of the variable. The answer provides a value for the variable polarity feature. This value is a focused polarity feature merged in the C domain in the answer. I assume the focused polarity value to specifically 62 merge in the specifier of the higher contrastive polarity head, i.e., specifier of Polfoc. The merged valued feature, in turn, values the variable on the IP-internal Pol head as the two polarity heads are in a feature sharing relationship. The Eans ellipsis feature is merged with the IP-internal Pol head as it is the first Pol head it encounters during bottom-up structure building. This Eans feature triggers an ellipsis of the complement of the Pol head, which effectively is υP ellipsis. As we now know υP ellipsis in Hindi is of the verb stranding kind, i.e., the V+υ complex has moved out of the υP to a higher head. As already mentioned, Manetta (2019), following Gribanova (2013a) for Russian, argues that the head to which the V+υ complex moves, is the Aspect head. However, if we look at (59), we observe that the ability modal can intervene between the verb and at least certain kinds of aspect (progressive aspect here). So, it is not always the case that the V+υ+Asp 0 form a complex head. A detailed study of the verbal structure of Hindi, is out of scope for this paper but for our purposes I assume that the higher head which attracts the V+υ complex is in fact the polarity head. The affirmative polarity feature in a clause is the default feature and is not overtly realized. It does require a lexical exponence and hence it attracts the lower V+υ complex head. In case of negative declarative constructions, it would attract the Neg head right below the polarity head and above the υ head. Since in Hindi the negation marker forms a syntactically complex head with the verb, the V+υ complex still moves out of the υP. Various analyses (Bhatt 2005, Kumar 2006, Bhatt & Dayal 2007 and Manetta (2019)) are present in the Hindi literature for the position that the V+υ complex raises to. However, where the V+υ complex raises to does not affect this proposal as long it is an IP-internal head. The motivation for the polarity head to be the head attracting the V+υ complex comes from the notion that the polarity head needs a lexical exponence. The lower polarity head is then realized by the V+υ complex. The higher polarity head hosts the focused 63 valued feature merged in the answer and gets its lexical exponence by the overt realization of the polarity particle. The structure of the verb stacking fragment answer in (60) is given in (61). 61. Another way of answering a polar question is by the polarity particles, though such answers are marked in Hindi they are made viable by with focus prosody or extra-linguistic features like a gestural node. To the extent that such fragment answers (as in 63) are acceptable, we can account for the underlying clausal structure with an IP ellipsis account. Recall that the polarity particles Figure 7: Verb stacking as υP Ellipsis 64 have been argued to have a clausal structure to maintain syntax-semantic mapping as they represent propositions semantically (Hamblin 1958). 62. Q: (kya) raaghav kitaab paDh paa rahaa thaa? Q raaghav book read able PROG.3Msg AUX.PST.3MSg ‘Was Raghav able to read the book?’ 63. A: ?haaN yes ‘Yes.’ Given the proposal of ellipsis of the complement being triggered by an Eans feature merged with the pol head, we can account for this answer pattern by positing that the Eans feature is merged with the higher Polfoc head which takes an IP complement. Hence in this case, the Eans feature is not merged with the first instance of the polarity head but waits for the Polfoc head to merge into the structure. This is motivated by the fact that polarity particles as answers are focused and hence occupy the higher position in the specifier of the Polfoc head. The structure is given in (64). Similar accounts for polarity particles have been presented in Holmberg (2015) and Kramer & Rawlins (2010). The idea that contrastive polarity ellipsis is in effect IP ellipsis has been argued for extensively in the verb ellipsis literature across languages (Vicente 2006, Kazenin 2006, Morris 2008, Barros 2014, Gribanova 2017). 65 64. The argument that Eans feature has to wait to merge with a higher Polfoc head when a lower Pol head is available seems to be a fallible violation of the Earliness Principle and which could account for the υP ellipsis pattern being the unmarked answer form whereas the IP ellipsis answer pattern gives us a marked fragment answer for polar questions in Hindi. Merging the Eans feature when it first encounters a polarity head is a natural consequence of bottom-up structure building. In this case the ellipsis feature will be merged with the lower IP-internal polarity head, thereby triggering υP ellipsis. Waiting for a higher polarity head to merge the ellipsis feature would then be a marked derivation thereby accounting for the fragment answers in Hindi which are derived by IP ellipsis to be rare/marked occurrences when compared to those derived by υP ellipsis. Another probable cause for the markedness of the fragment answer in (63) could be the identity condition on ellipsis. Ellipsis is argued to take place under semantic identity. The semantic identity of the interpreted proposition has further been argued to be determined from the υP domain (Rudin 2018). Anything above the υP domain thus has a potential to be non-identical but still respect semantic propositional identity. This could be argued to be the case with IP ellipsis in fragment answers with featural (non)identity. The IP-internal Pol head in the question antecedent is a variable while in the ellipsis Figure 8: Polarity particle as IP Ellipsis 66 site in the answer it is a valued feature. Hence, we could argue that υP ellipsis respects identity in a more stringent manner than IP ellipsis and thus is the unmarked form of fragment answer in Hindi 10 . Having said that it is relevant to note that both υP ellipsis and IP ellipsis respect the semantic identity constraint and hence both can be employed to derive fragment answers. Thereby we have argued for an ellipsis analysis of fragment answers wherein the ellipsis is triggered by an Eans feature and the ellipsis site is determined by the head which hosts this E ans feature. Only a polarity head is claimed to be able to host the ellipsis feature and two such polarity heads have been motivated in the Hindi clausal structure, one above υP and one lower in the C domain, giving way to two ellipsis sites. IP ellipsis would result in polarity particles as fragment answers. υP ellipsis, on the other hand, accounts for the obligatory verb stacking answer pattern. IP ellipsis in answers is however dis-preferred as it does not respect complete identity with its antecedent and also violates the Earliness Principle. The fragment answers obtained by υP ellipsis hence are the unmarked answer forms. This account also explains the obligatory verb stacking pattern in Hindi. Since only IP ellipsis or υP ellipsis are possible to derive the answer patterns, we either get the entire verbal sequence or none of the verbal elements as a fragment answer as the ellipsis site is strictly constrained by the ellipsis feature. The assumption here is that none of the verbal elements raise higher than I. The fact that Hindi does not attest verb echo patterns can then be explained following this assumption. Since a verbal element cannot raise higher than I, IP ellipsis cannot be used to derive the highest verb as a fragment answer. Concluding, fragment answers in Hindi are obtained by υP ellipsis or IP ellipsis. The obligatory verb stacking pattern is a result of υP ellipsis triggered by an Eans feature on the IP- 10 Refer to chapter 5, section 5.2.2 for another plausible reason for the markedness of polarity particles as fragment answers. 67 internal Pol head wherein the subject stays in spec,υP and the verb moves higher to the Pol head to provide lexical exponence to it. Polarity particles can also be answers to polar questions though they are the marked answer forms. They are derived by IP ellipsis triggered by the Eans feature merged with the higher Polfoc head in the C domain wherein the entire clause is inside IP and gets elided and the valued polarity feature is merged in spec,PolfocP thereby providing lexical exponence to the higher Polfoc head. 2.4 Conclusion In this chapter, I have attempted to give an ellipsis account of fragment answers to polar questions in Hindi. An ellipsis account is motivated under the assumption that the fragment answers are semantically propositional and respect the syntax-semantics mapping, and thereby have a clausal structure. Since the overt realization of this syntactic structure appears to be less/smaller than a full-fledged clause, syntactic mechanisms which could account for phonologically null constituents are at play. Hindi attests to various such mechanism which could in principle account for missing arguments in the fragment answer. I have argued against a pro drop analysis, an argument ellipsis analysis and a verb stranding VPE analysis as they are not able to account for the entire range of attested patterns. Based on Holmberg’s (2015) Big Ellipsis approach, a υP ellipsis analysis for Hindi has been motivated wherein the subject stays in spec,υP while the V+υ complex moves out of υP to a higher phrase. This gives us the obligatory verb stacking pattern in Hindi which is the unmarked answer form. Polar questions in Hindi can also be answered by polarity particles which are argued to have clausal structures as well. The occurrence of a polarity particle as an answer fragment is obtained by IP ellipsis. The empirical observation that only υP 68 or IP ellipsis is available to derive fragment answers suggests that the ellipsis site is constrained due to some reason. A Merchant (2005) style ellipsis feature is posited which is argued to be hosted by the polarity head thereby allowing ellipsis of only the complements of the polarity head. Two different positions for the polarity head are independently motivated in Hindi. One position is IP- internal, at the edge of the υP domain such that the complement of the polarity head is υP. The other position is high in the C domain which hosts a contrastively focused polarity head and takes IP as its complement. This derives that fact that the only viable fragments in Hindi are due to υP ellipsis or IP ellipsis. The syntax of fragment answers gives us an insight into the syntax of polar questions. On the basis of answer forms that occur, the polar questions in Hindi are argued to have a parallel structure with two polarity phrases wherein the polarity heads are in a feature sharing relationship. A study of the structure of questions in isolation would not have led us to the same conclusion, as there seems to be no motivation to posit two polarity heads if we exclusively look at polar question structures (‘or not’ questions do give us a hint though) at least in Hindi. However, we do have cross-linguistic evidence from negative biased questions in English, for which Romero (2006) argues a structure with two distinct positions for polarity. An interesting avenue of further research lies in the explanatory power of an ellipsis account over other accounts for fragment answers. An ellipsis account for a language like Hindi which independently attests various other syntactic phenomena which could account for unpronounced structure, leads us to an appealing generalization of fragment answers across all languages being derived via ellipsis. Such a unified analysis would definitely present a more attractive picture theoretically, however empirically it remains to be proved. Another dimension of future research is to combine Holmberg’s (2015) analysis for ‘verb echo’ language with the 69 analysis presented here for a hitherto unaccounted for pattern of ‘verb stacking’ languages and attempt a unified approach that could account for cross-linguistic variations. A complete paradigm can be seen if we consider the three languages: Welsh, Finnish and Hindi. For a polar question the fragment answer in Welsh is the highest verbal element, i.e., the verb echo answer pattern is attested and verb stacking is prohibited. In Finnish, verb stacking is optional and thus answers in Finnish can either have just the highest verbal elements or more verbal elements if the question has a complex verbal sequence. In Hindi verb stacking is obligatory whereas verb echo patterns are ungrammatical. Holmberg’s (2015) analysis of fragment answer seems to not be able to account for languages like Hindi which have IP ellipsis in the form of polarity particle fragments but do not attest ‘verb echo’ patterns. Some of these issues are further explored in chapter 5 of this dissertation. If my analysis is in the right direction, this account of fragment answers should be applicable to answers of all types of polar questions. In chapter 3 of this dissertation, I will extend the analysis to negative biased polar questions and in chapter 4 to other types of questions like alternative questions and polar questions with narrow focus. Dating back to Katz and Postal (1964), the idea that all questions and in a parallel way all answers share a common structure to some extent, has merit. Combining Merchant’s (2005) analysis of fragment answers to wh- constituent questions with Holmberg’s (2015) analysis of fragments answers to polar questions, we are closer to unifying the seemingly different constructions of different types of questions and answers. 70 Chapter 3: Biased Questions and Answers in Hindi 3.1 Negative Polar Questions Negative polar questions are, simply put polar questions with a negation in them. Given this characterization they are supposed to behave quite similarly to positive/neutral polar questions. The prevalent theory for polar questions is that they denote a set of a nucleus proposition and its negation (Hamblin 1973), i.e., both denote the same nucleus set and its negation {p, ¬p}. In positive polar questions the positive nucleus proposition is ‘highlighted’ and in negative polar questions the negation of the nucleus proposition is ‘highlighted’. Roelofsen and Farkas (2015) define ‘highlighting’ as making salient one possibility from the alternative set such that the possibility becomes highly accessible in the discourse. Consequently, in a positive polar question from the set of {p, ¬p} alternatives the [p] possibility/alternative is highlighted whereas in a negative polar question the [¬p] possibility/alternative becomes more salient in the discourse. Negative polar questions on account of highlighting the negative alternative end up differing from their positive counterpart in the bias they project. Dayal (2016) in her brief overview of negative polar questions argues that positive and negative polar questions primarily differ in the bias they hold and consequently occur in different contexts. She further argues that the bias is introduced by negation and NPIs itself rather than by the question component in biased questions. For instance, one can utter (1) in an out-of-the-blue context but (2) can only be uttered where the speaker was expected to be a member of the 71 communist party but the speaker negates the bias. The negative statement or question can only be introduced when there is ‘some sort of givenness on its prejacent’ (Dayal 2018). 1. I am a member of the communist party. 2. I am not a member of the communist party. Asher & Resse (2010) further establish that negative and positive polar questions can’t be interchangeably used as the given context would only felicitously be followed by a positive question rather than its negative counterpart in (3). Consequently, the established notion is that positive and negative statements and contexts do not arise from the same world view or set of expectations but are licensed in different contexts. Whereas positive statement/questions can be uttered without any specific expectations or bias; negative statements/questions presuppose a bias towards one alternative. 3. Context: I have no beliefs on the matter. I just want to know… Are you/#Aren’t you a member of the communist party? Asher & Reese (2010) This is seemingly in contrast with the generally accepted semantic analysis that positive and negative polar questions denote the same nucleus set and its negation {p, ¬p} as suggested by Hamblin (1973) and hence should be treated equably. Empirically, this is shown by the same set of possible answers to both positive and negative polar questions. The possible answers to both positive and negative polar questions in (3) are given in (4). 4. Yes, I am a member of the communist party. /No, I am not a member of the communist party. However, as it has been established in detailed studies of answers to negative polar questions in English (Asher & Reese 2005, Reese 2007, Roelofsen & Farkas 2015 among others), the answer 72 patterns in (4) are too simplistic and it is in fact problematic to assume that positive and negative polar questions should get the same semantic representation. The aim of this chapter is to explore answers to biased questions in Hindi to the eventual purpose of comparing and contrasting negative polar questions with their positive/neutral counterparts. I provide further support for the idea that biased questions are a beast of slightly different nature than the neutral positive polarity questions as attested by the different answer forms they exemplify. I extend the υP ellipsis account proposed for fragment answers in Hindi to answers to negative biased questions. On the basis of the specific answer patterns attested in negative polar questions, I argue for the possibility of a syntactic representation of performative projections in the extended CP layer in a cartographic approach consequently extending the understanding of the clausal structure of Hindi. This chapter has the following sections: the 2nd section of the paper deals with the major data patterns attested in Hindi as answers to negative polar questions. The 3rd section is adapted from Roelofsen & Farkas’s (2015) analysis of polarity particles and applies it to Hindi answer patterns. Section 4 of this paper describes the prosodic differences between two types of answer systems attested in answers to negative biased questions and section 5 details the proposed analysis of both these answer systems. Section 6 briefly outlines the primary semantic models in the literature that encode bias in negative polar questions. Finally, section 7 briefly introduces other types of biased questions attested in Hindi and section 8 concludes. 73 3.2 Data - Answer Patterns To review the previous chapter, the unmarked answer form to positive polar questions is a verb stranded υP ellipsis fragment following a polarity particle. 5. Q: kya raaghav-ne kitab paDhii thii Q raghav-ERG book read.PFV.3FSg AUX.PST.3FSg ‘Did Raghav read the book?’ 6. A: i. haaN paDhii thii yes read.PFV.3FSg AUX.PST.3FSg ii. nahi paDhii thii no read.PFV.3FSg AUX.PST.3FSg iii. ?haaN yes iv. ?nahi no Bare polarity particles are marked answer forms given that Hindi prefers to have the explicit fragment prejacent but still acceptable as answers to positive polarity questions as can be seen in (6iii and iv). However, the first point of difference between positive/neutral and negative polarity questions is the (un)acceptability of bare polarity particles as answers (8iii and iv). Bare polarity particles seem incomplete answer forms to the question in (7). However, if accepted somehow as answers both the haaN ’yes’ and nahi ‘no’ particle would be taken to confirm the highlighted negative proposition (¬p). Kramer & Rawlins (2012) have defined this phenomenon as ‘Negative 74 Neutralization’ and in another work (Kramer & Rawlins 2012) have experimentally concluded the same for English. This could suggest that there is a strong bias for the [¬p] alternative. 7. Q: kya raaghav-ne kitab nahi paDhii thii Q raghav-ERG book NEG read.PFV.3FSg AUX.PST.3FSg ‘Didn’t Raghav read the book?’ 8. A: i. (nahi) nahi paDhii thii no NEG read.PFV.3FSg AUX.PST.3FSg ii. (haaN) paDhii thii naa 11 yes read.PFV.3FSg AUX.PST.3FSg Prt iii. #haaN yes iv. #nahi no The most expected answer forms to the question in (7) is given in (8i and ii). The (8i) answer form selects the [¬p] alternative, which is the same as the nucleus proposition of the question whereas the (8ii) answer form selects the [p] alternative which is the negation of the nucleus proposition [¬(¬p)] in the question. At first glance it appears that Hindi is a polarity-based language like English as opposed to a truth-based language like Japanese. Holmberg (2015) follows Jones (1999) in classifying languages as polarity based or truth based depending on the meaning conveyed by the polarity particle. In polarity-based languages like English, Swedish and Hindi the polarity particles (haaN ‘yes’ and nahi ‘no’ in Hindi) convey the polarity of the prejacent. In (8i) the prejacent has negative polarity and hence the use of the negative polarity particle nahi whereas in (8ii) the prejacent has positive polarity and occurs with the positive polarity particle haan. Further 11 naa is a discourse particle used to convey obviousness. 75 evidence that the polarity particles convey the polarity of the prejacent is the optionality of the particles themselves. They do not convey any new information and hence the particles are optional in (8i and ii). Languages like Hindi and English are different from truth-based languages like Japanese, Korean and Chinese wherein the polarity particle conveys the (dis)agreement of the speaker with the bias in the question. In (9), the polarity particle in the answer conveys that the speaker agrees with the questioner that ‘he is not tired’ as that is the highlighted/salient alternative in the negative polar question in (9). Consequently, in truth-based systems the polarity particle ‘yes’ conveys agreement whereas the particle ‘no’ conveys disagreement with the highlighted proposition. 9. Q: kimi tukararete nai You tired Neg ‘Are you not tired?’ 10. A: un (tukarete nahi) Yes (tired Neg) Lit: ‘Yes, I am not tired.’ -Japanese Holmberg (2015) The characterization of Hindi as a polarity-based language is actually an incomplete picture based on the most unmarked answer forms to the question in (7). The question in (7) can also be answered by the answer forms presented in (11). 11. A: i. nahi, PADHII thii no read.PFV.3FSg AUX.PST.3FSg ‘No, you are wrong, he did read it.’ ii. haaN, NAHI paDhii thii yes NEG read.PFV.3FSg AUX.PST.3FSg ‘Yes, you are right, he did not read it.’ 76 To elaborate, the answer forms in (11) are both truth-based answer forms wherein the polarity particles agree or disagree with the [¬p] alternative salient in the question. Immediate evidence of this comes from the fact that the polarity particles are not optional in these answer forms. Furthermore, there is no feature sharing between the polarity of the prejacent and the polarity particle itself. They are in fact of opposite values. In (11i) the prejacent has positive polarity with verum focus and the polarity particle is negative whereas in (11ii) the prejacent has negative polarity with verum focus while the polarity particle is positive. In (11i), the polarity particle nahi conveys the disagreement with the [¬p] bias in the question and explicitly selects the [p] alternative as its prejacent and similarly in (11ii) the polarity particle haaN signifies agreement with [¬p] bias in the question and explicitly states [¬p] alternative as its prejacent. There are certain other answer forms attested as possible answers to the question in (7). 12. i. nahi, paDhii to thii no read.PFV.3FSg Prt AUX.PST.3FSg ii. nahi nahi, PADHII thii no no read.PFV.3FSg AUX.PST.3FSg iii. haaN yaar, nahi paDhii thii yes Voc NEG read.PFV.3FSg AUX.PST.3FSg The answer forms in (12) are further instances of the truth-based answer system wherein the polarity particle conveys (dis)agreement with the [¬p] bias in questions rather than feature matching with the polarity of the prejacent. (12i) is a version of the answer in (11i) with an additional discourse marker to. to has been argued in Hindi to be a topic marker with an obviousness implication. This answer conveys disagreement to the [¬p] bias in the question and further asserts that the [p] proposition is the obvious alternative. The to marker makes the answer 77 more felicitous as it enforces the obviousness of the [p] alternative on the face of a strong salience of the [¬p] alternative .The answer form in (12ii) simply asserts a strong opposition to the [¬p] alternative by repeating the negative polarity particle and finally the (12iii) answer form exclaims the complete agreement of the speaker with a vocative to the [¬p] bias in the question. The use of the vocative signifies a lack of surprise at the [¬p] alternative. Thus far we have seen that even though unmarked answer forms to negative biased questions in Hindi are of the polarity-based type wherein the polarity particle conveys the polarity value of the prejacent; Hindi also exhibits answer patterns of the truth-based type wherein the polarity particle signifies (dis)agreement with the salient alternative in the questions. This leads us to conclude that the polarity particles are anaphoric in nature. In one instance they agree in their features with the prejacent and in another instance they are sensitive to the polarity of the question and signify their (dis)agreement with it. The treatment of polarity particles as being anaphoric has been dealt with some detail in Roelofsen & Farkas (2015). In the next section we explore the functionality of polarity particles in Hindi in detail. 3.3 The nature of polarity particles in Hindi Polarity particles are used to respond to both assertions and questions. Roelofsen & Farkas (2015) argue that they are anaphoric in nature as they are sensitive to the polarity value of their antecedent and are used accordingly in the response. We already saw in the previous section how polarity particles are used to signify (dis)agreement with the highlighted alternative expressed in the negative polar questions. 78 Furthermore, following Pope (1976), Roelofsen & Farkas (2015) argue that polarity particles have two functions. In one instance they mark the absolute polarity value of the proposition and in another they signify whether that prejacent agrees or disagrees with/reverses the content and polarity of the antecedent/question. They define the two functionalities as: 13. i. Absolute polarity features – [+] or [-] – These features presuppose that the prejacent expresses a proposition containing a single possibility which is highlighted and has positive or negative polarity. ii. Relative polarity features – [AGREE] or [REVERSE] – These features presuppose that the prejacent highlights a unique possibility α and the discourse context contains a unique most salient antecedent possibility β such that α agrees with/reverses β, both in terms of polarity and content. Roelofsen & Farkas (2015) In (14) and (15), in addition to (8) and (11) we observe that the polarity particles in Hindi serve dual functionality just like that in English as will be made clear soon. 14. i. Positive Initiative: raaghav-ne kitab paDhii thii raghav-ERG book read.PFV.3FSg AUX.PST.3FSg ii. Agreement: a. haan, paDhii thii/ yes read.PFV.3FSg AUX.PST.3FSg/ b. *nahi, paDhii thii no, read.PFV.3FSg AUX.PST.3FSg iii. Reversal: a. *haan, nahi paDhii thii/ yes, NEG read.PFV.3FSg AUX.PST.3FSg b. nahi, nahi paDhii thii no, NEG read.PFV.3FSg AUX.PST.3FSg 79 15. i. Negative Initiative: raaghav-ne kitab nahi paDhii thii Raghav-ERG book NEG read.PFV.3FSg AUX.PST.3FSg ii. Agreement: a. haan, NAHI paDhii thii/ yes NEG read.PFV.3FSg AUX.PST.3FSg b. (nahi), nahi paDhii thii No NEG read.PFV.3FSg AUX.PST.3FSg iii. Reversal: a. (haan) paDhii thii/ yes read.PFV.3FSg AUX.PST.3FSg b. nahi, PADHII thii no, read.PFV.3FSg AUX.PST.3FSg When responding to a positive initiative in (14i), haaN marks the [AGREE] polarity feature in (14ii) signifying that the prejacent agrees with the polarity and content of the positive initiative. The absolute polarity feature [+] is conveyed by the main verb itself. In the reverse response to the positive initiative, the first nahi in (14iii) marks the [REVERSE] polarity feature whereas the second nahi marks the [-] polarity feature. Until now, we have seen the polarity particle haaN only marking relative polarity feature [AGREE] whereas the polarity particle nahi was seen to mark both the absolute polarity feature [-] and the relative polarity feature [REVERSE]. But as a reversing response to the negative antecedent in (15i), we see the positive polarity particle haaN which is optionally present in (15iii) marking the absolute polarity feature [+]. 80 Consequently, we can conclude that the polarity particles in Hindi serve a dual function. 16. i. Positive polarity particle – haaN can realize the absolute polarity feature [+] and the relative polarity feature [AGREE]. ii. Negative polarity particle - nahi can realize the absolute polarity feature [-] and the relative polarity feature [REVERSE]. Given this dual functionality we come to a paradigm of polarity particles with a four-way distinction as given in the table in (17). 17. Figure 9: Possible Combinations of polarity features (Roelofsen & Farkas 2015) Relation with antecedent Polarity of prejacent Combinations [AGREE] [+] [AGREE,+] 12 [AGREE] [+] [AGREE,-] [REVERSE] [-] [REVERSE,+] [REVERSE] [-] [REVERSE,-] [AGREE,+] and [REVERSE,-] combinations are used as responses to positive initiatives whether assertions (as in 14) or questions. And [AGREE,-] and [REVERSE,+] are used as responses to negative assertions (15) and questions (11). The answers to negative polar questions in (8) exemplify a [+, +] and [-,-] pattern wherein the optional polarity particles reiterate the absolute polarity feature of the prejacent. 12 The first member in this set representation corresponds to the function/value of the polarity particle, while the second member corresponds to the value of the prejacent. Thereby [AGREE, +] representation indicates a structure where the polarity particle agrees with the value and content of the polarity of the antecedent and the polarity of the prejacent is positive. 81 To further clarify, in Holmberg’s (2015) classification of languages, polarity-based systems use [+, +] and [-,-] as answers to negative polar questions whereas truth-based systems use the [AGREE,-] and [REVERSE,+] answer forms. It is clear that Hindi uses both though the unmarked system in Hindi is the polarity-based system. As a response to (7) and (15), (8ii) and (15iii-a) is a [+, +] response and (8i) and (15ii-b) is a [-,-] response. Herein the first (only in 8ii and 15iii-a) polarity particles realize the absolute polarity features, i.e., they reflect the polarity value of the prejacent and hence have to feature-match with the polarity of the prejacent. Since they do not provide any new information the first (only in 8ii and 15iii-a) are optional in these answer forms. This exemplifies a polarity-based system. Answers in a truth-based system are also attested in Hindi. (11ii) and (15ii-a) as responses to (7) and (15) respectively realize the [AGREE,-] answer form and (11i) and (15iii-b), the [REVERSE,+] answer forms. In these cases, the obligatorily present polarity particles realize the relative polarity features which signify whether the unique possibility conveyed by the prejacent agrees or disagrees with the highlighted unique possibility of the antecedent in its polarity and content (which is the negative alternative for negative antecedents). In addition to the difference in optionality of the polarity particle, the two answer systems also differ in their prosodic properties, which will be elaborated on a little later. To reiterate, Hindi exhibits both types of answer forms but the unmarked answer form is the polarity-based form. That is obvious if we take a look at the feature combinations. [+, +] and [-,-] are clearly more unmarked combinations when compared to [AGREE,-] and [REVERSE,+]. In the latter set not only is the combination that of two different types of polarity features (absolute and relative); the combination has features usually encoded by opposite particles ([AGREE] usually realized by haaN and [-] usually realized nahi and similarly [REVERSE] realized by nahi and [+] realized by haaN or the main verb). There are even more minute markedness constraints 82 at play which result in different degrees of acceptability of the attested answer patterns. Roelofsen & Farkas (2015) argue that [REVERSE,+] is more marked than [AGREE, -] because [REVERSE] in general is more marked than [AGREE] and an absolute feature with a [REVERSE] response is also marked as they always contrast with the polarity of the antecedent. They argue that this is the reason that in a lot of languages (French & German), the [REVERSE,+] response has a special dedicated polarity particle (si and doch respectively) and even in English these answers are marked by verum focus on the auxiliary verb when no other answers receive a similar treatment. Furthermore, recalling the pattern in (7 and 8iii-iv), where we see that simple polarity particles are not good answers to negative antecedent as they are ambiguous. Roelofsen & Farkas (2015) suggest this pattern is due to a markedness constraint. Given that agreeing with the highlighted alternative in the question is less marked than rejecting or reversing it, when the answer is not clear as is the case with bare polarity particles, both the particles are taken to confirm the highlighted alternative. 3.4 Prosodic Properties of the two answer systems As mentioned in the previous section the truth-based answer forms and the polarity-based answer forms in Hindi also differ in prosodic properties. When the polarity particles are realizing absolute features in (8ii), (15iii-a), (8i) and (15ii-b) not only are they optional they also form one prosodic unit with their prejacent. There is no pause in between the polarity particle and the prejacent as can be seen in the PRAAT diagram in (18). However, in the truth-based system answer forms in (11ii), (15ii-a), (11i) and (15iii-b) when the polarity particles are obligatory and realizing the relative polarity features; there is a pause in between the polarity particle and the next prosodic 83 unit which constitutes the prejacent as is clear in the PRAAT analysis in (19). The pause duration in (19) is of 0.15 seconds. 18. Figure 10: PRAAT waveform and spectrogram for (8i) – (nahi) nahi paDhii thii 19. Figure 11: PRAAT waveform and spectogram for (11i) – nahi // PADHII thii 84 In addition to the pause there is one other prosodic difference. In the answers forms of the truth- based system there is focus on the sentential polarity of the prejacent, i.e., there is verum focus. The main verb in (11i) which encodes the positive polarity of the prejacent is focused and the sentential negation in (11ii) which encodes the negative polarity of the prejacent is also in focus. Such focus marking is not necessary in polarity-based answer systems. The focus marking is reflected in the pitch of the main verb, the main verb in (19) is marked by a rising pitch whereas that in (18) is marked by falling pitch. There is in fact a third prosodic cue specific to truth-based answer forms. The vowel in the polarity particle in such answer forms seems elongated when compared to the vowel length of polarity particles in polarity-based answer forms. The polarity particle in (19) is measured to be of the duration of 0.58 seconds while the duration of the first polarity particle in (18) is 0.45 seconds. 3.5 Analysis of Biased Answers In the previous sections we concluded that Hindi attests to two kinds of answer forms – polarity- based and truth-based systems when answering negative polar questions. These answer forms can be differentiated on lexical/semantic and prosodic grounds. I propose a syntactic distinction as well which explicates the attested patterns and their properties. I propose that when polarity particles realize absolute polarity features they occupy a lower position in the clausal spine as opposed to when they realize relative polarity features as is manifested in from the prosodic evidence. 85 3.5.1 Holmberg (2015) – A Syntactic Analysis This proposed difference in height is also reflected in Holmberg’s (2015) analysis of polarity- based and truth-based systems. Holmberg (2015) does not make a distinction between the dual functionality of polarity particles, rather he derives the two different positions solely based on locations they occupy in the clausal spine. He argues that in polarity-based systems the C domain which hosts the polarity particle is closer to the negation head enforcing a feature match requirement (an example schematization in 20). And in a truth-based system the Neg head is lower in the clausal spine, consequently allowing the polarity particle on the polarity head to differ in value with the negation in the Neg head (an example schematization in 21). 20. Polarity-based languages 13 - VP > υP > NegP > IP > PolP > CP 21. Truth-based languages – VP >NegP > υP > IP > PolP > CP Consequently, languages which employ the polarity-based system the negation is higher, close to IP and in languages which attest truth-based answers, the negation is lower, somewhat close to υP. The position of the polarity head which hosts the polarity particle is the same. The polarity head as per Holmberg (2015) is a projection in the C domain above IP. The analysis gets a little complicated for languages which have both kinds of answer forms like English and Finnish. For English two position for negation have been argued for. Negation occupies a higher Neg head in a polarity-based answer form (as shown in 22) and occupies the lower position in a truth-based 13 Note that these structures are just general representations of Holmberg’s (2015) proposed analysis. Actual structures vary with languages, as is shown with the position of lower negation in English in (23). Holmberg (2015) argues that lower negation is hosted by the specifier of VP rather than being hosted by its own projection. 86 answer form (shown in 23). This is also how he accounts for Ladd’s ambiguity in negative polar questions in English. 22. Polarity-based answer form– middle negation υP > NegP (hosts negation) > IP > PolP > CP (hosts polarity particle) 23. Truth-based answer form – lower negation VP (hosts negation in its specifier) > υP > IP > PolP > CP (hosts polarity particle) For Finnish he proposes a single head for negation higher in the clausal spine but argues that Finnish still attests to truth-based answer forms. He suggests that such answer forms in Finnish are made up of two propositional clauses, both of them having undergone ellipsis, the first one undergoes IP ellipsis while the second clause undergoes VP ellipsis. To illustrate, in (25), ‘no’ is a part of the first clause which has undergone IP ellipsis and ‘he is’ is the remnant of the second clause which has undergone VP ellipsis. The full clauses are given in (26). 24. Q: ei-kӧ jussi tule-kaan mukaan? Neg-Q jussi comes-after all along ‘Isn’t Jussi coming, after-all?’ 25. A: ei, kyllӓ se tulee No indeed he comes No, he is. - Finnish Homberg (2015) 26. i. [CP no [PolP Jussi [ IP is not [VP coming]]]] → not (Jussi is not coming), i.e. the highlighted alternative is wrong ii. [CP[PolP he [ IP is [VP coming]]]] → reiterates the positive alternative which is true. 87 3.5.2 Roelofsen & Farkas (2015) – A Lexical Semantic Analysis Holmberg’s (2015) account is a purely syntactic account used to explain the two answer form systems attested across languages and within a language. On the other hand, Roelofsen & Farkas (2015) argue for a purely lexical account to explain the two attested patterns. According to these authors, polarity particles can realize both the absolute and relative polarity features. Though some languages with a richer system have dedicated particles for dedicated functions, for most languages yes is an exponent of both [AGREE] and [+] while no is an exponent of both [REVERSE] and [-]. Consequently, in polarity-based answer forms the polarity particles realize the absolute polarity features whereas in truth-based answer forms they realize the relative polarity features. Being a semantic account, they have no comment on the position these different features occupy on the clausal spine or any other syntactic distinctions that they might have. 3.5.3 Tang (2020) – A Cartographic Analysis Tang (2020) argues for an extended CP layer which also represents performative projections following Ross (1970) in a cartographic approach to CP. He follows Ross’s (1970) motivation for a syntactic representation of performative verbs and sentence final particles to capture grammatical and pragmatic functions of the speech act taking the distributional properties of the sentence final particle ho in Cantonese. The complex speech act functionality of ho is represented as the ho sentence final particle originating in one extended CP projection and moving up to higher performative projections. A similar analysis of the ho particle is present in Law, Li & Bhadra 88 (2018) wherein ho is analyzed as a higher speech act operator which operates over an assertion operator to form biased questions. Tang (2020) adopts and extends Wiltschko & Heim (2016) and Wiltschko’s (2017a, 2017b, 2018) cartographic layers. Essentially, as per Wiltschko & Heim (2016) and Wiltschko’s (2017a, 2017b, 2018) there are four different layers of the clausal spine: 27. proposition > grounding > response > discourse The propositional layer consists of the core of the proposition (υP layer), the grounding layer consists of focus etc. where the speaker’s attitude towards the proposition is encoded. The response layer consists of a CoA (call of addressee) projection, termed as such by Tang (2020). The CoA layer determines what the speakers requires of the addressee to do with the proposition. The final discourse layer conveys the emotional attitude of the speaker towards the discourse. Tang (2020)’s proposed extended clausal structure for Cantonese is schematized in (28). 28. 3.5.4 Asher & Reese (2005) – A Pragmatics -Semantic Analysis Asher & Resse (2005) have argued that all biased questions are complex speech acts of the type assertion • question. Simply put, all biased questions entail an assertion of the speaker’s belief or bias and in addition to that have a question component which basically asks the addressee for their confirmation of the speaker’s belief. For a parallel syntactic representation of complex speech acts or higher speech act operator operating over other speech act operators, one could potentially argue that extended projections in the CP layer to denote this higher level of operation. Modelling this in a syntactic cartographic 89 approach, the assertion would be encoded in the proposition and grounding layer while the question component would be a call on the addressee to confirm the speaker’s assertion encoded in the CoA layer. This may further explain the un-embeddability of biased questions. Dayal (2016) arguably for the first-time notes that negative polar questions cannot be embedded and hence constitute a root phenomenon. As complex speech acts this could be explained with a suggestion that the embedded CP layer lacks all speech act projections. 29. John knows whether Mary likes Bill/?? Mary doesn’t like Bill. Dayal (2016) 3.5.5 Proposed Analysis The account which I will propose here combines Holmberg’s (2015) idea of difference in height of polarity projections and Roelofsen & Farkas’s (2015) idea of polarity particles having dual functionality to arrive at an analysis for polarity particles in Hindi. I adopt Roelofsen & Farkas’s (2015) idea that polarity particles can realize both kinds of polarity features (absolute and relative) and further propose that different features are hosted at different heads in the clausal spine. I posit that the relative polarity features occupy a higher head than the absolute polarity features. Adopting a cartographic approach for performative projections in the Hindi clausal spine I argue that while the absolute polarity feature is hosted by a polarity head above IP, the relative polarity features are hosted by an even higher head in the extended performative projection. I adopt Tang (2020) and Wiltschko’s (2017a, 2017b, 2018) layers for an extended discourse layer and argue that the relative polarity features are hosted in the performative response layer. I propose that in the truth-based answer systems wherein the polarity particles are anaphors sensitive to the polarity and 90 content of their antecedent questions, i.e., they realize relative polarity features, they are hosted by the CoA/response head. Polarity particles when realizing absolute polarity features are hosted by a polarity head above IP but lower in the spine than the CoA/response head. Evidence for an extended projection comes from the answer patterns attested in Hindi which argue for a higher position in the left periphery which hosts the relative polarity feature. First, from the lexical meaning conveyed by the polarity particles it is clear that they are realizing two functions. In the polarity-based answer systems the polarity particle simply means a ‘yes’ or a ‘no’. But in the truth-based systems, they signify (dis)agreement with the polarity and content of the antecedent question. This becomes further clear with extensions to the polarity particles as shown in (31) and (32). In (31), which is a polarity-based answer form to the negative biased question, the extension which overtly (dis)agrees with the antecedent is out whereas in (32), which is an instance of the truth-based answer form the extension is perfectly acceptable, even desired. 30. Q: kya raaghav-ne kitab nahi paDhii thii Q raghav-ERG book NEG read.PFV.3FSg AUX.PST.3FSg ‘Didn’t Raghav read the book?’ 31. A: # haaN tum sahi keh rahe ho, paDhii yes you right say PROG.2MSg AUX, read.PFV.3FSg thii naa AUX.PST.3FSg Prt #‘ Yes, you are right, he did read it (the book).’ 32. A: haaN tum sahi keh rahe ho, nahi paDhii yes you right say PROG.2Msg AUX, NEG read.PFV.3FSg thii AUX.PST.3FSg ‘ Yes, you are right, he didn’t read it (the book).’ 91 Second, there are prosodic differences in the two answer forms. When the polarity particle realizes a relative polarity feature in a truth-based answer form there is a pause between the polarity particle and the prejacent. The polarity feature in the prejacent, whether it be the negation or the main verb for the positive feature, receives verum focus and finally the polarity particle undergoes vowel lengthening in truth-based answer forms. All these prosodic cues are missing from polarity particles in polarity-based answer forms. Consequently, the proposed analysis argues for a different head that hosts the relative polarity features realized by polarity particles in a truth-based answer form. The tree structures given below reflect the essential difference between polarity-based and truth-based answer forms. (33) illustrates a polarity-based answer form. The negation/polarity of the prejacent is hosted by the default polarity head merged above υP and the optional polarity particle which realizes the absolute polarity feature ([-] in this case) is hosted by the higher polarity head above IP. (34) on the other hand, illustrates a truth-based answer form. The polarity of the prejacent occupies the same position of polarity head above υP but the polarity particle which realizes the relative polarity feature is hosted by the CoA head in the extended CP projection. The prejacent,a s established in chapter 2, is derived by verb stranding υP ellipsis. The assumptions made for the structure is that negation is low in Hindi (NegP has been posited to be between ʋP and AspP following Dwivedi 1991, Kumar 2006, Bhatt & Dayal2007) and it forms a complex head with the verb. Though the internal order of the neg+υ complex head is varied nothing can intervene between the two consequently giving evidence for the formation of a complex head. This complex head is in turn attracted by the higher polarity head to be able to lexicalize the polarity head. The Eans ellipsis feature is merged with the IP-internal Pol head as it is the first Pol head it encounters 92 during bottom-up structure building. This Eans feature triggers an ellipsis of the complement of the Pol head, which effectively is υP ellipsis. 33. Figure 12: Polarity-based answer – (nahi), nahi paDhii thii 93 34. In essentials, Holmberg (2015) too argues for a height difference analysis. For polarity-based languages he claims that the polarity head is closer to the Neg head and that is why they have to share features but in truth-based languages the polarity head is farther away from the Neg head as the Neg head is lower and hence can have opposing features. He does not differentiate between absolute and relative polarity features and modulates the distance between the two heads by Figure 13: Truth-based answer – nahi // PADHII thii 94 varying the position of the Neg head. This explanation doesn’t easily account for a language like Hindi which attests to both the patterns and has an arguably single lower projection for negation. The linear position of negation in Hindi is such that it has to be adjacent to the main verb (Mahajan 1988). Examples from the previous chapter repeated in (43-45) show that negation in the unmarked order is always to the left of the verb and in some marked cases where there is additional material to the right of the verb it can also be right of the verb. Essentially, it forms a complex head with the main verb as nothing can occur in between negation and the main verb. Even very low adverbs like ‘easily’, ‘quickly’ etc. cannot intervene between negation and the verb. Assuming that all the verbal elements do not form a complex head and raise to I, we can conclude that negation is attached lower in the clausal spine right above υP. 35. raaghav-ne kitaab nahi paDhii raaghav-ERG book NEG read.PFV.3FSg 36. raaghav-ne kitaab paDhii nahi thii raaghav-ERG book read.PFV.3FSg NEG AUX.PST.3FSg 37. *raaghav-ne nahi kitaab paDhii (thii) raaghav-ERG NEG book read.PFV.3FSg AUX.PST.3FSg Given that the head which hosts negation is lower in the clausal spine, one would expect that Hindi will be a truth-based language as per Holmberg’s (2015) analysis. However, we have observed the opposite. Hindi prefers answering by using the polarity-based answer form even though truth- based answer forms are also attested. One possible solution to this would be to argue that there is an additional head that may host negation higher in the clausal spine. In fact, such a proposal has been extended in the previous chapter for Hindi where I posit that the default polarity head in Hindi is merged above υP which hosts the negation. But there is also a higher polarity head in the C 95 domain which hosts contrastive polarity features. It could be argued that when Hindi attests to polarity-based answer forms the higher polarity head hosts the negation. A similar analysis has been argued for English by Holmberg (2015). I do not adopt this analysis on two counts. First, English provides us with empirical evidence that there are two positions for negation in a negative polar question as can be seen in (38). Hindi negative polar questions can only be formed in one way as seen in (7) repeated here in (39). Second, theoretically an approach positing the occurrence of negation in a higher polarity head falls short of taking into account the functionality and pragmatic inferences of the polarity particles. There is no simple explanation of how the negation in the prejacent occupying different heads will account for the different functions that the polarity particle performs in the two answer systems. It is not clear how the negation of the prejacent occupying a higher head will account for the fact that the polarity particle realizes an absolute polarity feature signifying [+] or [-] whereas when the negation of the prejacent occupies a lower head the polarity particle signifies [AGREE] or [REVERSE]. 38. i. Is John not at home? ii. Isn’t John at home? 39. Q: kya raaghav-ne kitab nahi paDhii thii Q raghav-ERG book NEG read.PFV.3FSg AUX.PST.3FSg ‘Didn’t Raghav read the book?’ Concluding, I propose an analysis in which negation always occupies the same position in Hindi, which is the default polarity head above υP but the difference in height between the negation and the higher polarity head is a result of the two positions of the two types of polarity features. Absolute polarity features are hosted by the higher polarity head in the C domain whereas the 96 relative polarity head are even higher in the response layer hosted by the CoA head in a cartographic approach. This accounts for all the empirical facts attested: (i) two different functions of polarity features are reflected in two different syntactic heads, (ii) the optionality of the polarity particle in the polarity-based answer form is also reflected in their dual functionality and the redundancy of the polarity particle in this answer form, (iii) the response layer CoA head hosts relative polarity features which accounts for the anaphoric properties of the polarity particle hosted on this head, (iv) the prosodic differences in the two answer forms are also accounted for by different syntactic positions, wherein the discourse properties of the relative polarity feature are represented by prosodic cues and (v) Holmberg’s (2015) polarity featural anti-OCP constraint is also satisfied as in the truth-based answer form where opposing values of the polarity particle and negation in the prejacent is allowed, the polarity particle is further away from the negation. 3.5.5.1 Evidence from positive/neutral polarity questions As noted in the paper, Hindi generally does not attest answer forms which do not have an overt polarity particle, as can be seen in (41). However, initial observation suggests that this polarity particle can be optional in answer forms when an overt auxiliary is present in the answer, as in (43). 40. Q: (kya) raaghav-ne kitaab paDhii? Q Raghav-ERG book read.PFV.3Fsg ‘Did Raghav read the book?’ 41. A: *(haaN) paDhii yes read.PFV.3Fsg ‘Yes.’ 97 42. Q: (kya) raaghav kitaab paDh paa rahaa thaa? Q raaghav book read able PROG.3MSg AUX.PFV.3MSg ‘Was Raghav able to read the book?’ 43. A: (haaN) paDh paa rahaa thaa + gestural nod yes read able PROG.3Msg AUX.PFV.3MSg ‘Yes.’ Speakers vary in the acceptance of answers forms in (43) with most speakers preferring the overt realization of the polarity particle. For speakers who can omit the polarity particle the answer is accompanied by a gestural nod. It is important to note that no amount of gestural nodding can allow for omitting the polarity particle in (41). It could be argued that the gestural nod is a pragmatic realization of the polarity particle. However, the optionality of the particle itself in presence of the auxiliary presents an interesting challenge. For our current concerns, it is to be noted that there is no prosodic pause between the polarity particle and the verb in (41) whereas the polarity particle when present in (43) is followed by a pause. This leads to the hypothesis that the polarity particle in the two utterances might be in different positions with the particle in (43) being in a higher position than the particle in (40). Fleshing out this idea it would seem that the particle in (41) is purely for the lexical realization of the polarity feature and is merged in the C domain. When the auxiliary is present, we don’t necessarily need the polarity particle to lexicalize the polarity feature as the auxiliary can be suggested to raise to the C domain to serve the purpose. The optional polarity particle in (43) is then argued to be in a higher position and is a discourse particle serving the function of reinforcing the positive bias in the yes/no question. 98 I reformulate the argument for the nature and behavior of polarity particles, given that we now know of their dual functionality having observed the negative polar answer forms. In (41) the polarity particle realizes an absolute polarity feature and is obligatory because the main verb In Hindi doesn’t move high enough to the C domain to provide lexical exponence to the absolute feature. When an auxiliary is present it can act as the lexical exponent of the absolute polarity feature as in (43). Hence, when the polarity particle is present in (43) it does not realize an absolute polarity feature but realizes a relative polarity feature which occupies a higher position and consequently is followed by a pause. The relative polarity feature is optional in cases of positive/neutral polarity questions as they do not encode a strong bias and hence the answer forms are not required to respond to the bias asserted in the positive/neutral questions. Consequently (41) encodes a polarity-based answer form for a positive/neutral polar question whereas (43) with its polarity particle present exemplifies a truth-based system if the two systems were solely defined by the functions of the polarity particles. 3.6 Semantics of Bias The traditional semantic treatment of polar questions is that they denote a set with the nucleus proposition and its negation formalized first in Hamblin (1973) as {p, ¬p}. Given this definition, negative polar questions would be treated the same as positive/neutral polar questions. The nucleus proposition of the negative polar question would be [¬p], consequently the set that the question denotes would be [¬p, ¬(¬p)] which is equivalent to [¬p, p]. Dayal (2016), Bolinger (1978), Asher & Reese (2005) among others define and formalize negative polar questions differently than their positive/neutral counterparts. The main distinction comes from the bias conveyed by the negative 99 polarity questions which is missing in their counterparts. This bias has been defined in many different ways in the literature. Simply put a biased polar question is a question which has a preferred answer. In case of polar questions, the speaker expects one alternative more likely to be true over the opposing alternative. A syntactic account by Holmberg (2015) merely suggests that the unmarked alternative for a negative polarity questions is [¬p] rather than [p], Roelofsen & Farkas (2015) define it as the highlighted alternative but they make a distinction between the highlighted alternative and bias. Bolinger (1978) defines polar questions as a singleton set wherein the negative polar question denotes a singleton set containing the nucleus proposition [¬p]. Bhadra (2017) further provides us with a brief overview of the different approaches that have been used to define bias in the current semantic literature. Bias has been argued as a pragmatic presupposition (Rohde 2006, Caponigro and Sprouse 2007), as a semantic presupposition (Guerzoni 2003, 2004), as a conversational implicature (Ladd 1981, Buring and Gunlogson 2000), as compelling evidence (Krifka 1995, Van Rooy 2003, Romero and Han 2004) and as assertion (Reese 2007, Asher and Reese 2005, Asher and Reese 2007). Bhadra in her work on the Bangla evidential particle naki argues for a definition in along the lines of the assertion analysis but essentially different from the previous work under the same model. The purpose of this work is not to argue or provide evidence in favor of one semantic analysis of the other. As the study of biased answer forms in Hindi suggests, there is a component in the answer that is anaphoric to the question antecedent and agrees or disagrees with it. The second component of a biased answer is the explicit prejacent which selects the propositional alternative which is true. I leave the exact semantic modelling of the bias open as long as this two-part answer system is accounted for. 100 3.7 Other Biased Questions Until now we have explicitly dealt with negative polar questions which are biased, as opposed to positive polar questions which are neutral. Having a negation in the core proposition is not the only way to formulate a biased question. In fact, biased questions can be formulated at least in four other ways in addition to negative polar questions. In this section we briefly overview the other four types of biased questions available in Hindi. Borkin (1971) noted that minimizer NPIs like a red cent, a damn thing, lift a finger etc. in a polar question make the negative alternative as the unmarked alternative over its counterpart. The same is true for Hindi polar questions with minimizer NPIs like ratti bhar bhii ‘even a little’ as is seen in (44). As the answer forms suggest in (45), specially (45iii & iv), it is very clear that the negative alternative was expected to be true, and if the positive alternative is selected it needs further clarification or extension which makes it clear that this was the unexpected choice. 44. Q: kya raaghav rattibhar-bhii anu-ko pasand kartaa Q Raghav a little-even anu-DAT like do.HAB.3FSg hai AUX.PRES.3FSg ‘Does Raghav like Anu even a little?’ 45. A: i. nahi kartaa yaar Neg do.HAB.3FSg Voc ‘No he doesn’t.’ (it’s unsurprising) ii. haaN kartaa hai naa yes do.HAB.3FSg AUX.PRES.3FSg Prt ‘Yes, he does.’ 101 iii. sahi bol rahe ho yaar, nahi kartaa right say PROG.2Msg AUX Voc NEG do.HAB.3FSg ‘You are right, he doesn’t.’ iv. haaN kartaa hai naa, tumhe kisne yes do.HAB.3FSg AUX.PRES.3FSg Prt you-DAT who kahaa nahi kartaa say NEG do.HAB.3FSg ‘Yes he does, who told you he doesn’t.’ A third way to formulate biased questions is to have stressed weak NPIs like ‘any’ and ‘ever’. Hindi examples are given in (46). NPIs in Hindi are formed by the combination of wh-words and the ‘even’ particle bhii for example kuch bhii ‘anything’, koi bhii ‘anyone’, kahin bhii ‘anywhere’, kabhi bhii ‘ever’ etc. As suggested by the implications present in the answer forms in (47), it is clear that the negative alternative is the unmarked/expected answer. 46. Q: kya raaghav-ne anu ke-liye KUCH-bhii kiyaa hai Q Raghav-ERG anu for anything do.3FSg AUX.PRES.3FSg ‘Has Raghav done ANYTHING for Anu?’ 47. A: i. kahaaN, kabhii-bhii nahi yaar where 14 anytime NEG Voc ‘Never.’ ii. haaN kiyaa to hai yes do.3FSg Prt AUX.PRES.3FSg ‘Yes, he has.’ 14 kahan ‘where’ – asked as a rhetorical question loosely translated to – ‘why are you even wondering? it should be obvious.’ – is considered to be a reliable marker of a biased question. 102 A fourth way to formulate biased questions which are biased towards the negative alternative is through a combination of word order, intonation and stress. In (48) the preposed verbal complex is stressed and suddenly there is a bias towards the negative alternative. The speaker when asking this question has a belief of the negative alternative but still wants to confirm this as true. 48. Q: PADHII thii kya raaghav-ne kitab read.PFV.3FSg AUX.PST.3FSg Q Raghav-ERG book ‘Did Raghav READ the book?’ 49. A: i. nahi yaar, kahaaN no Voc where ‘No he didn’t.’ ii. haaN paDhii thii naa yes read.PFV.3FSg AUX.PST.3FSg Prt ‘Yes, he has.’ Hindi also attests to tag questions which introduce bias in polar questions. However, this fifth and final way to formulate biased questions is different from the ways mentioned earlier in that this introduces a bias for whatever is the nucleus proposition in the question. So if the polarity of the nucleus proposition is positive then the tag question has a positive bias, as seen in (50-51), and if the nucleus proposition has negative polarity, the bias is also towards the negative alternative as in (52-53). 50. Q: raaghav-ne kitab paDhii hai, hai naa? Raghav-ERG book read.PFV.3FSg AUX.PRES.3FSg AUX Prt ‘Raghav has read the book, hasn’t he?’ 103 51. A: i. haaN paDhii hai yes read.PFV.3FSg AUX.PRES.3FSg ‘Yes, he has.’ ii. nahi yaar nahi paDhii No Voc NEG read.PFV.3FSg ‘No, he hasn’t.’ 52. Q: raaghav-ne kitab nahi paDhii hai, hai naa? Raghav-ERG book NEG read.PFV.3FSg AUX.PRES.3FSg AUX Prt ‘Raghav hasn’t read the book, has he?’ 53. A: i. nahi yaar nahi paDhii No Voc NEG read.PFV.3FSg ‘No, he hasn’t.’ ii. haaN paDhii hai naa yes read.PFV.3FSg AUX.PRES.3FSg Prt ‘Yes, he has.’ The usual answer patterns to all these questions are the same as that to negative polar question with similar prosodic patterns. The additional answer forms given here just reinforce the bias in the question, consequently confirming that these questions prefer one alternative over another to be true. None of these answer forms will be felicitous as answers to a positive/neutral question in a out-of-the-blue context. Another test for biased questions is Sadock’s (1971, 1974) diagnostics for illocutionary force test. Sadock (1971, 1974) argues that certain co-occurrence restrictions hold between certain discourse words and the illocutionary force of a sentence. For instance, discourse markers like after all, yet can only occur with assertions while phrases like tell me, by any chance can only occur with questions. Asher & Resse (2005) use this diagnostic to conclude that biased questions are of the complex speech act type of assertion • question, i.e., biased 104 questions are both an assertion and a question. In (55) we see that both an assertion discourse marker and an interrogative discourse marker are felicitous with all kinds of biased questions attested in Hindi whereas the assertions discourse marker is not felicitous with the positive/neutral polar question as seen in (54). 54. i. # akhirkaar, kya raaghav-ne kitab paDhii thii after all Q Raghav-ERG book read.PFV.3FSg AUX.PST.3FSg ii. bataao mujhe, kya raaghav-ne kitab paDhii thii tell me Q Raghav-ERG book read.PFV.3FSg AUX.PST.3FSg 55. i. akhirkaar, kya raaghav-ne kitab nahi paDhii after all Q Raghav-ERG book NEG read.PFV.3FSg thii AUX.PST.3FSg ii. bataao mujhe kya raaghav-ne kitab nahi paDhii tell me Q Raghav-ERG book NEG read.PFV.3FSg thii AUX.PST.3FSg iii. akhirkaar, kya raaghav rattibhar-bhii anu-ko pasand after all Q Raghav a little-even anu-DAT like kartaa hai do.HAB.3FSg AUX.PRES.3FSg iv. bataao mujhe, kya raaghav rattibhar-bhii anu-ko pasand tell me Q Raghav a little-even anu-DAT like kartaa hai do.HAB.3FSg AUX.PRES.3FSg v. akhirkaar, kya raaghav-ne anu ke-liye KUCH-bhii after all Q Raghav-ERG anu for anything kiyaa hai do.3FSg AUX.PRES.3FSg 105 vi. bataao mujhe, kya raaghav-ne anu ke-liye KUCH-bhii tell me Q Raghav-ERG anu for anything kiyaa hai do.3FSg AUX.PRES.3FSg vii. akhirkaar, PADHII thii kya raaghav-ne after all read.PFV.3FSg AUX.PST.3FSg Q Raghav-ERG kitab book viii. bataao mujhe, PADHII thii kya raaghav-ne tell me read.PFV.3FSg AUX.PST.3FSg Q Raghav-ERG kitab book ix. akhirkaar, raaghav-ne kitab paDhii hai, after all Raghav-ERG book read.PFV.3FSg AUX.PRES.3FSg hai naa AUX Prt x. bataao mujhe, raaghav-ne kitab paDhii hai, tell me Raghav-ERG book read.PFV.3FSg AUX.PRES.3FSg hai naa AUX Prt 3.8 Conclusion Biased questions in Hindi can be formed in different ways. They are different from positive/neutral polar questions in that they convey a bias by virtue of which they cannot occur in neutral situations but presuppose a specific context. This chapter has presented a detailed study of negative polar 106 questions and their answer forms. The answer patterns in addition to the prosodic patterns and the discourse features that can attach to these questions confirm that they are different from positive polar questions. A detailed study of these three features has led us to various conclusions about Hindi clausal structure, the nature of these biased questions and the nature and function of polarity particles. Answers to biased questions provide us with evidence that polarity particles are anaphoric in nature and that they are sensitive to the polarity and the content of the antecedent. The syntactic representation of such anaphoric polarity particles has been suggested to be structurally high up in the extended left periphery as a performative head in the response layer. This led to a cartographic modelling which can faithfully capture the different distributions, functions and meanings associated with these particles. In addition, we have also concluded at the end of this chapter that the primary distinction between these questions is the bias inherent in these questions. Though we have not gone into detail on the semantic formulation of bias we have examined the effect that biased questions have on their answer forms which has further contributed to extending our understanding of a range of related facts. For instance, the dual functionality of the polarity particles, their anaphoricity and their multiple representation in the cartographic architecture would have all been impossible to detect had we just considered positive/neutral polar questions. In the next chapter we extend our understanding of polar questions and their answer forms to alternative ‘or not questions’ and then to free choice alternative questions by attempting a detailed study of their answer forms. 107 Chapter 4: Other Types of Question - Answers In this chapter we take the next step in the study of polar questions and their answers in Hindi. Following Hamblin (1958)’s interpretation of questions, we have adopted the idea that questions are interpreted as the set of focused alternatives which are possible answers to that question. In the previous two chapters we examined polar questions which in their alternative set have {p, ¬p} as their possible answers. There is yet another kind of polar question with the same alternative set, the ‘or not’ questions which I term ‘Polar Alternative Questions’ (PAQs). Syntactically, PAQs are different from general polar questions in the sense that they explicitly mention the exhaustive set of alternatives as possible answers in the question itself, as can be seen in (1). Interpretively, they have the same alternative set as possible answers as the polar questions in (2) and (3), though there is no bias or highlighted alternative in (1) as there is in (2) and (3). The answer to all these questions is selecting one proposition which is true from the alternative set. 1. Did John read a book or not? 2. Did John read the book? 3. Did John not read the book? Yet another type of question that explicitly mentions the exhaustive set of alternatives which constitute possible answers to the question are of the type (4) wherein the alternatives are independent propositions. I call them ‘Alternative Choice Questions’ (ACQs). 4. Did John read a book or a newspaper? 108 ACQs with PAQs are similar in that they both explicitly mention the set of alternatives and answering these questions merely is selecting the alternative which is true. They are different of course in the alternative set that are possible answer forms. While PAQs as their name suggests have the positive and negative polarity of the nucleus proposition as their answer set, with ACQs the answer set constitutes of alternative propositions explicitly mentioned with a disjunction in the question. ACQs can be considered to be one step away from wh-questions in terms of their interpretation. While ACQs are restricted in their alternative set to explicitly mentioned choices in the question, there is no such constraint on wh-questions. The set of alternatives invoked by a wh- question is often unbounded and hence not inherently defined though they could be bounded pragmatically by the discourse context in which those questions are set up. Thereby as also mentioned in Chapter 1, the categorization of different kinds of questions then is based on the alternative set they invoke as their possible answer forms. Polar questions are questions which have as their alternative set only {p, ¬p} as possible answer options. We then have alternative questions which explicitly mention the possible answer alternatives in their questions and finally we have wh-questions with a seemingly unbounded alternative set of possible answers. We do have overlap between these categories. As we saw, PAQs are polar questions in the sense that the alternative set consists of positive and negative polarity propositions as its possible answers but they are also alternative questions because both those alternatives are explicitly mentioned in the question. There are subtypes as well, as illustrated in (5). This is a polar question given that the alternative set has positive and negative propositions as its possible answer forms but the entire proposition is not under the scope of being questioned, only the contrastively focused constituent, 109 as can be seen in the alternative set given in (6). I call these ‘Polar questions with narrow focus’ (PFQs). 5. Was it John who read the book? 6. [Yes, it was john who read the book, No, it was not john who read the book] This chapter explores these different kinds of questions – PAQs, ACQs and PFQs in Hindi and the answers to these questions. The aim of this chapter is to provide an analysis for all these questions and their answer forms with the eventual objective of discovering what they may say about general aspects of Hindi clausal structure. Following Han & Romero (2004a) I argue for a Q operator movement + ellipsis hybrid approach to account for PAQs and ACQs. PFQs receive a straightforward account being derived by interrogative operator movement to spec,CP to get sentential scope though we observe certain interesting properties of the Hindi question particle kya interacting with such questions, which sheds more light on nature of kya. The answers to all these questions, specifically the fragment answer forms are argued to be derived uniformly from verb stranding υP ellipsis. The chapter is organized as follows: section 4.1 deals with PAQs and their answer forms in Hindi. 4.2 explores in detail the account of ACQs and their answer forms and proposes an analysis based on Han & Romero (2004a) for alternative questions. Section 4.3 then considers PFQs and section 4.4 concludes. 110 4.1 Introduction to Polar Alternative Questions (PAQs) in Hindi ‘Or not’ questions could be argued to be alternative questions rather than yes/no questions given that in such questions the alternatives are explicitly mentioned in a coordination structure with a disjunct. With ‘or not’ questions, the alternatives happen to be the affirmative and the negative polarity (as illustrated in 7). They have been argued in the literature to explicitly pose a choice between two polarity phrases joined by a disjunction with material elided from the second disjunct under identity with the first disjunct (Quirk et al. 1985, Huddleston & Pullum 2002, Han & Romero 2004). As stated earlier, logically they are equivalent to general polar questions by virtue of generating the same set of alternatives as possible answers. 7. Did John read the book or not? - [Did John read the book] or [Did John not read the book]? ‘Or not’ questions in Hindi can be argued to have a similar structure. The choice between positive and negative polarity phrase is explicitly stated in the question with the disjunction yaa or ki. ki is an interrogative disjunct in Hindi and cannot occur in declarative constructions as can be seen in (8). It can thus be argued to have a [+Q] feature. 8. raaghav-ne kitab yaa/#ki akhbaar paDhaa thaa Raghav-ERG book or newspaper read.PFV.3MSg AUX.PRES.3MSg ‘Raghav read a book or a newspaper.’ PAQs in Hindi occur in two forms. In (9), we have the ‘or not’ form where the positive polarity nucleus proposition is followed by or not. The second structure associated with PAQs in Hindi is given in (10) where the positive polarity nucleus proposition is followed by or not υP – where the 111 entire verbal sequence is present in addition to or not. At this point, the only difference between these two types of PAQs seems to be the material pronounced in the second disjunct. 9. ‘Or not’ PAQs Q: (kya) raaghav-ne kitab paDhii thii Q Raghav-ERG book read.PFV.3FSg AUX.PRES.3FSg yaa/ki nahi? Or not ‘Did Raghav read the book or not?’ 10. ‘Or not υP’ PAQs Q: (kya) raaghav-ne kitab paDhii thii Q Raghav-ERG book read.PFV.3FSg AUX.PRES.3FSg yaa/ki nahi paDhii thii? or not read.PFV.3FSg AUX.PRES.3FSg ‘Did Raghav read the book or not?’ A crucial difference between PAQs and general polar questions is that they cannot be answered with polarity particles yes or no. As seen in (12iii, 12iv, 14iii and 14iv) polar questions with disjunctions/PAQs in Hindi cannot be answered simply by polarity particles. This difference in answer patterns suggests that even though semantically these questions are interpreted as invoking {p, ¬p} as alternatives, there is a difference between answering PAQs and polar questions. Answering a polar question, entails the choice of one of the contextually salient alternatives from the alternatives set. A PAQ requires as its’ answer, one of the explicitly mentioned alternatives. Thereby, even if logically, in both cases, the alternative set is the same, syntactically the answer forms vary as the answer forms to alternative questions depend on explicitly mentioned alternatives in the questions structurally whereas this is not a requirement of answer forms to general polar questions. We further saw in the previous chapter how polarity particles are in a way 112 anaphoric, referring back to the nucleus proposition in the question and since PAQs do not have a singleton proposition to refer back to they do not serve as appropriate antecedents to polarity particles. This is why simple polarity particles as answers are acceptable for neutral polar questions but not for PAQs. The answer of an ‘or not’ question is the repetition of the affirmative or negative nucleus proposition. 11. Q: (kya) raaghav-ne kitab paDhii thii Q Raghav-ERG book read.PFV.3FSg AUX.PRES.3FSg yaa/ki nahi? or not ‘Did Raghav read the book or not?’ 12. A: i. (haaN) paDhii thii Yes read.PFV.3FSg AUX.PRES.3FSg ‘Yes, Raghav read the book.’ ii. (nahi) nahi paDhii thii no NEG read.PFV.3FSg AUX.PRES.3FSg ‘No, Raghav didn’t read the book.’ iii. #haaN yes ‘Yes, Raghav read the book.’ iv. #nahi no ‘No, Raghav didn’t read the book.’ 13. Q: (kya) raaghav-ne kitab paDhii thii Q Raghav-ERG book read.PFV.3FSg AUX.PRES.3FSg yaa/ki nahi paDhii thii? Or NEG read.PFV.3FSg AUX.PRES.3FSg ‘Did Raghav read the book or not?’ 113 14. A: i. (#haaN) paDhii thii Yes read.PFV.3FSg AUX.PRES.3FSg ‘Yes, Raghav read the book.’ ii. (#nahi) nahi paDhii thii no NEG read.PFV.3FSg AUX.PRES.3FSg ‘No, Raghav didn’t read the book.’ iii. #haaN yes ‘Yes, Raghav read the book.’ iv. #nahi no ‘No, Raghav didn’t read the book.’ Crucially, the two kinds of PAQs have one more difference in addition to the amount of material pronounced in the second disjunct. Though polarity particles by themselves are not viable answers, they can be a part of the answers to PAQs, although they are only attested in one kind of PAQ, - the ‘or not’ kind as can be seen in (12i) and (12ii). Polarity particles are not attested in answer forms to ‘or not υP’ PAQs as seen in (14i) and (14ii). In both cases the answers still entail choosing one of the explicitly mentioned alternatives in the question. Thereby this difference in answer forms can be argued to be a result of the difference in the explicitly mentioned salient alternatives in the questions. This leads us to conclude that the exact formulation of the {p, ¬p} alternative set varies in the two question forms. I take this difference in answer patterns to suggest that the two kinds of PAQs differ in how they are derived. In the next subsection I present an analysis for the two kinds of PAQs attested in Hindi. 114 4.1.1. Analysis of PAQs in Hindi One possible analysis of both the PAQs would be to suggest that the disjunction particle yaa/ki takes the higher PolfocP as its two disjuncts. Recall that in Chapter 2, I have argued for two polarity heads, the higher of which is merged above IP only when a contrastively focused polarity head is present. The two salient alternatives in both PAQs are the positive and negative counterparts of the nucleus proposition, i.e., {p, ¬p} along with the corresponding polarity particles, formalized as in (15). 15. {[Yes, Raghav read the book], [No, Raghav did not read the book]} Thereby the level of disjunction is PolfocP as the polarity is explicitly contrastively focused in the two disjuncts. The second disjunct in the question then undergoes ellipsis under identity with the first disjunct as its antecedent. An ellipsis account for PAQs has been proposed by Han & Romero (2004) for English. For Hindi, there are two options for ellipsis in the second disjunct. Either the entire IP is elided leaving a stranded Polfoc head in the second disjunct or the υP is elided in which case the second disjunct has the entire verbal sequence. The first option generates the ‘or not’ PAQs whereas the second options results in the ‘or not υP’ PAQs. The schematization follows. 115 16. 17. (kya) raaghav-ne kitab paDhii thii yaa/ki [PolfocP nahi, [ IP [υP raaghav-ne kitab ti] (nahi paDhii)i thii]] 18. (kya) raaghav-ne kitab paDhii thii yaa/ki [PolfocP nahi, [ IP [υP raaghav-ne kitab ti] (nahi paDhii)i thii]] Further evidence for this difference in the ellipsis site of two different kinds of PAQs attested in Hindi comes from adverb inclusion tests. ‘or not’ PAQs attest to obligatory adverb inclusion (as seen in 19) corroborating the fact that the a higher clausal ellipsis is attested in such PAQs. In (20) the selection of ¬p alternative can only be interpreted as ‘Raaghav didn’t read the book twice’ which means that the adverb is necessarily in narrow focus in the nucleus proposition. υP ellipsis is usually argued to have optional adverb inclusion and such is the case with ‘or not υP’ PAQs as well (as seen in 21-22). The selection of ¬p alternative as an answer to the question in (21) can be Figure 14: Or not (υP) PAQs 116 interpreted in two ways, one in which the adverb is in narrow focus in the nucleus proposition and one in which the adverb is not present in the nucleus proposition at all. Thereby, as is clear, while choosing the ¬p alternative, only in the case of ‘or not υP’ PAQs can we take it to have an interpretation where the adverb is not included. 19. Q: (kya) raaghav-ne kitab do baar paDhii thii Q Raghav-ERG book two times read.PFV.3FSg AUX.PRES.3FSg yaa/ki nahi? or no ‘Did Raghav read the book twice or not?’ 20. A: nahi, nahi paDhii thii no not read.PFV.3FSg AUX.PRES.3FSg ‘Raghav did not read the book twice.’ Possible Continuation – ‘He only read it once.’ *‘Raghav did not read the book at all.’ 21. Q: (kya) raaghav-ne kitab do baar paDhii thii Q Raghav-ERG book two times read.PFV.3FSg AUX.PRES.3FSg yaa/ki nahi paDhii thii? or NEG read.PFV.3FSg AUX.PRES.3FSg ‘Did Raghav read the book twice or not?’ 22. A: nahi paDHii thii NEG read.PFV.3FSg AUX.PRES.3FSg ‘Raghav did not read the book twice.’ Possible Continuation – ‘He only read it once.’ OR ‘Raghav did not read the book at all.’ 117 As per this analysis, the difference between the two PAQs is constrained by the ellipsis site though the level of disjunction is the same. However, if the only difference between the two questions is the size of ellipsis and they have the exactly the same disjuncts, it is difficult to account for the different answer patterns. Given that answers are simply selection of one of the explicit alternatives, there is no obvious way to rule out answer forms like (12i) and (12ii) to questions of the type in (13). But as we can see in (14i) and (14ii), those answer forms are not attested. The difference in answer forms suggest that the two kinds of PAQs have different alternatives as their explicit choices. Though they both exhibit difference in the size of the material left after ellipsis I argue that that is due to the difference in the size of the disjunct structures rather than the difference in ellipsis site exclusively. I propose that the structure that after ellipsis results in the ‘or not’ PAQs in (9 or 11) is given in (23) and the underlying structure for ‘or not υP’ PAQs like (10 or 13) is given in (25). The disjunctions are at different levels which determine the different ellipsis sites. In the case of ‘or not’ PAQs the disjunction yaa/ki takes the higher PolfocP as its two disjuncts, as the polarity is explicitly contrastively focused in the two disjuncts. The higher Polfoc head is motivated and merged into the structure and hosts the contrastively focused polarity particle. The second disjunct undergoes IP ellipsis as IP is the complement of the Polfoc head and can be elided under semantic identity. The ellipsis as proposed following Merchant (2005) is triggered by an E feature on the Polfoc head. The corresponding tree structure is in (24). 23. (kya) raaghav-ne kitab paDhii thii yaa/ki [PolfocP nahi, [ IP [υP raaghav-ne kitab ti] (nahi paDhii)i thii]] 118 24. The two alternatives in alternative questions in general can be argued to be clausal and thereby are at least at the IP level. For ‘or not’ PAQ cases the contrastively focused polarity licenses a higher polarity head and thereby licenses the ellipsis of a higher structure (IP) but when there is no contrastively focused polarity as is the case with ‘or not υP’ PAQ structures the two disjuncts are IP level disjuncts. So, in the case of ‘or not υP’ PAQs disjunction yaa/ki takes IP as its two disjuncts. The second disjunct undergoes verb stranding υP ellipsis triggered by an E feature on the lower IP-internal Pol head which takes υP as its complement as shown in (25 and 26). As I have motivated in chapter 2, the ellipsis feature can only be hosted by polarity heads (more on this in chapter 5). In ‘or not’ PAQ structures the higher Polfoc head carries the ellipsis feature and licenses the ellipsis of its IP complement. For ‘or not υP’ PAQ structures the higher polarity head is not merged, the disjunction is at the IP level and thereby the only possible host of the ellipsis Figure 15: ‘or not’ PAQs 119 feature is the IP-internal default polarity head which when it hosts the ellipsis feature licenses the ellipsis of its υP complement. 25. (kya) raaghav-ne kitab paDhii thii yaa/ki [υP raaghav-ne kitab ti] (nahi paDhii)i thii 26. Motivating the two kinds of PAQs from the difference in disjunct sizes rather than directly from the difference in ellipsis sites accounts for the different answer patterns that these question forms attest to. Furthermore, it is incongruent to account for ‘or not υP’ PAQs from a structure with the higher PolfocP as the disjunct level as any general constrain on ellipsis of the nature of MaxElide would argue against a υP ellipsis when a higher-level IP Ellipsis is possible. Hence, I adopt the analysis which differentiates between levels of disjunction rather than difference in ellipsis site only as potential differences between the two kinds of PAQs. Both disjunction structures have clausal disjuncts and in both the cases ellipsis is triggered by the two polarity heads. The structures Figure 16: ‘or not υP’ PAQs 120 of ‘or not’ questions in Hindi thereby provide us with additional evidence of two polarity phrases and consequently two levels of disjunctions with two potential ellipsis sites. 4.1.2. Analysis of Answers to PAQs in Hindi In Chapter 2 of this dissertation (refer to section 2.5.3) I proposed that fragment answer forms are derived from verb stranding υP Ellipsis in Hindi. As argued, fragment answers are derived by ellipsis which is triggered by an Eans feature and the ellipsis site is determined by the head which hosts this Eans feature. Only a polarity head is claimed to be able to host the ellipsis feature and two such polarity heads have been motivated in the Hindi clausal structure, one right above above υP and one in the C domain right above IP, giving way to two ellipsis sites. IP ellipsis would result in simple polarity particles as fragment answers. υP ellipsis, on the other hand, accounts for the obligatory verb stacking answer pattern. We can extend the analysis to answers to PAQs as well. For ‘or not’ PAQs repeated below in (27), the attested answer forms are as given in (28). 27. (kya) raaghav-ne kitab paDhii thii yaa/ki [PolfocP nahi, [ IP [υP raaghav-ne kitab ti] (nahi paDhii)i thii]] 28. A: i. (haaN) paDhii thii Yes read.PFV.3FSg AUX.PRES.3FSg ‘Yes, Raghav read the book.’ ii. (nahi) nahi paDhii thii no NEG read.PFV.3FSg AUX.PRES.3FSg ‘No, Raghav didn’t read the book.’ 121 And for the ‘or not υP’ PAQs in (29) the attested answer forms are again repeated in (30). 29. Q: (kya) raaghav-ne kitab paDhii thii yaa/ki [PolfocP nahi, [ IP [υP raaghav-ne kitab ti] (nahi paDhii)i thii]] 30. A: i. (#haaN) paDhii thii Yes read.PFV.3FSg AUX.PRES.3FSg ‘Yes, Raghav read the book.’ ii. (#nahi) nahi paDhii thii no NEG read.PFV.3FSg AUX.PRES.3FSg ‘No, Raghav didn’t read the book.’ There is straightforward derivation of answer forms to both types of PAQs if the analysis of the PAQs is as argued for in the previous section. The answers to both kinds of PAQs attest to a verb stacking pattern where the entire verbal sequence is repeated from the question antecedent. This calls for a verb stranding υP ellipsis analysis to derive the answer forms. The difference between the answer forms to ‘or not’ PAQs and ‘or not υP’ PAQs is that polarity particles are optionally present in the answer forms to the former types of questions. However, answer forms to ‘or not υP’ do not include polarity particles acting as lexical exponence of polarity features [+/-] though as we can see in (29ii), the sentential negation nahi is obviously present in the complex head (nahi paDhii) it forms with the verb. As mentioned earlier the answer to PAQs is simply one of the explicit choices given in the question antecedent, I argue that this difference in answer forms is a direct result of the difference in the two kinds of PAQs. Given that the two disjuncts of different sizes in the two kinds of questions, we can account for the different answer forms for both these questions, specifically, the absence of polarity particles in the answer form for ‘or not υP’ PAQs. For ‘or not’ PAQs the disjunction level is PolfocP 122 and since polarity particles are merged as the Polfoc head they are crucially present in both the disjuncts for ‘or not’ PAQs. This can be seen in the structure in (31). Polarity particles are not present in the answer forms for ‘or not υP’ PAQs as they are not present in the explicitly mentioned disjuncts in the question itself which are at the IP level as represented in (32). Furthermore, after one alternative from among {p, ¬p} is chosen from the question, the selected alternative undergoes ellipsis, specifically verb stranding υP ellipsis. The structural representation of the answer form (28ii) is shown in (31) and that of (30ii) is shown in (32). In both these cases the IP-internal Pol head bears an Eans ellipsis feature which triggers the ellipsis of its υP complement. 31. Figure 17: Answer to ‘or not’ PAQs – nahi, [υP raaghav-ne kitab ti] (nahi paDhii)i thii] 123 32. Thereby, the fact that answers to ‘or not υP’ PAQs do not include polarity particles is attested by the non-existence of polarity particle in the explicit disjuncts present in the questions. Furthermore, the fact that simple polarity particles are not attested as answer forms to PAQs in general is accounted for by the fact that the particles cannot be anaphoric to an unambiguous antecedent as both the alternatives are explicitly mentioned in the question. Concluding this sub-section on PAQs and their answer forms in Hindi, I have argued for an ellipsis analysis for the PAQs constrained by two distinct levels of disjunction. The answer forms are mostly similar to that of answers to general polar questions and hence a verb stranding υP ellipsis approach can adequately account for the attested patterns. In the next sub-section, we proceed to analyze the alternative choice questions and their answer forms in Hindi. Figure 18:Answer to ‘or not υP’ PAQs – [υP raaghav-ne kitab ti] (nahi paDhii)i thii] 124 4.2 Introduction to Alternative Choice Questions (ACQs) in Hindi Yet another type of question with a disjunction in it are alternative choice questions (ACQs) characterized by logically independent propositions in a coordinated structure conjoined by a disjunct like in (4) repeated in (33) below. Such constructions in English are ambiguous between a polar question reading and an alternative question reading as can be seen. 33. Did John read a book or a newspaper? y/n reading – Did John read something (either a book or a newspaper)? alt reading – Did John read a book or Did John read a newspaper? The two readings can be made clear by the answer forms that such questions attest as illustrated in (35). The y/n reading is corroborated by the availability of polar particles in answers and the alt reading is also substantiated given that the very same question can be answered by choosing one of the explicit stated alternatives as its answer. 34. Q: Did John read a book or a newspaper? 35. A: i. Yes, he did/No he didn’t (read either a book or a newspaper.) – y/n reading ii. He read a book./He read a newspaper. – alt reading Studies in the literature have argued for certain prosodic features which can disambiguate the two readings. Bartels (1999), Han & Romero (2004) and Hoeks (2019) have argued that both the disjuncts need to have pitch accents in order for the alt reading to be available. In addition to pitch accents on the disjuncts there is also another prosodic difference. For the y/n reading to obtain, there is a final rise contour on the question as is typical for questions but for the alt reading a final 125 fall contour is attested. Pruitt (2008) and Pruitt & Roelofsen (2013) present experimental data corroborating the same. 36. a. Did Mary drink MINERAL WATER or LEMONADE? - alt reading H*/L* H- H* L-L% b. Did Mary drink mineral water or lemonade? - y/n reading H*/L* H-H% - Hoeks (2019: 2) Questions of the type in (33) are also attested in Hindi (in 37) though predominantly they only have a y/n reading. Some speakers do attest to an alt reading with the same structure but certain prosodic features need to be strictly adhered to 15 , to make the alt reading available. 37. Q: (kya) raaghav-ne chai yaa kofi pii? Q Raghav-ERG tea or coffee drink.PFV.3FSg ‘Did Raghav drink tea or coffee?’ Dayal (2016) and Bhatt & Dayal (2020) argue that there are essentially three prosodic differences that can disambiguate the two readings with the same surface string in (37). First, as in English, the two disjuncts need to have pitch accents in order for the alt reading to be available in Hindi. Secondly, there is a pause in between the two disjuncts in this reading as in this reading the disjuncts are considered to be separate prosodic units. No such pause is required with a y/n reading. Finally, similar to English, the key difference is found to be in the final intonation pattern. The surface string when pronounced with a final rise contour is interpreted as a y/n reading whereas 15 Most native speakers find an alt reading difficult to attain even after the prosodic conditions are respected. The difficulty with an alt reading with this surface string could be due to the availability of another way to form alternative choice questions in Hindi. 126 the alt reading is obtained with a final falling contour. The prosodic differences are illustrated in (38). 38. a. (kya) raaghav-ne chai yaa/*ki kofi pii? ↑ - y/n reading b. (kya) raaghav-ne [chai]F // yaa/ki [kofi]F pii? ↓ -alt reading In addition to this prosodic featural differences between the two readings Hindi has other ways to form ACQs. Note that in (38a) the only possible disjunction marker is yaa. ki being an interrogative disjunction marker cannot occur in polar questions with a disjunction in them. This signals that the disjunction structure in (38a) is a local disjunction and does not get interrogative clausal scope. The analysis we adopt for ACQs in Hindi will support this observation. The unmarked way to formulate ACQs in Hindi is illustrated in (39), which is minimally different in surface form from their predominantly y/n reading counterpart in (38a). The surface string in (39) is crucially different from (37) in that that the second disjunct in (39) follows after the verbal sequence of the first disjunct. Crucially, this way of formulating alt reading is the unmarked way to do so as it provides an unambiguous alternative question reading. 39. Q: (kya) raaghav-ne chai pii yaa/ki kofi (pii)? - alt reading Q Raghav-ERG tea drink.PFV.3FSg or coffee drink.PFV.3FSg alt reading - ‘Did Raghav drink tea or coffee?’ Confirmation of the (un)availability of the reading(s) to these ACQs come from the answers attested. (37) can be answered with (40i), (40ii) or (40iii) but (39) can only have (40iii) as a viable answer option. 40. A: i. haan (chai/kofi) pii - y/n reading Yes tea/coffee drink.PFV.3FSg ‘Yes, Raghav drank something (either tea or coffee).’ 127 ii. nahi pii - y/n reading NEG drink.PFV.3FSg ‘No Raghav didn’t drink anything.’ iii. (#haan) chai/kofi pii - alt reading yes tea/coffee drink.PFV.3FSg ‘Raghav drank tea.’/‘Raghav drank coffee.’ The presence of polarity particles in answer forms (40i) and (40ii) suggest that the question which has a disjunction in it is not an alternative choice question but a polar question. And the fact that (40iii) is also an attested answer form clearly suggests that such questions also have an alternative choice reading. 4.2.1. Han & Romero (2004) There are essentially two main approaches to ACQs in the literature. Larson (1985) and von Stechow (1991) have essentially argued for a scoping mechanism to account for alternative questions where the operator associated with the disjunction takes a wider scope to force a choice between the two disjuncts as the answer to the question. Pruitt & Roelofsen (2011) and Uegaki (2014) on the other hand derive alternative questions from a disjunction of full polar questions. For the analysis of Hindi ACQs, I adopt an approach laid out by Han & Romero (2004) which is seemingly a combination of both these approaches. They, in their analysis of English ACQs propose an account which combines the scoping mechanism of a whether/Q operator laid out in Larson (1985) and an ellipsis account of either…or constructions by Schwarz (1999) to account for the properties exhibited by alternative questions. They argue that disjunction in questions is 128 always accompanied by an operator, an overt whether or a covert Q just like disjunction in declaratives is accompanied by an either operator. For an alt reading, this operator moves to spec, CP to take clausal scope and the second clausal disjunct undergoes ellipsis under strict identity with the first clausal disjunct. For a polar question reading, the disjunction operator remains in its local scope and essentially scopes under the polar interrogative operator which takes the clausal scope. The two analyses for the two readings are illustrated in (41). 41. Did John read a book or a newspaper? Y/n reading – PolQi ( ti) [did John read either [ a book or a newspaper] ] Alt reading – whetherQi did ti[John read a book] or [john read a newspaper] 4.2.2. Analysis of ACQs in Hindi For a polar question reading of ACQs in Hindi, I argue following Han & Romero(2004) that the disjunction operator remains in its local scope and essentially scopes under the interrogative operator which takes clausal scope. The schematization for a y/n reading of the ACQ structure is given in (42). The structure I have argued for is similar to the structure of neutral polar questions in Hindi motivated in chapter 1. There is an interrogative force operator in the Force domain which binds the question variable of the form [± Pol]. [± Pol] is originally merged in the IP-internal Pol head but moves to the C domain to get a sentential interrogative scope. The only trivial difference between neutral polarity questions and polar interpretation of ACQs is that in ACQs we have a local/DP level disjunction structure as shown in the figure below. 129 42. Having outlined the polar interpretation of ACQs, I now focus on the alt readings where the disjunction in ACQs plays a crucial role. I adopt Han & Romero’s (2004) analysis for the alt reading of ACQs in Hindi. Accordingly, I motivate that the whether/Q operator in Hindi ACQs which is a null operator moves to spec,CP to take clausal scope and the second disjunct undergoes ellipsis to derive the surface string attested in ACQs. Evidence for movement of the whether/Q operator comes from different island effects and answer forms for both readings in matrix and Figure 19: Account of y/n reading of ACQs in Hindi 130 embedded cases of ACQs. In (43) we have a CNPC structure which is considered to be an island for movement. When a disjunction structure is embedded inside a CNPC structure the disjunction cannot take wide scope and we do not get an alt reading. As we can see from the attested answer form in (44) the only available interpretation is that of a matrix polar question. 43. Q: tumhe kya ye baat pataa hai ki You-DAT Q this thing know AUX.PRES.3Sg that raaghav-ne chai yaa/ki kofi pii Raghav-ERG tea or coffee drink.PFV.3FSg ‘Do you know the fact whether Raghav drank tea or coffee?’ 44. A: i. haan pataa hai - matrix y/n reading Yes know AUX.PRES.3Sg ‘Yes, I know what Raghav drank.’ ii. #chai pii - alt reading - unavailable tea drink. PFV.3FSg ‘Raghav drank tea.’ Further evidence for movement of whether/Q comes from scope-marking constructions in Hindi. Han & Romero (2004) indeed use Hindi scoping constructions to prove that the whether/Q operator needs to take scope in the appropriate clause to get an alt reading at the clausal level. Getting to the appropriate clause might either involve movement or a scoping mechanism like in Hindi. For questions to take matrix scope Hindi employs the scope marker kya in the matrix clause as can be seen in (45). (46 and 47) shows us that only with the scope-marking kya can we get a matrix alt reading of the embedded disjunction as is clear from the attested answer form in (48). (48) is a viable answer to (46) but not to (47). Thus, we see that scope-marking kya is obligatory to get a matrix scope for an embedded ACQ as illustrated in (47). 131 45. Q: tumhe kya pataa hai ki sita kisse You-DAT what know AUX.PRES.3Sg that sita whom milii? met.PFV.3FSg ‘Who do you know Sita met?’ 46. Q: jaun kya sochtaa hai [ki Chandra-ne kofi pii thii John SM think be-PRES that Chandra-Erg coffee drink-Pfv Past yaa chai]? or tea ‘Does John think that Chandra drank coffee or tea?’ (both y/n and ACQ reading) (Han & Romero 2004: 540) 47. Q: Jaun sochtaa hai [ki Chandra-ne kofi pii thii John think be-PRES that Chandra-Erg coffee drink-Pfv Past yaa chai]? or tea *What does John think: that Chandra drank coffee or that Chandra drank tea?’ (alt-question) Available – ‘Does John think that Chandra drank tea or coffee?’ (matrix y/n) (Han & Romero 2004: 540) 48. A: jaun sochtaa hai ki Chandra-ne kofi/chai pii thii John think AUX.PRES that Chandra-Erg coffee/tea drink-Pfv Past ‘John thinks that Chandra drank coffee.’/‘John thinks that Chandra drank tea.’ Thus, we can conclusively prove that the null whether/Q operator needs to reach spec, CP position or have a scope marker in the clause that it is scoping over to get an ACQ interpretation. After having shown that ACQs require that the whether/Q operator undergo (covert) movement to 132 spec,CP for the operator to take a clausal scope and be interpreted as a question generating the two clausal disjuncts as alternatives, the next step is to show that the second disjunct undergoes ellipsis to derive the surface string of the attested ACQs. Though the surface string in English and Hindi ACQs is ambiguous, with the correct prosody the two interpretations can be disambiguated. Han & Romero (2004) take the specific prosodic properties that gives an ACQ interpretation to be evidence for clausal disjuncts one of which undergoes ellipsis. As mentioned earlier for an ACQ interpretation both the remnant in the second disjunct and its correlate in the antecedent need to receive pitch accents. A common use of such double focus is to represent two parallel propositions where the only thing different between them are the focused constituents while everything else remains the same (see Rooth (1985, 1992) for a formalization of contrastive focus). This suggests that the parallel disjuncts are clausal in nature. Furthermore, the unmarked structure for the unambiguous ACQ repeated here in (49), also suggests that the disjuncts are clausal (at least υP) rather than being NP/DP disjuncts. In (49) the remnant DP follows after the verbal sequence of the first disjunct clearly indicating that the disjunct size is not as small as a DP. Crucially with this word order, we do not get a y/n reading, only an alt reading is possible. 49. Q: (kya) raaghav-ne chai pii yaa/ki kofi (pii)? Q Raghav-ERG tea drink.PFV.3FSg or coffee drink.PFV.3FSg alt reading - ‘Did Raghav drink tea or coffee?’ Yet another piece of evidence comes from case marking patterns which can be used to distinguish between the two readings. As we can see in (50) if both the DPs are individually case-marked the only available reading is the alt reading but when only the second DP is case-marked in (51) a y/n reading is the predominant reading. Case marking on both the DPs in (50) suggest that the disjuncts 133 are at least at the VP/υP level. Independent case marking on both DPs would be harder to explain if they were only DP disjuncts conjoined at a DP level. The single case marking on y/n reading construction is explained if the two DPs are a part of a conjoined DP which is case-marked after the two DPs have been conjoined. 50. Q: (kya) raaghav sitaa-se yaa/ki geeta-se milaa? - alt reading Q Raghav sita-with or geeta-with meet.PST.3MSg Alt reading – ‘Did Raghav meet Sita or did he meet Geeta?’ 51. Q: (kya) raaghav sitaa yaa/ki geeta-se milaa? - y/n reading Q Raghav sita or geeta-with meet.PST.3MSg y/n reading - ‘Did Raghav meet Sita or Geeta?’ Having presented evidence and argumentation that the null whether/Q operator undergoes (covert) movement to spec,CP and that the two disjuncts are clausal we can adopt Han & Romero’s (2004) analysis for Hindi ACQs. A null whether/Q operator is generated at the edge of υP and undergoes covert movement to the spec,CP position to take clausal scope which gives an interrogative interpretation to the disjunction operator. Furthermore, a part of the second disjunct gets elided under identity with the first disjunct as its antecedent. In the case of a y/n reading the null disjunction operator either (non-interrogative version of whether/Q) is generated at the local DP coordinated structure and stays there. This also explains why the interrogative disjunction operator ki is not possible with this reading. ki is not attested as the local disjunction operator does not move higher to take a clausal scope and get interpreted as an interrogative. The polar interrogative operator takes the entire proposition under its scope and moves to the spec,CP position to receive an interrogative interpretation. The structures are illustrated in (52). 134 52. (kya) raaghav-ne chai yaa/ki kofi pii? Y/n reading – PolQi ( ti) [(kya) raaaghav-ne either [DP chai yaa kofi ] pii] Alt reading – whetherQi (kya) ti [ IP[υP raaghav-ne chaii pii]] yaa/ki [ IP[υP raaghav-ne kofii pii]] The surface string that unambiguously gives an alt reading is repeated in (53a and 53b). There are two different surface forms here constrained by the presence or absence of the verbal sequence in the second disjunct. (53a) can be derived by the verb stranding υP ellipsis of the second disjunct where the main verb has escaped the ellipsis site by virtue of head movement and the contrastively focused remnant has also escaped to a higher position allowing the υP to get elided under identity with the first disjunct as its antecedent. This is represented in (54). 53. a. (kya) raaghav-ne chai pii yaa/ki kofi (pii)? Q Raghav-ERG tea drink.PFV.3FSg or coffee drink.PFV.3FSg b. (kya) raaghav-ne chai pii yaa/ki kofi? Q Raghav-ERG tea drink.PFV.3FSg or coffee alt reading - ‘Did Raghav drink tea or coffee?’ 54. whetherQi (kya) ti [ IP[υP raaghav-ne chaii pii]] yaa/ki [IP (kofi)i [υP raaghav-ne ti tj ] (pii)j] However, (53b) cannot be due to verb stranding υP ellipsis as the verbal sequence also gets deleted in this case. There are two possible options that could potentially derive the surface form in (53b). The first option is that it is ellipsis of a higher structure, possible IP ellipsis. The material left behind which in this case is the contrastively focused remnant DP moves out of the ellipsis site, i.e., moves higher than IP following which the entire IP gets deleted (as shown in 55). 135 55. whetherQi (kya) ti [ IP[υP raaghav-ne chaii pii]] yaa/ki [(kofi)i [ IP [υP raaghav-ne ti tj ] (pii)j]] The second possible derivation is through a special kind of ellipsis called Gapping. Ross (1970) defines as deletion of a verb in the second coordinate in coordinate structure under identity with the first coordinate which may leave behind non-deleted contrastively focused remnants which have correlates in their antecedent is essentially different from υP/IP ellipsis as it does not necessitate deletion of one congruous constituent and is not licensed by a functional head. Traditionally, ellipsis like IP or υP/VP ellipsis have been argued to be ‘true’ ellipsis whereas other null anaphora structures like gapping (see Ross 1970; Johnson 1996, 2009, 2018; Steedman 1990; Hartmann 2000; Toosarvandani 2013, 2016) or stripping (see Reinhart 1991, Depiante 2000, Wurmbrand 2017, Johnson 2018, among others) have been argued to be restricted from this category of ‘true’ ellipsis. Traditionally, it has been argued that gapping leaves behind more than one remnant constituent whereas stripping deletes everything but one remnant constituent, though this distinction has been debated by Schwarz 1999, Hudson (1976), Neijt (1979), Reinhart (1991), and Johnson (1996) all of whom argue that gapping constructions may also leave one remnant. It is not the purpose of this work to detail the differences between the two kinds of null anaphora or to lay out the differences between gapping and stripping in Hindi. However, I do motivate a difference between IP ellipsis structures and stripping/gapping structures by virtue of their licensing heads in chapter 5. I essentially argue that IP ellipsis in Hindi can only be licensed by an explicit contrastively focused polarity head. Cases like (53b) schematized in (55) though they look like IP ellipsis are actually stripping/gapping constructions. The crucial difference between them 136 is that the stripping/gapping construction in Hindi is not licensed by a functional projection whereas IP ellipsis is licensed by the contrastively focused polarity head. However there seems to be no other empirical difference between the IP ellipsis patterns and the stripping/gapping structures. Crucially, I argue that the stripping/gapping structure in (53b) is also derived by movement of the remnant out of the ellipsis site followed by deletion of the ellipsis site. As we know movement of constituents is prohibited out of island structures. The prediction is that if a remnant from the island structure remains in an ACQ construction while rest of the island gets deleted the process involved is stripping/gapping but if constructing an island structure as one disjunct in an ACQ set up leads to ungrammaticality, the ungrammaticality can be explained by the prohibition against movement out of an island so that the remaining island constituent can get deleted in the second conjunct. In (56) we have a due to adjunct island, movement out of which leads to ungrammaticality as illustrated in (57). A reason adverbial clause is a nominalized clause with a gerundive structure out of which movement is prohibited. Thereby, we cannot move a DP from within the adjunct island outside of the island. 56. raaghav sunita-ke kolkata jaane ki vajah-se Raghav sunita-GEN Kolkata go-INF.OBL GEN reason-P gussa hai angry AUX.PRES.3Sg ‘Raghav is angry about Sunita going to Kolkata.’ 57. *kolkata raaghav sunita-ke jaane ki vajah-se Kolkata Raghav sunita-GEN go-INF.OBL GEN reason-P gussa hai angry AUX.PRES.3Sg Intended - ‘Raghav is angry about Sunita going to Kolkata.’ 137 But when we construct this island in an ACQ set up, i.e., as a part of a coordinated structure, we can have a remnant in the second disjunct while everything else gets deleted as exhibited in (58a). 58. a. (kya) raaghav sunita-ke kolkata jaane ki Q Raghav sunita-GEN Kolkata go-INF.OBL GEN vajah-se gussa hai yaa/ki bombay? reason-P angry PRES.3Sg or Bombay ‘Is Raghav angry with Sunita for going to Kolkata or for going to Bombay?’ This island amelioration effect could possibly be accounted for a by a non-movement approach to ellipsis. Arguably, the remnant would remain in its base merge position and everything around it would get deleted. But, even though, stripping/gapping structures do not always involve deletion of a congruous constituent it has been argued that movement out of ellipsis site takes place giving only an impression of non-congruity. Let us recall, that sluicing (deletion of IP following a stranded wh-phrase - Ross 1969; Chung, Ladusaw, and McCloskey 1995; and Merchant 2001) has also been argued to repair islands. Chomsky (1972) and Merchant (2001) among others have motivated an analysis of island sluicing structures to explain the impression of island repair that such sluicing structures give while maintaining that sluicing is essentially movement of a wh- phrase out of an IP followed by IP ellipsis. Chomsky (1972) argues that when the wh-phrase crosses an island node, it marks the island-node with some feature which if present in the PF leads to ungrammaticality. But in sluiced structures due to the deletion of IP, the feature is not present at the PF and this is how the effect of sluicing repairing islands is achieved. Merchant (2001) argues for a different approach. He argues that in most cases of seeming island repair in sluiced structures, the movement of the wh-phrase is actually out of a non-island structure. Merchant argues this for adverbial adjunct islands as well. Though Hindi does not provide any empirical 138 evidence in support of one analysis versus another, I posit that the derivation structure of (58) is as given in (59) following Merchant. By this analysis, the movement of the remnant Bombay is out of a non-island structure and hence no island violation is encountered giving an impression of island repair. 59. (kya) [raaghav sunita-ke Kolkata jaane ki vajah-se gussa hai] yaa/ki [Bombay jaane ki vajah-se gussa hai]? Thereby, concluding this section, I have proposed a whether/Q operator movement + ellipsis analysis for alt reading of ACQs in Hindi. The interrogative operator – the polar operator for the y/n reading of ACQs or the disjunction whether/Q operator for the alt reading of ACQs move to spec,CP to take clausal scope. For the y/n reading of ACQs there is no bi-clausal structure that undergoes ellipsis, instead the disjunction is a local DP disjunct and I treat this structure similar to that of neutral polar questions. However, for the alt reading of ACQ structures the disjunction is at a clausal level and hence a bi-clausal structure is attested, wherein the second clause undergoes ellipsis (of the kind of stripping/gapping rather than IP ellipsis). The structure for the y/n reading for ACQs was given in (42) and structure that I motivate for the alt reading of ACQs is given in (60). 139 60. 4.2.3. Analysis of Answers to ACQs in Hindi After having motivated the structure of questions, answers to ACQs in Hindi receive a straightforward explanation through the verb stranding υP ellipsis approach as the answer patterns attested are exactly the same as those attested in answers to general polar questions and answers to PAQs. Repeating the answer patterns below, (62i and 62ii) are the attested answer patterns for the y/n reading of ACQs of the type (61) and (64) is the attested answer pattern to the alt reading to ACQs of the type (63). Figure 20: Account of alt reading of ACQs in Hindi – Movement + Ellipsis Account 140 61. Q: (kya) raaghav-ne chai yaa kofi pii? ↑ - y/n reading Q Raghav-ERG tea or coffee drink.PFV.3FSg ‘Did Raghav drink tea or coffee?’ 62. A i. haan (chai/kofi) pii yes tea/coffee drink.PFV.3FSg ‘Yes, Raghav drank something (tea or coffee).’ ii. nahi pii NEG drink.PFV.3FSg ‘No, Raghav didn’t drink anything.’ 63. Q: (kya) raaghav-ne [chai]F // yaa/ki [kofi]F pii? ↓ -alt reading Q Raghav-ERG tea or coffee drink.PFV.3FSg Or Q: (kya) raaghav-ne chai pii yaa/ki kofi (pii)? Q Raghav-ERG tea drink.PFV.3FSg or coffee drink.PFV.3FSg ‘Did Raghav drink tea or coffee?’ 64. A: (#haan) chai/kofi pii Yes tea/coffee drink.PFV.3FSg ‘Raghav drank tea.’/‘Raghav drank coffee.’ The answer forms underlyingly are full clauses that then undergo verb stranding υP ellipsis under identity with the selected disjunct alternative as its antecedent. (61) is a polar question with the alternatives set being {[p(DP1 V DP2)], ¬[p(DP1 V DP2)]} of the kind {p, ¬p} which in this example is {[haan, raaghav-ne chai yaa kofi pii], [nahi, raaghav-ne chaai yaa kofi nahi pii]}. The answer then is a choice of one alternative proposition which then undergoes verb stranding υP ellipsis resulting in the attested surface form. The structure of this answer would be very similar to that of 141 answers to neutral polar questions laid out in chapter 2. (63), on the other hand is an ACQ with the alternative set being {p, q} which in this example means {[p raaghav-ne chai piiya], [q raaghav-ne kofi pii]}. The answer here too is the selection of one of the explicit choices mentioned which then undergoes verb stranding υP ellipsis under identity with the selected disjunct mentioned in the question antecedent. The structure of this answer is schematized in (65). 65. 4.2.4 Interim Conclusion: Section 4.1 and 4.2 In the previous two sections we have extensively examined two kinds of questions with disjunctions in them: PAQs and ACQs. Both these questions are alternative questions in the sense that they mention the salient alternative set in the questions explicitly and answering such questions effectively means selecting one of those explicitly mentioned alternatives (here I am comparing PAQs to alt reading of ACQ). The difference between these two types is the nature of the disjunct propositions which in the case of PAQs are of the form {p, ¬p} and in the case of Figure 21: Fragment answer for alt-reading of ACQs 142 ACQs are either of the form {[p(DP1 V DP2)], ¬[p(DP1 V DP2)]} or {p, q}. The two readings of ACQs in Hindi can be disambiguated by prosody, case-marking or word order in the surface form. Yet another similarity we observed between the PAQs and the alt reading of ACQs is that both involve clausal level disjuncts which leads to partial ellipsis of the second disjunct under identity with the first disjunct. There is no ellipsis in y/n reading of ACQs as the disjuncts therein are low level DP disjuncts. However, for all questions in general and for both PAQs and ACQs there is an interrogative operator that moves to the spec,CP position to take interrogative scope. The answer forms have a unified analysis of verb stranding υP ellipsis. After one of the explicit alternatives is chosen that propositional alternative undergoes verb stranding υP ellipsis to derive the attested surface forms under identity with the explicit alternative as its antecedent. In the final section of this chapter, I now discuss in detail a different kind of polar question which I call polar questions with narrow focus or PFQs. 4.3 Introduction Polar Questions with Narrow Focus (PFQs) in Hindi In this section we explore a specific sub-type of polar questions which has a narrow focus on one of its constituents. The way such questions are formed in English is through a (pseudo-)cleft construction like in (66) which generates the alternatives in (67). The answer to such a question is given in (68). In this question the DP constituent John is contrastively focused. Such questions carry a presupposition, in this case – someone read the book and what is at-issue is whether it was John who read it. By asking a polar question the addressee is given a chance to confirm or deny the at issue proposition. Holmberg (2015) posits that the speaker by asking such a question wants to know whether the value assignment ‘who read the book = john’ is true. 143 66. Q: Was it John who read the book? 67. {[Yes, it was John who read the book],[ No, it was not John who read the book]} 68. A: i. Yes/No. ii. Yes, it was. /No, it wasn’t. However, in Hindi, polar questions with narrow focus are formulated as shown in (69). And the attested answer forms are in (70). The constituent which has narrow focus is prosodically marked with focal stress. The unmarked order is for that constituent to occur in the immediately pre-verbal position though that is not obligatory. Also, the question particle kya occurs to the left of the focused constituent. 69. Q: yeh kitab kya [raaghav-ne]F paDhii? this book Q Raghav-ERG read.PFV.3FSg ‘Was it Raghav who read this book?’ 70. A: i. haan raaghav-ne paDhii yes Raghav-ERG read.PFV.3FSg Yes, it was Raghav who read the book.’ ii. nahi raaghav-ne nahi paDhii no Raghav-ERG NEG read.PFV.3FSg ‘No, it was not Raghav who read the book.’ iii. ?haaN yes ‘Yes, it was Raghav who read the book.’ iv. ?nahi no ‘No, it was not Raghav who read the book.’ 144 Hindi being a scrambling language, the unmarked order in (69) is not the only way to form the same question. All options in (71) are acceptable forms for the same question though (71c) is degraded for some speakers. 71. a. (kya) [raaghav-ne]F yeh kitab paDhii? b. (kya) yeh kitab [raaghav-ne]F paDhii? c. ?yeh kitab [raaghav-ne]F (kya) paDhii? The feature obligatory for a polar question with narrow focus reading is to have focal stress on the constituent which bears focus. The question particle kya is always optional in all questions we have encountered until now and such is the case here as well. Though until now we have only encountered the question particle in a clause initial position, in (69) we see that kya can occupy non-initial positions as well. Unmarked answers to such questions are laid out in (70). The usual answer form consists of the polarity particle followed by the focused constituent and finally the entire verbal sequence repeated as is. We have seen other answer forms also have the entire verbal sequence repeated but what is interesting to note is the obligatory presence of the polarity particle and the repetition of the focused constituent. Simple polarity particles as answers are not ungrammatical but are degraded. 4.3.1. Holmberg’s (2015) Analysis of Questions and Answers with Narrow Focus The polar questions with narrow focus receive a similar treatment to that of general polar questions in Hindi. As stated in Chapter 1, I adopt Holmberg’s (2015) formalization of question structure by 145 positing a Q force interrogative operator which takes CP as its complement and binds the question variable. The variable moves to the C domain to get sentential scope. ‘The Q force feature contributes a request to the addressee to assign a value to the variable’ (Holmberg 2015). For general polar questions, the question variable is the [±Pol] polarity variable. Holmberg (2015) further posits that for questions with narrow focus the question variable which undergoes overt or covert movement to the C domain is [±A] where A is the constituent which has narrow focus. I have presented here data from Finnish that Holmberg (2015) uses to explain the analysis he proposes. 72. Q: kahiva-ko marja haluaa? Coffee-[±] Marja wants ‘Is is coffee that Marja wants?’ 73. A: i. kahiva Coffee ii. kyllä yes ‘Yes.’ Holmberg (2015: 216) The structure that he posits for the example above is given below. He argues that the questions variable [±kahiva] here has undergone overt movement to the C domain to get sentential scope to derive the sentential disjunction set {[Marja wants coffee], [Marja want non-coffee]} and to enable the Q force feature to ask the addressee to provide a value to the question variable. When answering with kahiva the variable is valued positively, which means that the alternative [Marja wants coffee] has been selected. The PolP is elided and the focused valued [+ kahiva] is spelled out as [kahiva]. The full answer structure is in (75). 146 74. Q: [CP [± kahiva] Foc [PolP Marja POL [VP haluaa <[±kahiva]>]]] 75. A: [CP [+ kahiva] Foc [PolP Marja POL [VP haluaa <[+kahiva]>]]] As illustrated in (73ii), a polarity particle can also be used to answer such questions. Holmberg argues that ‘Kyllä is a generalized affirmative focus particle, a plus-value assigner to polar variables in questions.’ He further posits that kyllä can assign a [+] value to any [A] variable whether [A] be a polarity phrase or any other constituent like [kahiva] in this case. The answer pattern attested in Finnish provides support for such an analysis. 4.3.2. Analysis of PFQs in Hindi The PFQs broadly receive a similar treatment to that of general polar questions in Hindi. However, the answer patterns of Hindi are crucially different from those of Finnish and so the question structure should reflect these differences. I adopt the general structure of questions in which there is a Q force feature which asks the addressee to provide a value to the question variable in the C domain in the question. Furthermore, the question variable [A] in PFQs is the constituent which bears the narrow focus. In (69) [A] would be [Raghav]. As stated earlier such questions come with a presupposition, in this case ‘someone read the book’. The focused value over which alternatives are generated is [Raaghav]F with possible alternative being {[Raghav], [not Raghav]}. The alternative [not Raghav] could be constrained contextually. For instance, in the context provided in (76), [not Raghav] would be logically equivalent to [Dharmesh]. 147 76. Raghav and Dharmesh are the only two people in the library. The librarian sees a book out from its shelf and asks (69). I further argue that there is another variable in a feature sharing relationship with the [A] question variable in Hindi PFQs. The polarity variable in the IP-internal Pol head also needs to receive a value. Crucially, these two question variables are not independent and both do not undergo covert movement to the C domain to get sentential scope. But they get valued at the same time with the same feature. As per Holmberg (2015), the question variable(s) move to a position in the C domain to get sentential scope and receive a focused valued feature as answer. However, for Hindi, I have argued that there is no movement of the variable(s) to a higher position but a polarity variable merged in the C domain get a focused valued feature from the answer and in turn values the other variable(s) in its domain. Effectively, this general polarity variable after receiving a value, assigns whatever polarity value it receives to [± Raghav] and to the IP-internal Pol head which carries an [± Pol] feature. I do not commit to a specific position for [± Raghav] in the structure. In unmarked cases it has been argued that DP constituents move out of υP and occupy the specifier position of an IP-internal FocP. It has been argued that this preverbal position is an IP-internal focus position (Brody 1989, 1996; Jayaseelan1989,1995; Kidwai1995). But my analysis is not contingent upon this movement of the focused DP. The structure of an PFQ in Hindi is shown in (77). 77. [(kya)[CP [±Pol] [CP [ IP [± Raghav-ne] [PolP [±Pol] [υP yeh kitab <[± Raghav-ne]> ] paDHii ] ] ] ] 148 The motivation behind the suggestion of three question variables in this structure comes from the answer forms. Answers to general polar questions in (78) optionally include the polarity particle as seen in (79), the structure for which is given in (81). Refer to Chapter 2, section 2.5.3 for more details. 78. Q: (kya) raaghav-ne yeh kitab paDhii thii? Q Raghav-ERG this book read.PFV.3FSg AUX.PST.3FSg ‘Did Raghav read this book?’ Figure 22: PFQs in Hindi 149 79. A: i. (haaN) paDhii thii Yes read.PFV.3FSg AUX.PST.3FSg ‘Yes, he did.’ ii. (nahi) nahi paDhii thii no NEG read.PFV.3FSg AUX.PST.3FSg ‘No, he didn’t.’ 80. Q: [(kya)[CP [±Pol] [CP [IP [PolP [±Pol] [υP Raghav-ne yeh kitab] paDHii thii ] ] ] ] 81. A: i. [CP [+Pol] [CP [IP [PolP [+Pol] [υP Raghav-ne yeh kitab] paDHii thii ] ] ] ii. [CP [-Pol] [CP [ IP [PolP [-Pol] [υP Raghav-ne yeh kitab] paDHii thii ] ] ] The polarity variable in the C domain receives its value from the externally merged polarity particle. The polarity particle values the variable by assigning it a [+] value in (79i and 80Ai). This valued variable in turn values the IP-internal Polarity head which is lexically supported by the main verb which then makes the externally merged polarity particle redundant, them being the exact same feature and hence optional. However, as we can see in answers to PFQs in Hindi (70i and ii), there are three essential components: the obligatorily present polarity particle, the constituent which bears focus/[A] variable and the verbal sequence. This leads me to posit the structure in (77) for PFQs in Hindi. The general polarity variable in the C domain receives a value from the polarity particle in the answer and in turn values the constituent variable with narrow focus and the IP-internal Pol head. This follows the intuition that what (69) actually translates to is ‘Is it true that it was Raghav who read the book?’ Yet another piece of evidence for a general polarity variable comes from the occurrence of simple polarity particles as attested answers (70iii and 70iv). In this case since the 150 polarity particles are known to assign polarity values to the other two variables they have the potential to act as complete answers to PFQs. I consider cross-linguistic variations in answer forms to PFQs as further evidence for this structure. Holmberg in his cross-linguistic study of such questions proposes the structure in (74) for the Finnish PFQ in (72) essentially due to the answer pattern attested in (73i). Given that the constituent which bears focus by itself is an appropriate answer form to the question he argues that the constituent bearing narrow focus is the only question variable and moves , in fact overtly to the C domain to get sentential scope and a value in the answer. Such valuation in Finnish is more direct in comparison to English (refer to set 66 - 68) where the answer form as a polarity particle answers whether the value assignment ‘who read the book = john’ is true. From the answer pattern it seems that Hindi does both, the answer forms reflect that a value has been assigned to the [A] variable and the value assignment ‘who read the book=Raghav’ has been checked to be true or false. The reason for this seemingly redundant but obligatorily present structure could be same reason due to which Hindi does not exhibit verb-echo answer patterns and requires the presence of the entire verbal sequence in its answer forms. The IP-internal polarity head and (in this case) the position which is occupied by the [A] variable (IP-internal spec,FocP) is lower in the Hindi clausal structure. Neither the polarity head nor the focus position occupied by the constituent bearing narrow focus is close to the C-domain forcing a more elaborate structure after verb stranding υP ellipsis. As established in Chapter 2, the use of a simple polarity particle as answer is derived from IP ellipsis. Finally, in (77) I have merged kya the question particle in Hindi at a high position, higher than CP. In chapter 1, I briefly laid out the occurrence restrictions on kya and argued that kya is best treated as a focus sensitive operator and I assumed following Bhatt & Dayal (2020) that it 151 occupies a position in the Force projection, higher than CP due to certain embeddability restrictions. Refer to Chapter 1 for more details. 4.3.3. Analysis of answers to PFQs in Hindi If the analysis posited for PFQs is in the right direction the structure of answers to PFQs would be parallel to the structure of questions repeated in (82) with valued polarity variables and without kya question particle. 82. [(kya)[CP [±Pol] [CP [ IP [± Raghav-ne] [PolP [±Pol] [υP yeh kitab <[± Raghav-ne]> ] paDHii ] ] ] ] Accommodating those changes, the structure of answers to PFQs in Hindi would be: 83. a. [CP [+Pol] [CP [ IP [+ Raghav-ne] [PolP [+Pol] [υP yeh kitab <[+ Raghav-ne]> ] paDHii ] ] ] 84. b. [CP [-Pol] [CP [ IP [- Raghav-ne] [PolP [-Pol] [υP yeh kitab <[- Raghav-ne]> ] paDHii ] ] ] The answer forms which have the structure schematized above are repeated below in (85). 85. i. haan raaghav-ne paDhii yes Raghav-ERG read.PFV.3FSg ‘Yes, it was Raghav who read the book.’ ii. nahi raaghav-ne nahi paDhii no Raghav-ERG NEG read.PFV.3FSg ‘No, it was not Raghav who read the book.’ iii. haaN yes ‘Yes, it was Raghav who read the book.’ 152 iv. nahi no ‘No, it was not Raghav who read the book.’ As mentioned in the previous sub-section, (85i and 85ii) are derived from verb stranding υP ellipsis illustrated in (86) whereas (85iii and 85iv) are derived from IP ellipsis (87). IP ellipsis is degraded in this context but still considered a viable answer form by most speakers. Thereby the presence of the entire verbal sequence can be accounted for. The presence of the focused DP in the answer receives an explanation in the structure provided in (86). Since the focused DP moves to an IP- internal FocusP which is merged higher than the IP-internal PolP it escapes the ellipsis site and hence is present in the answer. The obligatory presence of the polarity particle has been explained in the previous sub-section. Generally, polarity particles are optional in answers to neutral polar questions because they encode the same information encoded by the verbal sequence – the polarity of the proposition. However, the polarity particle in the C domain in the answer to PFQs is not optional as not only does it value the sentential polarity variable in IP-internal PolP but it also assigns a value to the question variable encoded in the focused XP here [Raghav]F. 86. [CP [+Pol] [CP [ IP [+ Raghav-ne] [PolP [+Pol] [υP yeh kitab <[+ Raghav-ne]> ] paDHii ] ] ] 87. [CP [+Pol] [CP [ IP [+ Raghav-ne] [PolP [+Pol] [υP yeh kitab <[+ Raghav-ne]> ] paDHii ] ] ] Below I provide substantial proof that the focused subject in such PFQs occupies a higher position than its internal merge position to escapes the ellipsis site. 88. Q: yeh kitab kya kisii-ne paDhii? this book Q someone-ERG read.PFV.3FSg ‘Did someone read this book?’ 153 89. A: haan kisii-ne #(TO) paDhii yes someone-ERG TO read.PFV.3FSg ‘Yes.’ -to in the Hindi literature has been argued to be a topic marker (Bayer et al. 2014). Though I have argued in a recent collaborated paper (Dash et. al 2022) that -to is a discourse marker with a complex common ground management function, it still stands that -to marked constituents may occupy a higher position and are given information. As we can see in (89) that an indefinite like someone cannot escape the ellipsis site unless explicitly marked with -to suggests that the focused constituents in answers to such PFQs may occupy a higher position and hence may escape the ellipsis site. However, there is an interesting challenge in the derivation of the attested answer forms. The unmarked structure for PFQs in Hindi (repeated in 90) as mentioned earlier is formed by moving the focused DP to a preverbal focus position. The question particle kya in the unmarked order is to the left of the focused DP. All the other/non-focused arguments are then to the left of the kya and essentially have a specific interpretation. 90. Q: yeh kitab kya [raaghav-ne]F paDhii? this book Q Raghav-ERG read.PFV.3FSg ‘Was it Raghav who read this book?’ As mentioned in Chapter 1, while discussing the nature of kya, I argued that kya demarcates a clause into not-at-issue and at-issue content. Bhatt & Dayal (2014) have similarly argued that everything to the left of kya is given information and hence not-at-issue whereas everything to the right is new/at-issue content. Though in more recent work Bhatt & Dayal (2020) posit that only 154 the constituent to the immediate left of kya can also be at issue content whereas Biezma et al. (2017) argue that all constituents can be questioned with the right prosodic properties. Overall, there is no restriction on questioning material to the right of kya, whereas it could be concluded that special constraints need to be met for the material on the left to be questioned. However, given that Bhatt & Dayal (2020) argue that kya is merged in ForceP, and generally material to the left of kya are definite DPs which are mostly considered given information, it has been posited that they raise higher than kya (higher than ForceP). If the not-at-issue content occupies such a high position in these constructions where there is a clause-medial kya, a non-trivial question arises. Why does such material not occur in answer forms, as materials placed in such high positions should escape both IP and verb stranding υP ellipsis. The presence of the focused argument in the unmarked fragment answer can be accounted for by the focused argument moving outside the ellipsis site to a preverbal focus position, just like in the questions. Still a υP ellipsis account (schematized in 91) seems to not be able to account for the missing object DP in the fragment answer in (85i). 155 91. A υP ellipsis account can explain the missing object in the fragment answer, only if the object stays in its base position as a complement of V. However, as we can notice, the object has moved to a topic position and is higher than the subject in the contrastive focus question in (90). First, there is no motivation to restrict the object in its base position in the answer when it has clearly moved higher in the question. If the object DP is in the base position, then υP ellipsis can easily account for the absence of this DP in the answer. We only need to ensure that the deletion of the ʋP happens before the object DP inside ʋP moves to a higher position escaping ellipsis. Assuming Figure 23: Fragment answer to FPQs 156 a PF deletion approach to ellipsis, this is difficult to account as the movement of the object DP to a topic position is a movement in the narrow syntax whereas the deletion of ʋP would happen later in the PF. This allows for material to escape ʋP in the narrow syntax and be overtly realized in the elided structure. This is exactly what happens with the focused subject DP. The subject DP originates in spec,ʋP which is inside the domain of the ellipsis site but before PF deletion the subject moves higher, outside of the ellipsis site to a pre-verbal focus position and hence is overtly realized in the answer. The question then is, why the object cannot move out if the subject can. The answer would be the object can indeed move out and be overtly realized in the answer as we can see in (92). The next question then would be if the subject can stay inside the ellipsis site as the object does in (93). If such a pattern as given in (93) is attested, then it becomes a general issue which is not specific to object DP. 92. A: haan yeh kitaab raaghav-ne paDhii yes this book raghav-ERG read.PFV.3FSg ‘Yes.’ 93. A: haan paDhii yes read.PFV.3FSg ‘Yes.’ It turns out that (93) is indeed an acceptable fragment answer to the question in (90). That suggests that even though the subject has moved to a focus position and the object to a topic position, both of which are higher than ʋP in the question; they need not move out in the answer. The answer form in (93) suggests that the focus subject DP and the topic object DP remain in the base generated position allowing for ʋP ellipsis to result in no phonetic realization for both. Furthermore, this is not a pattern specific to question-answer pairs. This is attested in general ʋP ellipsis as well, as can 157 be seen in (94). In (94) the object DP has moved to a higher topic position in the antecedent clause but has remained in the base position in the ellipsis clause to undergo ʋP ellipsis. In fact, the English equivalent in (95) is found acceptable by native speakers if a context has been set such that that book is established as a topic of the utterance. 94. vo kitab raaghav-ne paDhii this book raghav-ERG read.PFV.3FSg ‘That book, Raghav read.’ aur Dharmesh-ne bhi paDhii and Dharmesh-ERG also read.PFV.3FSg ‘and Dharmesh too.’ 95. That book, Raghav read and Dharmesh did too. We can thereby conclude that in structures marked for ellipsis, other syntactic operations, at least optional movement operations, need not apply. In the antecedent clause, i.e., the question in (90) the object has a topic feature which results in it moving to a higher topic position and the subject has a focus feature which leads to its focus movement to a focus position. Both these movements are optional and in the answer form, both the DPs may stay in their base generated position in the ellipsis clause thereby resulting in their ellipsis when ʋP is elided licensed by the ellipsis feature. This makes the necessity of an ellipsis feature licensing ellipsis more relevant as there is no other way to mark an ellipsis structure in the narrow syntax if there is no ellipsis feature. This is of course assuming a PF deletion approach to ellipsis. On interesting restriction is that, though both the topic movement of the object DP and the focus movement of the subject DP are optional, if the object DP moves out of the ellipsis site the subject DP also has to escape ellipsis as well as (96) is not an attested answer form to the question in (90) but the vice versa is not true (85i). This 158 intuitively makes sense, that if a not-at issue content is to be repeated in the target clause the at- issue content is also obligatorily present as no ellipsis has taken place. 96. A: # haan yeh kitaab paDhii yes this book read.PFV.3FSg ‘Yes.’ However, not-at issue content is not generally repeated but at issue content is, hence the unmarked answer form in (85i and 85ii). The focused subject DP escaping the ellipsis site does not trigger a necessary movement of the topical object DP out of the ellipsis site. A formal account of answers to PFQs will need to account for this uni-directional entailment. I leave this for future research. 4.4 Conclusion In this chapter, I have detailed three different kinds of questions in Hindi and their answer forms. Polar alternative questions/PAQs have a bi-clausal structure wherein the second disjunct is derived by ellipsis under identity with the first disjunct. The two kinds of PAQs attested in Hindi are argued to be derived by two different levels of clausal disjuncts. Evidence for such a treatment comes from the different answer patterns attested for both these types of questions. The next type of questions are the alternative choice questions/ACQs. They also receive a similar bi-clausal treatment. Such questions, following Han & Romero’s (2004) analysis are also argued to be clausal level disjuncts having undergone movement of the whether/Q operator to get sentential interrogative scope and also exhibiting ellipsis of the second disjunct under identity with the first. The final type of question that I explored in this chapter are polar questions with narrow focus (PFQs) and we saw evidence for the treatment of the question particle kya in chapter 1. I proposed an analysis for such questions in line with Holmberg’s (2015) general analysis of polar questions 159 detailed in Chapter 1 but modified the analysis to account for the different answer patterns in Hindi. For all three types of questions, I have argued for a uniform analysis of fragment answers in accordance with the analysis for similar answer patterns in Chapters 2 and 3. Overall, fragment answers throughout Hindi, for all kinds of questions can be derived by verb stranding υP ellipsis. 160 Chapter 5: Conclusion Hindi employs numerous mechanisms to delete structures that are nevertheless interpreted in the LF. Chapter 2 gave us a brief overview of how Hindi independently attests pro drop, argument ellipsis, verb stranding ʋP ellipsis and even IP ellipsis. The missing argument in (1A) can then have multiple sources. However, in (3) based on missing adjunct patterns we conclude that the attested process has to be verb stranding ʋP ellipsis or IP ellipsis and not pro drop or argument ellipsis as pro drop and argument ellipsis cannot account for missing but interpreted adverbs. And furthermore, I have extensively argued in Chapter 2 why (4) is a result of verb stranding υP ellipsis rather than IP ellipsis. The major motivation for such an argument being that the verbal sequence does not form a complex head to raise higher than the IP in Hindi. Even so there seems to be no way of knowing for sure if the missing argument in (2) is due to pro drop or argument ellipsis or verb stranding ʋP ellipsis. Extending this analysis for fragment answers, I have argued that fragments answers to different types of questions, can all be accounted for by this analysis. 1. Q: (kya) raaghav-ne kitaab paDhii? Q Raghav-ERG book read.PFV.3Fsg ‘Did Raghav read the book?’ 2. A: haaN paDhii yes read.PFV.3Fsg ‘Yes, Raghav read the book.’ 161 3. Q: (kya) raaghav-ne do baar kitaab padhii Q raghav-ERG two times book read.PFV.3FSg ‘Did Raghav read the book twice?’ 4. A: haaN paDhii yes read.PFV.3Fsg ‘Yes, Raghav read the book (twice).’ In the final chapter of this dissertation, I depart from discussing fragment answers to questions to explicate ellipsis patterns in general. As we saw in Chapter 2, Hindi attests verb stranding υP ellipsis fragment answers and fragment patterns derived by IP ellipsis as well. In section 5.2 I briefly reiterate Manetta’s (2019) arguments for verb stranding υP ellipsis being generally attested in Hindi, independent of the question-answer discourse structure. Next in section 5.3 I make a case for IP ellipsis also being generally attested in the language and provide an analysis for the phenomenon. In section 5.4, I discuss in detail the relevance of the two kinds of ellipsis and their licensing functional heads in contributing towards the debate on duality of functional projections. In section 5.4 I review the relevance of this research, specifically its contribution to cross-linguistic research in fragment answer patterns in particular and ellipsis in general. Section 5.5 closes off this dissertation with the conclusion. 5.1 Verb stranding υP Ellipsis in Hindi: Manetta (2019) Verb stranding verb phrase ellipsis as defined by Goldberg (2005) is independently attested in Hindi, as was mentioned in Chapter 2. To reiterate the arguments, in (5) provides evidence for a 162 verb phrase ellipsis account based on adjunct inclusion and availability of sloppy interpretation as opposed to a pro drop or an argument ellipsis account of the missing argument (Bhatt & Dayal (2007), Simpson et al (2013), Manetta (2019). 5. raam-ne apanaa nayaa lekh do baar paDhaa raam-ERG self new article two times read.PFV.3MSg ‘Rami read hisi/selfi new article twice.’ raaj-ne bhi paDhaa raaj-ERG also read.PFV.3MSg ‘Rajj also read (hisi/j article twice).’ Manetta (2019) 6. Q: kya Ram-ne Sita-ko santaraa yaa Mina-ko amruud Q Ram-ERG Sita-DAT orange or Mina-DAT guava diy-aa thaa give-PFV.M AUX.PST.M ‘Had Ram given an orange to Sita or a guava to Mina?’ 7. A: haaN, Ram-ne __ diy-aa thaa Yes, Ram-ERG give-PFV.M AUX.PST.M ‘Yes, Ram had given (an orange to Sita or a guava to Mina)’ [crucially true even if Ram only gave a guava to Mina.] - Manetta (2019) Manetta (2019), following Gribanova (2013), argues that the ultimate test of υP/VP ellipsis is to examine deletion patterns in conjunction/disjunction structures. In (6) the larger verb phrase containing the disjunct of the two coordinated verb phrase has been elided. Though Hindi attests to independent argument ellipsis/pro drop and even missing adjuncts or PPs, independent deletion of a disjunction marker is not attested. The missing disjunct in (7) thereby argues for a larger constituent being elided, which in this case is the conjoined υP. 163 Manetta has a detailed discussion of verb stranding verb phrase ellipsis in Hindi where she argues that the main verb forms a complex head with υ and raises out of the υP layer following which υP gets elided with its remaining content. One piece of evidence she presents is from verb phrase ellipsis cases with complex predicate constructions. In (8), assuming that the light verb of the complex predicate is base merged at υ and the main verb is merged at V, verb phrase ellipsis attests only those structures in which the entire V+υ complex as moved out of the elided constituent. As we can see, in Hindi the light verb cannot be stranded eliding only the main verb. This suggests that the main verb and the light verb form a complex head and furthermore this complex head raises higher than little υ to escape ellipsis. The claim that (8ii) is an instance of verb stranding υP ellipsis and not pro drop or argument ellipsis can be proved by adjunct inclusion in (8ii). 8. i. kabir-ne us kitaab-ko pahli baar paR lii-yaa. kabir-ERG this book-ACC first time read take-PFV.M ‘Kabir managed to read this book for the first time.’ ii. Meena-ne bhi __ paR lii-yaa. Meena-ERG also read take-PFV.M ‘Meena also read (this book for the first time).’ iii. ?*Meena-ne bhi _ lii-yaa. Manetta (2019) Manetta further argues that the ellipsis site for the verb phrase ellipsis is higher than VP and lower than AspP and considers it to be υP following Merchant 2013 (for English), McCloskey 1991 (for Irish), Gribanova 2017 (for Russian) among others. 164 9. Shyaam-ne Mina-ko xat ya Sita-ko kavitaaN likh Shyaam-ERG Mina-ACC letters or Sita-ACC poems write diy-ee thee? give-PFV.PL AUX.PL NahiiN, par Ali abhi likh de raha hai No, but Ali now write give PROG AUX.Sg ‘Had Shyaam written letters to Mina or poems to Sita? No, but now Ali is writing (letters to Mina or poems to Sita).’ Manetta (2019) In Hindi ellipsis constructions, aspectual heads can be different for the antecedent and the elided constituent (in 9) suggesting that the ellipsis site is lower than AspP. Given that verbal identity is obligatory in VP/υP ellipsis (Goldberg 2005), the different aspects in (9) provide evidence tha the ellipsis site is lower than AspP. Hence, I assume following Manetta (2019) that Hindi verb ellipsis patterns can be characterized as verb stranding υP ellipsis (see also Simpson et al. 2013 for verb stranding υP ellipsis in Hindi). In this section I briefly mentioned the arguments that Manetta considers analyzing the patterns attested in (6A) as verb stranding υP ellipsis. In this dissertation I have extensively argued that fragment answer patterns to different kinds of questions can be derived by verb stranding υP ellipsis as well. In the next section, we briefly describe another kind of ellipsis attested in Hindi. 165 5.2 IP Ellipsis in Hindi In the previous section we saw that Hindi independently attests verb stranding υP ellipsis independent of question-answer constructions and we further reviewed its properties. I have claimed that verb stranding υP ellipsis is the unmarked approach to deriving fragment answer patterns in Hindi. Furthermore, though simple polarity particles are not the unmarked answer forms, I have followed Holmberg (2015) in positing that they are full clausal structures having undergone IP ellipsis. IP ellipsis is also independently attested in Hindi as can be seen in (10). The first evidence that it is a higher-level verbal ellipsis is the obligatory deletion of the auxiliary which is merged at the I/tense head in Hindi (Mahajan 1990, Bhatt 2005). 10. raaghav-ne vo kitab do baar paDhii thii, raghav-ERG that book two times read.PFV.3FSg AUX.PST.3FSg par dharmesh-ne nahi but dharmesh-ERG NEG ‘Raghav read that book twice but Dharmesh didn’t (read that book twice)’ Also, in (10) there is evidence for adverb inclusion as the target clause can be interpreted as containing the adverb - ‘Dharmesh did not read the book twice’. Adverb inclusion is considered to be a characteristic property of verbal/clausal ellipsis as opposed to pro drop mechanisms. Yet another property characteristic of such ellipsis is the availability of sloppy interpretation which is also attested in examples of the type in (10), as we can see in (11). The target clause in (11) can be interpreted as ‘Dharmesh did not wash his own car.’ 166 11. raaghav-ne apnii gaDii dhoyii thii, raghav-ERG self car wash.PFV.3FSg AUX.PST.3FSg par dharmesh-ne nahi but dharmesh-ERG NEG ‘Raghavi washed hisi car but Dharmeshj didn’t (wash hisj car)’ A crucial property of IP ellipsis in Hindi is that it is licensed only when the antecedent and target clauses have contrastive polarity as can be seen from comparing (10) with (12). In (12) both the clauses have positive polarity and the target clause with IP ellipsis is fairly marked. Similarly, even when both the antecedent and the target clause have negative polarity (like in 13), IP ellipsis is not licensed as the two clauses do not have contrasting polarity. Ideally, the opposing set up of (10) should be attested in the language, a set up wherein the antecedent clause has negative polarity and the target clause has positive polarity as in (14). If the licensing condition is contrastive polarity, (14) should be a grammatical structure without the verbal sequence in the target clause. I argue that even though IP ellipsis in (14) is bad, it is not because it doesn’t satisfy the licensing requirement for IP ellipsis. (14) obligatorily needs the verbal sequence because verbal sequence acts as the lexical exponent of sentential positive polarity in Hindi. Hindi does not have a dedicated particle to express positive sentential polarity. In fact, we have motivated the head movement of the main verb outside of υ to the IP-internal Pol head for the purposes of providing a lexical exponence to positive polarity. Thereby we can maintain that the licensing condition for IP ellipsis is contrastive polarity though the only attested case is a [p but ¬q] structure. 167 12. ??raaghav-ne vo kitab do baar paDhii thii, raghav-ERG that book two times read.PFV.3FSg AUX.PST.3FSg aur dharmesh-ne bhi and dharmesh-ERG also ‘Raghav read that book twice and Dharmesh did too (read that book twice).’ 13. *raaghav-ne vo kitab do baar nahi paDhii thii, raghav-ERG that book two times NEG read.PFV.3FSg AUX.PST.3FSg aur dharmesh-ne bhi nahii and dharmesh-ERG also NEG ‘Raghav didn’t read that book twice and Dharmesh didn’t either (read that book twice).’ 14. raaghav-ne vo kitab do baar nahi paDhii thii, raghav-ERG that book two times NEG read.PFV.3FSg AUX.PST.3FSg par dharmesh-ne *(paDHii thii) but dharmesh-ERG read.PFV.3FSg AUX.PST.3FSg ‘Raghav didn’t read that book twice but Dharmesh did (read that book twice).’ Further evidence for the proposal that licensing condition of IP ellipsis is contrastive polarity we observe that when the two clauses do have contrasting polarity, a structure with stranded verbal sequence like in (15 and 16) does not exhibit the characteristic properties of clausal ellipsis – no adverb inclusion as illustrated in (15) and no sloppy interpretation as illustrated in (16). 15. raaghav-ne vo kitab do baar paDhii thii, raghav-ERG that book two times read.PFV.3FSg AUX.PST.3FSg par dharmesh-ne nahi paDhii thii but dharmesh-ERG NEG read.PFV.3FSg AUX.PST.3FSg ‘Raaghav read that book twice but Dharmesh didn’t read (that book even once).’ 168 16. raaghav-ne apni kitab do baar paDhii thii, raghav-ERG self book two times read.PFV.3FSg AUX.PST.3FSg par dharmesh-ne nahi paDhii thii but dharmesh-ERG NEG read.PFV.3FSg AUX.PST.3FSg ‘Raaghav read his own book twice but Dharmesh didn’t read Raghav’s book at all.’ In (15) and (16) since the verbal sequence is present, these structures are clearly not derived by IP ellipsis though initially we may hypothesize that they are derived by verb stranding υP ellipsis but the absence of adverb inclusion and sloppy interpretation lead us to conclude that the structure above are derived by argument ellipsis or pro-drop which are also independently attested in Hindi. We can further account for the unavailability of verb stranding υP ellipsis when IP ellipsis is licensed by a MaxElide condition (Merchant 2004, Gribanova 2017). Contrasting (15) with (10) and (16) with (11) leads me to propose that IP ellipsis in Hindi is licensed when the antecedent clause and target clause have contrasting polarity. In such cases adverb inclusion reading is preferred and sloppy interpretation is available only when the verbal sequence is elided. And with cases where both clauses have the same polarity adverb inclusion reading and sloppy interpretation is attested only when the verbal sequence is overt. In addition to having contrastive polarity, an IP ellipsis structure also obligatorily has a contrastively focused remnant XP with an overt correlate in the antecedent clause. The XP can be any constituent in the target clause. Yet another property of IP ellipsis structures in Hindi is that it is bounded, i.e., the antecedent and the target clause should be at the same level as we can glean from the contrast in (10) and (17). Furthermore, IP ellipsis in Hindi seems to require an overt conjunction. Constructions like (10) and (11) would be bad without the overt but. Furthermore, IP ellipsis in Hindi do not allow more than one remnant as can be seen in (18). These properties have 169 been attested in similar structures for other languages like French (Morris 2008) and Spanish (Vicente 2006) though there is cross-linguistic variation in how languages behave with respect to these properties. 17. *raaghav-ne vo kitab do baar paDhii thii raghav-ERG that book two times read.PFV.3FSg AUX.PST.3FSg par meine sunaa ki dharmesh-ne nahi but I-ERG hear.PFV.3MSg that Dharmesh-ERG NEG Intended – ‘Raghav read that book twice but I heard that Dharmesh didn’t.’ 18. *raaghav-ne geeta-ko kitab dii thii raghav-ERG gita-DAT book give.PFV.3MSg AUX.PST.3FSg par dharmesh-ne sitaa-ko nahi but dharmesh-ERG sita-DAT NEG ‘Raghav gave a book to Gita but Dharmesh didn’t give one to Sita’ IP ellipsis is also sensitive to islands as we can see in (19) and (20). (19) is an adjunct island, wherein the remnant XP needs to move out of the nominalized gerund structure to not be inside the ellipsis site. And (20) is a relative clause island wherein the remnant XP needs to move out of the relative clause. 19. *raaghav-ne ghar jaane ke baad kitab Raghav-ERG home go.INF.OBL GEN after book paDhii thii par skul nahii read.PFV.3FSg AUX.PST.3FSg but school NEG Intended – ‘Raghav read the book after reaching home and not after reaching school.’ 170 20. *raaghav-ne vo kitab paDhii hai jo Raghav-ERG that book read.PFV.3MSg AUX.PRES.3Sg that amit-ne likhii thii par sumit-ne nahi amit-ERG write. PFV.3FSg AUR.PST.3FSg but sumit-ERG not Intended - ‘Raghav has read the book that Amit wrote but not the one that Sumit wrote.’ 5.2.1 Analysis for IP Ellipsis in Hindi As we observed, the two constructions given in (19) and (20) are ungrammatical in Hindi. This suggests a movement + deletion approach for IP ellipsis. This is indeed what I propose - IP ellipsis in Hindi, just like verb stranding υP ellipsis is derived from movement of the remnant outside of the ellipsis site followed by deletion of IP triggered by an ellipsis feature E à la Merchant (2001). Furthermore, as already proposed for simple polarity particles as answers in Chapter 2, I argue that the head that carries the ellipsis feature is the Polfoc head merged only when the polarity feature has a [Foc] feature. The Polfoc head takes IP as its complement and when hosting the ellipsis feature licenses the ellipsis of its IP complement. This explains why all attested cases of IP ellipsis need the antecedent and the target clause to have contrasting polarity. The Merchant style E feature which licenses deletion has the lexical specification as [*uFoc]. It is a strong feature (EPP) which can be satisfied only by a DP with a [iFoc] feature. This is why one and only one remnant XP which is in contrast with its correlate in the antecedent clause escapes the ellipsis site and moves to spec,PolfocP. There cannot be more than one contrastively focused remnant because the necessity 171 of a contrastively focused element as a remnant is a by-product of IP ellipsis and not the cause. If a contrastively focused remnant were an independent licensing condition on IP ellipsis, there would be no explanation for a ban on multiple such elements. The structure for IP ellipsis is illustrated in (21). The idea that contrastive polarity ellipsis is in effect IP ellipsis has been argued for extensively in the verb ellipsis literature across languages (Vicente 2006, Kazenin 2006, Morris 2008, Barros 2014, Gribanova 2017). 21. Figure 24: IP Ellipsis in Hindi 172 5.2.2 Fragment Answers to polar Questions Holmberg (2015) argues for an IP ellipsis analysis for fragment answers to polar questions. In languages like Welsh and Finnish which attest to verb-echo patters, fragment answers are derived by the highest verbal element moving out of IP followed by IP ellipsis (refer to Chapter 1 for an in-detail literature review). Furthermore, he argues that simple polarity particles as answers are full clauses having undergone IP ellipsis as well. Chapter 2 of my dissertation is focused on why verb-echo patterns are not attested in Hindi, motivating instead a verb-stacking pattern derived by verb stranding υP ellipsis. In that chapter I have also mentioned how simple polarity particles as answers to polar questions are marked in Hindi. A great deal of head-nodding gesture needs to accompany simple polarity particles for them to act as viable answers. In section 1.5.3 I have briefly hypothesized why simple polarity answers like in (22) are marked answer forms. As per the proposed structure Eans feature has to wait to merge with a higher Polfoc head when a lower Pol head is available. I posited that this seems to be a fallible violation of the Earliness Principle and which could account for the υP ellipsis pattern being the unmarked answer form whereas the IP ellipsis answer pattern gives us a marked fragment answer for polar questions in Hindi. 22. Q: kya raaghav-ne vo kitab do baar paDhii? Q raghav-ERG that book two times read.PFV.3FSg ‘Did Raghav read that book two times?’ 23. A: #haan / #nahi yes / NEG ‘Yes, he did.’/‘No he didnt.’ 173 However, if the posited analysis for IP ellipsis is on the right track it might provide another explanation for why simple polarity particles as answers are not attested in Hindi. I follow Holmberg (2015) in assuming that a polar question has a variable polarity and the answer gives us that value of polarity which is either positive or negative. These two options belong to the focused alternative set generated by the neutral polar question. This leads me to posit that polarity particles in (22A) do not occupy the contrastive polarity head but are focused alternatives just like the focused alternative DPs in verbal IP ellipsis. If such is the case, that polarity particle occupy spec,PolfocP and not the Polfoc head then a Polfoc head is not merged. And since contrasting polarity in the Polfoc head licenses IP ellipsis, IP ellipsis answer patterns are not licensed in Hindi. 5.2.3 Conclusion: IP Ellipsis in Hindi To reiterate, I have argued that IP ellipsis is licensed in only those structures where a contrastive polarity head can be merged. A contrastive polarity head thereby licenses the ellipsis of its complement IP and furthermore requires a focused XP in its specifier due to the presence of a strong EPP feature. I have extended the analysis of IP ellipsis to simple polarity particles as fragment answers and posited that such structures are marked as fragment answers because a contrastive polarity head is not licensed in this case, given that polarity particles occupy a specifier position being members of the focused alternative set rather than being hosted by the Polfoc head. There is a potential way to extend this analysis to sluicing structures as well. Crosslinguistically, sluicing has been analyzed as being derived from IP ellipsis (Ross 1969, Merchant 2001). Hindi attests to sluicing as can be seen in (24). A case could be made that 174 contrastive polarity in the matrix clause licenses the deletion of an embedded IP after movement of a contrastively focused DP outside of the ellipsis site. 24. (mujhe pataa hai ki) raaghav-ne kal koi kitab i-DAT know AUX.PRES.3Sg that raghav-ERG yesterday some book paDhii thii par mujhe nahi pataa kaunsii read.PFV.3FSg AUX.PST.3FSg but I-DAT NEG know which ‘(I know) Raghav read some book but I don’t know which (book he read). ’ This movement + deletion approach accounts for the form-identity generalizations (Merchant 2016) that such IP ellipsis exhibits. There is strict case connectivity and obligatory ad-position pidepiping. Further evidence for a movement + deletion approach comes from island effects. As we saw, movement out of an island is ungrammatical. The proposal that a Polfoc head is merged only when contrasting polarity is attested between the antecedent clause and the target clause further finds support from the empirical evidence below. In (25), (26) and (27) fragments are not licensed because there is no overt contrastive polarity to license a Polfoc head which would host the E feature which in turn would license the ellipsis of its complement IP. Negating an asserted statement in (25), asking a wh question in (26) and answering a wh-constituent question with a fragment in (27) are all marked constructions as in none of these constructions is IP ellipsis licensed. This is assuming that all the fragments uttered by B were full clauses having undergone ellipsis. However, for all these cases and whenever else the contrastive polarity requirement is not satisfied, the E feature can only be merged with the default polarity head IP-internally which then licenses υP ellipsis. All the marked IP ellipsis constructions can be repaired with υP ellipsis. The B utterances given below are perfectly fine if the verbal sequence is repeated along with the fragments. 175 25. A: raaghav-ne kitab paDhii thii Raghav-ERG book read.PFV.3FSg AUX.PST.3MSg ‘Raghav read a book.’ B: #nahi. Neg ‘No.’ 26. A: raaghav-ne kal kisii-ko dekhaa tha Raghav-ERG yesterday someone saw.PFV.3MSg AUX.PST.3MSg ‘Raghav saw someone here yesterday.’ B: #kisko Whom ‘Whom did Raghav see?’ 27. A: raaghav-ne kal kya paDhaa? Raghav-ERG yesterday what read.PST.3MSg ‘What did Raghav read yesterday?’ B: #ek kitaab a book ‘A book.’ There is one seeming exception to the analysis posited here. Cases in which no contrastive polarity is present but there is a contrastively focused element as a fragment are attested in contexts where those contrastively focused elements are D-linked, i.e., when the focused alternative set is in the common ground. Recall that the marked structure in (12) led us to conclude that the target clause 176 cannot undergo IP ellipsis when the antecedent and target clause do not have contrasting polarity. However, (12) is not ungrammatical. In fact, the markedness of (12) can be improved upon if the sentence is uttered in a context as shown in (28) and if the remnant XP and its correlate bear contrastive focal stress. 28. Context – Raghav and Dharmesh are the two students in this class who consistently perform really well. ?[raaghav-ne]F vo kitab do baar paDhii thii, raghav-ERG that book two times read.PFV.3FSg AUX.PST.3FSg aur [dharmesh-ne]F bhi and dharmesh-ERG also ‘Raghav read that book twice and Dharmesh did too (read that book twice).’ Furthermore, as we see in (29) a sluiced fragment is better with a D-linked wh phrase and also in (30) we see that a viable answer to a D-linked wh-phrase can be a fragment without the verbal sequence present. 29. A: raaghav-ne kal koi kitab paDhii thii raaghav-ERG yesterday some book read.PFV.3FSg AUX.PST.3FSg ‘Raghav read some book yesterday.’ B: kaunsi kitab which book ‘Which book?’ 30. A: raghav-ne kal kaunsii kitab paDhii? raaghav-ERG yesterday which book read.PFV.3FSg ‘Which book did Raghav read yesterday?’ 177 B: chomsky-ki (kitab paDhii) Chomsky-GEN book read.PFV.3FSg ‘Chomsky’s book.’ Pesetsky (1987) has argued that D-linked wh-phrases refer to ‘familiar’ sets. The range of felicitous answers is limited to a set that both the speaker and the hearer have in mind. But no such requirement is imposed on wh-phrases like who or what. In all cases of remnant XPs we have seen here, they are focused constituents which are members of an alternative set. For instance, in (28), having a clearly defined alternative set and with focal stress on both the alternatives, the remnant XP and its correlate are easily identified as members of the alternative set. I posit that when remnants are contrastively focused thereby bearing focal stress, the null anaphora is a result of stripping/gapping. Vicente (2006) while making a distinction between IP ellipsis and stripping in Spanish argues that stripping is characterized by focal stress on remnants which need not be necessarily the case in IP ellipsis constructions. Other Romance languages like French (Morris 2008) have attested similar differences between IP ellipsis and stripping. Romance languages Like French, Spanish, Catalan among others have well motivated distinctions between the two process thereby motivating a categorization previously mentioned in chapter 4 of ‘true’ anaphora. Hankamer (1979) made a distinction between υP/VP ellipsis and IP ellipsis, V/s gapping and stripping constructions based on their varying properties and divided deletion processes into Type A and Type B. One distinction between the two kinds of deletion processes is whether or not they delete a constituent. Type A kind υP and IP ellipsis are argued to delete one continuous constituent (though a remnant may move out of it) whereas Type B kind of deletion processes like gapping and stripping delete non-constituents. Another difference is that ‘true’ anaphora like IP or υP/VP 178 ellipsis is argued to be licensed by a functional head while that is not the case with stripping/gapping constructions. In the previous sub-section, I have argued that IP ellipsis in Hindi is licensed by the Polfoc head which is merged only when there is an explicit contrastive polarity set up. In the constructions above (28-30), this licensing requirement is not fulfilled and hence the B structures in those examples cannot be derived from IP ellipsis. I posit that those structures are a result of stripping. Recall that I posited a similar derivation for the alt reading of ACQs in the previous chapter. The licensing requirement for stripping instead of a functional head is that there can only be one remnant and this remnant and its parallel correlate are D-linked phrases, i.e., the phrases that are members of a salient set of alternatives. The licensing differences between IP ellipsis and stripping is the only observational difference between the two structures in Hindi. Since stripping is a different process from IP ellipsis and does not follow the licensing conditions of IP ellipsis, I suggest that all the cases in which D-linked XPs are remnants are derived from stripping. Moreover, I have mentioned earlier how simple polarity particles are very marked answer forms to polar questions. The only context where they are acceptable as answers are when there is an explicit instruction given: Answer with either yes or no; for instance, in official questionnaires or during some interrogation. We could argue that in such case, the polarity particles are members of an explicitly salient familiar set and hence are D-linked. And as we have seen D-linked focused constituents are licensed via stripping. However, I maintain that cases which have clauses with contrasting polarity license IP ellipsis in the target clause by simultaneously leaving a focused remnant which has moved out of IP. Further endeavors need to be made to establish stripping as a viable phenomenon in Hindi to account for the structures mentioned in this section and to elaborate on other differences between IP ellipsis and stripping. I leave this for future research. 179 In this section I have presented a preliminary analysis for IP ellipsis in Hindi. Further properties of IP ellipsis like its boundedness and requirement of local conjunction at the same level of structure still need to be accounted for and I leave this for future research. However, I have motivated a system of ellipsis in this work, where ellipsis is triggered by an ellipsis feature which can only be hosted by the two polarity heads that I have motivated in the Hindi clausal structure. The default polarity head is merged in an IP-internal position and licenses verb stranding υP ellipsis. A higher polarity head is merged only if the polarity has a [Foc] feature, above IP which when merged may host the ellipsis feature and license the ellipsis of its complement IP. In the next section of this chapter, I discuss the motivations and implications of double functional projections of polarity. 5.3 Doubling of Functional Projections After Chomsky’s (2001) Phase theory which argues that CP and ʋP are the two phases of a clause and are two independent domains of syntactic operations, there have been many proposals in the literature which have tried to establish a parallelism between these two phases. Belletti (2004) among others have argued that the ʋP periphery might have a similar structure to that of the CP periphery. Belletti (2001, 2004) gives an account of the pronoun doubling phenomena and postverbal subject in subject inversion structures in Italian and other Romance languages by positing an IP-internal topic and focus phrase just above ʋP in the clausal spine. Evidence comes from different interpretations and intonations typically associated with focus and topic positions in a way significantly parallel to the positions available in the CP left periphery. Furthermore, for case stacking configurations in Korean, Schütze (2001) has argued for a structure with an IP- 180 internal focus phrase adjoined to VP along with a focus phrase in the left periphery. Both V and I are argued to be focus licensors. Evidence comes from the fact that case stacking is only possible in focus sensitive environments like wh-phrases, correlation contexts etc. The stacked case markers are reanalyzed as discourse particles marking focus and occupying the IP-internal and left peripheral positions. And an IP-internal Focus phrase has been extensively argued for many Indo Aryan languages such as Malayalam and Hindi among others (Brody 1989, 1996; Jayaseelan 1989,1995; Kidwai1995). Additionally, Brody and Szabolsci (2000) have argued for a slightly different kind of doubling of functional projections. Examining Hungarian structures which have more than one quantificational phrase they argue for multiple scope-establishing functional projections above every verbal projection following Hallman’s (1998) Reiterative Syntax. Each series of scopal projections is argued to be strictly ranked and contains various heads for different types of quantifiers. More specific to polarity type functional elements, many different languages exhibit what has been termed in the literature as ‘negative concord’. Romance languages and Slavic languages like Spanish, Greek, French, Russian, Bulgarian etc all attest to multiple negative words in a single utterance even though semantically this results in a single occurrence of negation. Across these varied languages and the varied analyses posited in the literature for their treatment, it has often been proposed that there is a doubling of the negation projection to host the different n-words in a clause. Multiple negation has been analyzed in various ways over the years, for example with the n-words as negative quantifiers or NPIs or realizations of syntactic agreement. Without detailing the finer points of this continuing debate, one conclusion that has been reached is that since there are multiple expressions of negation in a single utterance, these multiple expressions should be hosted in various positions in the clausal spine resulting in the doubling of the negation functional 181 projection. Even a language like English which does not have negative concord is argued to have multiple projections for negation. Holmberg (2013, 2015) posits a structure with one NegP adjoined at the VP level with a VP scope and termed lower negation whereas another NegP is projected above ʋP (but lower than IP) and has sentential scope. This negation is termed as the middle negation. There is yet another negation merged at C termed as high negation. There are syntactic, morphological and scopal differences separating these three projections of negation. After having discussed the different types of functional projections that have been argued to have multiple instantiations in the clausal spine, we now explicate the nature of polarity. Holmberg’s (2013, 2015) arguments for multiple positions for negation in an English clause could be reanalyzed as multiple positions for polarity. It is just that with negation there is an overt phonological realization of the polarity head/phrase but there is no overt marker for positive polarity. This argument can be further substantiated by the nature of the verum operator which can be argued to be the semantic counterpart of the polarity head in syntax. Gutzman et al (2017) argue for a verum operator inherently associated with focus. The intonation marker traditionally argued to be associated with verum focus (Hohle 1992) is reanalyzed as marking just the verum operator and not necessarily focused verum. An implication would then be that verum and focused verum are marked differently and have different positions in the syntactic structure. Romero in a series of work (Romero and Han (2002), Han and Romero (2006), Romero (2006) among others) has argued for two different positions for verum. This research gives an account of Ladd’s p/¬p ambiguity and epistemic biases in negative biased questions by arguing for interpretive differences and scopal interactions based on two different positions for polarity. Holmberg (2013, 2015) presents the same evidence to argue for multiple positions of negation. For instance, Romero (2006) presents a discussion of negative biased questions and the biases corresponding to those 182 questions. She gives an analysis for a commonly observed fact in English that in (53) when the negation is lower in the question there is not necessarily an inherent bias for a particular response, whereas in (54) when the negation has been preposed/moved higher the bias is for the affirmative proposition. 31. Is John not coming? Bias: none in particular. 32. Isn’t John coming? Bias: John is coming. The claim made in these works is that preposed negation in polar questions has the discourse function of marking focus on polarity. This focused verum gives rise to the epistemic implicature. Since non-preposed negation need not be focused no such implicature arises. Drawing a parallel to syntax one could argue there are two dedicated positions for the verum operator with the higher position being the one wherein the verum operator is focused. Constructions with emphatic/focused do in English have also been associated with verum focus. Though the particulars vary, many syntactic accounts have been posited for languages like Nupe, Spanish, Catalan, Hungarian, Vietnamese and English among others with a projection on the left periphery of clausal structure to encode focus on sentential polarity (Fisher & Alexiadou 2001, Holmberg 2007, Hernanz 2011, Poletto 2008, Zanuttini 1997, Martins 2006, Kandybowicz 2008 among others). And a relationship has been established through different ways between these projections in the left periphery and the lower projection of a default polarity head of the nature of Laka’s (1994) ΣP. Furthermore, for all intents and purposes, Gribanova’s (2017) analysis of clauses in Russian does indeed posit two polarity phrases. Though Gribanova (2017) argues that the IP- 183 internal phrase named as NegP is to host the morphological realization of sentential negation she also states that the positive valued polarity feature receives a zero exponent. This projection in actuality then acts as the default polarity head while the higher head outside of IP which Gribanova (2017) terms as PolP is the polarity head specific to contrastive polarity as she argues that polarity is inherently focused and verbal movement to this Pol head is only possible in discourse contexts which involve contrastive polarity focus. Therefore, I have proposed that there are indeed two distinct positions for polarity wherein the lower position is the default polarity position with the higher position being specific to contrastive polarity focus. The two polarity heads are in a feature sharing relationship much like a concord relationship. Languages then simply differ in the height of the polarity phrases in the clausal spine. If this reasoning is in the right direction, then the negative biased questions in (53) and (54) provide direct evidence for two polarity phrases in questions. A schematic structure is for (53) is represented in (33) and one for (54) is represented in (34). 33. [CP [PolfocP Is [ IP John [PolP not [VP coming ]]]]] 34. [CP [PolfocP Isn’t [ IP John [PolP [VP coming ]]]]] Adding to the direct evidence from negative biased questions in English for two different positions for polarity in questions, we have contrastive focus polar question constructions in Hindi and another Indo-Aryan language Odia which could potentially illustrate the two polarity positions. Holmberg (2015) argues that questions contain a variable – the polarity variable, the disjunction of which needs to move to C to get a sentential scope. In Finnish, this movement can be seen in the overt syntax. Holmberg (2015) further argues that in languages which have Q particles, the Q particles mark the question disjunction. In contrastive focus questions in Finnish, the contrastively 184 focused constituent along with the Q particle moves to an initial position in the C domain. In Odia (another Indo Aryan language) polar questions with narrow focus are marked by the Q particle to the left to the focused constituent and the focused constituent moving to a preverbal position argued (in earlier work Syed and Dash(2017)) to be the IP-internal focus position (example in (35)). Furthermore, we can also get another Q particle at the edge of the clause. This Q particle - ki is in general attested in clause final positions in polar questions in Odia. If we assume Holmberg’s (2015) idea of disjunction of the polarity feature moving from a lower position to the C domain then the two Q particles in the example below could be considered to be two instantiations of the polarity feature. Hindi can also have two Q particles in a similar construction as can be seen in (36) (please note that the Q particle in this example is in the clause final position whereas elsewhere I have a clause initial question particle 16 - the question particle in Hindi can either be in the clause initial or clause final position). In previous work (Syed and Dash 2017) we had analyzed the Q particles to adjoin to the two FocusPs but it could be easily reanalyzed as being adjoined to the two PolPs specially if we assume Q particles to be morphological realization of the polarity variable disjunction. 35. bohi kiso raaghav poDhilaa (ki) -Odia book Q raghav read.PST.3Sg Q 36. kitab kya raaghav-ne paDhii (kya) -Hindi book Qprt Raghav-ERG read.PST.3Sg Q ‘Was it Raghav who read the book?’ 16 In the first chapter I have argued that kya is not a Q particle in Hindi, though I have glossed it as Q and I refer to it as a question particle. Whether or not the Odia -ki is similar to Hindi kya, I leave for future research. For current purposes the evidence that in some languages we may get dual Q/question particles can be taken as evidence for the potential existence of dual projections. 185 Given such assumptions, examples (35) and (36) may provide cross-linguistic evidence of the overt morphological realization of both the polarity heads in questions. As positive polarity is usually very rarely marked overtly, it is difficult to see further clear evidence of these two positions. To the extent that negation is in effect [-Pol], different characteristics of the English negation explicated above give sufficient evidence for different positions for polarity. Further research needs to be carried out, ideally in languages which overtly mark positive polarity, to check for evidence of the two polarity positions. In addition, clear evidence that the verb in Hindi (and Russian among other languages) does move to a Polarity head and not any other head between I and ʋ would further substantiate the claim of an IP-internal PolP. Gribanova (2017) argues the very same for Russian following Fowler (1994) and Bailyn (1995) among others. She states that the verb in Russian moves out of ʋP to a head lower than I which she posits to be the Neg head. As already stated, this Neg head effectively functions as a Pol head for Gribanova (2017). Yet another potential piece of evidence for an IP-internal polarity position in questions could come from scope-marking wh- constructions in Hindi and Kashmiri (yet another Indo-Aryan language). Manetta (2010) following Rackowski and Richard (2005) argues that wh-material when moving to spec,CP moves through spec,ʋP. She argues that ʋP is a phase and the embedded wh- phrase in (37) moves to spec,ʋP so that it can stay accessible to the matrix C for its wh features to be interpreted and valued. The wh-expletive in the matrix clause is argued to not have interpretable features. 37. sita-ne kya socaa ki Ravii-ne kis-ko dekhaa? sita-ERG SM 17 thought that Ravi-ERG who-ACC saw ‘Who did Sita think that Ravi saw?’ - (Manetta 2010) 17 Manetta glosses kya as expletive (expl) though I have glossed it as SM (scope marker). 186 38. [CP [C[Q] … [ʋP kya [ʋ [CP [C [ʋP kis-ko [ʋ kis-ko ]]] [wh-f] [wh-f-u] [wh-f] [wh-f] [wh-f] Keeping the insight from Holmberg (2015) that wh questions also have a question variable with a pragmatically varied disjunctive set of alternatives, we could potentially argue that the wh-phrase in (37) moves to the spec,PolP (IP-internal PolP) and not spec,ʋP assuming Manetta’s (2010) arguments for movement of the wh phrase to the edge of the ʋP phase. Concluding this section, doubling of functional projections is a widely argued claim in the literature. To the extent that there is a polarity phrase in every clause (Laka 1990), it is not too surprising to posit that there are indeed two distinct positions for polarity. There have been various arguments that the polarity phrase high in the structure above IP is associated with focus. What remains is a need to successfully argue that polarity need not be inherently focused and hence there is distinct position for default polarity somewhere lower in the clausal structure. Research in semantic interpretations already point in that direction and many syntactic properties can be modelled in such a way in support of this hypothesis. 5.4 Crosslinguistic Implications A popular assumption in the literature concerning the syntax of fragment answers, both for wh- constituent questions and polar questions is that they are semantically propositional and have a full-fledged clausal syntactic structure (Hankamer 1979; Morgan 1973, 1989; Stanley 2000; Reich 2002, 2003; Merchant 2005; Holmberg 2015; Gribanova 2017). Most of the clause is then argued 187 to be phonetically null thereby resulting only in fragments as answers. Merchant (2005) has argued for a movement and ellipsis analysis to argue that the fragment answers to wh questions are full- fledged clauses having undergone ellipsis. Holmberg (2015) and Gribanova (2017) among others argue the same for fragment answers to polar questions. As mentioned in chapter 1, Holmberg’s (2015) analysis for fragment answers in Welsh and Finnish is based on a theory of lexical exponence. It is argued that all polar questions contain a polarity variable which moves to the C domain for sentential scope. The value of this polarity variable is merged as a focused feature in the C domain in the answer. This focused value needs a lexical exponence, which is either provided by the polarity particles polar or by the highest verbal element moving from I to the C domain. The rest of the answer which effectively is the IP (PolP for Holmberg 2015) of the answer clause, being similar to the question, is elided. The example below from Welsh in (39-40) and (41) illustrates the derivation. 39. Q: all mair aros can mair stay ‘Can Mair stay?’ 40. A: gall (*aros) - Verb Echo can stay ‘Yes.’ Welsh (Holmberg 2015) 188 Figure 25: IP Ellipsis in Welsh 41. Gribanova (2017) provides a similar IP ellipsis analysis for fragment answers in Russian. The highest verbal element moves higher than IP and IP is elided under identity with the question. Though the intuition is the same the implementation is different from Holmberg’s (2015) analysis. For Gribanova (2017) the movement of the highest verbal element in Russian is an independent movement and is not motivated by the need of lexical exponence though it serves that purpose. Gribanova (2017) argues that even in declarative contexts the verb can move higher than IP. The head to which the verbal complex moves is argued to be a Pol head. Hence this movement is still motivated by a polarity feature, specifically the discourse function fulfilled by this movement is that of marking emphatic contrastive polarity focus. Gribanova (2017) argues there to be a PolP higher than IP and the verb moves to this Pol head to mark the discourse function of contrastive polarity. An illustration of the account is shown in in (44). 189 Figure 26: IP Ellipsis in Russian 42. Q: evgenija otpravila posylku v moskvu Evgenija send.PST.3FSg package.ACC to Moscow.ACC ‘Did Evgenija send the package to Moscow?’ 43. A: (Da,) otpravila - Verb Echo yes send.PST.3FSg ‘Yes.’ Russian (Gribanova 2017) 44. Both Holmberg (2015) and Gribanova (2017) essentially account for verbal fragments as polar answers by analyses which motivate the verb moving higher than IP followed by IP ellipsis. 190 Further, both accounts potentially allow ʋP ellipsis in fragment answers to account for answer patterns when something more than the highest verbal element or the polarity particle constitutes the answer (like in 19). The motivation for the verb to move higher than IP is different in both accounts. 45. Q: on-ko marja voinut kantaa tuon kiven yksin rannalta has-Q Marja could carry that stone from shore ‘Could Marja have carried that stone from the shore alone?’ 46. A: on (voinut (kantaa)) has could carry ‘Yes.’ - Finnish (Holmberg 2015) Hindi varies from Welsh, Finnish, Russian and other languages already analyzed in the literature primarily on one major account: it does not attest the verb echo pattern abundantly prevalent in relevant literature. Holmberg (2015) defines ‘verb echo’ answers as the answer pattern wherein the highest verbal element is repeated as a fragment answer to the question, as attested in Welsh and Russian here in (40) and (43). In fact, the unmarked fragment answer to a polar question in Hindi is what I have defined to be the ‘verb stacking’ pattern illustrated in (48ii). 47. Q: (kya) raaghav kitaab paDh paa rahaa thaa? Q raaghav book read able PROG.3MSg AUX.PST.3MSg ‘Was Raghav able to read the book?’ 48. A: i. haaN Yes ii. (haaN) paDh paa rahaa thaa - ‘Verb Stacking’ yes read able PROG.3MSg AUX.PST.3MSg 191 iii. #paa rahaa thaa able PROG.3Msg AUX.PST.3MSg iv. #rahaa tha PROG.3Msg AUX.PST.3MSg v. #padh read vi. #thaa - Verb echo (Holmberg 2015) AUX.PST.3MSg ‘Yes.’ Consequently, analyses which might posit movement of the verb higher than IP followed by IP ellipsis fail to account for the Hindi answers as Hindi does not manifest the relevant pattern. Simply adopting Holmberg’s (2015) PolP/IP ellipsis analysis for Hindi incurs certain non-trivial issues. With this account there is no easy explanation for why the valued polarity feature cannot be lexicalized by the highest verbal element in the Hindi verbal spine (i.e. (47vi) is not an acceptable answer form). Holmberg’s (2015) analysis enforces the movement of the highest verbal element to C but those patterns are not attested in Hindi. Given that IP ellipsis is possible in general in Hindi as we saw in section 5.2 and in this case if the fragment answer in (48i) is assumed to have a clausal structure having undergone IP ellipsis, it is unclear as to why Hindi does not attest ‘verb echo’ patterns. Also, unlike Finnish, Hindi does not manifest the optional occurrence of more than one verbal element and thereby we cannot explain why (48iii-vi) are unattested answer forms. There is no easy explanation for verb stacking being obligatory and it also being the unmarked answer pattern. I thereby conclude that the analysis discussed in Holmberg (2015) for Finnish and Welsh cannot be applied as it is for Hindi fragment answers. 192 Let us recall that Holmberg’s (2015) ‘verb echo’ patterns are motivated by a need to provide lexical exponence to the focused polarity value in the C domain. It is not clear as to how the entire verbal complex would move up to the C domain to provide lexical exponence. For Holmberg’s (2015) account to work all the verbal elements in Hindi would have to form a complex head via successive head movement and raise higher than I. In contrast, there is reason to believe that the verbal elements in Hindi do not form a complex head. Though Holmberg (2015) allows for ʋP ellipsis deriving fragment answers it is unclear as to how the particular derivation plays out. His analysis works for languages which attest ‘verb echo’ patterns and may also attest ‘verb stacking’ but not for languages like Hindi which obligatory attest ‘verb stacking’ and a ‘verb echo’ pattern is ungrammatical. Gribanova’s (2017) analysis for fragment answers in Russian is based on the interaction between independent mechanisms of head movement of the verb and attested ellipsis patterns in Russian. For Russian, Gribanova (2017) argues that the verb moves to the high Pol head for discourse reasons and since IP ellipsis is independently attested in the language, it applies under identity to give us the fragment answer patterns. If such an approach were to be adopted, one would have to argue that in Hindi the verb cannot move to a position higher than IP and thereby IP ellipsis in Hindi does not give us the ‘verb echo’ pattern. The only difference then would be parametric in the sense that in Russian the verbs can move outside of IP while in Hindi they can’t. The analysis presented in the paper for Hindi is amenable to Gribanova’s (2017) analysis if we take into account the language-specific differences – subject in Hindi may stay in spec,vP and the verb does not move to the high Pol. Gribanova’s (2017) analysis further revolves around the assumption that the subject in Russian has to be in spec,TP which is argued to not be the case with Hindi, so while a ʋP ellipsis analysis can account for the missing subject in Hindi it can’t account for the same in 193 Russian. That is, if we take the language particular properties into account Gribanova’s analysis can account for the Hindi fragment answer patterns. However, I will try and show that the analysis presented here might better capture the insights involved in fragment answer patterns cross- linguistically. The analysis presented in the paper for Hindi is a verb stranding ʋP ellipsis approach to account for the ‘verb stacking’ pattern in Hindi. Hindi furthermore does not attest the ‘verb echo’ pattern as exhibited by Welsh, Finnish, Russian among other languages. As per this analysis ʋP ellipsis is triggered by an ellipsis feature Eans which is merged with the Pol head. The Pol head in Hindi is merged in an IP-internal position such that it takes ʋP as its complement. The ellipsis feature leads to the deletion of ʋP, its complement. This analysis I argue can better explain the Hindi pattern of ‘verb stacking’ as the fragment is a result of verb stranding ʋP ellipsis triggered by an ellipsis feature. Importantly this is not an account which enforces (Holmberg 2015) or assumes (Gribanova 2017) the movement of the verbal elements higher than I, though by no means it disallows said movement. All the three approaches mentioned here have similar insights in the sense that all of them adhere to an ellipsis approach somehow constrained by the polarity feature. For Holmberg (2015) it is the lexicalization of the focused polarity value which triggers the verbal movement from I to C. This movement is then enforced by the analysis as it is considered obligatory for polar answers. IP ellipsis follows under identity with the question. For Gribanova (2017) the verb head movement to a higher position is independent of the analysis itself, though still constrained by polarity. Gribanova (2017) argues this head movement to a high Pol head to be a result of marking the discourse function of contrastive polarity focus. The ellipsis feature on Pol then results in the 194 ellipsis of the IP complement. The analysis presented in this paper is simply an ellipsis analysis triggered by an ellipsis feature on a Pol head. We have already discussed how Holmberg’s (2015) analysis would not work for verb stacking languages like Hindi. Gribanova’s (2017) account is merely an interaction of two independent processes and can be employed for Hindi, however I will argue that it lacks certain insights that can be better explained by the analysis presented in this paper. Gribanova (2017) claims that the SV order as illustrated in (50) is not attested in Russian answers but such an order is in fact predicted to occur. 49. Q: evgenija otpravila posylku v moskvu Evgenija send.PST.3FSg package.ACC to Moscow.ACC ‘Did Evgenija send the package to Moscow?’ 50. A: (Da,) (#/?ona) otpravila yes she send.PST.3FSg ‘Yes.’ - Russian This order is a result of a case where the verb hasn’t moved as high as to Pol but stays in the IP- internal Neg head and ʋP ellipsis which is independently attested in Russian is implemented. This word order should be attested. In fact, this order is available in declarative contexts. There is no simple explanation for why this order is not attested in answer patterns. One way of accounting for this would be to say that in answers the verb obligatorily has to move to the high Pol head. If this is true, we will find only VSO order in answers to polar questions. I am unaware if that is indeed the case. Gribanova (2017) argues that the verb complex only moves to the high Pol head resulting in a VSO word order when the polarity is contrastively focused. The fragment answers to polar questions then have to be argued to be always have contrastively focused polarity. Gribanova (2017) further posits that when the verb has moved high to Pol, the VS order pattern 195 resulted by ʋP ellipsis is unattested as it violates MaxElide. The empirical conclusion irrespective of the explanation is that we never get ʋP ellipsis in polar answers in Russian. This insight however is not explicitly clear in Gribanova’s (2017) analysis. Let us add to this another language which attests to a similar pattern. Welsh as argued by Holmberg (2015) only has the ‘verb echo’ pattern and not the ‘verb stacking’ pattern. Hence answers in Welsh are not derived by ʋP ellipsis. However, ʋP ellipsis is independently attested in the language (Rouveret 2012). What we have here are two ‘verb echo’ languages which do not attest ʋP ellipsis in answers. As illustration of another language type, we have Hindi which attests to ʋP ellipsis in answers. The analysis for fragment polar answers should ideally allow for this parametric variation in the (un)availability of ʋP ellipsis as suitable phenomena for derivation of fragment answers. I believe the analysis I have argued for, can be modelled to account for this variation. If the analysis I have proposed is on the right track, ellipsis is constrained by an ellipsis feature merged with the Pol head and all we need to do to capture the cross-linguistic variation is to say that languages differ in where the Pol head is merged in the clausal structure. It can be suggested that the Pol head in Welsh and Russian is merged IP externally and when the ellipsis feature is hosted on this IP external Pol head it results in IP ellipsis giving the ‘verb echo’ answer pattern. Holmberg (2015) and Gribanova (2017) do posit an IP external high Pol head in Welsh, Finnish and Russian. Contrastively in Hindi, the Pol head is merged right above ʋP such that when the ellipsis feature is hosted on this Pol head it results in the ellipsis of the ʋP complement. Thereby variation in ‘verb echo’ and ‘verb stacking’ patterns are merely a result of the height of the Pol in a specific language. This allows us an explanation of why ʋP ellipsis though independently attested is not a possible ellipsis option in deriving answer forms in Welsh and Russian. McCloskey (2017) gives a similar argument while comparing responsive (IP) ellipsis in Irish and VP ellipsis in English. He also 196 suggests that ellipsis size is constrained across languages by the height at which a Pol head is merged. This extension to the analysis presented here would easily account for the different types of languages mentioned in the literature. The attested patterns in Finnish however are a little problematic. Finnish is a ‘verb echo’ language which could be accounted for by a high Pol head resulting in IP ellipsis for the answer fragments. However, Finnish also attests ʋP ellipsis patterns as can be seen in (19). One possible way to explain the Finnish patterns would be to suggest that all the verbal elements in the language can form a complex head and can raise to a position higher than I and all answer patterns are derived by IP ellipsis. Such an account was indeed argued by Holmberg (2001). However later Holmberg (2015) favors a ʋP ellipsis approach. Indeed, the optionality of the presence of multiple verbal elements makes the theory of a complex verbal head formation untenable. Yet another way to deal with the Finnish patterns would be to suggest that a Max Elide constraint is in play and while it is ranked very high in Welsh and Russian, Finnish has a low ranked Max Elide and hence we see lower constituent ellipsis patterns. We need additional data as to the well formedness and (un)markedness of different options in (19) to be able to better support this idea. Concluding, the three analyses mentioned here are all couched in the same philosophy, just the implementation is slightly different. It is agreed that the polarity head plays an important role in the derivation of polar answers. The point of contention is as to how exactly it constrains the observed answer patterns. The analysis presented in this dissertation, I believe can account for both ‘verb echo’ and ‘verb stacking’ languages by constraining the kind of ellipsis attested in answers which seems to be a subset of the ellipsis attested in general in that particular language. This subset is modulated by an ellipsis feature hosted by a polarity head whose height parametrically varies from language to language thereby giving us languages with different fragment answer patterns. 197 5.5 Conclusion Bošković (2014) posits that ellipsis is constrained by phases. He argues that only phases and complements of phase heads can undergo ellipsis. Chomsky (200, 2001) has argued that CP and υP are phases. This seems coherent with the system of ellipsis we have developed here in Hindi. As per Chomsky, proposition-hood plays an important role in determining phases. Chomsky argues that phases are syntactic counterparts of the semantic notion of a proposition, thereby positing that CP and υP are syntactic reflections of propositions. If we assume that only CP and υP are phases, following Chomsky, and that ellipsis is constrained by phases, we can account for the two kinds of ellipsis we have seen in this work. Here I have argued that fragment answers specifically and clausal ellipsis more generally is derived by either verb stranding υP ellipsis or by IP ellipsis. I have further argued that υP and IP are complements to the two polarity phrases projected in the Hindi clausal structure. Since clauses are the syntactic representations of propositions and given that answers are always full propositions, it makes sense to constrain clausal ellipsis, particularly in answers, to the two clausal representations in syntax. And it seems plausible to have two polarity heads as the edges of these two clausal representations. When initially posited by Laka (1994), a polarity head was argued to host the force of the proposition- whether it be assertion or negation. 198 References Asher, N., & Reese, B. (2005). Negative bias in polar questions. In Proceedings of Sinn und Bedeutung (Vol. 9, pp. 30-43). Asher, N., & Reese, B. (2007). Intonation and discourse: Biased questions. Interdisciplinary studies on information structure, 8, 1-38. Bailey, L. R. (2013). The syntax of question particles. PhD Dissertation. Newcastle University. Bailyn, J. F. 1995. A Configurational Approach to Russian “Free” Word Order, unpublished Ph. D (Doctoral dissertation, dissertation, Cornell University, Ithaca). Barros, M. (2014). Sluicing and identity in ellipsis. Rutgers The State University of New Jersey-New Brunswick. Bartels, C. (1999). The intonation of English statements and questions: A compositional interpretation. Routledge. Bayer, J., Dasgupta, P., Mukhopadhyay, S., & Ghosh, R. (2014). Functional structure and the Bangla discourse particle to. Retrieved July, 19, 2015. Belletti, A. (2001). Inversion as focalization. Subject inversion in Romance and the theory of Universal Grammar, 60-90. Belletti, A. (2004). Aspects of the low IP area. The structure of CP and IP. The cartography of syntactic structures, 2, 16-51. Bhadra, D. (2017). Evidentiality and questions: Bangla at the interfaces (Doctoral dissertation, Rutgers University-School of Graduate Studies). Bhatt, R. & Homer, V. (2014). The Interpretation of Infinitival Negation in Hindi-Urdu. Presented at Fasal 4, Rutgers University, NJ. 199 Bhatt, R. (2005). Long distance agreement in Hindi-Urdu. Natural Language and Linguistic Theory 23: 757–807. Bhatt, R., & Dayal, V. (2007). Rightward scrambling as rightward remnant movement. Linguistic Inquiry, 38(2), 287-301. Bhatt, R., & Dayal, V. (2020). Polar question particles: Hindi-Urdu kya. Natural Language & Linguistic Theory, 38(4), 1115-1144. Bhatt, R & Dayal, V. (2014). Polar questions and disjunction: clues from hindi-urdu polar kya. Talk presented Israel Association for Theoretical Linguistics: 30th Annual Meeting Ben Gurion University of the Negev Bhattacharya, T and Simpson, A. (2012). Sluicing in Indo-Aryan: An investigation of Bangla and Hindi. In Sluicing: Cross-linguistic perspectives, ed. by Andrew Simpson and Jason Merchant, 183–218. Oxford: Oxford University Press. Biezma, María, Miriam Butt, and Farhat Jabeen. 2017. Interpretations of Urdu/Hindi polar kya. Slides of talk at the Workshop on Non-At-Issue Meaning and Information Structure at the University of Konstanz. Bolinger, D. (1978). Yes—no questions are not alternative questions. In Questions (pp. 87- 105). Springer, Dordrecht. Borkin, A. (1971). Polarity items in questions. Bošković, Ž. (2014). Now I’ma phase, now I’m not a phase: On the variability of phases with extraction and ellipsis. Linguistic inquiry, 45(1), 27-89. Breitbarth, A., De Clercq, K., & Haegeman, L. (2013). The syntax of polarity emphasis. Lingua, 128, 1-8. Brody,M. & Szabolsci, A. (2000). Overt scope: a case study in Hungarian. Retrieved from: http://homepages.nyu.edu/~as109/papers.html. Brody, M. (1989). “Some Reflections on the Focus Field in Hungarian”. University College London Working Papers in Linguistics 2.201–225. 200 Brody, M. (1996). “Focus in Perfect Syntax”. Paper presented at the Table ronde internationale sur la grammaire du focus, University of Paris II & University of Paris X, February1996. Buring, D., & Gunlogson, C. (2000). Aren't positive and negative polar questions the same? Krifka, M. (1995). The semantics and pragmatics of polarity items. Linguistic Analysis 25, 209–257. Butt, M, Bögel, T & Jabeen, F. (2017). Urdu/Hindi questions at the syntax- pragmatics[1]prosody interface. Slides of talk at Formal Approaches to South Asian Languages (FASAL) 7 at MIT Cable, S. (2007). The grammar of Q: Q-particles and the nature of Wh-fronting, as revealed by the Wh-questions of Tlingit (Doctoral dissertation, Massachusetts Institute of Technology). Caponigro, I. and J. Sprouse (2007). Rhetorical questions as questions. In Proceedings of Sinn und Bedeutung, Volume 11, pp. 121–133. Cheng, L. L. S. (1991). On the typology of wh-questions. MIT Dissertation. Chomsky, N. & Lasnik, H.(1993). Principles and parameters theory. In Syntax: An international handbook of contemporary research. Eds. Joachim Jacobs, Arnim vonStechow, Wolfgang Sternefeld, and Theo Vennemann, 506–569. Berlin: Walter de Gruyter. Chomsky, Noam. (2001). “Derivation by Phase”, in M. Kenstowicz (ed.), Ken Hale: A Life in Language, Cambridge (Mass.): MIT Press, 1-52. Chung, S. (2006). Sluicing and the lexicon: The point of no return. In Proceedings of the annual meeting of the Berkeley Linguistics Society (Vol. 31, pp. 73-91). Berkeley Linguistics Society Berkeley, CA. Chung, S., Ladusaw, W. A., & McCloskey, J. (1995). Sluicing and logical form. Natural language semantics, 3(3), 239-282. Dash, B., Datta, M., & Simpson, A. (2022). A Unified Analysis of the Hindi & Bangla discourse particle-to. Formal Approaches to South Asian Languages, 1(1). 201 Dayal, V. (1993). Scope marking as indirect wh-dependency. Natural language semantics, 2(2), 137-170. Dayal. (2016). Questions. Oxford University Press. Depiante, M. A. (2000). The syntax of deep and surface anaphora: A study of null complement anaphora and stripping/bare argument ellipsis. University of Connecticut. Diesing, M. (1992). Bare plural subjects and the derivation of logical representations. Linguistic Inquiry, 23(3), 353-380. Dwivedi, V. (1991). Negation as a functional projection in Hindi. In WECOL (Vol. 4, pp. 88- 100). Fiengo, R., & May, R. (1994). Indices and identity (Vol. 24). MIT press. Fischer, S., Alexiadou, A., 2001. On stylistic fronting: Scandinavian vs. Romance. Working Papers in Scandinavian Syntax 68, 117—145 Fowler, G. (1994). Verbal prefixes as functional heads. Studies in the Linguistic Sciences 24(1–2): 171–185. Ginzburg, J and Sag, I. (2000). Interrogative Investigations: The Form, Meaning, and Use of English Interrogatives. Center for the Study of Language and Information, Stanford, CA. Goldberg, L. M. (2005). Verb-stranding VP ellipsis: A cross-linguistic study (Doctoral dissertation, McGill University). Gribanova, V. (2013a). Verb-stranding verb phrase ellipsis and the structure of the Russian verbal complex. Natural Language & Linguistic Theory, 31(1): 91–136. Gribanova, V. (2017). Head movement and ellipsis in the expression of Russian polarity focus. Natural Language & Linguistic Theory, 35(4), 1079-1121. 202 Gribanova, V., & Manetta, E. (2016). Ellipsis in wh-in-situ languages: Deriving apparent sluicing in Hindi-Urdu and Uzbek. Linguistic Inquiry, 47(4), 631-668. Guerzoni, E. (2003). Why even ask?: on the pragmatics of questions and the semantics of answers. Ph. D. thesis, Massachusetts Institute of Technology. Guerzoni, E. (2004). Even-npis in yes/no questions. Natural Language Semantics 12(4), 319– 343. Gutzmann, D., Hartmann, K., & Matthewson, L. (2017). Verum focus is verum, not focus. Cross-linguistic evidence. ms., available at https://www. danielgutzmann. com/work/verum-focus-is-verum-not-focus. Hagstrom, P. A. (1998). Decomposing questions (Doctoral dissertation, Massachusetts Institute of Technology). Haider, H. (2013). Symmetry breaking in syntax (No. 136). Cambridge University Press. Hallman, P. (1998). Reiterative syntax. In Clitics, pronouns and movement. Ed. J. Black and Hamblin, C. (1958). ‘Questions’, Australasian Journal of Philosophy 36 (3): 159-168. Hamblin, C. L. (1973). Questions in Montague grammar. Foundations of language, 10(1), 41-53. Han, C. H., & Romero, M. (2004). Disjunction, focus, and scope. Linguistic Inquiry, 35(2), 179-217. Han, C. H., & Romero, M. (2004). The syntax of whether/Q... or questions: Ellipsis combined with movement. Natural Language & Linguistic Theory, 22(3), 527-564 Hankamer, J. (1979). Deletion in Coordinate Structures. Garland, New York. PhD thesis, Yale University, 1971. 203 Hankamer, J.(1979). Deletion in Coordinate Structures. Garland, New York. PhD thesis, Yale University, 1971. Hartmann, K. (2000). Right Node Raising and Gapping: Interface Conditions on Prosodic Deletion. John Benjamins Publishing. Hernanz, M.L., 2011. Assertive bien in Spanish and the left periphery. In: Benincà, P., Munaro, N. (Eds.), Mapping the Left Periphery. Oxford University Press, Oxford, pp. 19--62. Hoeks, M. (2019). Focus alternatives in alternative questions. European Summer School for Logic, Language and Information (ESSLLI), 1-12. Höhle, T. N. (1992). Ueber verum-fokus im Deutschen. In Informationsstruktur und grammatik (pp. 112-141). VS Verlag für Sozialwissenschaften, Wiesbaden. Holmberg, A. (2013). The syntax of answers to polar questions in English and Swedish. Lingua, 128, 31-50. Holmberg, A., (2007). Null subjects and polarity focus. Studia Linguistica 61 (3), 212--236. Holmberg, A. (2015). The syntax of yes and no. Oxford University Press. Huddleston, R. & Pullum, G. (2002). The Cambridge grammar of the English Language. Cambridge: Cambridge University Press Hudson, R. A. (1976). Conjunction reduction, gapping, and right-node raising. Language, 535-562. Jayaseelan, K. A. (2008). Question particles and disjunction. Ms., Hyderabad, English and Foreign Languages University. Jayaseelan, K.A. (1989). “Question-Word Movement to Focus and Scrambling in Malayalam”. Ms.,Central Institute of English & Foreign Languages, Hyderabad. Johnson, K. (1996). In search of the English middle field. Ms., University of Massachusetts, Amherst, 516. 204 Johnson, K. (2004). How to be quiet. In Proceedings from the Annual Meeting of the Chicago Linguistic Society (Vol. 40, No. 2, pp. 1-20). Chicago Linguistic Society. Johnson, K. (2009). Gapping is not (VP-) ellipsis. Linguistic inquiry, 40(2), 289-328. Johnson, K. (2018). Gapping and Stripping. In Jeroen van Craenenbroeck & Tanja Temmerman (eds.), The Oxford Handbook of Ellipsis, 561–604. Oxford University Press. Jones, B. M. (1999). The Welsh answering system (Vol. 120). Walter de Gruyter. Kandybowicz, J., 2008. The Grammar of Repetition. John Benjamins, Amsterdam. Kazenin, K. (2006). Polarity in Russian and typology of predicate ellipsis. Ms., Moscow State University. Kidwai, Ayesha. (1995). Binding and Free Word Order Phenomena in Hindi and Urdu. Doctoral dissertation, Jawaharlal Nehru University, New Delhi. Kim, S. (1999). Sloppy/strict identity, empty objects, and NP ellipsis. Journal of East Asian Linguistics, 8(4), 255-284. Kishimoto, H. (2005). Wh-in-situ and movement in Sinhala questions. Natural Language & Linguistic Theory, 23(1), 1-51. Kramer, R., & Rawlins, K. (2010). Polarity particles and ellipsis: A (somewhat) cross- linguistic perspective. In UCSC workshop on polarity particles. Kramer, R., & Rawlins, K. (2012, June). An ellipsis approach to answer particles in positive and negative contexts. In Workshop on the Syntax of Answers to Polar Questions, Newcastle University. Kumar, R. (2006). Negation and Licensing of Negative Polarity Items in Hindi Syntax. New York: Routledge. 205 Ladd, D. R. (1981). A first look at the semantics and pragmatics of negative questions and tag questions. In Papers from the Regional Meeting. Chicago Linguistics Society, Chicago, Ill., Number 17, pp. 164–171. Ladusaw, W. A. (1992, June). Expressing negation. In Semantics and linguistic theory (Vol. 2, pp. 237-260). Lahiri, U. (2002). Questions and answers in embedded contexts (Vol. 2). Oxford University Press on Demand. Laka, I. (1994). On the Syntax of Negation. Garland, New York/London. Lipták, A., (2003). The expression of sentential emphasis in Hungarian. Ms. University of Leiden. Langacker, R. W. (1974). Movement rules in functional perspective. Language, 630-664. Larson, Richard K. 1985. On the sy/ntax of disjunction scope. Natural Language and Linguistic Theory 3(2). 217–164 Mahajan, A. K. (1988). Word order and negation in Hindi. Ms. Massachusetts institute of Technology. Mahajan, A. K. (1990). The A/A-bar distinction and movement theory. Doctoral Dissertation. MIT. Manetta, E. (2010). Wh-expletives in Hindi-Urdu: the vP phase. Linguistic Inquiry, 41(1), 1- 34. Manetta, E. (2019). Verb-phrase ellipsis and complex predicates in Hindi-Urdu. Natural Language & Linguistic Theory, 37(3), 915-953. Martins, A.M., (2006). Emphatic affirmation and polarity: contrasting Portuguese with Brazilian Portuguese, Spanish, Catalan and Galician. In: Doetjes, J., Gonzalez, P. (Eds.), Romance Languages and Linguistic Theory 2004. John Benjamins, Amsterdam & Philadelphia, pp. 197--223. 206 McCloskey, J. (1991). Clause structure, ellipsis and proper government in Irish. Lingua, 85(2-3), 259-302. McCloskey, J. (1997). Subjecthood and subject positions. In Elements of grammar (pp. 197- 235). Springer, Dordrecht. McCloskey, J. (2017). Ellipsis, polarity, and the cartography of verb-initial orders in Irish. Elements of comparative syntax: Theory and description, 99-151. Merchant, J. (2001). The syntax of silence: Sluicing, islands, and the theory of ellipsis. Oxford University Press on Demand. Merchant, J. (2005). Fragments and ellipsis. Linguistics and philosophy, 27(6), 661-738. Merchant, J. (2013). Voice and ellipsis. Linguistic Inquiry, 44(1), 77-108. Merchant, J. (2016). Ellipsis: A survey of analytical approaches. Ms., University of Chicago, Chicago, IL. Miyagawa, S.(2001). The EPP, scrambling, and wh-in-situ. In Ken Hale: A life in language, ed. by Michael Kenstowicz, 293–338. Cambridge, Mass.: MIT Press. Miyagawa, S., & Arikawa, K. (2007). Locality in syntax and floating numeral quantifiers. Linguistic Inquiry, 38(4), 645-670. Morgan, J. (1973). Sentence Fragments and the Notion “sentence”’, in Braj Kachru. Robert Lees, Yakov Malkiel, Angelina Pietrangeli, and Sol Saporta (eds.), Issues in Linguistics, pp. 719–751, University of Illinois Press, Urbana. Morgan, J.(1989). Sentence Fragments Revisited. in Bradley Music, Randolph Graczyk, and Caroline Wiltshire (eds.), CLS 25, Parasession on Language in Context, pp. 228–241, Chicago Linguistic Society, Chicago. Morris, A. (2008). Polarity ellipsis and negative stripping. Unpublished manuscript. http://babel. ucsc. edu/~ hank/PolarityEllipsisandNegStripping. pdf. 207 Neijt, A. (1979). Gapping: A Contribution to Sentence Grammar, Foris, Dordrecht. NeijtGapping: A Contribution to Sentence Grammar1979. Nevins, A. (2008, April). Sluicing≠ stripping: evidence from P-stranding. In 3rd Annual Moscow Student Conference on Linguistics, Moscow. Oku, S. (1998). A theory of selection and reconstruction in the minimalist perspective. Otani, K., & Whitman, J. (1991). V-raising and VP-ellipsis. Linguistic Inquiry, 345-358. Pesetsky, D. (1987). WH-in-situ: Movement and unselective binding. The representation of (in) definiteness, ed. by Eric Reuland and Alice ter Meulen, 98–129. Poletto, C., 2008. The syntax of focus negation. In: Working papers in Linguistics. University of Venice, Vol. 18, pp. 179--201. Pope, E. (1976). Questions and Answers in English. The Hague: Mouton. Pruitt, K. (2008). Prosody and focus in alternative questions: Accounting for interpretation. Experimental and Theoretical Advances in Prosody. Cornell, Ithaca, NY. Pruitt, K., & Roelofsen, F. (2013). The interpretation of prosody in disjunctive questions. Linguistic inquiry, 44(4), 632-650. Pruitt, K & Roelofsen, F. (2011). Disjunctive questions: Prosody, sy/ntax, and semantics. Ms., University of Massachusetts, Amherst and ILLC, University of Amsterdam Quirk, R., Greenbaum, S., Leech, G., Svartvik, J. (1985). A comprehensive grammar of the English language. Essex: Longman. Rackowski, A., & Richards, N. (2005). Phase edge and extraction: A Tagalog case study. Linguistic Inquiry, 36(4), 565-599. Reese .(2007). Bias in Questions. PhD Dissertation at The University of Texas, Austin. 208 Reese, B., & Asher, N. (2010). Biased questions, intonation and discourse. Information Structure: Theoretical, Typological, and Experimental Perspectives, 139-73. Reich, I. (2002). Question/Answer Congruence and the Semantics of wh-phrases. Theoretical Linguistics, 28, 73–94. Reich, I. (2003,). Frage, Antwort und Fokus. Akademie Verlag, Berlin. Reinhart, T. (1991). Elliptic conjunctions-non-quantificational LF. Rizzi, L. (1997). ‘The fine structure of the left periphery’, in Liliane M. V. Haegeman (ed.), elements of grammar. Dordrecht: Kluwer. Roelofsen, F., & Farkas, D. F. (2015). Polarity particle responses as a window onto the interpretation of questions and assertions. Language, 91(2), 359-414. Rohde, H. (2006). Rhetorical questions as redundant interrogatives. San Diego Linguistics Papers 2, 134–168. Romero, M. (2006). Biased polar questions: The role of verum. Sprache und Datenverarbeitung, 30(1), 9-24. Romero, M., & Han, C. H. (2002). Verum focus in negative polar questions and Ladd's p/¬ p ambiguity. In Semantics and Linguistic Theory (Vol. 12, pp. 204-224). Romero, M. & Han, C.H. (2004). On negative yes/no questions. Linguistics and Philosophy 27(5), 609–658. Rooth, Mats. 1985. ‘Association with Focus’, Ph.D. thesis, University of Massachusetts at Amherst. Rooth, Mats. 1992. ‘A Theory of Focus Interpretation’, Natural Language Semantics 1, 75– 116. Ross John, R. (1969). Guess who. In 5th Annual Meeting of the Chicago Linguistic Society (CLS 5) (pp. 252-286). 209 Ross, J. R. (1970). On declarative sentences. Readings in English transformational grammar, 222, 272. Ross, J. R.(1970). ‘Gapping and the Order of Constituents’, in M. Bierwisch and K. E. Heidolph (eds.), Progress in Linguistics, Mouton, The Hague, pp. 249–259. Rouveret, A. (2012). VP ellipsis, phases and the syntax of morphology. Natural Language & Linguistic Theory, 30(3), 897-963. Ruangjaroon, S. (2005). The syntax of wh-expressions as variables in Thai. PhD diss., The University of British Columbia. Rudin, D. (2018). Head-based syntactic identity in sluicing. Linguistic Inquiry, 1-53. Sato, Y., & Hayashi, S. (2018). String‐Vacuous Head Movement in Japanese: New Evidence from Verb‐Echo Answers. Syntax, 21(1), 72-90. Sadock, J. M. (1971, April). Queclaratives. In Seventh Regional Meeting of the Chicago Linguistic Society (Vol. 7, pp. 223-232). Sadock, J. M. (1974). Toward a linguistic theory of speech acts. Academic Pr. Schutze, C. T. (2001). On Korean" Case stacking": The varied functions of the particles ka and lul. Linguistic review, 18(3), 193-232. Schwarz, B. (1999). ‘On the Sy/ntax of either . . . or’, Natural Language and Linguistic Theory 17(2), 339–370. Schwarzschild, R. (1999). GIVENness, AvoidF and other constraints on the placement of accent. Natural language semantics, 7(2), 141-177. Şener, S., & Takahashi, D. (2009). Argument ellipsis in Japanese and Turkish. In sixth Workshop on Altaic Formal Linguistics (WAFL-6), Nagoya University, Japan. 210 Simpson, A. (2014). Verbal answers to polar questions, focus and ellipsis. Chinese syntax in a cross-linguistic perspective, 299-333. Simpson, A., Choudhury, A., & Menon, M. (2013). Argument ellipsis and the licensing of covert nominals in Bangla, Hindi and Malayalam. Lingua, 134, 103-128. Stanley, J. (2000). Context and Logical Form. Linguistics and Philosophy 23,391–434. Stechow, A. V. (1996). The different readings of wieder ‘again’: A structural account. Journal of semantics, 13(2), 87-138. Steedman, M. J.(1990). Gapping as Constituent Coordination. Linguistics and Philosophy 13(2), 207-263. Syed, S., & Dash, B. (2017). A unified account of the yes/no particle in Hindi, Bangla and Odia. Proceedings of GLOW in Asia XI, 1, 201-212. Takahashi, D. (2006). Apparent parasitic gaps and null arguments in Japanese. Journal of East Asian Linguistics, 15(1), 1-35. Tang, S. W. (2020). Cartographic syntax of performative projections: evidence from Cantonese. Journal of East Asian Linguistics, 29(1), 1-30. Toosarvandani, M. (2008). Wh-movement and the syntax of sluicing. Journal of Linguistics 44:677–722. Toosarvandani, M. (2013). Corrective but Coordinates Clauses Not Always but Sometimes. Natural Language and Linguistic Theory 31(3), 827–863. Toosarvandani, M. (2016). Embedding the Antecedent in Gapping: Low Coordination and the Role of Parallelism. Linguistic Inquiry 47(2), 381–390 Uegaki, W. (2014, August). Japanese alternative questions are disjunctions of polar questions. In Semantics and linguistic theory (Vol. 24, pp. 42-62). Van Rooy, R. (2003). Negative polarity items in questions: Strength as relevance. Journal of semantics 20(3), 239–273. 211 Vicente, L. (2006). Short negative replies in Spanish. Linguistics in the Netherlands, 23(1), 199-211. von Stechow, A. (1991). Focusing and backgrounding operators. In W. Abraham (ed.), Discourse particles. (pp. 37–84). Amsterdam: John Benjamins Williams, E. S. (1977). Discourse and logical form. Linguistic inquiry, 101-139. Wiltschko, M. (2017a). Ergative constellations in the structure of speech acts. In The oxford handbook of ergativity. Wiltschko, M. (2017b). Response markers as a window into linguistic modularity. Edited by Clemens Mayr Leibniz-Zentrum Allgemeine Sprachwissenschaft Edwin Williams Princeton University, 303. Wiltschko, M. (2018). Response particles beyond answering. Order and structure in syntax I, 241. Wiltschko, M., & Heim, J. (2016). The syntax of confirmationals. Outside the Clause: Form and function of extra-clausal constituents, 178, 305. Wurmbrand, S. (2017). Stripping and topless complements. Linguistic Inquiry, 48(2), 341- 366. Yaisomanang, S. (2012). The syntax of yes-no questions and answers in Thai. PhD dissertation, Newcastle University. Zanuttini, R., 1997. Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford University Press, Oxford.

Abstract (if available)

Abstract Merchant (2005) and Holmberg’s (2015) research on fragment answers posits that answer forms have structures similar to the questions they are responses to, thereby arguing that the fragment answer forms are derived from the non-pronunciation/ellipsis of significant portions of a full clausal sentence. Answers to yes/no questions like (1) below are argued to have an underlying sentential structure which is only partially pronounced (non-pronounced material is represented with strike-through across words). Welsh, Irish, Finnish, Mandarin Chinese, Japanese and Russian are ‘verb echo’ languages which show a pattern wherein the structurally highest verbal element is repeated as an answer. Holmberg (2015) analyzes ‘verb echo’ patterns as a structure derived by IP Ellipsis with the highest verb having raised to a position higher than IP. Interestingly, in Hindi, the entire verbal sequence needs to be repeated in an answer as can be seen in (2). I term the pattern in Hindi as ‘verb stacking’.
1. Q: all mair aros
can mair stay
‘Can Mair stay?’

A: gall Mair aros
can Mair stay
‘Yes.’ verb echo -Welsh (Holmberg 2015)

2. Q: (kya) raaghav kitaab paDh paa rahaa thaa?
Q raaghav book read able PROG.3MSg AUX.PST.3MSg
‘Was Raghav able to read the book?’

A: haaN paDh paa rahaa thaa raaghav kitaab
yes read able PROG.3Msg AUX.PST.3MSg raaghav book
‘Yes.’ verb stacking - Hindi

The primary objective of this dissertation is to provide an account of the novel fragment answer patterns attested in Hindi. Any account of fragment answers in Hindi should ideally be able to account for obligatory verb stacking given that the verbal sequence does not form a complex head and does not raise higher than IP. We also need to account for the fact that IP ellipsis is independently attested in the language but as we do not see verb echo patterns, fragment answers cannot be derived by IP ellipsis. Yet another empirical observation that needs to be accounted for is that in Hindi the verb stacking pattern is the unmarked answer pattern whereas answering a yes/no question simply with a polarity particle is a marked response. I propose that ellipsis in fragments answers is triggered by a Merchant (2005) style Ellipsis feature hosted by the polarity head which causes the complement of the polarity head to be elided under semantic identity. The default polarity head in Hindi is merged IP-internally just above the ʋP, leading to the deletion of ʋP after the main verb has raised out of it, resulting in the verb stacking pattern. I show that this fragment answer pattern follows from an independently motivated ellipsis process of υ-stranding υPE (Bhatt & Dayal 2007, Manetta 2019) in Hindi. The proposed analysis assumes head movement of the verb outside of ʋP to the IP-internal polarity head. Bhatt & Dayal (2007) and Manetta (2019) among others have all argued for the main verbal element in Hindi to move out of ʋP, though there is not full agreement on the identity of the head to which the verbal element is assumed to move to. I propose that the verb moves to a polarity head projected immediately above the ʋP and below the aspect head, for the purpose of lexical exponence of the polarity feature. Furthermore, I suggest that this head movement is syntactic as it feeds into the syntactic process of ellipsis and is not completely determined by morphological considerations.
In addition to arguing for an IP-internal default polarity head I motivate a higher polarity phrase specified for contrastively focused polarity. Romero (2006) among others has argued for two different positions of verum focus which may syntactically correspond to two distinct polarity phrases, one being a default projection and the other involving contrastive focus. Most recent literature on polarity emphasis presents polarity in clauses by giving a distributed structure, a projection on the left periphery which encodes focused polarity along with a lower polarity projection (Breitbarth et al. 2013). The analysis proposed for Hindi here supports this generalization. In Hindi, the higher projection is only licensed when parallel constructions have opposing polarity making contrastive polarity explicit in the discourse. I argue that when merged, the higher polarity head can host an ellipsis feature and licenses ellipsis of its complement IP. I further propose that verbal ellipsis in Hindi is constrained and licensed by polarity heads. The relevance of polarity comes into play only when it is being questioned, as in polar questions, or opposed, as in contrastive polarity ellipsis structures. I postulate that the IP-internal polarity phrase licenses verb stranding ʋP ellipsis and the contrastive polarity phrase above IP licenses IP ellipsis. Given that the only possible verbal ellipsis options are ʋP ellipsis or IP ellipsis, I further provide support for a general account of ellipsis being constrained by the two phases: ʋP and CP and licensed by polarity (Bošković 2014, McCloskey 2017).
Besides the two dedicated positions for polarity in Hindi, I further motivate yet another projection which could potentially host a polarity particle assuming a cartographic approach to performative projections in Hindi. While examining fragment answer patterns to negative biased polar questions, I argue, following Roelofsen & Farkas (2015) that polarity particles may be anaphoric to the antecedent questions and signify (dis)-agreement of the fragment answer with the highlighted alternative of the question. Based on the prosodic properties of polarity particles in fragment answer patterns, I argue that when they have such an anaphoric function, they occupy a higher position in a performative projection in the extended left periphery.
The dissertation, consequently, argues for three different clausal representations for polarity of a sentence. The default polarity head carrying the sentential polarity of a clause is merged in an IP-internal position. When in discourse there is explicit contrastively focused polarity, sentential polarity occupies a higher position and the dedicated contrastive polarity head takes an IP complement. Finally, when a polarity particle is anaphoric to a previous sentence and references the polarity of the previous sentence, it occupies a position in the extended left periphery. Such a clausal representation is shown to be able to model fragment answers to all kinds of questions in Hindi.
Additionally, I posit that typological variation in fragment answer patterns is determined by the position of polarity phrases in different languages. While an IP-internal default polarity head accounts for verb stacking ellipsis pattens in fragment answers, a higher polarity phrase in other languages results in the different ‘verb echo’ (Holmberg 2015) pattern exhibited in those languages (Irish (McCloskey 2017), Finnish (Holmberg 2015), Welsh (Holmberg 2015), Russian (Gribanova 2017), Mandarin (Wei 2016) and Japanese (Sato et al. 2018). Since polarity phrases may occur in different positions in the clausal structure in different languages, I propose that this variation in the physical location of a polarity phrase directly results in differences in the amount of sentential material that can be left unpronounced in an answer, producing fragments of differing types and sizes as semantically parallel answer forms. Thereby, in principle, all an answer-form requires is a valued polarity feature in a language and as soon as that requirement is fulfilled, ellipsis of all subordinate material takes place. Hence the position of the polarity head in a clause determines the size of ellipsis and consequently the size of the answer fragments. A stronger generalization I posit is that cross-linguistically fragment answers are always derived by verbal ellipsis, even in languages like Hindi which attest to mechanisms like pro drop and argument ellipsis. These other mechanisms could independently account for some silent syntactic structure but not that of fragments as answers as can be consistently observed in Hindi fragment answer patterns.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Towards the unity of movement: implications from verb movement in Cantonese

PDF

Syntax-prosody interactions in the clausal domain: head movement and coalescence

PDF

The morphosyntax of states: deriving aspect and event roles from argument structure

PDF

Perspective in Turkish complementation

PDF

Copy theory of movement and PF conditions on spell-out

PDF

Building phrase structure from items and contexts

PDF

The case of a person: The person case constraint in German

PDF

Where number lies: plural marking, numerals, and the collective-distributive distinction

PDF

Number marking and definiteness in Bangla

PDF

Interaction between prosody and information structure: experimental evidence from Hindi and Bangla

PDF

Subject and adjunct island effects through the lens of selective opacity in Bangla subjects and adjuncts

PDF

The theory of empty noun in Chinese: with special reference to the right node raising construction

PDF

Processing the dynamicity of events in language

PDF

The grammar of individuation, number and measurement

PDF

Towards a correlational law of language: three factors constraining judgement variation

PDF

A reduplicative analysis of sentence modal adverbs in Spanish

PDF

The converb -j in Alashan Mongolian

PDF

Building adjectival meaning without adjectives

PDF

Syntactic and non-syntactic factors in reflexive pronoun resolution in Mandarin Chinese

PDF

Decomposing Slavic aspect: the role of aspectual morphology in Polish and other Slavic languages

Asset Metadata

Creator Dash, Bhamati (author)

Core Title Silence in answers: a study of ellipsis in Hindi

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Linguistics

Degree Conferral Date 2022-08

Publication Date 07/28/2022

Defense Date 06/06/2022

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

Tag fragment answers,Hindi syntax,OAI-PMH Harvest,polar questions,polarity,syntax,verb phrase ellipsis

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Simpson, Andrew (committee chair), Goodman, Jeremy (committee member), Keine, Stefan (committee member), Li, Audrey (committee member), Pancheva, Roumyana (committee member)

Creator Email bhamati.dash@gmail.com,bhamatid@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC111375558

Unique identifier UC111375558

Legacy Identifier etd-DashBhamat-11036

Document Type Dissertation

Format application/pdf (imt)

Rights Dash, Bhamati

Type texts

Source 20220729-usctheses-batch-963 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu