UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The syntax of WH-expressions as variables in Thai Ruangjaroon, Sugunya Dec 23, 2005

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
831-ubc_2005-105623.pdf [ 6.33MB ]
Metadata
JSON: 831-1.0099823.json
JSON-LD: 831-1.0099823-ld.json
RDF/XML (Pretty): 831-1.0099823-rdf.xml
RDF/JSON: 831-1.0099823-rdf.json
Turtle: 831-1.0099823-turtle.txt
N-Triples: 831-1.0099823-rdf-ntriples.txt
Original Record: 831-1.0099823-source.json
Full Text
831-1.0099823-fulltext.txt
Citation
831-1.0099823.ris

Full Text

THE SYNTAX OF WH-EXPRESSIONS AS VARIABLES IN THAI by SUGUNYA RUANGJAROON B.A., Tharnmasat University, 1992 M.A., Portland State University, 1996 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES Linguistics THE UNIVERSITY OF BRITISH COLUMBIA August 2005 © Sugunya Ruangjaroon, 2005 Abstract This dissertation presents an in-depth examination of Thai wh-expression as variables. I claim that wh-expressions are variables with no inherent interrogative force. As variables, they acquire different interpretations in different contexts. A syntactic relation between the operator and the variable is implemented in terms of the probe-goal relation (Chomsky 2000). The probe-goal relation is established by the operation of Match. In Thai, a goal (as a variable) is "underspecified" for featural content. A feature specified on the probe is copied onto the ^ underspecified goal, thereby satisfying feature matching. In wh-contexts, the probe is identified as a covert interrogative Q[Wh]. The [wh] feature of the probe Q is copied onto the underspecified goal. I argue that the probe-goal relation is established via Match (without Move). The covert Q[wh] probe is base-generated in C. In the context of negation, a goal matches the [neg] feature on the Neg probe, hence functioning as a Negative Polarity Item (NPI). In a yes-no construction, the goal matches the [polarity] feature on the Q[p0iarity] probe, functioning as an Existential Polarity Item (EPI). The probe-goal relation is predictably constrained by the c-command relation and locality conditions. In addition to wh in-situ, it appears there is also partial and long-distance wh-movement in Thai. However, contrary to superficial appearances, I argue that the probe-goal relation in Thai does not involve Move. In particular, I show that apparent cases of movement actually involve two distinct types of cleft constructions. The first I analyze as reduced contrastive wh-clefts with bare wh-expressions. The second I analyze as reduced identificational wh-clefts with D-linked wh-expressions. Semantically and syntactically, contrastive wh-clefts differ from identificational wh-clefts. Contrastive wh-clefts are formed with the overt copula (pen), a definite marker (thv) and a nominalized clause. Identificational wh-clefts, on the other hand, are formed with the overt copula (A^i:) and a relative clause. ii Table of contents Abstract ii Table of Contents iiAbbreviations viAcknowledgements viii Chapter 1 Inroduction 1 1.1 Three construals of the operator-variable relation: wh in-situ, NPI and EPI 1 1.2 The operator-variable relation as a probe-goal relation 2 1.3 Outline of the thesis 7 Chapter 2 Wh-expressions as variables 11 2.1 Deriving the properties of wh-questions in Thai 12.2 Underspecified goals in a wh-construction 3 2.2.1 LF abstract/covert movement analysis 14 2.2.2 A'binding analysis 15 2.2.3 The probe-goal relation 7 2.3 Identifying the probe: Q[Wh] 22.3.1 Comparing Thai, Japanese and Mandarin: matrix Q[wh] 27 2.3.2 Selectional restrictions: embedded Q[Wh). 29 2.4 Matching the Q[wh) probe 32 2.4.1 Deriving the wh-construal 32.4.2 The goal as an underspecified variable 33 2.5 The structural properties of the probe-goal relation 36 2.5.1 The lack of an asymmetry between wh-subjects 37 and wh-objects 2.5.2 Long-distance probe-goal relation 37 2.5.3 No island effects with wh in-situ arguments 39 2.5.4 No island effects with wh in-situ adjuncts 42 2.6 The interpretive properties of the probe-goal relation 48 2.6.1 Multiple wh-questions 42.6.2 Wh-Quantifier interaction 52 2.6.3 Some complications 5 iii Chapter 3 Variables as polarity items 58 3.1 Matching probes for polarity3.2 Subject/non-subject asymmetry with NPI-construal 63 3.2.1 The position of negative may 5 3.2.2 (The absence of) NPI-construal in subject position 67 3.2.3 NPI-construal in object position 70 3.3 Subject/non-subject asymmetry with EPI-construal 1 3.3.1 The position of the yes-no question marker (Qpoiarity) may 74 3.3.2 (The absence of) EPI-construal in subject position 77 3.3.3 EPI-construal in object position 80 3.4 An asymmetry between wh-arguments and wh-adjuncts 1 3.5 Matrix/embedded scope asymmetries 85 3.5.1 Selectional restrictions3.5.2 Embedded NPI and EPI subjects 87 3.5.3 Embedded NPI and EPI adjuncts 9 3.6 Extending the analysis: comparatives, modals, if-clauses 91 3.7 Summary 93 Chapter 4 Wh-expressions at the left edge of the clause: constrastive wh-clefts 95 4.1 Introduction 94.2 The semantics of contrastive clefts with the overt copula pen 99 4.2.1 Contrastive clefts have an existential presupposition 99 4.2.2 Contrastive clefts have a uniqueness presupposition 101 4.2.3 Contrastive clefts have focus 102 4.3 The syntax of contrastive cleft with overt pen 106 4.3.1 hl'on as a subject oriented nominalizer 107 4.4 The semantics of reduced contrastive wh-clefts Ill 4.4.1 Reduced contrastive wh-clefts have an existential 112 presupposition 4.4.2 Reduced contrastive wh-clefts have a uniqueness 113 presupposition 4.4.3 Reduced wh-clefts have a contrastive focus 114 4.5 The syntax of reduced contrastive wh-clefts 116 4.6 Consequences of the contrastive wh-cleft analysis 119 IV 4.6.1 An asymmetry between wh-subjects and wh-objects 119 4.6.2 Wh-object can occur as a cleftee with the presence of passive markers 120 4.6.3 Only [+human] subj ects are allowed 121 Chapter 5 Wh-expressions at the left edge of the clause: identificational wh-clefts 122 5.1 Introduction 125.2 The semantics of identificational clefts with overt copula l^i: 124 5.2.1 Identificational clefts have an existential presupposition 125 5.2.2 Identificational clefts have a uniqueness presupposition 126 5.2.3 Identificational clefts do not have a contrastive focus 127 5.3 The syntax of identificational clefts with overt l^i: 128 5.3.1 Identificational l^i: copula 130 5.3.2 Cleftee is a relative clause 131 5.3.2.1 pro as a head noun in relative clauses 131 5.3.2.2 The morpheme thi: is a complementizer in a..133 relative clause 5.4 The semantics of reduced identificational wh-clefts 135 5.4.1 Reduced identificational wh-clefts have an existential 135 presupposition 5.4.2 Reduced identificational wh-clefts have a uniqueness 137 presupposition 5.4.3 Reduced Identificational wh-clefts lack a contrastive focus... 138 5.5 The syntax of reduced identificational wh-clefts 140 5.5.1 Evidence for non-movement analysis: 143 identificational wh-clefts/relative clause parallel 5.6 Consequences of the identificational wh-cleft analysis 144 5.6.1 The lack of an asymmetry between wh-subjects 144 and wh-objects 5.6.2 Both [+human] and [-human] are allowed 146 in a cleftee position 5.7 Predictions of the identificational wh-cleft analysis 147 5.7.1 Identificational wh-clefts rather than wh-intervention effects. 147 v 5.7.2 Wh-argument locatives as D-linked wh-expressions 152 Chapter 6 Conclusions 156 6.1 Summary of major findings 156.2 A note on wh-adjuncts and Move 158 6.2.1 Wh-focus adjuncts as topicalization? 161 6.2.2 "Wh-focus adjuncts as a cleft? 162 Bibliography 163 vi Abbreviations Abbreviations of grammatical category labels asp aspect marker cl classifier comp complementizer dem demonstrative def definite det determiner neg negation nom nominalizer pass passive prog progressive Other abbreviations distributive marker existential polarity item focus negative polarity item topic wh-operator yes-no operator dist EPI foe NPI top Q[wh] Q[yes-no] Vll Acknowledgements The first person that I would like to thank is Rose-Marie Dechaine, chair of my thesis committee, my supervisor, and my role model. When I first started the linguistics program at UBC, Rose-Marie was the first linguist that I met and the first instructor I. had. Rose-Marie is an excellent researcher, an attentive supervisor, and a caring teacher. There were times when I did not understand my language and was entangled in being too familiar with the problems of my language. Rose-Marie helped me see the problem more clearly and taught me how to ask the right questions, which gradually led me to an understanding of the language. Thai is still an under-analyzed language and my work on this topic is original (as far as I know). Rose-Marie drew my attention to some articles about Thai which have influenced my analysis. Anytime I discovered something new about my language, I would tell Rose-Marie and she would give me a hand-shake as a signal that I am on the right track. In Thailand you normally don't get a chance to shake hands with your teacher. Rose-Marie also guided me with respect to linguistic exposition. She has trained me well these past six years. It is because of Rose-Marie, I am confident that I will contribute a great deal to Thai linguistics and return home well prepared to pass on the skills I have acquired here to others. The thesis would not have turned out the way it did today without the help of Martina Wiltschko. She is a remarkable committee member who is intelligent and hard-working. Her comments were always very insightful and thought-provoking and helped me improve my work. She helped clarify points that were problematic, and she also made interesting observations. She took time for explanation when I understood things during the meetings. She was tireless in her efforts to make sure I understand my work. I cannot imagine myself being trained by someone else but Martina. I would also like to thank Henry Davis for his helpful comments and his insightful suggestions. Henry has been very nice and supportive, despite his hectic schedules and I thank him for that. I would like to extend my sincere gratitude to Hotze Rullmann and Andrew Simpson for constructive comments and insightful observations. Their thoughts helped strengthen the hypothesis of this dissertation and provided sage guidance for future work. I was very fortunate to have Lisa Matthewson as the Chair of my generals paper committee. She showed interest in my language and recognized that I was working under time pressure and helped me speed up my defense for both generals papers. I am very grateful to her for this. I viii also worked closely with Patricia Shaw who has a keen interest in tonal languages. I was very impressed by how deeply and thoroughly she understands tonal phenomena. I also want to thank all the professors in the Department of Linguistics who have played such a vital role in my education: Felicia Lee, Kimary Shahin, Laura Downing, Strang Burton, Hidekazu Tanaka and especially Douglas Pulleyblank. Doug is not only a great teacher but a very good graduate adviser. He has been very helpful, supportive and positive. He makes me feel that every, problem has a solution and that is why he is an excellent graduate adviser. It has been an enjoyable journey at UBC. It was within the Linguistics department that I made a life-long friend Kayono Shiobara. We danced (Salsa and free-style) sang (karaoke and Christmas songs) drank (vodka and wine) laughed (at our wittiness and sense of humor that we both share), talked (about everything), complained (that Linguistics is tough but challenging) and spent Christmas together while attending UBC, a lonely time for international students like me. Kayono made this journey of mine more memorable. Several of my classmates have also become my close friends and I will never forget them. Noriko Yamane-Tanaka (thanks for the great friendship), Masaru Kiyota (a nice friend from the beginning), Atsushi Fujimori (funny and a great friend), Yunhee Chung (best buddy ever), Sun-Young (who always understands me), Rachel Wojdak (super nice, she invited me to her wedding and this made up all the wedding parties that I missed in Thailand), Carrie Gillon, Clare Cook, Jeff Muehlbauer (funny), Leora Bar-el (she shows me how to be a good TA), Scott Shank, Andrea Gormley, Naomi Sawai, Oladiipo Ajiboye, Olga Steriopolo, Gessiane Picanco, Tyler Peterson, Solveiga Armoskaite, and Gene Chu. I could not possibly list all of my fellow graduate students here but I have enjoyed their friendship and I thank them all as well. I was also fortunate to have good friends like Edna Dharmaratne (who has been listening to me and caring for me all these years), Keying Wu (who is a true friend), and Makiko Mimura. I am especially grateful to Clare Cook who helped me proofread the first draft and also provided me with data on Plains Cree and Yoruba. Most importantly, she has been a good friend and a great listener to me all these years. I also owe a lot to good colleaques of mine, Jason Brown who agreed right away to help me proofread in such a very short notice and did it within a day. I thank Rachel Wojdak who also helped me with proof-reading and has also been a very good friend to me all these years. I have also been tremendously encouraged by my best friend Supattra Tongkalaya who has been a data consultant and a mentor to me all these years. I ix would also like to thank Ann Jose, my forever bestfriend, who reminds me how important Education is to me and encourages me to gracefully stick with it no matter what happens in my life. I am very grateful to my mother and sister for their love and their confidence in me. I would like to thank my mother Sasirat Ruangjaroon for taking good care of herself and being healthy all these years. During my study, she, never caused me to worry that I was away from home. I would also like to express my love to my sister Supaporn Ruangjaroon who is the best sister anyone could have. She has been taking good care of our mother for ten years while I have been studying abroad. I also owe a debt of gratitude to my father Paiwong Ruangjaroon who supported me and financially brought me to this point. I am deeply thankful to the Royal Thai Government for awarding me a full scholarship for five years. Thanks to the financial support from my country, I was able to focus fully on my studies and research. And I thank all my supervisory committee members for financially supporting me in the last three months. The last person that I would like to thank is myself for making the choice to study Linguistics. It brings me great joy to see how my native language works. x Chapter 1 Introduction 1.1 Three construals of the operator-variable relation: wh in-situ, NPI and EPI This dissertation argues that Thai wh-expressions are variables with no inherent interrogative force. As variables, they acquire an interpretation by "variable assignment". That is, a variable is assigned an interpretation by "a feature copy" operation. This predicts that variable expressions that appear in wh-contexts will appear in other contexts, and their interpretations are constrained by the syntactic context in which they occur, as in (1). In wh-contexts, variables are assigned a wh-construal by copying the [wh] feature of the operator Q, as in (la). In the context of negation, variables are assigned a Negative Polarity Item (NPI) construal by copying the negative feature of a Neg operator (a kind of Negative Concord), as in (lb). In a yes-no construction, a variable is assigned an Existential Polarity Item (EPI) construal by copying the polarity feature of a yes-no question marker, as in (lc). (1) a. Q[wh] [ variable ] wh-construal b. Neg [ variable ] NPI-construac- Q[yes-n0] [ variable ] EPI-construal Examples given in (2) illustrate the claim that variable expressions that are construed as wh-expressions in wh-contexts also have the status of polarity items in other contexts. In the absence of an overt operator, kl'ray is interpreted as [+wh, +human], equivalent to 'who', as in (2a). In the presence of negative may, the variable expression is interpreted as [+Neg, +human], equivalent to 'anyone' or 'nobody', as in (2b). And in the presence of the yes-no question marker may, the variable expression if ray is interpreted as as [-Neg, +human] equivalent to 'someone', as in (2c). WH-CONSTRUAL (2) a. Nit hen [khray] see VARIABLE. +HUMAN Who did Nit see? 1 NPI-CONSTUAL b. Nit may hen [khray] neg see VARIABLE. +HUMAN *(0 Who did Nit not see? = (») Nit did not see anyone / nobody. EPI-CONSTRUAL c. Nit hen [khray] may see VAR1ABLE.+HUMAN Q[poiarilyj * (i) Who did Nit see? = (ii) Did Nit see someone or did not see nobody? The examples in (2) confirm that Thai wh-expressions are in fact variables in that they get interpreted relative to the syntactic context that hosts them. In this dissertation, I propose that the operator-variable relation in Thai is implemented as a syntactic probe-goal relation (Chomsky 2000). The next section discusses how the probe-goal analysis captures the syntactic restrictions that hold between the operator and the variable. 1.2 The operator-variable relation as a probe-goal relation Wh-questions are often analyzed in terms of an operator-variable structure (e.g. Cheng 1991, Aoun and Li 1993, Tsai 1994, Cole and Hermon 1998). The wh-operator takes scope over the whole sentence and binds a variable, as in (3). (3) [OP, [ variable; •']. ] The grammar provides two ways to derive the operator-variable pair found in wh-questions (Tsai 1994). The in-situ analysis has OP[Q] base-generated and the wh-operator binds a variable, as in (4a). The movement type involves overt wh-movement and the wh-operator binds the variable, as in (4b). 2 WH IN-SITU (4) a. [ 0Pi[Q] [ variable; ] ] WH-MOVEMENT b. [ wh; [ t, ] ] The question that arises is that "why is the probe-goal relation relevant for modeling the operator-variable relation?" As we saw in (2), Thai wh-expressions have the status of polarity items in some contexts. This indicates that wh-expressions are variables: to be interpreted, they must be syntactically "bound" and "coindexed with a c-commanding antecedent (Grodzinsky and Reinhart 1993). Along the same lines, it has been proposed in the literature that the operator-variable relation involves a licensor-licensee relations such that wh-expressions are treated as polarity items that require a licensor for interpretation (e.g. Huang 1982, Nishigauchi 1990, Cheng 1991, Li 1992a, Lin 1996 and Sigrid & Kim 1997). Licensors for polarity items are usually formed by the same set of licensors—be it negation, an existential quantifier or a universal quantifier. While such licensor-licensee analyses account for languages where wh-expressions are polarity items, they do not account for languages where wh-expressions are not polarity items. In this study, the operator-variable relation will be implemented as a probe-goal relation. The operator (as the probe) is related to the goal (as the variable) by the operation Match. The proposed analysis provides a unified analysis for both polarity item and non-polarity item languages (See section 2.2.3 for discussion). I argue that a "probe" is identified as Q[wh], Neg or Q[yes-no] and a "goal" is an underspecified variable. In particular, the probe and the goal interact via "feature matching". Notice that the operator-variable relation, as the probe-goal relation, is reversed from the usual kind of probe-goal dependency (i.e. Agreement), where the (j) features on a verb match features of the goal DP. For "agreement", Chomsky (2000) argues that a <j) feature on a verb (as a probe) is seeking for the closest matching goal (the DP), namely "matching features that establish agreement" (Chomsky 2000: 122). Under the present anlysis, the goal is seeking for a matching probe. This "reverse" dependency seems to me to be specific to the operator-variable relation that is relevant for wh-questions. 3 Match is defined by Chomsky (2000: 122) as in (5). (5) Matching is feature identity The first question that arises is how "feature identity" is satisfied with respect to Match. In as much as identity, requires the presence of the same feature, Match is always satisfied if the Probe and the Goal have exactly the same feature specification, as in (6a). But there are at least three other logical possibilities to consider, namely those in (6b-d). POSSIBLE MATCHING RELATIONS (6) Probe Goal a. Q [whj [wh] b. Q [wh, F] [wh] c. Q [wh] [wh, F] d. Q [wh] [u, F] As already mentioned, (6a) satisfies Match because the probe and goal have an identical feature, namely [wh]. The Matching relation in (6b) and (6c), on the other hand, can be established through a superset or a subset relation. That is, either the probe or the goal has an additional feature besides the [wh] feature. If such feature specification satisfies Match, then this implies that the relevant notion of "feature identity" requires that the Probe and Goal share at least one feature. If Match requires that the entire feature specification of the Probe and Goal be identical, then (6b-c) would not satisfy Match. As we shall see in Chapter 2, (6b-c) do in fact satisfy Match. Finally, there is the question of the status of (6d), where the Goal is unspecified for the feature of the Probe. At first glance, (6d) does not satisfy Match, contradicting the definition given in (5) that Match is feature identity. The analysis of Thai wh-expressions that I propose claims that Match can be satisfied in (6d) via feature "copying". In particular, I argue that, in Thai, the operator-variable relation, as a probe-goal relation, satisfies Match through feature copying. Thai wh-expressions are variables; as such they are "underspecified" goals whose featural [u] content needs to be filled in. The underspecified goal in (7a) is filled in by the [wh] feature on the Q probe which is copied onto the underspecified goal. When, the "underspecified" goal is left unfilled, as illustrated in (7b), the sentence is ill-formed. (7b) cannot be interpreted as a command because there is no 4 available probe in an imperative sentence. The goal remains thus uninterpretable due to the lack of an appropriate probe. WH-CONSTRUAL (7) a. khun kin [?aray] you eat VARIABLE.-HUMAN What did you eat? IMPERATIVE b. * kin [?aray] eat VARIABLE. -HUMAN [Eat what!] The underspecified goal is constrained by the domain in which it occurs. The syntactic domain determines which features are copied onto the underspecified goal. The feature copy operation is restricted to the following features: [wh], [neg] and [polarity]. Copying wh-feature, therefore, yields a wh-construal, while copying a negative feature yields a negative construal. Along the same line, copying a polarity feature yields a positive construal. The feature copy analysis captures the fact that Thai variable expressions are invariant forms, regardless of their different interpretations. MATCHING RELATIONS IN THAI (8) Probe Goal [wh] [p, F] [neg] [p, F] [polarity] [p, F] Domain a wh-question a negative clause a yes-no question This analysis predicts that the goal that is "underspecified" for a feature will have a fixed interpretation—be it wh-construal or polarity construal—depending on the feature of the probe copied onto the goal. This is illustrated in (9). We see that in addition to the wh-construal, the goal may have an NPI construal. This is due to a [Neg] feature of the probe that is copied onto the goal, creating feature identity for Match. The goal may also have an EPI construal. The polarity feature on the Q[yes-no] probe is the one that is copied and filled in for the underspecified goal. Hence, Match is satisfied. 5 PROBE GOAL [WH] [NEG] [POLARITY] kt'ray 'who' 'anyone' 'someone' [u, +human] ?aray 'what' 'anything' 'someone' [n,-human] thi:ndy 'where' 'anywhere' 'somewhere' [n, +place] rriiiaray 'when' 'anytime' 'sometime' [u, +time] ya:rjray 'how' 'anyhow' 'somehow' [u, +way] tham-may 'why' 'any reason' 'some reason' [u, +reason] I argue that in Thai, the probe-goal relation is only established via Match (without Move). There are, however, some languages where the probe-goal relation is established via Move, as illustrated in the table in (10). (10) Feature Match Move Move Specifications Probe [±wh] Probe [±wh] Probe [+wh] Goal [+wh] Old Chinese * * Goal [+wh, F] * Yoruba Goal [u, F] Thai English, French * Given that Move is a by-product of Agree, the question that arises is 'why do some languages need Agree?' Agree is taken to be an operation that deletes uninterpretable features that render the probe and goal active in order for Agree to apply (Chomsky 2000: 123). If Agree is feature deletion, we need Agree to delete an uninterpretable feature prior to LF to avoid a crash of the derivation (by definition). Take English as an example. The unintepretable [wh] feature is on the C head and is copied onto the underspecified goal. (See a detailed discussion of "the underspecified goal" in English in 2.2.3). After feature copying, the probe and goal match in [wh] features. Agree then triggers overt wh-movement to satisfy the EPP feature of C in case of A' movement. The goal will have to move to the probe, forming a specifier of the probe. The uninterpretable feature on the probe and the goal needs to be deleted before LF via Agree. 6 This is where the system is different between Thai and English, in that Thai only needs Match, while in English Match and Move are both required. I delay the detailed discussion of how the probe-goal analysis accounts for typological differences in other languages until chapter 2. I have set out some core assumptions of this work regarding how probe-goal relations are established in (11). The probe-goal relation is constrained by the following conditions in that the probe and the goal must satisfy feature identity (Match), c-command and locality. (11) (i) Feature identity: The probe-goal satisfies Match, which requires probe and goal to have identical features. (ii) C-command condition: The probe must c-commands the goal. (iii) Locality: Match is satisfied by the most local probe. In this study, I implement the operator-variable relation as a probe-goal relation. Wh-expressions are treated as a goal underspecified for a feature. This follows from the claim that the goal is a variable that needs to be licensed. Within the probe-goal dependency, the "underspecified" goal needs to be filled by featural content. This is done through "feature copying". Then, the probe and the goal enter into a Matching relation. In order to Match, the goal must be in the domain of the probe and must satisfy locality conditions. Discussion of the probe-goal relation is spread through the following four chapters. Each chapter considers the probe-goal relation as it occurs in different domains. 1.3 Outline of the thesis This thesis is organized into a first and second chapter that set out the general position that will be adopted and expanded into the three other chapters. I propose that the syntactic relation between the operator and the variable is implemented in terms of the probe-goal relation. The probe-goal relation requires the underspecified goal enter a matching relation (Chomsky, 2000:122). I claim that the feature-matching requirement in Thai is met by the "feature copying" operation. The feature of the probe is copied onto the goal that is underspecified for a feature. After feature copying, the probe and the goal are identical in feature identity, as required by Match. Chapter 2 discusses the probe-goal relation in wh-contexts such as (12). 7 WH-CONSTRUAL (12) Nit hen [khray] see VARIABLE. +HUMAN lit = Nit saw who = Who did Nit see? I argue that the probe is a covert Q[Wi,], and the goal, on the other hand, is a variable expression. The [wh] feature of the probe is copied onto the underspecified goal. The goal matches with the probe in [wh] features, and hence is interpreted as an interrogative. I claim that the Matching relation between the probe and the goal in Thai is established without Move. Much of chapter 2 is spent discussing the two major consequences of the claim that the covert Q[wh] probe is base-generated in C from where it takes scope and matches with the goal. First, the proposed analysis correctly predicts the structural properties of Thai wh in-situ. More specifically, it derives the absence of an asymmetry between wh-subjects and wh-objects, as well as the absence of an asymmetry between wh in-situ arguments and wh-adjuncts with respect to island effects. Second, the proposed analysis also correctly predicts the interpretive properties of Thai wh in-situ. In particular, it derives the absence of pair-list readings in multiple wh-questions, as well as the absence of list-readings in wh-constructions that contain a quantifier. Chapter 3 discusses variable expressions in polarity contexts such as (13). NPI-CONSTRUAL (13) a. Nit may hen [khray] neg see VARIABLE.+HUMAN Nit did not see anyone. EPI-CONSTRUAL b. Nit hen [khray] may see VARIABLE. +HUMAN Q[poiarity] Did Nit see someone? The probe-goal analysis extends to NPI and EPI contexts. In the context of negation, the [neg] feature on the Neg probe is copied onto the underspecified goal. The goal matches with the probe in [neg] features, and is interpreted as a Negative Polarity Item (NPI). In a yes-no 8 construction, the goal matches the [polarity] feature on the Q[yes-no] probe, interpreted as an Existential Polarity Item (EPI). In chapter 3, I argue that the interpretation assigned to a variable expression is predictably constrained by the c-command relation that holds between the probe and the goal. The proposed analysis correctly predicts the presence of subject/object asymmetries, and captures the fact that complement/adjunct asymmetries with respect to NPI-and EPI-construals only hold in matrix clauses. Thus, while NPI- and EPI-construals are unavailable with adjuncts in matrix clauses, they are available in embedded clauses. This is because a matrix negative probe or a matrix yes-no question probe is available for the embedded goal. The availability of NPI- and EPI-construals in embedded clauses are a side-effect of the locality condition that requires that the closest c-commanding probe is the one that enters into the probe-goal relation. Chapter 4 is concerned with an apparent case of wh-movement, as examplified in (14a), where a wh-expression is found in sentence-initial position. I argue that the Matching relation between the probe and the goal does not induce Move, despite surface appearances. (14a) is analyzed as a reduced contrastive wh-cleft with the same structure as the contrastive wh-cleft in (14b). The latter are called pen clefts because they have the overt copula pen. (14) REDUCED CONTRASTIVE WH-CLEFT a. [ khray thi: ] tham ca.ii ta?:k VARIABLE. +HUMAN def cause plate brake Who broke a plate? CONTRASTIVE WH-CLEFT b. [khray thi:] pen khon tham cam ta>:k VARIABLE.+HUMAN def be nom cause plate break Who was the one that broke a plate? I show that the semantics and syntax of reduced contrastive wh-clefts have the same properties as contrastive wh-clefts with the overt copula pen. I close the chapter by considering the three major consequences of analyzing such apparent cases of wh-movement as reduced wh-clefts. First, only wh-subjects can occur as a cleftee (due to the inability of the nominalizer k^on to nominalize the object). Second, wh-objects can occur as a cleftee only when passive markers are present (the object can only nominalize if it is promoted to subject by passivization). Last, 9 only 'who' can occur as a cleftee (due to [+human] specification of the nominalizer khon). These restrictions fall out naturally from the contrastive wh-cleft analysis. Chapter 5 looks at another case of apparent wh-movement, which involves discourse-linked wh-expressions such as 'which man', as in (15a). Such constructions are analyzed as reduced identificational wh-clefts with the same structure as identificational wh-clefts such as (15b). The latter are called kf'i: clefts because they have the overt copula T^i:. (15) REDUCED IDENTIFICATIONAL WH-CLEFT a. [phu:cha.y khon nay thi:] Nit ?a>p ch3p man cl VARIABLE comp hide like Which man does Nit secretly have a crush on? IDENTIFICATIONAL WH-CLEFT b. [phu:cha.y khon nay] kht: khon thi: Nit ?a>.p ch3.p man cl VARIABLE be cl comp hide like Which man is the one that Nit secretly has a crush on? I show that the semantics and syntax of reduced identificational clefts have the same properties as identificational wh-clefts with the overt copula ht'i:. This leads to the conclusion that the thi: clauses with D-linked wh-expressions such as (15a) are derived from identificational wh-clefts such as (15b). The chapter ends by considering two predictions that fall out from the identificational wh-cleft analysis. Thai wh-intervention effects are explained by the reduced identificational wh-cleft analysis. To satisfy the probe-goal relation that the goal must be at the left edge of the clause in order to match with the [wh] of the Q probe, rather than with a [Neg] feature of an intervening probe. Another consequence of analyzing thi: clauses with D-linked wh-expressions as reduced wh-clefts is that it reveals the internal structure of wh-argument locatives. It suggests that they may be inherently D-linked wh-expressions such as 'which place', in contrast to the bare wh-expressions such as 'where'. The last chapter concludes, tying together the results of the whole thesis. The chapter has a final note on wh-adjuncts and how they interact with Move. 10 Chapter 2 Wh-expressions as variables 2.1 Deriving the properties of wh-questions in Thai In Thai, when wh-questions are formed the wh-expression—be it subject (henceforth wh-subject), object (wh-object), indirect object (wh-indirect object) or possessor (wh-possessor) — occurs in the same position as the corresponding non-wh-expression in declarative sentences. This is illustrated in (1) for wh-subjects, in (2) for wh-objects, in (3) for wh-indirect objects, and in (4) for wh-possessors. WH-SUBJECT (1) Q: [k ray] si: nans?: m?:awa.Tini; VARIABLE.+HUMAN buy book yesterday Who bought a book yesterday? SUBJECT A: [Nit] si: nans?: m?:awa.Tini: buy book yesterday Nit bought a book yesterday. WH-OBJECT (2) Q: Nit s?: [2aray] m?:awa.-nni: buy VARIABLE. -HUMAN yesterday lit = Nit bought what yesterday? What did Nit buy yesterday? OBJECT A: Nit si: [nans?:] m?;awa.Ttni: buy book yesterday Nit bought a book yesterday. 11 WH-INDIRECT OBJECT (3) Q: Nit hay narjst: [khray] give book VARIABLE. +HUMAN lit = Nit gave a book to whom yesterday? To whom did Nit give a book yesterday? INDIRECT OBJECT A: Nit hay nans?: [Lek] m?:awa.Tini: give book yesterday Nit gave a book to Lek yesterday. WH-POSSESSOR (4) Q: Nit ?a.n nans?: kh5:n [khray] read book of VARIABLE. +HUMAN lit = Nit read a book of who yesterday? Whose book did Nit read yesterday? POSSESSOR A: Nit ?a.n nans?: kh5.Tj [Lek] m?:awa:nni: read book of yesterday Nit read Lek's book yesterday. The examples in (1) through (4) establish that Thai wh-expressions occur in their base-generated position, i.e. Thai is a wh in-situ language. This chapter discusses wh in-situ licensing. I propose that in Thai wh in-situ constructions, wh-expressions are variables with no inherent interrogative force, i.e. they are underspecified for the wh-feature. As variables, they acquire a wh-construal by virtue of being in the scope of an interrogative operator. In particular, I propose that the syntactic relation between the operator and the variable is implemented in terms of the probe-goal relation (Chomsky 2000). The probe-goal relation requires the goal to match with the probe, where Match is defined as feature identity. I argue that the probe is a covert Q[Wh] morpheme specified with a [wh] feature, and that this covert Q[wh] is base-generated in C from where it takes wh-scope and is matched with the goal. m?:awa.-nni: yesterday mi:awa.Tini: yesterday 12 For Thai, analyzing the relation between the covert Q[wh (in C) and the in-situ variable as an instance of the probe-goal relation has two major consequences. First, the proposed analysis derives the structural properties of Thai wh in-situ. In particular, it derives the absence of an asymmetry between wh-subjects and wh-objects, as well as the absence of an asymmetry between wh in-situ arguments and wh-adjuncts with respect to island effects. Second, the proposed analysis derives the interpretive properties of Thai wh in-situ. In particular, it derives the absence of pair-list readings in multiple wh-questions, as well as the absence of list-readings in wh-constructions that contain a quantifier. The chapter is organized as follows. Section 2.2 discusses how goals, as variables, are matched in a wh-construction, and considers three alternative analyses: covert feature-movement, A'-binding, and feature-matching forced by the probe-goal relation. I argue that implementing an operator-variable relation as a probe-goal relation yields the best results. Section 2.3 presents evidence for the presence of a covert Q[wh] probe; it is this probe that provides the wh-feature that is copied onto the underspecified goal. Section 2.4 presents evidence that wh-expressions in Thai are underspecified variables whose construal is constrained by the syntactic context in which they occur. In section 2.5 I go on to argue that the structural constraints of the probe-goal relation account for the lack of asymmetry between wh-subjects and wh-objects with respect to how variables are construed in a wh-context. The proposed analysis also correctly predicts that there will be no differences between wh in-situ arguments and wh in-situ adjuncts with respect to island effects. In section 2.6, the chapter closes with a discussion of the interpretive properties of Thai wh in-situ constructions, as they relate to multiple wh-questions and to wh-questions that contain quantifiers (henceforth wh-quantifier interaction). 2.2 Underspecified goals in a wh-construction I begin by considering how wh-expressions, as underspecified goals, are matched with the covert Q[W|,] probe in a wh-construction. Consider again the following example of an in-situ wh-object: 13 (5) Nit si: [?aray] mluwajini; buy VA RIA BLE. -HUMA N yesterday lit = Nit bought what yesterday? What did Nit buy yesterday? One question that arises is the extent to which a wh in-situ construction has the same properties as a wh-movement construction. In the syntactic literature, two approaches have been pursued: (i) the abstract movement analysis, also called covert movement (e.g. Huang 1982); (ii) the A' binding analysis (e.g. Aoun 1985, Aoun&Li 1993b, Chang 1995). I consider each in turn. 2.2.1 LF abstract/covert movement analysis It has been claimed that wh-expressions that occur in their base-generated position undergo LF movement (e.g. Huang 1982, Tsai 1994). This is illustrated in (6). (6a) is a structure where there is no movement of the wh-expressions in the surface form, while in (6b) the wh-expression undergoes covert movement from the in-situ position to the clause periphery. (6) LF abstract/covert movement analysis a. S-structure [rp Subject Verb [rjp WH ] ] b. LF [CP [DP WHj ] [jp Subject Verb tj ] ] As established above, Thai is a wh in-situ language. At first glance, covert movement appears to be a possible analysis. The argument for this kind of analysis is primarily based on the fact that wh-expressions cannot be contained within a syntactic island. Assuming that islands diagnose a movement relation (Ross 1967), covert movement must also be constrained from moving out of the islands, in the same way as the overt movement is. (7) is an example of a wh in-situ language, Sinhala. Hagstrom (1998) takes the data in (7) as evidence for a covert movement analysis for this language due to its sensitivity to syntactic island constraints. The examples in (7) illustrate that wh-expressions cannot occur inside a complex noun phrase island, as in (7a), and an adjunct island, as in (7b). According to Hagstrom, the Q morpheme dd covertly moves across the islands, hence yielding ungrammaticality. 14 SINHALA (7) a. * oyaa [kay da liyapu pota] kieuwe you who Q wrote book read-E [You read the book that who wrote?] (Kishimoto 1992:56) [kau da ena kota] Ranjit paadam karaminhihe? who Q came time Ranjit study doing was-E [Ranjit was studying when who came?] (Kishimoto 1992:58) However, in Thai, wh-expressions—both wh-arguments and wh-adjuncts—are allowed to occur inside islands and show no island effects (cf. section 2.5). If there were a covert movement in Thai, we would predict island effects, which in fact do not occur, as shown in (8). We see that the examples in (8) are perfectly well-formed. Wh-expressions can occur inside a relative clause island, as in (8a), and an adjunct island, as in (8b), with no island violation I take this as evidence for not adopting a covert movement analysis for Thai. (8) a. k un c'a.p ple.-rj t'i: [k ray] ra.-rj you like song comp VARIABLE, -hhuman sing Who did you like the song such that x sang? khaw t"u:k lay?3;k phra? (khaw;) khamo.y he pass fire becausefhe stea What was he fired because he stole? [?aray] VARIABLE, -human 2.2.2 A' binding analysis In an A' binding analysis, the relation between a wh-operator and a wh-expression is treated as an antecedent-anaphor/pronoun relation (Aoun 1985, Chang 1995, Sloan 1991). It has been observed that wh-expressions behave like anaphors (Aoun 1985, Chang 1995) or pronouns (Sloan 1991). They are subject to binding principles in the same way that anaphors and pronouns are. They, however, have a different binding domain in that they are A' bound by an A' binder—a covert operator OP[wh] in specifier of CP—for interpretation, rather than A bound by an A-antecedent. This is illustrated in (9). 15 (9) A' binding analysis [CPOP,wh|i [rp Subject Verb [DP WH ]j ] In (9), the wh-expression is bound by a wh-operator; this binding relation crucially does not involve movement. As such, it is not subject to constraints on movement, such as subjacency. This analysis predicts the absence of island effects since wh-expressions can be bound by an A' binder OP[Wh] generated in C position even when they occur inside an island. As we shall see below, Thai wh in-situ does not show island effects, so at first glance an A' binding analysis seems promising. The A' binding analysis, however, cannot account for the fact that wh-expressions in Thai can have other interpretations. In addition to having a wh-construal, as in (10a), (10b) and (10c) illustrate that variable expressions can also be construed as negative polarity items (NPIs) in the context of negation, and as existential polarity items (EPIs) in a yes-no construction. WH-CONSTRUAL (10) a. Nit hen [khray] see VARIABLE.+HUMAN Who did Nit see? NPI-CONSTRUAL b. Nit may hen neg see Nit did not see anyone EPI-CONSTRUAL c. Nit hen [khray] may see VARIABLE. +HUMAN Q[polarity] Did Nit see someone? Thus, in-situ variable expressions are not inherently interrogative. Rather, they are variables that acquire their interrogative, negative and existential force by being in the scope of the relevant operators. As underspecified goals, they automatically acquire the feature of the most [khray] VARIABLE.+HUMAN 16 local c-commanding operators. In this way, the probe-goal analysis need not posit a dedicated operator position in Spec CP. Moreover, since the non-wh-operators—Neg and Q[yes-no] are generated lower in the tree (see Chapter 3 for discussion of the position of Neg and Q[yes-no])> they are not be appropriate binders in the A' binding analysis which locates the operator in Spec CP. Also, we will see in Chapter 2, section 2.5.4 that wh-rationale adjuncts in Thai undergo overt movement, unlike other wh-expressions. A concern that arises given such a unique characterization of wh-rationale adjuncts is how the A' binding analysis will capture the movement of the wh-adjuncts? To conclude, the A' binding analysis does not account for the properties of Thai wh in-situ. It cannot explain wh-expressions' status as polarity items. 2.2.3 The probe-goal relation In this section, I introduce the probe-goal relation (Chomsky 2000) and motivate why I adopt this probe-goal relation to analyze variable expressions in Thai. Note that all three approaches —the covert movement analysis, the A' binding analysis, and the probe-goal anlysis—treat the relation between the wh-feature and the wh-expression as an operator-variable relation. By hypothesis, Thai wh-expressions are variables whose interpretations are constrained by the syntactic domain in which they occur. I treat the operator-variable relation as a probe-goal relation. I propose that probes (as operators) and goals (as variables) are mediated through the Matching relation. Match is defined as feature identity between a probe and a goal (Chomsky 2000). I consider the logical possibilities of how Match is satisfied by the probe-goal relation, and argue that Match in Thai is satisfied by feature "copying". There are (at least) four possible ways that the probe and the goal can enter into a Matching relation. First, the probe and the goal are featurally identical, as in (11a). Second, the probe and the goal are matched through a superset relation, as in (1 lb). That is, the probe has two features: Fl and F2 but the goal has one feature [Fl]. Only Fl of the probe matches with the Fl of the goal. The probe and the goal in (11c), on the other hand, are matched through a subset relation. The probe has one feature [Fl], while the goal has two features [Fl] and [F2]. It is [Fl] of the probe and the goal that matches. If feature specifications such as (llb-c) satisfy Match, then this implies that the 17 relevant notion of "feature identity" requires that the Probe and Goal share at least one feature. The last case in (lid) is of particular interest. Here, the probe has one feature [Fl], while the goal has two features: one underspecified for a feature, and the other [F2]. POSSIBLE MATCHING RELATIONS (11) Probe Goal a. [Fl] [Fl] b. [Fl, F2] [Fl] c. [Fl] [Fl, F2] d. [Fl] [p, F2] 18 I apply the possibilities established in (11) to actual wh-cases, as listed in (12). (12) Feature specifications: Probe and Goal Matching Relation Probe Goal Predictions Language a. probe = goal Fl: [WH] Fl: [WH] A language with a single general purpose wh-expression Old Chinese b. probe is a superset of goal Fl: [WH] F2 : ?? Fl: [WH] probe: Fl restricted to wh-contexts F2: ?? goal: Fl restricted to wh-contexts ?? c. probe is a subset of goal Fl: [WH] Fl: [WH] F2: [+human] [-human] probe: Fl restricted to wh-contexts goal: Fl restricted to wh-contexts F2: restricted to semantic features that are composed of wh-expressions (e.g. +human, +entity, +location, +time etc.) Yoruba d. goal is underspecified for feature Fl: [WH] Fl: [u] F2: [+human] [-human] probe: Fl restricted to wh-contexts goal: Fl is not restricted to wh-contexts F2: restricted to semantic features that are composed of wh-expressions (e.g. +human, +entity, +location, +time etc.) Thai English French Let us walk through (12) step by step. The first case in (12a) illustrates that matching is feature identity. The probe and the goal are featurally identical. This type of matching relation predicts a language with an invariant morpheme which generalizes for all wh-expressions. Old Chinese appears to be such a language that uses a single morpheme he for all general purpose wh-19 expressions'. According to Wu (2005), he can be interpreted as 'what' (13a), 'where' (13b), 'when' (13c), 'how' (13d) and 'why' (13e). The interpretations are determined by syntactic positions and contexts in which it occurs. OLD CHINESE2 (13) a. yu [he] yan I what say What can I say? (Shangshu, Gao Yao Mo) b. Zi yu [he] wang you want where go Where do you want to go? (Zhanguoce, Qince) c. fang [he] wei qi particle when to be time When will be the time? (Shijing, Qinfeng, Xiaorong) d. ru [he] sheng zai shang you how live at up How will you be able to live upon the earth? (Shangshu, Pan Geng) e. wo du [he] hai I along why harm Why am I along harmed? (Shijing, Xiaoya, Lu'e) he3 occurs in a preverbal object position in (13a), and it has the status of 'what'. But when it occurs as an object of directional verbs, it functions as 'where'. For 'when', 'how' and 'why', he occurs in a preverbal adverbial position. We see that Old Chinese illustrates Match as feature identity between the probe and the goal. 1 See Keying Wu's dissertation (2005) titled "Interrogatives in Old Chinese" (in preparation). 2 The data in (13) is from different periods of time. 3 he, however, cannot be interpreted as 'who'. 20 A second case to consider is where Match takes place when the probe is a superset of the goal, as in (12b). The probe has two features: [Fl: WH] and [F2]. The second feature [F2] can be anything. The goal, on the other hand, has only one feature [WH] which is restricted to wh-contexts. We predict that such a language may use different question morphemes for different clauses (e.g. matrix or embedded clauses). To my knowledge, no such language is attested. The third case in (12c) is that the probe is a subset of the goal. The probe carries only one feature which is [Fl: WH], while the goal is given two features: [Fl: WH] and [F2]. The [F2] feature can be a feature such as [+animacy] or [+human], etc. Languages that illustrate this kind of Matching relation will have wh-expressions functioning only with interrogative force but have variant morphemes for different wh-questions, unlike Old Chinese. Yoruba is an example of such a language. The data (14a-b) is taken from Jones (2004) and (14c-f) from Cook (2004). YORuBa (14) a. [ta] ni 6 ra iwe [+wh, +human] who foe4 3sg buy book Who bought a book? b. [ki] ni Ade ra [+wh, +entity] what foe buy What did Ade buy? c. [nibo] ni 6 ti joro [+wh, +location] where foe 3sg em eat.mango Where does she eat mangoes? d. [igba] wo ni 6 maa time when foe 3sg hab When does she eat mangoes? n joro prog eat.mango [+wh, +time] Yoruba wh-words are accompanied by the presence of a focus marker ni (Dechaine 2001). 21 e. [bawo] ni Ade se how foe em How did Ade sell drums? talu sell.drum [+wh, +way] f. [nitori] ki ni Ade se sere [+wh, +reason] reason wh foe em play Why did/does Ade play? The last case to consider is (12d) where the probe is restricted to wh-contexts since it only carries the [Fl: WH] feature. On the other hand, the goal is not restricted to wh-contexts since Fl of the goal is underspecified for a feature in the sense that [u] needs to be filled by feature content. The underspecified goal is seeking for a probe that is specified for a feature. In this case, the probe has a [Fl: WH] feature. The [wh] feature is then copied from the probe onto the underspecified goal. Now, the goal has a featural content. Not only does it have a feature, but its feature matches with [Fl: WH] of the probe, creating feature identity for Match. Since the goal is underspecified for a feature, it is not restricted to a [wh] feature, or to any feature. This predicts that the goal can have interpretations other than wh-construal, depending on the feature of the probe that is copied onto the underspecified goal. I argue that Thai is such a language, i.e. the interpretation of the goal is constrained by the feature of the probe. The following examples illustrate how the interpretation of wh-expressions such as /fray [+human] can be interpreted as 'who', as in (15a), as 'anyone', as in (15b) or as 'someone', as in (15c). (15) THAI WH-CONSTRUAL a. Nit hen [khray] see VARIABLE. +HUMAN Who did Nit see? NPI-CONSTRUAL b. Nit may hen [khray] neg see VARIABLE. -/-HUMAN Nit did not see anyone. 22 EPI-CONSTRUAL c. Nit hen [khray] may see VARIABLE. +HUMAN Q[poiarity] Did Nit see someone? The above examples establish that variable expressions as goals are featurally underspecified in Thai. They enter the Matching relation by copying a feature specified on the probe. The underspecified nature of Thai variable expressions is illustrated, in (16), where we see that the full range of argument, locative, temporal, manner and reason expressions may have a wh-construal, an NPI-construal, or an EPI-construal. PROBE GOAL [WH] [NEG] [POLARITY] lixray 'who' 'anyone' 'someone' [u, +human] laray 'what' 'anything' 'someone' [u,-human] thi:nay 'where' 'anywhere' 'somewhere' [u, +place] miiaray 'when' 'anytime' 'sometime' +time] yairjray 'how' 'anyhow' 'somehow' [H, +way] tham-may 'why' 'any reason' 'some reason' [n, +reason] This type of matching relation, whereby the goal is underspecified for a feature, is not specific to Thai. If we consider English and French, particularly in relative clauses and free relative constructions, we see that wh-expressions do not always have a wh-construal. That is, the interpretation of wh-expressions in English and French is also contextually determined. This is illustrated in (17) for English. (17) ENGLISH a. Who did John see? [+wh, +human] b. The man who John saw [-wh, +human] c. Whoever John saw [-wh, +human] 23 In (17a), the wh-expression has a wh-construal in a wh-context. The wh-expression as a goal is underspecified for a feature, while the probe Q has a [+wh] feature. The goal is seeking for a probe and match in a [+wh] feature. Hence, the wh-expression in (17a) is construed as a wh-interrogative. In (17b), since the goal is underspecified for a feature, it is also looking for a probe that is specified for a feature to match with, in which case, the probe relative operator has a [-wh] feature. The goal matches with the probe specified with the [-wh] feature, and hence is not interpreted as an interrogative. While in (17c), the probe has a [-wh] feature, the goal copies [-wh] feature of the probe. They featurally match. Hence, the goal is construed as a free relative. This establishes that wh-expressions in English are also underspecified goals whose construal is constrained by the probe whose context they occur in. French shows a similar pattern to English and Thai in that wh-expressions are not always interrogatives. In (18a) and (18b), the wh-expressions are underspecified goals. They are looking for a probe specified with a feature to be filled in by that feature. Then, they copy the [+wh] feature of the Q probe in a wh-context. In the context of a relative clause, the wh-expressions are not interrogatives (18c). This is because the probe has a [-wh] feature and they match with it. In a free relative construction (18d), the goal is interpreted as a free relative reading. This is due to the presence of a [-wh] feature on the probe that it matches with. FRENCH (18) a. Jean a vu [qui] has seen who Who did Jean see? [+wh, +human] [Qui] est-ce que who is-it that Who did Jean see? Jean a has vu seen [+wh, +human] c. L'homme [qui] a vu the man who has seen The man who saw Jean Jean [-wh, +human] 24 d. [QuiconqueJ a vu whoever has Whoever has seen Jean Jean seen [-wh, +human] So far, I have claimed that the probe and the goal enter a Matching relation through feature identity, more specifically through feature copying in Thai. The goal in Thai is underspecified for a feature, in which case, probes are not restricted to [+wh] feature contexts. As we have seen, the feature of the probe determines the interpretation of the goal. The feature of the probe is copied onto the underspecified goal, be it [+wh] or [-wh]. The above examples show that this property is not specific to Thai, but is also attested in English and French. In those languages, wh-expressions have interrogative force (in wh-contexts) or non-interrogative force (in relative clause contexts) depending on the feature of the probe that they match with. A difference between Thai and English/French is that the probe-goal relation in Thai stops at Match, while in English the probe-goal relation also satisfies Agree and Move. Agree is taken to be an operation that deletes uninterpretable features that render the probe and goal active in order for Agree to apply (Chomsky 2000: 123). It is Agree that leads to a deletion of an uninterpretable feature prior to LF. After the uninterpretatble feature is deleted, Agree can be (but need not be) accompanied by Move. Move is triggered by the EPP feature associated with the probe. The goal then will move to the probe, forming a specifier of the probe. The question at this point is 'how does the analysis proposed here apply to other languages cross-linguistically?' Recall that in Old Chinese, as in (13), the wh-exprssion he can only have an interrogative reading. Hence, the probe-goal relation in Old Chinese enters a Matching relation by both probe and goal being specified for a [wh] feature. Hence, the probe and goal featurally match. Similar to Old Chinese, wh-expressions in Yoruba are always interrogative. Wh-expressions in Yoruba occur in initial position, as illustrated in (14). Under the proposed analysis, the probe has a [wh] feature, while the goal carries two features: [wh] and another feature such as [+human], [-human], [+location], [+time], [+way] or [+reason]. The probe and goal featurally match in [wh] features. Then, the EPP property of the probe in C triggers overt movement of the goal to Spec CP. 25 In this dissertation, I argue that in Thai, the probe-goal relation is established via Match (without Move). (19) illustrates how other languages fit into the proposed analysis. (19) Feature Specifications Match Probe [±wh] Move Probe [±wh] Move Probe [+wh] Goal [+wh] Old Chinese * * Goal [+wh, F] * * Yoruba Goal [p, F] Thai English, French * I argue that the probe-goal relation is best represented as an operator-variable relation in Thai. The goal is featurally underspecified and not restricted to [wh] contexts and the underspecified goal must be filled in by the featural content can match with any feature of the probe through feature copying. This captures the fact that Thai wh-expressions are variables with no inherent interrogative force. They acquire different interpretations (wh-, NPI- and EPI-construals) by matching with the features on the probe, namely [Q: wh], [Neg] and [Q: polarity]. The Probe-goal relation is represented schematically below. (20) Probe-goal analysis WH-CONSTRUAL a. [CP probe : Q [wh] [jp Subject Verb [ goal: p ] ] ] a' [CP probe : Q [wh] [pp Subject Verb [ goal: [wh] ] ] ] NEG-CONSTRUAL ' b. [CP Q[wh] [jp Subject [ probe: NEG [ Verb [goal: p ] ] ] ] ] b' [CP Q[wh] [pp Subject [ probe: NEG [ Verb [goal: NEG ] ] ] ] ] EPI-CONSTRUAL c. [CP Q[wh] [rp Subject [probe: Q|poiarityi [ Verb [goal: p] ] ] ] c' [CP Q[wll] [pp Subject [probe: Q|poiarity| [ Verb [goal: polarity ] ] ] ] 26 The proposed analysis, however, not only captures the status of Thai wh-expressions as variables, but also accounts for the absence of island effects. A probe matches in features with a goal such that the closest c-commanding probe—which need not be in the same clause—is the one that enters into the probe-goal relation. The present analysis predicts no island effects anywhere because the probe as Q[W|,] that is introduced higher up in a matrix clause is the closest c-commanding probe that is copied onto the underspecified goal generated down below. To summarize, I have claimed that the operator-variable relation in Thai is best implemented as the probe-goal relation. I have set out some core assumptions regarding how the probe-goal relation is established, and what conditions are imposed on this relation. I have discussed data from other languages that support the analysis. The next section motivates the presence of a covert Q[wh] morpheme that carries a wh-feature; it is this feature that is copied onto the underspecified goal and gives rise to wh-construals in Thai. 2.3 Identifying the probe: Q|Wh| I argue that Thai wh-constructions contain an abstract Q morpheme that is specified for a [wh] feature. It is this Q[wll] operator that forces in-situ variables to be construed as wh-expressions. As we shall presently see, in Thai, the presence of this covert Q[wh] can be detected in both matrix and embedded clauses. 2.3.1 Comparing Thai, Japanese and Mandarin: matrix Q(wh] The presence of an abstract Q[wh] can be motivated by comparing Thai to other wh in-situ languages like Japanese and Mandarin. To form wh-questions, Japanese requires wh-words and the sentence-final particle ka, (21a). In contrast to this, with Mandarin, the sentence-final particle ne that occurs with wh-questions is optional, (21b). As for Thai, only the wh-expression is present: there is no overt Q[wh] morpheme in wh-questions, (21c). 27 (21) a. JAPANESE (Hagstrom 1998:15) John-ga [nani-o] kaimasita ka John-nom what-acc bought Q What did John buy? b. MANDARIN (Cheng 1991:30) Qiaofeng mai-le [shenme] buy-asp what What did Qiaofeng buy? c. THAI Nit si; [?aray] buy VARIABLE. -HUMAN What did Nit buy? Even though there is no overt Q[Wh] morpheme in Thai5, wh-expressions nevertheless receive an interrogative interpretation in a parallel fashion to wh-expressions in Japanese and Mandarin. This is summarized in (22)6. (22) THE REALIZATION OF Q[wh] IN THREE WH IN-SITU LANGUAGES Q[wh] a. Japanese ka b. Mandarin (ne) c. Thai 0 5 Andrew Simpson pointed out that Thai has the optional occurrence of Id for use in emphatic-insistent wh-question. I, however, treat this particle as an emphatic marker, rather than a Q[Wh] particle. 6 As shown in (21a) and (21b), overt Q[wl)] ka and nemark interrogative clauses in Japanese and Mandarin respectively. Note that when Q[wh] is overt, it appears sentence-finally. The position of Q[wh] is not significant to my claim but there may be a correlation between the position of Q[wh] and word order. The fact that Q[Wi,] ka appears as a question-final particle in Japanese reflects its word order as being consistent with a head-final language. As for Mandarin, despite its appearance of being head-final in some structures (i.e. ne appears as a clause-final particle), Mandarin is a head-initial language (Cheng 1991). One analysis that has been pursued for Mandarin is to generate the question particle in a head-initial position and front the IP to derive the sentence-final position of the question particle. (ne) (Q) 28 The parallel between Japanese, Mandarin and Thai can be accounted for by positing a covert Q[wh] morpheme. On this view, one would describe the distribution of Q[wh] as follows: in some languages it is overt (e.g., Japanese), in other languages it may be covert or overt (e.g., Mandarin), and in other languages it is always covert (e.g., Thai). In all three types of languages, it is the presence of Q[wh]—whether overt or covert—that forces a wh-construal. In the next section, I present independent evidence for the presence of a covert Q[wh] in Thai. 2.3.2 Selectional restrictions: embedded Q|Wh| In addition to the covert Q[Wh] that occurs in matrix clauses, there are reasons to think that Q[Wh] also occurs in embedded clauses in Thai. The evidence comes from the restrictions imposed by verbs on the clauses that they select. It is well-known that certain verbs require certain kinds of complements (Grimshaw 1979, Huang 1982). It is assumed that such selectional requirements are listed with each verb in its lexical entry. For example, in English, the verb 'ask' selects for [+wh] complements, as in (23). This contrasts with 'know', which selects for both [+wh] complements and [-wh] complements, as in (24). And yet other verbs such as 'think' select exclusively for [-wh] complements, as in (25). ENGLISH (23) a. He asked who read the book. b. * He asked that Mary read the book. (24) a. He knew who read the book. b. He knew that Mary read the book. (25) a. * He thought who read the book. b. He thought that Mary read the book. In Thai, the verb thd:m 'ask' takes either a [+wh] NP or CP complement (i.e., an indirect question), (26a-b), but not a [-wh] CP complement. Just as the source of the interrogative force in matrix clauses is (covert) Q[wh] (26c), similarly in embedded clauses the presence of (covert) Q[wh] is forced by the selectional requirement of the verb 'ask', and the embedded wh-question construal in (26b) naturally follows. 29 (26) a. khaw tha.m [NP khamtha.Tn ] he ask question He asked a question. b. khaw f/'a.m [Cp wa:7 [Cp khray ?a.-n nans?: ] he ask comp VARIABLE. +human read book He asked who read the book. c. * khaw tha.-m [CP wa: [IP Nit ?a.n nans?:] he ask comp read book He asked that Nit read the book. 7 The embedded clause meets the selectional requirement of the matrix verb faim 'ask' which selects [+wh]. A wh-variable is bound by the most local operator available, in this case a covert Q[Wh] operator. Since the complementizer wa: co-occurs with wh in-situ which by hypothesis needs a null Q[Wh], this suggests that there are two C projections as shown schematically in (i). (i) [VP... 'ask'... [CP [c Q|Wh| • • • [CP [C wa:... [,P ... variable ... ]]]]]] As a result, the CP domain must be split into at least two projections: one projection that specifies the force of the sentence (ForceP) and another projection that determine the finiteness (FinP) following Rizzi's articulated CP structure (1997). Since Q[Wy contributes the interrogative force to the sentence and wa: introduces finite clauses, Q[wh] and wa: occupy Force and Fin respectively. The following question arises that how can we determine that Q[Wh] precedes wa: ? In (ii), the overt Q[yes-no] morpheme is appended after the matrix verb yielding matrix yes-no questions. I argue in the next chapter that both Q[yes-no] and Q[wh] are m complementary distribution, the position of the overt Q[yes-no] morpheme suggest that ForceP should precede FinP. (ii) khaw ni; may [CP wa: [CP khray ?a.n narjs?: yu: ]] he know Q[y^-„0j comp VARIABLE. +HUMAN read book prog Did he know who was reading the book? 30 Like its English counterpart, the Thai verb ru: 'know' selects for either a [+wh] or a [-wh] complement. The presence of (covert) Q[wh] in the embedded C yields the construal in (27a-i), while the presence of (covert) Q[wh] in the matrix C yields the construal in (27a-ii). The latter reading predictably arises when the complement of ru: 'know' is [-wh]. That this verb can introduce a [-wh] complement is confirmed by the examples in (27b-c), which show that [-wh] NP and CP complements are possible. (27) a. khaw ru: [cpwa: [Cpkhray ?a.n nans?:] he know comp VARIABLE. +HUMAN read book = (i) He knew who read the book. = (ii) Who did he know read the book? b. khaw ru: [NP k'Va.-mcirj] he know truth He knew the truth. khaw ru: [Cp wa: [IP Nit ?a.-n nans?:] he know comp read book He knew that Nit read the book. The verb kl'it 'think', on the other hand, selects for only [-wh] complement CPs, (28a). It prohibits NP complements, (28b), as well as embedded wh-questions (28c-i). In (28c-i), because the verb 'think' does not select embedded wh-question, the embedded wh-question construal is not possible. However, the matrix wh-questions remains available, as (28c-ii), because there is always the possibility of howing a covert Q[wh] in the matrix clause. (28) a. khaw k'*it [CP wa: [IPNit ?a.n nans?:] he think comp read book He thought that Nit read the book. b. * khaw khit [DPnans{:] he think book [* He thought the book.] 31 c. khaw khit [cpwa: [Cpkhray ?a.n nans?:] he think comp VARIABLE. +-HUMANread book ^ (i) He thought who read the book. = (ii)Who did he think read the book? (29) provides a summary of the selectional restrictions requirement of the verbs behave in Thai. (29) embedded [+wh] matrix [+wh] a. tha.Tn 'ask' X b. ni: 'know' V c. khit 'think' X V To sum up, Thai and English show a a parallel behavior in how verbs select their complements and both languages have a covert Q[wh] in embedded [+wh] clauses. The data above supports the claim that there is a covert Q[wh] in Thai, and that this covert Q[wh] occurs in both matrix and embedded clauses. 2.4 Matching the Q|Wh| probe In the previous section, I motivated my claim that there is a covert Q[wh]. In this section I argue that the goal in Thai is a variable underspecified for a wh-feature, and it matches in feature with this covert Q[wh] as the probe. 2.4.1 Deriving the wh-construal A key claim of the present analysis is that in-situ expressions that are found in wh-constructions are not inherently specified for wh-features. Rather, as a variable, the interpretation of the in-situ expression is constrained by the operator that c-commands it. For example, in (30), it is the abstract Q[wh] in C that determines the wh-construal of the in-situ expression ifray in object position. 32 (30) tray AS IN-SITU WH-OBJECT Nit hen [khray] see VARIABLE. +HUMAN Who did Nit see? Evidence in favor of analyzing tray as a variable comes from the fact that its interpretations are constrained by the feature of the probe it is copied onto. By copying the [wh] feature of the abstract Q probe, tray gets a wh-construal. By copying the [Neg] feature of the negative probe, tray is interpreted as a negative polarity item, (31). And by copying the [polarity] feature of the yes-no question marker, tray is interpreted as an existential polarity item, (32). (31) tray AS NEGATIVE POLARITY ITEM (NPI) Nit may hen [khray] neg see VARIABLE. -/-HUMAN Nit did not see anyone. (32) tray AS IN EXISTENTIAL POLARITY ITEM (EPI) Nit hen [k'Vay] may see VARIABLE. +HUMAN Qfpoiarity] Did Nit see someone? These data establish that the in-situ expression is a variable, i.e. it is not inherently specifed for wh-features (in contrast to English who), for negation (in contrast to English nobody) or for existential force (in contrast to English someone). 2.4.2 The goal as an underspecified variable Regardless of whether variable expressions are construed as interrogatives, as negative polarity items, or as existential polarity items, their morphological composition is invariant. This is illustrated in (33) for the four main dialects of the Thai language, namely the Standard, Southern, Northeastern, and Northern variants. 33 (33) INVENTORY OF VARIABLE EXPRESSIONS IN DIALECTS OF THAI FEATURE MAKE-UP CONSTRUAL Standard Southern Northeastern Northern [+human] Who, anyone, someone kh-ray kh-ray ph-ay ph-ay [-human] what, anything, something ?a-ray (?ay)-ray (?i)-yarj ?a-yan [+location] where, anywhere, somewhere (thi:)-nay (rt.O-nay say or m)rj-day ti-nay [+time] when, anytime, sometime m?:a-ray ml:a-ray m1:-day m?:-day [+reason] why, any reason, some reason tham-may Say het-yarj ya-yarj [+way] How, anyhow, somehow ya.Tj-ray or(yarj)-rjay tham-phrf: or (yarj)-ray can- day ca-day (33) reveals that all variable expressions in Thai are composed of two morphemes. They all share the same second morpheme, which is some form of -(X)ay: -ay, -ray, -nay, -day, -may, -nay. We have already established, that.these \\x-(X)ay\ forms have the status of variable expressions that are underspecified for a feature. In terms of their morphosyntax, they are composed of two features: Fl and F2. Specifically, the first morpheme spells out the relevant F2 features, and I propose that the invariant second morpheme -(X)ay instantiates the variable underspecified for Fl feature, as in (34). (34) GENERAL FORM OF THAI VARIABLE EXPRESSIONS [ [F2 ] [Fl u (X)ay ] ] I now illustrate how the morphosyntactic analysis in (34) accounts for the surface forms of Thai variable expressions. Consider first the elements which instantiate the F2 [+human] and [-human] values; they are analyzed as (35) and (36) respectively. (35) [F2 [+HUMAN ] [Fl p (X)ay ] ] a. kh-ray Standard b. kh-ray Southern c. ph-ay Northeasterh d. p -ay Northern 'who, anyone, someone' 34 (36) [F2 [—HUMAN ] [Fl u. (X)ay ] ] a. ?a-ray Standard b. (?ay)-ray Southern c. (?i)-yarj Northeasterd. ?a-yan Northern 'what, anything, something' The F2 of locative and temporal variables are the [+place] and [+time] features, in which the morphemes can occur independently as prepositions (e.g. thi: talait 'at market' or mv.a chaw 'at morning'). (37) [F2 [P at.place ] [Fl u (X)ay ] ] a. (thi:)-nay Standard b. (thi:)-nay Southern c. say or m3rj-day Northeastern d. ti-nay Norther'where, anywhere, somewhere' (38) [F2 [P at.time ] [Fl u (X)ay ] ] a. m1:a-ray Standard b. m?:a-ray Southern c. m?:-day Northeasterd. ml-day Northern 'when, anytime, sometime' The F2 of wh-rationale and wh-manner are [+reason] and [+way] respectively. The first morpheme of rational variable-expressions both variables can occur independently as a verb, while the first morpheme of manner expressions can occur independently. 35 (39) [F2 [v do .reason ] [Fl u (X)ay ] ] a. tham-may Standard b. s-ay Southern c. het-yan Northeasterd. ya-yan Northern 'why, any reason, some reason' (40) [F2[N way] [Fl uf^av]] a. yarn-ray or (yarj)-rjay Standard b. (yan)-ray8 Southern c. can- day Northeasterd. ca-day Northern 'how, anyhow, somehow' The proposed analysis is supported by morphosyntactic evidence above that, in Thai the goal has two components: the first is a semantic constant (+human, -human, +place, +time, +reason, +way); the second component is the underspecified variable -(x)ay. 2.5 The structural properties of the probe-goal relation In the previous section I introduced the claim that the wh-construal is constrained by probe-goal relation. In this section, I argue for a non-movement analysis of the probe-goal relation. That is, the probe-goal relation is established via Match without Move. In Thai, the probe Q[wh] is based-generated in C, c-commanding all wh-variables in the clause. The position of the probe predictably interacts with the Matching relation. This accounts for the lack of asymmetry between wh-subject and wh-object (with respect to how variables are construed in both positions and also the lack of asymmetry between wh in-situ arguments and wh in-situ adjuncts with respect to the absence of island effects). Another way to say 'how' in the southern dialect is tham-phri: which literally means 'do how'. 36 2.5.1 The lack of an asymmetry between wh-subjects and wh-objects Variable expressions are structurally matched with the feature of the probe is covert Q[wh] operator in C. This predicts that there should not be an asymmetry between wh-subjects and wh-objects because the goal in both positions is in the domain of the Q[wh] probe in C. WH-SUBJECT (41) a. [probe: Q [ wh ] [goal: [ u, ] [ verb object ] ] ] a' [probe: Q [ wh ] [goal: [ wh ] [ verb object ] ] ] WH-OBJECT . . b. [probe: Q [ wh ] [subject verb [goal: [ u, ] ] ] ] b' [probe: Q [ wh ] [subject verb [goal: [ wh ] ] ] ] This prediction is borne out, as we see in (42). (42) a. [khray] hen Nit VARIABLE. +HUMAN see Who saw Nit? b. Nit hen [khray] see VARIABLE. +HUMAN Who did Nit see? The goal matches in feature with the closest probe (a covert Q[Wh]) and is in the c-command domain of the probe. 2.5.2 Long-distance probe-goal relation In the previous section, I illustrated the probe-goal relation for contexts where both the probe and the goal are contained in the same clause. In this section, I consider long-distance probe-goal relations where the probe is in a matrix C, while the goal is in an embedded clause. I then argue that the present analysis predicts no island effects anywhere in Thai. 37 Examples of long-distance probe-goal relations are given below. All wh-expressions in embedded clauses—wh-subjects (43), wh-objects (44), wh-indirect objects (45) and wh-possessors (46)— only allow matrix wh-construals. WH-SUBJECT (43) John khit wa: [khray] st: nans?: ma: think comp VARIABLE. -/-HUMAN buy book come Who did John think bought a book? WH-OBJECT (44) John k''it wa: Nit si: [?aray] ma: think comp buy VARIABLE. -HUMAN come lit = John thought Nit bought what. What did John think Nit bought? WH-INDIRECT OBJECT (45) John khit wa: Nit hay nans?: [khray] think comp give book VARIABLE. +HUMAN lit = John thought Nit gave a book to whom. Whom did John think Nit gave a book to? WH-POSSESSOR (46) John khit wa: Nit ?a.n nans?: kh5.T) [khray] think comp read book of VARIABLE. +HUMAN lit = John thought Nit read a book of who. Whose book did John think Nit read? 38 We have already seen in section 2.2.2 that, in Thai, the verb 'think' selects exclusively for [-wh] complements. Since the verb 'think' does not select embedded wh-questions, an embedded wh-construal is not possible. One question that arises is how the goal in (43-46) gets a matrix wh-construal. It is the covert Q[wh] probe in the matrix clause that makes the matrix wh-construal possible. The underspecified goal in the embedded clause looks for a probe to match in [wh] feature; in this case, it is the matrix Q[wh]. A key claim of the present analysis is that such long distance probe-goal relations are possible as long as no other operator intervenes. 2.5.3 No island effects with wh in-situ arguments The present analysis correctly predicts an absence of island effects. It also correctly derives the absence of asymmetry between wh in-situ arguments and wh-adjuncts (see section 2.5.4) with respect to island effects9. The data given from (47) through (49) show that wh in-situ arguments are able to occur in any island without creating island effects. Because the probe Q[wh] is base-generated in the matrix C, the goal copies the [wh] feature from the probe thereby satisfying Match. Note that such long distance probe-goal relations allow the goal to occur in an island. If the goal were to undergo covert wh-movement from a relative clause island, we would expect (47) to be ungrammatical. However, (47) is a grammatical sentence. This is consistent with the claim that there is no covert wh-movement of the goal in Thai. RELATIVE CLAUSE ISLAND (47) Q: khun ch3.p [Np ple.Tj thi: [CP khray r5.Tj]] you like song comp VARIABLE. +HUMAN sing Who did you like the song such that x sang? A: Britney Spears If there were covert wh-movement, then extraction of the goal from a sentential subject island is expected to be ungrammatical. However, (48) is well-formed with the goal inside the island 9 Note that the term "island" is used for ease of exposition. No node is assumed to be a "barrier" intrinsically. 39 matching the [wh] feature of the probe base-generated in the matrix C. This suggests that no covert movement of the goal takes place. The goal matches in feature with the probe through a long-distance relation. (48) SENTENTIAL SUBJECT ISLAND Q: [Npkarn10 thi: [ip Nit man kap khray]] thamhay ma?: kh5.Tj Nit may pjxay nom comp engage with VARIABLE. -/--HUMAN cause mother of neg please That Nit got engaged with who upset his mother. A:John If there were covert wh-movement, it would not be possible for the goal to move out of an adjunct island. However, (50) is a grammatical sentence having the goal take scope outside the island. This supports the claim that the probe Q[wh] is base-generated high up in the clause, and there is no covert movement. The probe-goal relation is only mediated by Match. ADJUNCT ISLAND (49) Q: k'Wj thu:k lay?3:k [phr5? (khaw() khamo.y ?aray ] he pass fire becausefhe) steal VARIABLE. -HUMAN What was he fired because he stole? A: rjan Money Now let us consider a wh-island in (50). There appears to be a wh-island effect in Thai: the wh-island prevents the wh-expression ?aray 'what' from being to be construed as questioning an embedded clause as shown in (50-ii). The only available interpretation in (50) is yes-no matrix question, where the verb 'know' takes [+wh] complement, as in (50-i). 6 Sentential nominalizer and complementizer are optional. ' 40 (50) WH-ISLAND" Q: khun ru: may12 wa: Dang hen ?aray you know yes-no comp see VARIABLE. -HUMAN = (i) Do you know what Dang saw? & (ii)What do you know whether Dang saw? A: ru: know Yes, I do. The proposed analysis predicts no island effects. (50) seems to contradict such claim. However, I show schematically below in (51) that my analysis predicts what looks like a wh-island effect in (50). Recall that the verb 'know' takes a [+wh] or a [-wh] complement. We expect that the presence of the probe in the embedded C is the closest probe where the goal matches (by feature copying) in [wh] feature yielding wh-embedded scope construal in (50-i). The reason that wh-matrix scope construal in (51 -ii) is not available is because the goal cannot skip the closest c-command probe to agree with the probe in the matrix C due to locality conditions. (51) a. [IP Subject Q[yes-no] Verb [CP Q[+wh] [IP Subject Verb [variable] ] ] ] b. [IP Subject Qlyes-no] Verb [CP Q[-wh] [IP Subject Verb [variable] ] ] ] 11 Andrew Simpson pointed out the data in (i) as a wh-island effect example because (i) cannot have a matrix wh-construal. (i) khaw ni; khray s»: ?aray he know who buy what = He knew who bought what. * What did he know who bought? The fact that (i) cannot have a matrix wh-construal suggests that there is a wh-island effect in Thai, just as in Japanese. I leave this for future research. 12 See chapter 3 for a detailed discussion of the internal structure of the yes-no marker may. 41 This example illustrates how the goal matches with the closest probe. This data supports my claim that probe-goal relation in Thai is constrained by a locality condition such that the closest c-commanding probe is the one that enters into the probe-goal relation. Overall, the point of this section was to show that there is no evidence for movement of the goal in Thai. Rather the goal is base-generated and can enter into a long-distance relationship with the probe. And this is supported by the lack of island effects. 2.5.4 No island effects with wh in-situ adjuncts I have shown that wh-expressions in Thai are in-situ in argument positions—be it subjects, objects, indirect object and possessors. I show, in this section, that wh-adjuncts also occur in their base-generated position and match with the covert probe Q[Wh] in the matrix C, the same way wh-arguments do. The proposed analysis predicts that , like wh-arguments, wh-adjuncts will not show island effects. As before, the long-distance probe-goal relation is established through Match. In Thai, wh-adjuncts also occur in the same position as the corresponding non-wh expression in declarative sentences. This is shown in (52) for wh-locative adjuncts, in (53) for wh-temporal adjuncts, and in (54) for wh-manner adjuncts. (52) WH-LOCATIVE ADJUNCT Q: Nit st: wa.y [(thi:) nay] buy wine VARIABLE. PLACE Where did Nit buy a bottle of wine? LOCATIVE ADJUNCT A: Nit si: wa.y thi: ha.T) buy wine at mall Nit bought it at a mall. (53) WH-TEMPORAL ADJUNCT Q: Nit ri.ancop [mt:aray] graduate VARIABLE. TIME When did Nit graduate? 42 TEMPORAL ADJUNCT A: Nit rirancop rri?:a pi: thi:loB.,w graduate at year last Nit graduated last year. (54) WH-MANNER ADJUNCT Q: Nit khap rot drive car How did Nit drive? MANNER ADJUNCT A: Nit khap rot rew drive car fast Nit drove fast. The examples above establish that all wh-expressions in Thai occur in their base-generated position. I assume that wh-adjuncts are right-adjoined to IP corresponding to their surface forms. (55) [CP Q(WH] [IP SUBJECT VERB [IP WH-ADJUNCT ] ] ] If the probe Q[wh] is base-generated in the matrix C, this predicts that wh-adjuncts should be able to occur inside an island without creating any island effects, the same way wh in-situ arguments do. The data turn out to be just as predicted. The example (56)13 shows a wh-locative adjunct in a relative clause island. Crucially, a matrix wh-reading is possible. If there were covert wh-movement, the sentence should not allow a matrix wh-reading because the probe Q[Wh] would have to move out of an island to the matrix C. pen [yannay] be VARIABLE. WAY 13 When wh-adjuncts are in the island, as in (56), long answer-forms seem to be required. Because the attached sites can be ambiguous, the verbs in the matrix or embedded clause need to be included in the answer to disambiguate the two readings: a matrix or an embedded wh-construal. 43 We would expect the sentence to be ungrammatical. However, (57) is grammatical. This suggests that no covert movement of the goal occurs. Instead, the goal matches in feature with the probe through a long-distance probe-goal relation. (56) RELATIVE CLAUSE ISLAND: WH-LOCATIVE ADJUNCT a. Q: k]lun ca; [NP krapaw [cp thi: Nit thamha.y ] ] t't.nay you find purse comp lose VA RIABLE. PLA CE Where did Nit lose her purse that you found? A: thamha.-y bon rotme; lost on bus (he) lost (it) on the bus. Similarly, a wh-locative adjunct can occur in a sentential subject island. As before the goals matches in the [wh] feature of the probe Q[Wh]> which lies outside the island. If there were a covert movement of the goal, we would expect (58) to be ungrammatical. However, (58) is a well-formed sentence. (57) SENTENTIAL SUBJECT ISLAND: WH-LOCATIVE ADJUNCT Q: [khun man kap Nit thi.-nay ] thamhay you engage with VARIABLE.PLACE cause Where did you get engaged to Nit upset Nit's mother? A: thi: ra.n ?a:ha.-n at store food That I got engaged with Nit at a restaurant upset Nit's mother. Wh-temporal. adjuncts show a similar pattern for both relative clause islands and sentential subject islands. If the goal were to move from inside an island, we would expect (58) and (59) to be ungrammatical. However (58) and (59) are well-formed. Therefore, no movement of the goal crosses the islands. ma>: Nit kro.-t ] ] mother angry 44 (5 8) RELATIVE CLAUSE ISLAND: WH-TEMPORAL ADJUNCT Q: khun cd: [NP krapaw [CP thi: Nit thamha.y m?:aray ]] you find purse comp lose VARIABLE. TIME When did Nit lose her purse that you found? A: Nit thamha.y ml'awa.-nni: lose yesterday Nit lost it yesterday. (5 9) SENTENTIAL SUBJECT ISLAND: WH-TEMPORAL ADJUNCT Q: [khaw klap thi.T) ba.n ml'aray ] thamhay he come arrive home VARIABLE. TIME cause When did that he got back upset your mother? A: ti: si: at four That he came home at 4am upset his mother. Wh-manner adjuncts show the same pattern as the wh-adjuncts discussed above in that they can be contained inside the islands, as illustrated in (60) for relative clause islands and in (61) for sentential subject islands. (60) RELATIVE CLAUSE ISLAND: WH-MANNER ADJUNCT Q: khun ca: [NP krapaw [CP thi: Nit thamha.y yannay ]] you find purse comp lose VARIABLE. WA Y How did Nit lose the purse that you found? A: Nit pay li.-m way thi: h3.-nnam go forget at restroom Nit lost (it) at a restroom. ma?: kro.t mother angry 45 (61) SENTENTIAL SUBJECT ISLAND: WH-MANNER ADJUNCT Q: [khaw khap rot yangay ] thamhay rot chon he drive car VARIABLE.WAY cause car crash How did that he drive the car cause the car crashed? A: khap may rawarj drive neg careful That (he) drove carelessly cause the car crashed. All the examples above illustrate a long-distance relation between the probe and goal which parallels the case of wh in-situ arguments. This is consistent with the claim that the relation between the probe and the goal does not involve movement with respect to island effects; rather the goal matches the probe through a long-distance relation. Considering wh in-situ arguments and wh in-situ adjuncts together, the evidence in the section strongly suggests that the syntactic position of the probe Q[wh] in Thai is base-generated in the matrix C. Notice that I have not yet discussed wh-rationale adjuncts. It is because they behave differently from the rest of the wh-adjuncts. On the one hand, like other adjunct expressions, wh-rational adjuncts can occur in the same position as the corresponding non-wh expression. This is illustrated in (62). (62) WH-RATIONALE ADJUNCT Q: Nit la:23:k [thammay] quit VARIABLE.REASON Why did Nit quit? A: Nit la:?3:k phr5? bt;a quit because bore Nit quit because (she is) bored. However, wh-rationale adjuncts can occur in embedded clause, as in (63), in which case there may be in-situ (63a), undergo partial movement (63b), or undergo long-distance movement (63c). 46 (63) WH IN-SITU a. khaw khit wa: Nit la:?j:k [thammay] he think comp quit VARIABLE.REASON Why did you think Nit quit? INTERMEDIATE SPEC CP b. khaw k'lt wa: [thammay] Nit la:?j;k he think comp VARIABLE.REASON quit Why did you think Nit quit? MATRIX SPEC CP J> .1 iJu... iJl. c. [t ammay] knaw knit wa: Nit la:?a:k VARIABLE.REASON he think comp quit Why did you think Nit quit? Moreover, wh-rational adjuncts, unlike other adjuncts, appear to be sensitive to island effects, as in (64). (64a) is ungrammatical when wh-rationale adjuncts occurs inside the relative clause island. (64b) also does not allow 'why' to extract out of the relative clause island. (64c), on the contrary, is grammatical but it only has a wh-matrix construal. (64) RELATIVE CLAUSE ISLAND: WH-RATIONALE ADJUNCT a. * khun ch3.p [Npnarjs!: [cpthi: Nit you like book comp [Why did Nit write the book that you liked?] k ia.n t ammay ] ] write VARIABLE.REASON b. * khun ch3.p thammayi [Np nans?: you like VARIABLE.REASON book [Why did Nit write the book that you liked?] [CPri: Nit knia.Ti tj ] ] comp write c. [thammayi khun ch3.p tj [NP nans?: VARIABLE.REASON you like book = (i) Why did you like the book that Nit wrote? ^ (ii) Why did Nit write the book that you liked? [CPthi: Nit khia.n t( ] ] comp write 47 The data in (64) contradicts what we saw earlier for other adjunct expressions where the probe-goal relation is established via Match rather than Move. Taken together, (63) and (64) show that wh-rationale adjuncts in fact undergo overt movement, as evidenced by their sensitivity to island effects. Two questions arise: (i) 'what triggers the movement?', (ii) 'why is Move only specific to wh-rationale adjuncts?' Although I do not discuss wh-adjuncts in this dissertation, it is likely that an EPP-feature requires the goal to move to the specifier of CP. But why does EPP only force the movement of wh-rationale adjuncts? This is beyond the scope of this dissertation. I leave this area for future research. 2.6 The interpretive properties of the probe-goal relation In this section, I discuss the interpretive properties of Thai wh in-situ construction, as they relate to multiple wh-questions and to wh-questions that contain quantifiers (wh-quantifier interaction). I show that the present analysis (i.e. Match without Move) correctly accounts for the absence of pair-list readings14 in Thai. I have already argued that Thai wh-construction contain an abstract Q[wh] probe that is base-generated in C. By base-generating the probe in C, we predict that multiple wh-questions cannot have a pair-list readings in Thai because both goals are necessarily in the scope of Q[Wi,j. This accounts for the fact that Thai multiple wh-questions only allow a single-pair answer reading. Using the same generalization on multiple wh-questions with questions that contain quantifiers, we also correctly predict that when Q[wh] takes scope over both a quantifier and a variable, a list answer is not possible. The present analysis correctly accounts for the fact that Thai only allows single answers in wh-questions that contain quantifiers. 2.6.1 Multiple wh-questions The proposed analysis correctly derives the interpretive correlates of Thai multiple wh-questions which are consistent with Hagstrom (1998)'s generalizations on how single-pair and pair-list readings are derived in multiple wh-questions. His generalizations are as follows. A multiple wh-question can be answered with a single-pair answer, answered by a single proposition referred as 'the single-pair reading' or a list of pair referred as 'the pair-list reading'. 48 (65) MULTIPLE WH-QUESTIONS: SINGLE-PAIR READING A multiple wh-question gets a single-pair reading when all wh-expressions are in the scope of Q[wh] (adapted from Hagstrom 1998:72). [PROBE: Q,wh, [WH-SUBJECT VERB WH-OBJECT ] ] (66) MULTIPLE WH-QUESTIONS: PAIR-LIST READING A multiple wh-question gets a pair-list reading when not all wh-expressions are in the scope of Q[wh] (adapted from Hagstrom 1998: 72). [WH-OBJECTj [PROBE: Q,wh | [ WH-SUBJECT VERB tj ] ] Adopting Hagstrom's generalizations, this predicts that a pair-list reading should be unavailable in Thai because both goals are in the scope of the probe Q[wh] generated in C, i.e. all wh-arguments are always in the scope of Q[wh]. Thus, the structure of Thai multiple wh-questions predictably satisfy the conditions for a single-pair reading only. (67) THAI MULTIPLE WH-QUESTION: SINGLE PAIR READING ONLY [PROBE:Q|wh, [VARIABLE VERB VARIABLE]] The data in (68) turns out just as predicted. (68) only has a single-pair answer reading. A pair-list reading that asks for specific things to be exhaustively paired with people is not available. 49 (68) V SINGLE-PAIR READING * PAIR-LIST READING Q: [khray] si: [?aray] ma: VARIABLE.+HUMAN buy VARIABLE.-HUMAN come Who bought what? A: Nit si: khanom '• ma; buy snack come Nit bought some snack. # A: Nit si: phdnlamay ci:ap kh5.T)wa.-n Kom k^.-arjdJLtn buy fruit dessert beverage Nit bought some fruits, Ciap bought some desserts and Korn bought beverages. The probe-goal correctly predicts that only single-pair answers are possible with multiple wh-questions in Thai. This is further supported by the behavior of multiple wh-questions when they are inside islands. Again, we predict that pair-list readings be unavailable, as confirmed by (69-71). The data below support the claim that there is no covert movement of the probe. The probe and goal relation is entered via Match (without Move). If there were a covert movement, we would expect such examples to be ungrammatical because the goal would move out of an island (i.e. a relative clause island, a sentential subject island and adjunct island). However, all the examples below are grammatical, crucially only the single-pair reading is available, as predicted. (69) V SINGLE-PAIR READING: RELATIVE CLAUSE ISLAND * PAIR-LIST READING Q: khun ch3.p [NP ple.T) t'l: [IP khray r3.T) thammay ] ] you like song comp VARIABLE.+HUMAN sing VARIABLE. REASON What is the reason such that you liked the song x sang? A: chan ch3.p ple.-n thi: Satirj r3.Tj phr3? (khaw) mi: ?ekkalak chaph3? tua / like song comp sing because (he) has character specially self I liked the song that Sting sang because he has a unique style of his own. 50 (70) V SINGLE-PAIR READING: SENTENTIAL SUBJECT ISLAND * PAIR-LIST READING Q: [NP karn t'l; [,Pkhray thal5? kap khray kan]] thamhay Nit krort nom comp VARIABLE. +HUMAN fight with VARIABLE. -/-HUMAN dist cause angry lit = Who fought with who made Nit angry? Who fighting with who made Nit angry? A: Nirj thal5? kap R3.n thamhay Nit krort fight with cause angry lit = Ning fought with Ron made Nit angry. Ning fighting with Ron made Nit angry. (71) V SINGLE-PAIR READING: ADJUNCT ISLAND * PAIR-LIST READING Q: [IPNit kro;t [iPphr5? Nirj kh5: hay khray s?: 2aray kan]] angry because ask give VARIABLE.+HUMAN buy VARIABLE.+HUMAN dist Nit was angry because Ning asked who to buy what? A: Nit kro:t phr5? Nirj kh5: hay R3.11 si: buri: angry because ask give buy cigarette Nit was angry because Ning asked Ron to buy a cigarette. To conclude, the present analysis does not allow a pair-list interpretation because wh-expressions are always in the scope of the probe. What this section showed us is that my analysis correctly predicts that multiple wh-questions in Thai can only receive a single-pair reading. And this supports the claim that no movement of the goal ever takes place in Thai. The goal matches in feature with the probe in the matrix C. This explains why no pair-list reading is available because all wh-expressions are in the scope of the Q[Wh]- In the next section, I turn to wh-constructions that contain both a wh-expression and a quantifier. I show that the present analysis also makes the right prediction that only single answers are possible when wh-expressions and quantifiers interact. 51 2.6.2 Wh-Quantifier interaction In this section, I look at scope interpretation between wh-expressions and quantifiers in Thai. How do they interact under the assumption that the probe Q[wh] is base-generated? The claim that Hagstrom makes is that the distribution of a single answer and list answer are derived in quantifier/wh-questions of this kind have a structure very much like those of multiple wh-questions. (72) WH-QUANTIFIER INTERACTION: SINGLE ANSWER A wh-construction that contains a quantifier gets a single answer when both question and quantifier are in the scope of Q[wi,]. (73) WH-QUANTIFIER INTERACTION: LIST ANSWER A wh-construction that contains a quantifier gets a list answer when the quantifier is not in the scope of Q[Wh] Consider how the probe-goal relation holds in Thai under the system Hagstrom proposes. When the universal quantifier is a subject and the wh-expression is an object, observe that only a single answer is possible (74). Similarly, only a single answer is available in (75) where the wh-expression is a subject and the universal quantifier is an object. (74) V SINGLE ANSWER * LIST ANSWER Q: thukkhon si: every-cl buy What did everyone I ma: come # A: Nit si: phdnlamay ci:ap kh3.-nwa.-n Kom k'l.-arjdi.-m buy fruit dessert beverage Nit bought fruits, Jiap bought desserts, Korn bought beverages. [2aray] ma: VARIABLE. -HUMAN come buy? A: thukk on si: k i:and?.Tn every-cl buy beverage Everyone bought beverages. 52 (75) V SINGLE ANSWER * LIST ANSWER Q [khray] kh3.Tj thukyan VARIABLE. +HUMAN buy thing evety-cl Who bought everything? A; Nit The present analysis correctly accounts for the fact that only the single answer is available. The single-answer reading is associated with a structure where the probe Q[wh] takes scope over both the wh-expression and the universal quantifier, as illustrated in (76). (76) SINGLE ANSWER a. [CP probe : Q,w„, [IP V [VP VERB WH-OBJECT ] ] ] b. [CP probe : Q,WH, [IP WH-SUBJECT [VP VERB V ] ] ] The lack of asymmetry between the wh-object/QP-subject and wh-subject/QP-object with respect to ambiguity/non-ambiguity in Thai is expected under a non-movement probe-goal relation, consistent with Hagstrom's generalization of how single-answer vs. list-answer readings can be derived. Now let us consider wh-quantifier interaction in relation to islands. In (77), the quantifier and wh-expression both are inside the adjunct island and it only gives rise to a single answer reading. If we were to assume that the quantifier undergoes covert movement, we would expect either a list-answer reading or ungrammaticality assuming that movement in general is island sensitive15. But this is not the case. (77) does not show an ambiguity and is a well-formed sentence. Therefore, I conclude that neither quantifiers nor wh-expressions not undergo covert movement. They are in the scope of the probe Q[wh] in C, and predictably allow only a single-answer reading to the question. 15 It has been observed in the syntactic literature that a scopal ambiguity arises in a wh-construction where they contain both a wh-expression and a quantifier, in particular, when the wh-expression is a object and a quantifier is an subject (e.g. English: May 1985, Chinese: Aoun and Li 1993). They are ambiguous in that there are two possible readings: single and list answer reading. To disambiguate the scope distinction between the two readings, the quantifier undergoes covert movement (quantifier raising) to have scope over the Q allowing the list answer reading. 53 (77) V SINGLE ANSWER * LIST ANSWER Q: khaw kro.t phr5? Nitkh5:hay thukkhon si: [2aray] he angry because ask give every-cl buy VARIABLE.+HUMAN What is he angry because Nit asked everyone to buy? A: khaw kro.t p'V5? Nit kh5: hay thukkhon si: wa.y he angry because ask give every-cl buy wine He is angry because Nit asked everyone to buy wine. # A: khaw kro.-t phr5? Nit kh5; hay Nik si: phdnlamay he angry because ask give buy fruit ci:ap kh3.Tjwa:n Korn k'l.-arjdi.m dessert beverage He is angry because Nit asked Nick to buy fruits, Jiap desserts, Korn beverages. The following example is provided in support of my claim that no quantifier raising (no QR) takes place in Thai. Notice that in (78), the universal quantifier is outside the adjunct island. We would not expect any island effects, even though it raises. If the quantifier were to raise to have scope over Q[Wh], we expect an ambiguity in that both single-answer and list-answer readings should be possible. (78) however only has single-answer reading. This confirms that quantifiers in Thai do not undergo QR. Since quantifers do not raise, the quantifier and the wh-expression are both in the domain of the probe Q[wh] in the matrix C. This is why the single-answer reading is the only available. We see that proposed analysis correctly predicts the absence of such ambiguity. (78) V SINGLE ANSWER * LIST ANSWER Q: thukkhon kro.-t p"r5? Nit kh5: hay khaw si: [2aray] every-cl angry because ask give he buy VARIABLE. +HUMAN What is everyone angry because Nit asked them to buy? 54 A: thukkhon krd.t p'V5? Nit kh5: hay khaw si: law every-cl angry because ask give he buy alcohol Everyone is angry because Nit asked them to buy alcohol. To summarize, Thai fits into the system Hagstrom (1998) has developed. These examples have shown that using Hagstrom's generalization regarding the distribution of single-answer versus list-answer readings is compatible with my claim that the goal does not undergo covert movement. Rather, the goal matches in feature with the probe in the matrix C without Move. 2.6.3 Some complications This section considers how pair-list answer readings arise in Thai. We see, in (79), that Thai requires the overt distributive operator kan to give rise to such readings. In fact, it can only receive a pair-list answer where each person must pair with an object they purchase16. (79) * SINGLE-PAIR READING V PAIR-LIST READING Q: [khray] si: [?aray] kan ma: VARIABLE.+HUMAN buy VARIABLE.-HUMAN dist come Who bought what? # A: Nit si: khandm ma: buy snack come Nit bought some snack. 16 D-linked wh-expressions in multiple wh-questions only allow the pair-list reading. Q: lu.k khon nay ch3.p kin ?aham ba*.p nay child cl variable like eat food kind variable Which child likes to eat which kind of food? A: Nit ch3.p kin khanom ci.ap k^.Tjwa.-n Korn phdnlamay like eat snack dessert fruit Nit likes to eat snacks, ciap likes to eat desserts, Korn likes to eat fruits. Following Hagstrom (1998), pair-list readings should not be possible because both D-linked wh-expressions are in the scope of Q[Wi,]. This suggests that pair-list readings will result just as long as the D-linked wh-expressions can be understood to be plural sets, and it may not have anything to do with being under the scope of Q[Wh], contra with Hagstrom's generalization. The difference between D-linked and bare wh-expressions remains to be accounted for. 55 A: Nit si: phdnlamay ci:ap kh5.Tjwa.n Korn k^.'arjdt.-m bring fruit dessert beverage Nit bought some fruits, Ciap bought some desserts and Korn bought beverages. Wh-expressions in (79) only have a pair-list answer reading (i.e. plural and distributive readings). It is the morpheme kan that gives the reading to (79). kan occurs in a wide range of syntactic environments with differing semantic interpretations (e.g. as an adverb equivalent to 'together', as a reciprocal pronoun equivalent to 'each other', as a plural marker or as a distributive marker) Following Stein (1981), I assume that kan is a plural and a distributive marker, equivalent to 'all' and 'each' in English. This morpheme takes scope over the entire proposition and gives pair-list readings to 'who' and 'what' such that 'who' and 'what' have more than one member, even though the wh-expressions are in the scope of the Q[wh] probe. However, the interesting observation is that the pair-list reading that arises with kan occurs even when wh-expressions are in the scope of Q[wh]. This suggests that wh-expressions in Thai may be inherently singular, as such, they would be compatible only with Hagstrom's generalization on the single-pair reading. Wh-quantifier interactions show a similar pattern in that in Thai, list-answer readings are derived by appending the distributive marker kan adjacent to the wh-object, the same way that pair-list answer readings are derived in multiple wh-questions. kan has scope over the universal quantifier and the wh-expression allowing a list-answer to occur as shown in (80). (80) * SINGLE ANSWER V LIST ANSWER Q: thukkhon si: every-cl buy What did everyone 1 A: Nit si: phonlamay ci:ap kh3.T)wa.-n Kom k^.'andi.m buy fruit dessert beverage Nit bought fruits, Jiap bought dessert, Kom bought beverage. [2aray] kan ma: VARIABLE.-HUMAN dist come buy? 56 However, when the quantifier 'each' is used, list answers are allowed even without the distributive marker kan. This may be because the quantifier is inherently distributive, while 'everyone' is inherently collective. We saw that this is not predicted by Hagstrom's generalization and this remains to be accounted for. I leave this for future research. (81) * SINGLE ANSWER A/ LIST ANSWER Q: ta?:Iakhon si: [2aray] ma: every-cl buy VARIABLE.-HUMAN come What did each person buy? A: Nit si: phdnlamay ci:ap kh5.-nwa.Ti Korn k^.-andi.-m buy fruit dessert beverage Nit bought fruits, Jiap bought dessert, Korn bought beverage. In this chapter, I have presented a non-movement analysis of wh-constructions in Thai and I have considered some of its consequences. I have analyzed the syntactic relation between the operator and the variable in term of the probe-goal relation. The probe-goal relation requires the goal as a variable to seek for the closest probe (an abstract Q[Wh]) and enter into a Matching relation through feature copying, here a [wh] feature. I have presented evidence for the claim that there is an abstract Q[wh] probe. I have also shown that wh-expressions in Thai are underspecified variables whose construal is constrained by the operator whose domain they occur in. Lastly, I have discussed the two major consequences of my claim that the covert Q[wh] probe is base-generated in C position from where it assigns wh-scope. First, the present analysis correctly predicts the distribution of Thai wh in-situ. Particularly, it derives the absence of an asymmetry between wh-subjects and wh-objects, as well as the absence of an asymmetry between wh in-situ arguments and wh-adjuncts with respect to island effects. Second, its accounts for restrictions on interpretation in multiple wh-questions, as well as in wh-questions that contain a quantifier. 57 Chapter 3 Variables as polarity items 3.1 Matching probes for polarity In the previous.chapter, I proposed that a wh-construal arises via a probe-goal relation between an invisible Q[wh] (in C) and a variable (in situ). I further argued that the relation between the Q[Wh] and the variable does not involve movement, as evidenced by the parallelism between wh-subjects and wh-objects on the one hand, and by the absence of island effects on the other hand. Moreover, base-generating the probe in C correctly predicts that, with multiple wh-questions, only the single-pair reading is available. Finally, we also make the right prediction that only single-answer readings arise in questions that contain wh-words and quantifiers. This is because both the variable and the quantifier are in the scope of the Q[wh] probe. A central claim of the proposed analysis developed in Chapter 2 is that in-situ wh-expressions are not inherently interrogative. Rather, they are variables that acquire their interrogative force by copying the [wh] feature of a covert Q. This analysis predicts that the same elements which appear in wh-contexts will appear in other contexts, and that their interpretation will be constrained by the operator whose scope they occur in. This prediction is confirmed, in that the same elements that are construed as wh-expressions in wh-contexts have the status of polarity items in other contexts. This chapter explores how the goal (as a variable expression) matches the relevant probe in NPI and EPI environments. In particular, in the context of negation, variable expressions function as Negative Polarity Items (NPIs). NPIs are matched with the [neg] feature on the probe. In a yes-no construction, variables function as Existential Polarity Items (EPIs). EPIs are matched with the [polarity] feature on the probe. The following examples illustrate how the interpretation of a variable expression such as tray [+human] is contextually determined. In the absence of an overt operator, tray is interpreted as a wh-expression equivalent to 'who', as in (la). As argued in the previous chapter, the wh-construal reflects the presence of a covert Q[wh] in C. In the presence of negation (may), the variable expression tray is interpreted as a negative polarity item17 equivalent to 'anyone', as in 17 The contexts that license English NPI are found to license Thai EPIs (i.e. yes-no questions, modals). There is no clear-cut distinction between NPIs and EPIs except that EPI 'someone' in Thai have narrower scope than NPIs 'anyone' in English. 58 (lb). And in the presence of a yes-no question marker (may), the variable expression ifray is interpreted as an existential polarity item equivalent to 'someone', as in (lc). WH-CONSTRUAL (1) a. Nit hen [khray]18 see VARIABLE, -/-human Who did Nit see? * WH-CONSTRUAL V NPI-CONSTRUAL b. Nit may du.thu:k neg insult * (i) Who did Nit not insult? = (ii) Nit did not insult anyone * WH-CONSTRUAL V EPI-CONSTRUAL c. khaw hen [khray] he see VARIABLE * (1) Who did he see? = (ii) Did he see someone? The availability of the NPI and EPI construals with variable expressions is syntactically conditioned. Thus, while the wh-construal is freely available, the NPI/EPI construal shows a subject/non-subject asymmetry. As we have seen in (1), in object position a variable expression such as if ray may have a wh-construal, an NPI- construal or an EPI-construal. In contrast to 18 Wh-expressions can also be interpreted as EPI and not being under scope of the probes mentioned. They, however, are accompanied by an overt existential quantifier ban 'some', in which case, c-command relation does not hold between the probe and the goal. (i) Nit hen [khray] barjkhon see VARIABLE.-/-human some-cl * Who did Nit see? = Nit saw someone. (ii) [kVayl barjk"on hen Nit VARIABLE.-f-human some-cl see * Who saw Nit? = Someone saw Nit. [khray] VARIABLE, -/human may . -hhuman Q[Poiarity] 59 this, in subject position if ray only has a wh-construal (2a); the NPI- and EPI-construals are unavailable, (2b-c). V WH-CONSTRUAL (2) a. [khray] hen Nit variable, -/-human see Who saw Nit? V WH-CONSTRUAL * NPI-CONSTRUAL b. [khray] may ch3.p Nam variable, -/-human neg like = (i) Who does not like Nam? * (ii) Anyone does not like Nam. * WH-CONSTRUAL * EPI-CONSTRUAL c. *[khray] hen khaw may variable, -/-human see he Qfpoianiy] ^ (i) Who saw him Nam? * (ii) Has someone seen him? We will see that the interpretation assigned to a variable expression is predictably determined by whether or not the c-command relation holds between the operator (as probe) and the variable expression (as goal). In order to enter a Matching relation, the goal must be in the domain of the probe. Note that the Matching relation is defined through feature identity when the feature of the probe is copied onto the underspecified goal. A domain is defined by the c-command relation. The goal must be c-commanded by the probe in order to be in the domain of the probe. Example (2b) clearly illustrates that the goal (in subject position) is not c-commanded by the negative probe, hence it is not in the domain of the probe. An NPI construal is unavailable. Sections 3.2 and 3.3 examine the details of the subject/object asymmetry with NPI- and EPI-construals in negative and yes-no question contexts. 60 Another way in which the availability of NPI- and EPI-construals with variable expressions is syntactically conditioned concerns complement/adjunct asymmetries. Recall that a variable expression in complement position (of a matrix clause) may have a wh-, an NPI-, or an EPI-construal, (3). This contrasts with variable expressions in adjunct position, which only have a wh-construal, (4). V WH-CONSTRUAL (3) a. Nit hen [khray] see VARIABLE, -f-human Who did Nit see? V NPI-CONSTRUAL b. Nit may du.lhu:k [khrayl neg insult VARIABLE, -hhuman Nit did not insult anyone V EPI-CONSTRUAL c. khaw hen [khray] may he see VARIABLE, -hhuman .[polarity] Has he seen someone? V WH-CONSTRUAL (4) a. Ra.ri ra.Tjha.y [thammay] cry VARlABLE.reason Why did Ron cry? 61 * NPI-CONSTRUAL b. * Rdj\ may r5.Tjha.y [thammay] neg cry VARIABLE, reason [Ron did not cry for any reason.] * EPI-CONSTRUAL c. * Ra.-n r5:rjha.y [thammay] may cry VARIABLE.reason Qipoiarity] [Did Ron cry for some reason?] I show that such complement/adjunct asymmetries arise from the interaction of the syntactic position of the probe and goal. The present analysis predicts that manner and reason wh-adverbials never receive an NPI- or EPI-construal because they are not c-commanded by the probe. This follows from the position of the relevant probes: negative may and the yes-no question marker may are generated lower than I and higher than V. This is discussed in section 3.4. ' . Another way in which NPI- and EPI-construals are syntactically conditioned concerns their distribution in embedded clauses. Depending on the selectional properties of the matrix verb, a variable expression in an embedded clause can have either a wh-construal (5a), an NPI-construal (5b), or an EPI-construal (5c). V WH-CONSTRUAL (5) a. Lek may day tha:m wa: [khray] durtNrk Ca.n not past ask comp VARIABLE, -hhuman insult John Lek did not ask who insulted John. V EMBEDDED NPI-CONSTRUAL b. Lek may khit wa: [khray] durtNrk Con not think comp VARIABLE. +human insult John Lek did not think that anyone insulted John. 62 V EMBEDDED EPI-CONSTRUAL c. Lek khit may wa: [khray] du.tVk Cb.n think yes-no comp VARIABLE, -thuman insult John Did Lek think that someone insulted John? Wh-, NPI- and EPI-construals in embedded clauses are a side-effect of locality conditions. I show that the goal and its matching probe are constrained by a locality condition such that the closet c-commanding probe is the one that enters into the probe-goal relation. This is discussed in section 3.5. 3.2 Subject/non-subject asymmetry with NPI-construal As discussed in the preceding chapter, the proposed analysis correctly predicts that, in wh-contexts, there is no asymmetry between wh-subjects and wh-objects since both goals are in the domain of the probe Q[Wh] in C. On the other hand, we expect a subject/non-subject asymmetry in NPI- and EPI-contexts: wh-subjects are outside the c-command domain of the probe, and thus do not receive NPI- and EPI-construals. This section examines the distribution of NPI-construals, and the next section examines the distribution of EPI-construals. As predicted, variable expressions in subject position only have a wh-construal as shown in (6). In (6), the subject is not in the domain of the negative probe because Neg is generated lower than the goal. Therefore, the [Neg] feature of the probe cannot be copied onto the underspecified goal and cannot receive an NPI-construal. However, the variable is inside the c-command domain of the Q[Wh] probe. The [wh] feature of the Q probe can thus be copied onto the goals which receives a wh-construal. WH-SUBJECT: V WH-CONSTRUAL NPI-CONSTRUAL (6) a. [khray] may c 3.p Nam VARIABLE, -fhuman neg like = (i) Who does not like Nam? & (ii) Anyone does not like Nam. 63 POSSESSOR SUBJECT: V WH-CONSTRUAL * NPI-CONSTRUAL b. ma: kh5.Tj [khray] may haw say Nit dog of VARIABLE. +human neg bark at = (i) Whose dog did not bark at Nit? * (ii) Anyone's dog did not bark at Nit. In object position, on the other hand, variable expressions receive only an NPI-construal, as in (7). Though the variables are in the domain of both negation and the Q[wh] probe, only the closest c-commanding probe, in this case the negative probe, is qualified to provide a feature for the goal to copy. This explains the unavailability of the wh-construal. We will see that the matching relation of the probe and goal in Thai is syntactically constrained by a locality restriction. WH-OBJECT: * WH-CONSTRUAL V NPI-CONSTRUAL (7) a. Nit may du.thu:k [khray] neg insult VARIABLE. +human * (i) Who did Nit not insult? = (ii) Nit did not insult anyone. POSSESSOR OBJECT: * WH-CONSTRUAL V NPI-CONSTRUAL b. Nit may yi.-m kh5.T) kh5.Tj [khray] neg borrow stuff of VARIABLE, -/-human * (i) Whose belongings did Nit not borrow? = (ii) Nit did not borrow anyone's belongings. We correctly predict the same for indirect objects in that they can have an NPI-construal. The data is given in (8). 64 (8) INDIRECT OBJECT: * WH-CONSTRUAL V NPI-CONSTRUAL Nit may hay rjdn [khray] neg give money VARIABLE, -hhuman & (i) To Whom did Nit not give the money? = (ii) Nit did not give the money to anyone. The above confirms that there is a subject/non-subject split with respect to the NPI-construal. The next step in understanding what makes an NPI-construal possible is to identify the syntactic position of negation. 3.2.1 The position of negative may In Thai, negative may occurs in preverbal position. I show in the examples below that may is generated closest to the verb relative to tense, modality and aspect marking. Note that Thai is a language that lacks overt tense marking. In (9a), the negative may simply appears before the predicate. In (9b), the sentence contains a modal (i.e. the future marker), and negation immediately precedes the verb. Examples (9c) and (9d) further confirm the preverbal position of negative may. We see that some aspect markers (e.g. imperfective) appear preverbally, while others (e.g. progressive) appear postverbally but negation still precedes the predicates. TENSE: PRESENT (9) a. khaw may ch3;p Nam he neg like He does not like Nam. MODALITY: FUTURE MARKER b. khaw ca? may kin kha.w he fut neg eat rice He will not eat the rice. 65 ASPECT MARKER: PROGRESSIVE c. khaw may day19 kin kha.-w yu: he neg past eat rice prog He-is not eating the rice. ASPECT MARKER: IMPERFECTIVE d. k'aw yan may day kin kha.-w he imperf neg past He hasn't eaten the rice yet. eat rice This establishes that negation occupies a position somewhere between the Subject and the Predicate (VP). Assuming that the Subject sits in Spec IP, then Neg is positioned between I and V, as in (10a). With respect to the potential c-command relations, note that while negative may c-commands the object, it does not c-command the subject. Contrast the structural position of the overt negative with that of the covert Q[wh] in C: the latter c-commands both the subject and the object, as in (10b). Finally, consider the structure in (10c), which has both Q[wh] and negative may: here the object is c-commanded by two operators (Neg and Q[Wh]), while the subject is c-commanded by only one operator (Q[Wh])-(10) a. [jp Subject [Infl] [NegP [Neg may ] [Vp V Object ]] ] b- [CP tc Q[wh] ] [TP Subject [Infl] [VP V Object ] ] ] c- [CP [C Q[wh] ] UP Subject [Infl] [NegP [Neg may ] [Vp V Object ] ] ] ] With these structures in mind, consider the predictions made by the proposed analysis regarding the construal of variable expressions. The essential claims of the probe-goal relation are such that a goal enters into a relation with a probe if and only if the following three conditions are met. (i) Feature identity: The probe-goal satisfies Match, which requires probe and goal to have identical features. (ii) C-command condition: The probe must c-command the goal. (iii) Locality: Match is satisfied by the most local probe. 19 The morpheme day can be treated as a past tense marker (Kanchanawan 1978), a verb (Sookgasem 1990) or as a modal (Warotamasikkhadit 1996). See Visonyanggoon (2000) for details. 66 Applying this to the structures in (10), we predict the following construals. First, in the presence of an overt negative probe, a goal will have an NPI-construal in object position, but not in (matrix) subject position, (1 la). This follows from the c-command restriction on the probe-goal relation. Second, in the presence of the covert Q[wh] probe, goals will have a wh-constfual in both subject and object position, as in (1 lb). Again, this follows from the c-command restriction on the probe-goal relation. Third, in the presence of overt negation and a covert Q[wh] probe, a goal in object position will have an NPI-construal, while a goal in subject position will have a wh-construal, as in (11c). The unavailability of the NPI-construal for the subject position follows from the c-command restriction. The unavailability of the wh-construal for the object position follows from the locality restriction. (11) a. [rp *NPI-subiect [Infll [NegP [Neg may ] [Vp V VNPI-obiect ] ] ] b- [CP tC Q[wh] ] []P ^wh-subiect [Infl] [Vp V Vwh-obiect ] ] ] c- [CP tC Q[wh] ] [fP Vwh-subiect [Infl][NegP [Neg may ItVP v VNPI-object ]]]] As we shall see, these predictions are borne out. 3.2.2 (The absence of) NPI-construal in subject position In the context of negation, goals (as variables) function as NPIs when they match the [Neg] feature of an overt Neg probe. However, in the presence of negation, subjects receive only a wh-construal, the NPI-construal is unavailable. This is illustrated in (12). V WH-CONSTRUAL * NPI-CONSTRUAL (12) a. [k'Vay] may ch3.p Nam VARIABLE. +human neg like = (i) Who does not like Nam?' ^ (ii) Anyone does not like Nam.' 67 WH-CONSTRUAL NPI-CONSTRUAL [?aray] may day Ion say Nam VARIABLE.-human neg past fall at = (i) What didn't hit Nam? & (ii) Anything didn't hit Nam. For goals in subject position, the analysis correctly predicts the absence of an NPI-construal (due to the c-command restriction) and the possibility of a wh-construal (again due to the c-command restriction). To see this, consider the structure in (13). In (13a), Neg does not c-command the subject position, so the goal cannot be construed as an NPI. In (13b), the goal is closest to and c-commanded by Q[wh], so it can be (and must be) construed as a wh-expression. (13) a. rCp lr O^h] 1 frp *NPI-subiect Until rN»gP rN?g may 1 Typ V Object 1111 b- [CP tc Q[wh] ] [IP Vwh-subiect [Infl] [NegP [Neg may ] [Vp V Object ] ] ] ] The question that arises is 'how do we get a NPI-construal in subject position?' My analysis predicts that a negative probe needs to be introduced higher up than the variable in subject position in order to license it as schematically illustrated in (14b). (14) a. rjp *NPI-subiect [Infl] [NegP [Neg may ] [VPV Object]]] b. Neg [jp VNPI-subiect [Infl] [VP V Object ] ] ] ] The data in (15) turns, out just as predicted. The goal in subject position indeed matches in feature with the Neg probe introduced higher. Therefore, the goal is in the c-command domain of the Neg probe and can receive an NPI-construal. Morever, this expression maymi:khray can only occur with the subject but, not the object, as shown by the contrast between (15a) and (15b). 68 VARIABLE-SUBJECT: * WH-CONSTRUAL V NPI-CONSTRUAL (15) a. maymi;[khray]20 ch3.p Nam neg- exist- VARIABLE, -/-human like ^ (i) Who does not like Nam? = (ii) No one likes Nam. (lit. 'There isn't anyone who likes Nam') VARIABLE-OBJECT: * WH-CONSTRUAL * NPI-CONSTRUAL b. * Nam ch3.p maymi:[khray] like neg- exist- VARIABLE, -/-human [Nam likes no one.] The fact that (15b) is ungrammatical suggests that with morphologically complex negation mdymi:khray), the Neg probe is introduced outside IP. If it were introduced inside the IP preceding the existential verb mi: 'have' in the same way as in morphologically simplex negation may is, an NPI-construal should be available in object position. We saw from the example (15a) above that the availability of the NPI-construal for the subject position follows from the c-command restriction. The proposed analysis also correctly predicts that the locality restriction will force the NPI reading on the variable khray 'who' in both subject and object position because the probe Neg mdymi:khray 'not-have-who' is closer to the variable than the probe Q[Whj. This is confirmed by (17). (16) a. [CP [c Q[wh] ] [Neg [jp VNPI-subiect [Infl] [yp V *wh-object ] ] ] ] ] b- [CP [C Q[wh] ] [Neg [rp VNPI-subiect [Infl] [yp V VNPI-obiectl ] ] ] ] mdymi:khray has a bi-clausal structure. 69 * WH-CONSTRUAL (17) V NPI-CONSTRUAL maymi:[khray] [khray] neg- exist- VARIABLE. -t-human like VARIABLE, -fhuman * (i) Who does nobody like? = (ii) Nobody likes anyone. To summarize, regarding goals in subject position, we see that the locality condition holds between the probe and goal relation in Thai. The (overt) Neg operator may cannot be a probe for the subject because Neg does not c-command the subject. The (covert) Q[wh] operator in C can be a probe for the subject because Q[wh] does c-command the subject position. Consequently, the subject of a matrix clause (predictably) gets a wh-construal, but not an NPI-construal. We return in section 3.5 to a discussion of the construal of subjects in embedded clauses. 3.2.3 NPI-construal in object position With goals in object position, we predict that only the NPI-construal is available. This is because the Neg operator is the closest c-commanding probe for the goal. This prediction is borne out, as shown by the examples in (18)21. * WH-CONSTRUAL V NPI-CONSTRUAL (18) a. Nit may du.tVk [khray] * (i) Who did Nit not insult? = (ii) Nit did not insult anyone. Echo wh-construal is possible but it is contextually restricted. A pause before the variable must be present, and extra morpheme ri: 'or' is required sentence-finally. In the present analysis, this would be an instance of local displacement of the variable expression so as to satisfy the locality restriction with respect to the scope of Q[Whj. I leave formalizing this aspect of my analysis to future research. neg insult VARIABLE, -hhuman 70 * WH-CONSTRUAL V NPI-CONSTRUAL b. Nit may hen [?aray] neg see VARIABLE.-human *(i) What did Nit not see? = (ii) Nit did not see anything. The NPI construal in (18) is illustrated structurally in (19). The generalization is that a wh-construal is not available in an NPI context because the closest c-commanding probe is Neg. Thus, the goal cannot match in [wh] feature with the probe Q by skipping Neg, as this would violate the locality restriction. (19) a. rrp *NPI-subiect rinfll |N»gP rN?g may 1 hrp V VNPI-obiect 111 b- [CP tc Q[wh] HrP Vwh-subject [Infl][NegP [Neg may ][Vp V *wh-object ]]]] The licensing condition22 between the probe and the goal is (predictably) syntactically conditioned by the c-command relation. 3.3 Subject/non-subject asymmetry with EPI-construal The proposed analysis correctly predicts that, in NPI-contexts, there is a subject/non-subject asymmetry in that goals in subject position do not receive NPI-construals because they are outside the c-command domain of the probe Neg. While goals in non-subject position get NPI-construals because they are in the c-command domain of the probe Neg. This section examines the distribution of EPI-construals. I observe that in Thai D-linked questions 'which' receive NPI-construals 'any' from being licensed by • the negative probe in a parallel fashion to bare variables that variables in object position can only be construed as an NPI. Variables in subject position, however, receive only a wh-construal. (i) ph6m may hen fararj khon nay ma; ta.-wa: praf^e.-t f^ay he neg see foreigner cl which come blame country Thai He don't see any foreigner blame Thailand. (ii) faran, cha.-t nay may t3.-wa: prat^e.-t t^ay foreigner country which neg blame country Thai Which country did not blame Thailand? 71 We shall see in the data below that, in EPI-contexts, goals in object position show the same pattern of how variables are construed as those in NPI-contexts. However, goals in subject position behave differently. In NPI-contexts, they are construed as a wh-construal, while there is no interpretation available for them in EPI-contexts. We see in (20) that the ungrammaticality results when goals occur in subject position in a yes-no construction. VARIABLE-SUBJECT: * WH-CONSTRUAL * EPI-CONSTRUAL (20) a. * [khray| ch3.p Nam may VARIABLE, -hhuman like Q[Poianty] [Does someone like Nam?] VARIABLE-SUBJECT: * WH-CONSTRUAL * EPI-CONSTRUAL b. * [?aray] Ion say Nam may VARIABLE.-human fall at Q[poiarity] [Did something fall down and hit Nam?] POSSESSOR SUBJECT: * WH-CONSTRUAL * EPI-CONSTRUAL C. * ma; kh5:rj [khray] haw say Nit may dog of VARIABLE, -hhuman bark at Q[Poiarity] [Did someone's dog bark at Nit?] Variable expressions in non-subject position, on the other hand, are able to be construed as EPI-construals in the same way as those of NPI-contexts. In (21), all variables in non-subject position—variable-object, variable-possessor or variable-indirect object—are in the domain of the Q[Poiarity]- Therefore, goals as variables match with the [polarity] feature of the probe Q[yes-no] and receive EPI-construals. 72 VARIABLE-OBJECT: * WH-CONSTRUAL V EPI-CONSTRUAL (21) a. Nit du.thu:k [khray] may insult VARIABLE, -hhuman Q[Poiarity] * (1) Who did Nit not insult? = (ii) Did Nit insult someone? b. Nam kVy c'3.p [?aray] may ever like VARIABLE.-human Q[poiarity] * (i) What did Nam ever like? = (ii) Did Nam ever like something? POSSESSOR SUBJECT: * WH-CONSTRUAL V EPI-CONSTRUAL c. Nit yi.m kh3.T) kh3.T) [khray] may borrow stuff of VARIABLE. +human Q[poiarity] * (i) Whose belongings did Nit not borrow? = (ii) Did Nit borrow someone's belongings? INDIRECT OBJECT: * WH-CONSTRUAL V EPI-CONSTRUAL d. Nit day hay rjdn [khray] may give money VARIABLE, -hhuman Q[polarity] * (i) To Whom did Nit not give the money? = (ii) Did Nit give the money to someone? The above confirms that there is a subject/non-subject split with respect to the EPI-construal. The next step in understanding what makes an EPI-construal possible is to identify the syntactic position Of Q[polarity]-73 3.3.1 The position of the yes-no question marker (Q[poiarityi) may I argue that the yes-no question marker may is an allomorph of negative may. Yes-no questions in Thai can be formed by appending a disjunctive phrase ri: may 'or not'23 immediately after the (affirmative) predicate (22a). Negation, however, can be dropped, leaving the disjunctive morpheme clause-finally24. The disjunctive marker ri: can also be dropped leaving the negation morpheme may realized with a rising tone may or a high tone may (22b). (22) a. khaw ch3.p Nam rt: may (ch3.p) he like or neg like Does he like Nam or not? b. khaw ch3.p Nam may he like Qfpohriiyj Does he like Nam? When the conjunction ri: 'or' is absent, as in (22b), negative may is instead marked by either a rising tone may (in a literary form) or a high tone may (in a colloquial speech) in order to be identified as a yes-no question. Thus, the may/may that occurs in yes-no questions is an allomorph of negative may. Another piece of evidence supporting the treatment of may and may as allomorphs comes from the fact that polar may cannot co-occur with negative may (23a). This implies that may and may are in complementary distribution, and occupy the same syntactic position. In (23b), negative may can co-occur with the disjuctive marker ri: 'or', in which case it is interpreted as a taq question. This type of yes-no question is described as an alternative question (A-not-A question) where two disjunctive alternatives, an affirmative proposition and the corresponding a negated propostion, are questioned. 24 In a colloquial speech, the vowel i: is lowered to 3:. (i) khaw ch3.p Nam rS: he like or Does he like Nam? 74 (23) a. * khaw may ch3.p Nam may he neg like Qfpoiarityj [Does he like Nam?] b. khaw may ch3.p Nam rS: he neg like or Doesn't he like Nam? If this analysis is correct, it implies that polar may in (23a) is generated in the same position as negative may. Despite the surface appearance of Q[poiarity] in postverbal position, Neg and Qlpoiarity] occupy the same position above the predicate (VP) but below the subject as illustrated in (24). NegP Neg Q[polarity] object With this structure, VP fronting is required in yes-no constructions. The prediction made by the syntactic position of an (overt) yes-no question marker may is that goals will have an EPI-construal in object position, but not in (matrix) subject position. The unavailability of the EPI-construal for the subject position follows from the c-command restriction, as illustrated in (25). (25) a. [n> *NPI-subiect flnfll [NegP [Neg may ] [yp V VNPI-object ] ] ] b. [jp *EPI-subiect rinfll [NegP [Q[Poiarity] may ] [Vp V VEPI-obiect ] ] ] A possible alternative analysis would be that a yes-no question has an adjoined coordinate structure, in particular a conjoined VP, as in (26). The surface form would be derived by eliding the whole VP in a negative conjunct. The disjunct 'or' is omitted and the negation may is marked by a high tone may instead of a falling tone. 75 (26) a. k'aw c'3p Nam rt: may [vpe] he like or not Does he like Nam him or does he not like Nam? b. khaw ch3.p Nam (rt:) may [Vpe] he like or not Does he like Nam him or does he not like Nam? (26b) would have the structure shown below: c'3.p Nam may like This structure predicts the same restriction on the Matching relation in that a goal in subject position will not have an EPI-construal because the conjoined phrase does not c-command the subject. It gives the same prediction as my analysis does since both negation may and may are generated in the Neg position. However, the conjoined VP analysis cannot provide the reason why the negation may changes its tone from falling to a high tone after the elision of the disjunction 'or'. In the next section, I discuss how the position of the probes correctly predicts the asymmetry between variable-subjects and variable-objects in an EPI environment where only goals in object position can match with the [polarity] feature of the Q[yes-no] probe. 76 3.3.2 (The absence of) EPI-construal in subject position We saw from the examples above that EPI-construals are not available in subject position. This can be explained by a syntactic matching condition that goals in subject position are not in the domain of the probe Q[poiarity] and thus cannot receive EPI-construals. This follows from the position of the yes-no question marker may which I claim is generated lower than the subject. This raises the question of why a wh-construal is not available. We would expect Q[Wh] which is generated higher than the subject and is the closest c-commanding probe available to provide a feature to be copied and yields a wh-construal. The sentences, however, rum out to be ungrammatical and there is no interpretation available. This data suggests that Q[Wh] and Q[poiarity] cannot co-occur. While Q[Wh] is [+Q, +wh], Q[poiarity] is [+Q, -wh]. They are both Q in that they type a clause (Cheng 1991) as a wh-question or a yes-no question. However, they cannot co-occur. When Q[Poiarity] is present, it blocks Q[wh] in C position. There can only be one Q operator—be it with either a wh- or a polarity-feature. Since Q[poiarity] (a yes-no question marker) is already present in (28), no Q[Wh] is allowed in C position. The goal in subject position, as a result, is left unspecified. This explains why there is no interpretation available for the goal. And it supports the central claim of the proposed analysis that wh-expressions are not inherently interrogative. Rather, they are pure variable expressions that need to be matched. And the matching relation between the probe and goal in Thai is syntactically conditioned. With this claim, the present analysis predicts that no wh-construals are available in an EPI-context. Just as expected, the data in (28), where Q[poiarity] is present in a multiple wh-question, is ungrammatical. The lack of the interpretation of the goal in subject position accounts for the ungrammaticality of (28). (28) * [khray] hen [?aray] may VARIABLE. +human see VARIABLE, —human Q[Poiarity] [who saw something?] 77 The question that arises is 'can goals in subject position ever get an EPI-construal in an EPI-context?' The proposed analysis predicts that they cannot get EPI-construals because they are not in the c-command domain of the Q[p0]arity] probe in order to be matched. Another way in which EPI-construals will become available for goals in subject position is by introducing another probe higher than subject position thus allowing them to copy its [polarity] feature and be construed as EPI-construals. The data below turns out as predicted. The morpheme mi: equivalent to 'exist' is introduced preceding the goal in subject position. This suggests that mi: behaves as an existential operator which is generated higher than IP and thus provides a feature for the goal to copy. The goal subject in (29a) now receives an EPI-construal yielding a well-formed sentence. In (29b), we see that the existential operator (as probe) c-commands both variable-subject and variable-object. The proposed analysis predicts that (29b) should be possible without a need to introduce another probe (i.e. Q[p0iarity])- Since there is no Q[poiarity], (29b), as expected, is interpreted as EPI construals in a declarative sentence. (29) a. mi:[khray] hen [?aray] may exist- VARIABLE, -hhuman see VARIABLE.-human Q[Poianty] Did someone see something? (lit. 'was there someone who saw something?') b. mi:[khray] hen [?aray] exist- VARIABLE, -hhuman see VARIABLE.-human Someone saw something. The data in (30) confirms that the existential operator is indeed generated outside the IP and this expression can only occur with subject in the same way as maymi:khray 'neg-exist-variable. +human' have an NPI-construal in subject position. 78 (30) * Nam hen mi: [?aray] see exist-VARIABLE.-human [Did Nam see something?] That the existential operator mi: is generated outside the IP is confirmed by the example in (30), which show that mi: cannot occur with a goal in object position25. m:i is an existential verb that takes an IP. (31) a. frp *EPI-subiect fulfil rN?gP fN?s mavl fyp V VEPI-obiect 111 b. [yp [v mi.irrpV EPI-subiect [Infl][NegP [Neg may][Vp vVEPI-obiect ]]]] The data below illustrates how subjects (32) receive EPI-construals. The existential verb mi: c-commands the subject and thus provides a feature to a goal to copy. This accounts for the availability of the EPI-construal for the subject position and it is consistent with the c-command restriction on the probe-goal relation. VARIABLE-SUBJECT: * WH-CONSTRUAL V EPI-CONSTRUAL (31) a. mi:[khray] ch3.p exist- VARIABLE, -hhuman like Did someone like Nam? Nam may Qlpolarity] mi:[?aray] Ion say Nam may exist- VARIABLE.-humanfall at Qlpolarity] Did something fall down and hit Nam? This is similar to what Cheng (1991) discusses about you 'have' in Mandarin Chinese. In order for indefinite subjects to be construed as an EPI, you 'have' must occur preceding them. 79 POSSESSOR SUBJECT: * WH-CONSTRUAL V EPI-CONSTRUAL c. mi; ma: (kh3:n) [khray] haw say Nit may exist dog (of) VARIABLE, -hhuman bark at Qfpoiarityj Did someone's dog bark at Nit? To sum up, because the Q[poiarity] probe is generated lower than the subject position, variable expressions in subject position cannot be filled by featural content. As we saw from the examples above, the unavailability of the EPI-construal results from the absence of a probe to provide a feature for the goal to copy in subject position. This reflects the fact that wh-expressions in Thai are pure variables that need to be filled in order to be interpreted. The variable-subjects, however, are able to be interpreted as EPIs when an overt existential operator is present (higher than the subject position). This confirms the c-command relation does hold in Thai. 3.3.3 EPI-construal in object position With variable expressions in object position (both direct and indirect objects), the analysis correctly predicts that only the EPI-construal is available. This is because Q[p0iarity] is the only c-commanding probe for the goal. There is no need for the overt existential operator mi: to be introduced. This is illustrated in the data below. VARIABLE-OBJECT: * WH-CONSTRUAL V EPI-CONSTRUAL (33) a. Nit du.-thu:k [khray] may insult VARIABLE, -hhuman Q[Poianiyj * (1) Who did Nit not insult? = (ii) Did Nit insult someone? b. Nam kVy ch3.p [?aray] may ever like VARIABLE.-human Q[Poiariiy] * (i) What did Nam ever like? = (ii) Did Nam ever like something? 80 POSSESSOR OBJECT: * WH-CONSTRUAL V EPI-CONSTRUAL c. Nit yi.-m kh3.Tj kh3.Tj [khray] may borrow stuff of VARIABLE, -hhuman Q[poiarity] & (i) Whose belongings did Nit not borrow? = (ii) Did Nit borrow someone's belongings? INDIRECT OBJECT: * WH-CONSTRUAL V EPI-CONSTRUAL d. Nit day hay rjdn [khray] may give money VARIABLE, -hhuman Qfpoianiyj (i) To Whom did Nit not give the money? = (ii) Did Nit give the money to someone? To summarize, I have shown that subject/non-subject asymmetries predictably arise from the interaction of the syntactic position of the probe and goal. The availability of NPI-and EPI-construals with variable expressions is syntactically restricted by the c-command relation and the locality condition between the probe and goal. Moreover, the goal needs to be filled by featural content in order to be interpreted. If no probe is available to provide a feature for the goal to copy, ungrammaticality results. 81 3.4 An asymmetry between wh-arguments and wh-adjuncts The previous two sections have established that the construal of variable expressions is syntactically conditioned. We have seen that while the (covert) Q[wh] has sentential scope, negative may and the yes-no operator may have VP scope. Because Q(wh] can scope over any argument, any argument can get a wh-construal. In contrast to this, because negative may and the yes-no operator may attach at the VP-level, only VP-internal arguments can get NPI- and EPI-construals. There is yet another way in which syntactic restrictions on variable expressions manifests itself, namely when variable expressions occur in adjunct position. In particular, adjunct rationale and manner expressions predictably fall within the domain of the (covert) Q[wh]> but outside the scope of negative may and the yes-no operator may . This is illustrated in (34) and (35), which show that only the wh-construal is available for rationale and manner adjuncts respectively. V WH-CONSTRUAL (34) a. Ra.Ti r5.*rjha.y cry Why did Ron cry? * NPI-CONSTRUAL b. * RJ.TI may n5:rjha:y [thammay] neg cry VARIABLE, reason [Ron did not cry for any reason.] [thammay] VARIABLE, reason * EPI-CONSTRUAL c. * Ra.n r5.T)ha.y [thammay] may cry VARIABLE, reason Q[polarity] [Did Ron cry for some reason?] V WH-CONSTRUAL (35) a. khaw khap rot [yannay] he drive car VARIABLE, way How did he drive a car? 82 * NPI-CONSTRUAL b. * khaw may khap rot [yarjnay]26 he neg drive car VARIABLE, way [He did not drive a car anyhow] * EPI-CONSTRUAL c. * khaw khap rot [yarjnay] may he drive car VARIABLE, way Q[Poiariiyj [Did he drive a car somehow?] The data in (34) and (35) raise the question of what blocks variable expressions in adjunct position from having an NPI- or EPI-construal. The absence of the NPI- or EPI-construal straightforwardly follows the probe-goal analysis, in particular from the c-command restriction on the probe-goal relation. The relevant structure of wh-adjuncts is illustrated in (36). Manner and reason wh-adverbials are adjoined to IP—outside the VP domain—and thus do not receive an NPI- or EPI-construals because they are not c-commanded by the Neg and Q[poiarity] probe, which are generated lower than wh-expressions. This also accounts for why adjuncts can have a wh-construal since they are in the c-command domain of the Q[Wh] probe. Although yanyay 'how'cannot occur in this context, it is possibly for it occur in the context below. However, it cannot be interpreted as an NPI. 'how' only receives a wh-construal because it is not c-commanded by the probe Neg. (i) khaw may day t3.p yarjrjay he neg past answer how = (i) How did he not reply? * (ii) He did not reply in any way. 83 (36) Neg Q[yes-no| adjunct27 V WH-CONSTRUAL * NPI-CONSTRUAL * EPI-CONSTRUAL VP V To sum up, the unavailability of NPI- and EPI-construals of adjuncts follow from the probe-goal analysis. Because they are not within the c-command domain of Neg and Q[poiarity], they then cannot receive NPI- and EPI-construals. This explains why there is an argument/adjunct split with respect to polarity construals. As we shall see, in the next section, subject/non-subject and complement/adjunct asymmetries no longer hold when we examine variable expressions in embedded clauses. This is due to the availability of a probe in a matrix clause and the side-effect of locality restrictions that holds between a probe and a goal in Thai. There is independent evidence in Thai that wh-adverbials are adjoined to IP. The sentence in (i) in English is ambiguous. (i) John didn't leave Mary because he loved her. = John loved Mary, so he didn't leave her. = John left Mary, but not because he loved her. (with the focus intonation on LOVED) (i) has two interpretations. The 'because' clause has scope over negation in the first reading suggesting that it is adjoined to IP, while negation has scope over the 'because' clause in the second reading suggesting it is adjoined to VP. Then, we would predict that in Thai only the first reading is allowed. The prediction is borned out (i) khaw may day thirj Nit pay phr5?wa: khaw rak rV' he neg past leave go because he love she He loved Nit, so he did not leave her. 84 3.5 Matrix/embedded scope asymmetries This section shows how the locality restriction applies to the probe-goal relation when the goal (i.e. variable expression) is in an embedded clause. I briefly review how matrix verbs impose selectional restrictions on embedded clauses (already discussed in Chapter 2). I then show how variable expressions in embedded clauses may be in a probe-goal relation with the closest operator. As we shall see, many of the NPI- and EPI-construals that are unavailable in matrix clauses, become possible in embedded clauses. In particular, variables in embedded subject position can get NPI- and EPI- construals under the scope of a matrix negative operator or matrix yes/no question operator. Similarly, variables in embedded adjunct position can get NPI-and EPI-construals under the scope of a matrix negative operator or matrix yes/no question operator. Thus the subject/non-subject asymmetry that holds of matrix clauses predictably does not hold of embedded clauses. And the argument/adjunct asymmetry that holds of matrix clauses predictably does not hold of embedded clauses. 3.5.1 Selectional restrictions Recall that Thai verbs select their complements in the same way as English verbs do. The verb 'ask' selects for a [+wh] complement, but not a [-wh] complement. This contrasts with 'know', which selects for both [+wh] and [-wh]-complements. And yet other verbs such as 'think' select exclusively for [-wh] complements. This is illustrated in (37). V EMBEDDED WH-QUESTION CONSTRUAL * MATRIX WH-QUESTION CONSTRUAL (37) a. khaw tha.Tn [Cp wa: [Cp khray ?a.n nans?:] he ask comp VARIABLE, -hhuman read book = (i) He asked who read the book. ^ (ii) Who did he ask read the book? 85 V EMBEDDED WH-QUESTION CONSTRUAL V MATRIX WH-QUESTION CONSTRUAL b. khaw ni: [cpwa: [Cpkhray ?a.n nags?:] he know comp VARIABLE, -hhuman read book = (i) He knew who read the book. = (ii) Who did he know read the book? * EMBEDDED WH-QUESTION CONSTRUAL V MATRIX WH-QUESTION CONSTRUAL c. khaw khit [cpwa: [Cpkhray he think comp VARIABLE, -hhuman ^ (i) He thought who read the book. = (ii) Who did he think read the book? In (37a), the wh-expression can only have an embedded wh-question construal, just as expected. The (covert) Q[wh] that is forced by the selectional requirement of the verb 'ask' provides a feature for the goal to copy. The wh-expression in (37b), on the other hand, has both an embedded wh-question and a matrix wh-question construal. The matrix wh-question reading arises when the complement of the verb 'know' is [-wh]. In (37c), the verb 'think', mentioned, selects only [-wh] complement CPs, so an embedded wh-question construal is thus not possible. The wh-expression receives a matrix wh-question construal from the presence of the (covert) Q[wh] probe in the matrix C instead. The examples above show how a wh-expression in an embedded clause can have either an embedded wh-scope or a matrix wh-scope depending on the selectional properties of the matrix verb. So far, we have only seen that embedded wh-subjects can only receive a wh-construal. Why are embedded NPI- and EPI-subject construals not available? The reason simply is that there is only one c-commanding probe, namely the covert Q[Whj. As we shall presently see that many of the NPI- and EPI-construals that are unavailable in a matrix clause, become possible in embedded clauses. In particular, variables in embedded subject position can get NPI- and EPI- construals under the scope of a matrix negative operator or matrix yes/no question operator. ?a.n nans?:] read book 86 3.5.2 Embedded NPI and EPI subjects This section shows how variable expressions in embedded clauses are in a probe-goal relation with the closest operator. The proposed analysis correctly predicts that NPI- and EPI-construals are possible for embedded subjects when the Neg and Q[poiarity] probes are introduced in a matrix clause. To see this, consider the examples below where the matrix verb 'think' occurs with Neg or Q[Poiarity]. A matrix negative operator or matrix yes/no question operator will (predictably) forces an NPI- or EPI-construal on embedded subjects because they are closer to the variables. Note that 'think' only selects for [-wh] complements. As a result, an embedded wh-construal should not be possible in this case. This is illustrated in (38) and (39). VARIABLE-SUBJECT: * WH-CONSTRUAL V EMBEDDED NPI-CONSTRUAL V EMBEDDED EPI-CONSTRUAL (38) a. khaw may khit wa: [khray] he neg think comp VARIABLE, -hhuman He did not think anyone liked Nam. b. khaw khit may wa: [khray] he think Quality] comp VARIABLE, -hhuman Did he think that someone liked Nam? (39) POSSESSOR SUBJECT: * WH-CONSTRUAL V EMBEDDED NPI-CONSTRUAL V EMBEDDED EPI-CONSTRUAL a. khaw may khit wa: nan kh5.T) [khray] ca? si: Nit day he neg think comp money of VARIABLE, -hhuman fut buy He did not think that anyone's money could buy Nit. b. khaw khit may wa: nan kh5.T) [khray] ca? si: Nit day he think Q[poinrity] comp money of VARIABLE, -hhuman fut buy Did he think that someone's money could buy Nit? ch3.p Nam like ch3.p Nam like 87 The above examples confirm that NPI- and EPI-construals are indeed available for variable expressions in embedded subject position. We have already seen that in embedded clauses, where two probes are available and both c-commands the goal, the closest probe is the one that enters into the probe-goal relation, consistent with the locality restriction. The relevant structures for a verb such as 'think' are given in (40). [CPtc Q[wh]] [IP Subject [NegP [Neg may ][Vp think [rp VNPI-subiect [VP V Object ] ] ] ] ] [CP tc Q[wh]][lP Subject [NegP [Neg Q|p0i.rfiy |][yp thinkfrp VEPI-subiectryp V Object ] ] ] In contrast to the verb 'think', we predict the opposite for the verb 'ask', namely only wh-construal are possible for variable expressions in an embedded subject position. Why is this so? Since 'ask' selects exclusively for [+wh] complements, Q[wh] is forced (due to the selectional restriction) to be closer to the variable in subject position than the Neg and Q[poiarity ] probe in the matrix clause, as illustrated in (41). [rp Subject [NegP [Neg may ][Vp ask [CP [C Qrwh]] [rp VWH-subiect [Vp V Object ] ] ] ] ] [jp Subject [NegPuN Cg Qlpolarity | ] [VP ask [CP tC Q[wh]] [IP V WH-subiect [Vp V Object ] ] ] The data below supports this prediction, and show the interpretation of variable expressions in Thai is constrained by the locality condition. VARIABLE-SUBJECT: V WH-CONSTRUAL * EMBEDDED NPI-CONSTRUAL * EMBEDDED EPI-CONSTRUAL (42) a. khaw may day tha.Tn wa: [khray] ch3.p Nam (40) (41) he neg ask comp VARIABLE, -hhuman like He did not ask who liked Nam. b. k 'aw day t'a.m may wa: [khray] Nam he ask Qfpomiiity] comp VARIABLE, -fhuman like Did he ask who liked Nam? 88 (43) POSSESSOR SUBJECT: V WH-CONSTRUAL * EMBEDDED NPI-CONSTRUAL * EMBEDDED EPI-CONSTRUAL a khaw may day tha.m wa: nan kb5.Tj [khray] si: Nit day he neg ask comp money of VARIABLE, -hhuman buy He did not ask whose money could buy Nit. b khaw day tha.Tn may wa: nan kha.T) [khray] si: Nit day he ask Q[Poiarity] comp money of VARIABLE, -hhuman buy Did he ask whose money could buy Nit? To summarize, we see that the matching restriction on the probe-goal relation, on the one hand, accounts for the subject/non-subject asymmetry that holds of matrix clauses, and on the other hand accounts for the absence of such an asymmetry in embedded clauses. Variables in an embedded suject position can get NPI- and EPI-construals under the scope of a matrix negative operator or a matrix yes/no question operator. 3.5.3 Embedded NPI and EPI adjuncts The main question, for this section, is what blocks wh-adjuncts from polarity licensing. Why can only arguments be polarity items? It is not that adjuncts can never be polarity items, but in matrix clauses, the only available probe that c-commands the adjuncts is Q[Wh]. This explains why adjuncts can only get wh-construals. The proposed analysis predicts that adjuncts can receive polarity construals in embedded clauses in the same way as arguments in embedded subject positions can. As illustrated in (44), wh-adjuncts in embedded clauses are matched with the feature of a matrix overt Neg or a matrix Q[poiarity] probe. Wh-adjuncts that are adjoined to IP are in the c-commanding domain of those two probes. They thus are and (must be) construed as NPIs and EPIs. Recall that we saw above (section 3.4) that, in matrix clauses, the Neg and Q[poiarity] operators are generated below wh-adjuncts, and this is why they are excluded from having NPI-and EPI-construals in those contexts. 89 (44) CP Neg VP Q|yes-no| wa: adjunct * WH-CONSTRUAL V NPI-CONSTRUAL V EPI-CONSTRUAL Neg Q[yes-no] VP V The data turns out as predicted: in embedded contexts, adjuncts can have NPI- and EPI-construals. * WH-CONSTRUAL V NPI-CONSTRUAL (45) a. khaw may k"it wa: Nit he neg think comp hurt person-other VARIABLE, way He did not think that Nit could hurt the other people anyhow. thamra.y khon?i.n day [yannay] * WH-CONSTRUAL A! EPI-CONSTRUAL b. khaw khit may wa: Nit thamra.y khon?i.n day lyannay] he think Q[polanty] comp hurt person-other VARIABLE, way Did he think that Nit could hurt the other people somehow? 90 (46) a. khaw may kVy khit wa; Nit ca? ri;an pay [thammay] he neg ever think comp fut study go VARIABLE, reason He has never thought that Nit will study for any reason. b. khaw kha.y khit may wa; Nit ca? ri;an pay [thammay] he ever think Q[Poiarity] comp fut study go VARIABLE, reason Has he ever thought that Nit will study for some reason? The proposed analysis correctly predicts that the availability of the NPI- and EPI-construal with variable expressions is syntactically conditioned. 3.6 Extending the analysis: comparatives, modals, if-clauses Wh-expressions can have NPI- and EPI-construals when they occur in other contexts such as comparatives, modals and if-clauses. They show the same pattern with respect to a subject/object asymmetry as those in negative and yes-no question contexts. The proposed analysis can thus be extended to account for variable expressions in such contexts. In the presence of the comparative kwa: 'more than', the variable expression in (47a) tray has an NPI-construal. Because the goal is in the domain of the probe, the goal copies presumably a [degree] feature from the probe. An NPI construal, on the other hand, is unavailable in (47b) because the goal in subject position is not c-commanded by the comparative probe, hence it is not in the domain of the probe. * WH-CONSTRUAL V NPI-CONSTRUAL (47) a. Nit ke;rj kwa: [khray] nay h3.Tj smart more than VARIABLE.+HUMAN in room & (i) Anyone is smarter than Nit in class. = (ii) Who is smarter than Nit in class? 91 V WH-CONSTRUAL * NPI-CONSTRUAL b. [khray] ke:rj kwa: Nit nay h3.T) VARIABLE.+HUMAN smart than in room ^ (i) Anyone is smarter than Nit in class. - (ii) Who is smarter than Nit in class? As predicted, in the presence of a modal tuiancd? 'should', tray has an EPI-construal in object position. In this case, the modal probe provides a [modal] feature for the goal to copy. The goal in subject position, on the other hand, cannot copy the [modal] feature from the probe introduced lower. V WH-CONSTRUAL * EPI-CONSTRUAL (48) a. Nit khu:an ca? ha: [khray] ma: cVay should will find VARIABLE.+HUMAN come help - (i) Nit should find someone to help her. * (ii) Who should help Nit? V WH-CONSTRUAL * EPI-CONSTRUAL b. [k'Vay] khu:an ca? ma: c^.-ay Nit VARIABLE. +HUMAN should will come help * (i) Someone should help Nit. = (ii) Who should help Nit? In the if-clause context, we correctly predict the absence of a subject/object asymmetry because the probe is introduced higher than the goal in both object and subject position, illustrated in (49). The conditional probe thd: 'if c-commands both the object (49a) and the subject (49b). This explains an availability of EPI-construals of the goal in both positions. 92 * WH-CONSTRUAL V EPI-CONSTRUAL (49) a. Ma?ri: du.t'u.'k [khray] k3: khu:an ca? kh5.-th6.t if Mary insult VARIABLE. +HUM'AN should will If Mary insults someone, (she) should apologize. apologize * WH-CONSTRUAL V EPI-CONSTRUAL b. tha: [khray] du.-fhu.'k kV k3: kh5:hay b3:k if VARIABLE. +HUMAN insult If someone insults you, (you) can tell (me). you ask give tell The above data shows that NPI- and EPI-construals in comparatives kwa:, modals ifuian ca2 and conditional clauses headed by t''d: are also captured by the proposed analysis. 3.7 Summary In this chapter, I argued that the elements which are construed as wh-expressions in wh-contexts do not have inherent interrogative force in Thai. I analyzed these elements as variable expressions whose interpretation was structurally determined by the probe-goal relation. As underspecified goals under the domain of a Q[wll] operator, the [wh] feature of the probe Q is copied onto the goal, yielding as a wh-construal. Under the domain of negation, the goal matches the [Neg] feature, functioning like a Negative Polarity Item (NPIs) equivalent to 'any'. And as underspecified goals under the domain of a yes-no question marker, these variable expressions behave like Existential Polarity Items (EPIs) equivalent to 'some'. Finally, we have seen that the probe-goal relation is subject to a c-command restriction (the probe must c-command the goal) and to a locality restriction (the goal matches with the closest probe). The proposed analysis correctly predicts the presence of subject/non-subject asymmetries, as well as complement/adjunct asymmetries in matrix clauses. It also captures the fact that such asymmetries with respect to NPI- and EPI-construals only hold in matrix clauses. Thus, while NPI- and EPI-construals are unavailable with subject and adjunct in matrix clauses, they are available in embedded clauses. This is because a matrix negative probe or a matrix yes-no question probe c-commands the embedded goal. The availability of NPI- and EPI-construals in 93 embedded clauses is a side-effect of the locality condition that requires that the closest commanding probe is the one that enters into the probe-goal relation. 94 Chapter 4 Wh-expressions at the left edge of the clause: contrastive wh-clefts 4.1 Introduction Consider the wh-questions in (1), which all involve the subject of an embedded clause getting a wh-construal. In (la), we see that the variable expression if ray appears in-situ. As discussed in the previous two chapters, in the absence of any other overt operator, a variable expression such a if ray is under the scope of a (covert) Q[wh] operator in the matrix C. As argued for in Chapter 2, this operator-variable relation is an instance of the probe-goal relation. Crucially, the Matching relation between the probe and the goal does not induce Move. With this in mind, consider the examples in (lb) and (lc), where the variable expression appears to have undergone movement: in (lb) to an intermediate position (called "partial wh-movement"), and in (lc) to a clause-initial position (called "long-distance wh-movement"). THAI WH IN-SITU (1) a. khun khit wa: [ khray ] rak Nit you think comp VARIABLE. +HUMAN love Who do you think loves Nit? (lit. 'You think who loves Nit?') THAI APPARENT PARTIAL WH-MOVEMENT b. khun tha.m wa: [khraythi:] Nam khit you ask comp VARIABLE. +HUMAN det think You asked who Nam thinks loves Nit. THAI APPARENT LONG-DISTANCE WH-MOVEMENT c. [ khray thi: ] khun tha.TTi wa: Nam khit wa; rak Nit VARIABLE.+HUMAN def you ask comp think comp love Who did you say that Nam thinks loves Nit? At first glance, taken together, the examples in (la-c) suggest that, in addition to wh in-situ, there is also partial and long-distance wh-movement in Thai. This apparent three-way split might seem unsurprising, given that there is independent cross-linguistic evidence for the wa: rak Nit comp love 95 distinction between wh in-situ, partial wh-movement and long-distance wh-movement. For example, languages such as Mandarain exhibit wh in-situ, in that wh-words remain in their base-generated position, (2a). In contrast to this, other languages such as German permit wh-words to move to an intermediate position, as in (2b): this is called partial wh-movement. And finally, in languages like English, wh-words move to a clause-initial position, as in (2c): this is called long-distance wh-movement. MANDARIN WH IN-SITU (2) a. Huangrong xiangxin Guojing believe What did Huangrong believe that Qiaofeng bought? mai-le [shenme] buy-asp what (Cheng 1991:194) GERMAN PARTIAL WH-MOVEMENT b. Was meinst du [wen] (da[3) sie [+wh] think you who that she Who do you think that she really loves? wirklich liebt really loves (Muller & Sternefeld 1993: 486) ENGLISH LONG-DISTANCE WH-MOVEMENT c. [What] did you think (that) Joey bought? Despite the surface parallels between the Thai examples in (lb-c) and the corresponding examples of partial and long-distance wh-movement in (2b-c), in this chapter and the next, I will argue that all wh-expressions in Thai are in-situ. Crucially, apparent cases of partial and long-distance wh-movement involve the obligatory presence of the determiner thi: immediately following the wh-expression. This is illustrated in (3) and (4) for Thai "partial" and "long distance" wh-movement respectively. The determiner thi: must co-occur with the wh-expressions in apparent cases of partial and long-distance wh-movement, as in (3a) and (4a). Without the presence of thi:, the sentences become ungrammatical as shown in (3b) and (4b). (3) a. khun tha.m wa: [ khray thi: ] Nam khit wa: rak Nit you ask . comp VARIABLE.+HUMAN def think comp love You asked who Nam thinks loves Nit. 96 b. * khun tha.Tn wa: [ khray ] Nam khit wa: rak Nit you ask comp VARIABLE.+HUMAN think comp love [You asked who Nam thinks loves Nit.] (4) a. [ khray thi: ] khun tha.m wa: Nam khit wa: rak Nit VARIABLE. +HUMAN def you ask comp think comp love Who did you say that Nam thinks loves Nit? [ khray ] khun tha.Tn wa: Nam khit wa: rak Nit VARIABLE.+HUMANyou ask comp think comp love [Who did you say that Nam thinks loves Nit?] The remainder of this chapter is devoted to showing that thi: clauses such as those in (3a) and (4a) involve clefting. In particular, I argue that thi: clauses are contrastive wh-clefts, and as such are the reduced counterparts of contrastive pen clefts. Contrastive pen clefts have an overt nominalizer (id'on), an overt copula (pen), and an overt determiner (thi:). The clefted element may be a nominal expression, as in (5a), or a wh-expression, as in (5b). The reduced counterparts of contrastive clefts appear in (6). I adopt the convention of referring to reduced contrastive clefts such as (6a) as "reduced clefts", and reduced contrastive wh-clefts such as (6b) as "reduced wh-clefts". LONG-DISTANCE CONTRASTIVE CLEFT (5) a. Nik thT: chan khit wa: pen khon tham cam ta»:k def I think comp be nom cause plate break Nick was the one that I think broke a plate. LONG-DISTANCE CONTRASTIVE WH-CLEFT b. [k'Vay] thi: khun khit wa: pen khon tham cam ta?:k VARIABLE. +HUMANdef you thinkcomp be nom cause plate break Who was the one you think that broke a plate? 97 REDUCED LONG-DISTANCE (CONTRASTIVE) CLEFT (6) a. Nik thi: chan khit wa: tham ca.n ta?:k def I think comp cause plate break Nick was the one I think that broke a plate. REDUCED LONG-DISTANCE (CONTRASTIVE) WH-CLEFT b. [khray thi:] khun khit wa: tham ca.n ta*:k VARIABLE.+HUMAN def you think comp cause plate brake Who was the one you think that broke a plate? I argue that (5) and (6) have the same structure. On this view, the reduced counterparts in (6) lack an overt copula, and lack an overt nominalizer. A notable property of reduced wh-clefts is that they can only allow bare wh-expressions equivalent to English 'who' or 'what'. Discourse-linked wh-expresions equivalent to English 'which N', on the other hand, are infelicitous with contrastive wh-cleft and their reduced counterparts. This is illustrated in (7). We will return to this in chapter 5. (7) a. # [ khon nay thi: ] pen khon tham ca.n ta?:k cl which def be nom cause plate brake Which one was the one that broke a plate? b. # [khon nay thi:] tham ca.n ta?:k cl which def cause plate brake Which one was the one that broke a plate? The remainder of the chapter shows that apparent instances of partial and long-distance wh-movement are in fact reduced wh-clefts. Section 4.2 discusses the semantics of contrastive clefts with the overt copula pen. Section 4.3 looks at the syntax of contrastive pen clefts and shows that the cleft clause is a nominalized clause marked by ifon. Sections 4.4 and 4.5 show that the semantics and syntax of reduced wh-clefts have the same properties as contrastive wh-clefts. Section 4.6 closes the chapter by considering the consequences of analyzing apparent cases of wh-movement as reduced wh-clefts. 98 4.2 The semantics of contrastive clefts with the overt copula pen In Thai, contrastive clefts have the following semantic properties: (0 (ii) (iii) The cleft clause is associated with an existential presupposition. The cleftee is associated with a uniqueness presupposition. The cleftee is associated with contrastive focus. Consider how these properties are exemplified in (8). (8) [DPNikthi:] pen deft be [>,p khon tham cam ta*:k ] nom cause plate break Nick was the one that broke a plate. Here the cleft clause corresponds to nP, which as we shall see below is in fact a nominalized clause. This cleft clause has an existential presupposition in the sense that it presupposes the existence of someone who broke the plate. In (8), the cleftee is a DP, here Nik thi:. This DP has a uniqueness presupposition: in this example Nick is the unique individual that satisfies the existential presupposition introduced by the cleft clause. As we shall see below, the source of the uniqueness presupposition is the definite marker thi:. The cleftee, in addition to satisfying a uniqueness presupposition, is also contrastively focused, and so Nik is here contrasted with any other accessible discourse referent. Thus, (8) can be uttered in situation where Nik, and noone else, broke a plate. I will show that it is the pen copula that is the source of this contrastive focus. 4.2.1 Contrastive clefts have an existential presupposition One of the semantic properties of contrastive clefts in Thai is that that the cleft clause is associated with an existential presupposition. To illustrate this, the example (8) is repeated below. 99 (9) [DPNik thT; ] pen [nP khon tham ca.n ta*:k] def be nom cause plate break Nick was the one that broke a plate. —> Existential presupposition: x broke a plate One of the clearest ways of showing the presence of an existential presupposition is to create a context where such a presupposition is explicitly denied. In such contexts, we expect that contrastive clefts will be infelicitous. The following context is provided to exemplify the presence of an existential presupposition in Thai clefts. (10) SI maymi:khray tham ca.n ta?:k nobody cause plate break Since nobody broke a plate,... a. nan ma.-ykhwa.Tnwa: [RO.TI may day tham cam ta£:k ] that mean neg cause plate break .. .it follows that Ron did not brake a plate. b. # nan ma.ykVa.Tnwa: may chay [ Ran thi: pen khon tham cam ta*:k ] that mean neg yes def be nom cause plate break .. .it follows that it was not Ron that broke a plate. (10a) is felicitous because an existential presupposition is unexpected in a non-clefted clause. In contrast to this, (10b) is infelicitous because the cleft induces an existential presupposition due to the open variable in the cleft clause. As such, the contradiction between the absence of the existential presupposition and a cleft clause results in infelicity. * WH-CONSTRUAL V NPI-CONSTRUAL c. Nit ke.Tj kwa: [khray] nay h3.T) smart than VARIABLE.-t-HUMAN in room Nit is smarter than anyone in class. 100 4.2.2 Contrastive clefts have a uniqueness presupposition Another semantic property of contrastive clefts in Thai is that the cleftee is associated with a uniqueness presupposition. In a cleft focus, only the unique value can be substituted for the focus position x. (11) [DP Nik thi:] pen [nP khon tham cam ta»:k] def be nom cause plate break Nick was the one that broke a plate. —> Existential presupposition: x broke plate —> Uniqueness presupposition: x = only Nit To illustrate, I use the sentence in the context (11). In a situation containing a contextually given set of individuals, a unique individual has to be picked out from the set (12). INDIVIDUALS IN DOMAIN: {Nick, Nat, New} (12) CONTEXT: Nick broke a plate, but Nat and New didn't; i.e. only Nick broke a plate. a. x tham cam ta?:k cause plate break X broke a plate. b. alternative propositions {Nick broke a plate, Nat broke a plate, New broke a plate} focus value of the open proposition c. [DP Nik thi:] pen [xpkhon tham cam ta?:k] def be nom cause plate break Nick was the one that broke a plate. I use an additive particle test to show that Thai contrastive clefts are associated with a uniqueness presupposition. We expect that contrastive clefts should not be compatible with additive particles due to their uniqueness presupposition. In the focus position of a cleft, there is a presupposition that the proposition holds true of no other individual than the set referred to in the focus position. Shank (2004) claims that in a focus construction, when a speaker does 101 not aim to contradict another speaker by asserting an alternative proposition, but merely wants to add more information, the speaker can do so by adding an additive particle such as 'also' or 'too' as long as the alternative set of individuals is taken as a true proposition in the given context. (13) INDIVIDUALS IN DOMAIN: {Nick, Nat, New} CONTEXT: Nick and Nat each broke a plate, but New didn't, a. # Nik thi: pen k''on tham ca.n ta>:k def be nom cause plate break Nat k3: dii:ay thi: pen khon tham ca.n ta?:k also together def be nom cause plate break Nick was the one that broke a plate. Nat was also the one that broke a plate. b. # Nik ti pen khon tham ca.n ta>:k de/ be nom cause plate break la?? Nat k3: tham mi.an kan and also do same each Nick was the one that broke a plate and so did Nat. (13) however demonstrates that only a unique value is acceptable in a cleft focus construction. When an additive particle is inserted, the sentences become infelicitous as seen in (13a) and (13b). This is what we predict if Thai contrastive clefts are indeed associated with a uniqueness presupposition. (13) supports this claim and also suggests that focus and cleft focus differ in this respect. In a cleft focus, there is a presupposition that the proposition holds true of no other individual(s) than the set referred to in the focus position. And in Thai, the uniqueness presupposition is triggered by the morpheme thi: which I argue in the next section introduces a definite interpretation on the NP it associated with. And it is this morpheme thi: that imposes a uniqueuess requirement on the cleftee. 4.2.3 Contrastive clefts have a contrastive focus Contrastive pen clefts encode a contrastive focus reading which comes from the definite marker thi: that introduces a definite reading on the NP that it marks in a nominal predicational 102 construction. I begin this section by introducing thi: as a definite maker19. thi: functions as a definite marker in characterizational copula sentences (Kuno and Wongkhomthong 1981). When t''i: is acting as a definite marker, it is only restricted to characterizational pen copula sentences which take a nominalized clause. The example in (15) illustrates the relevant context. (15) CONTEXT: B is talking to A about Nik, but there are two Nicks in their high school who they both know: one is a university lecturer, and the other is a high school drop out. A is not sure if they are talking about the same Nik, so he asks B to identify which Nik is under discussion. A: khun ' p'urt thi.Tj [Nik nay] you talk about VARIABLE Which Nick are you talking about? (lit. You are talking about Nick which?) Bl: Nik thi: pen ?acam def be teacher The Nick who is a university lecturer (as opposed to the Nick who is a high school drop out) 19 This departs from the traditional view, which treats as a complementizer introducing complement and relative clauses (cf. Hoonchamlong 1991, Visonyanggoon 2000, Wannapaiboon 2001). I also treat this morpheme as a complementizer in relative clauses in the next chapter. There are two complementizers in Thai that can be used to introduce complement clauses: fi: and wa:. While ti: introduces infinitive clauses, wa: introduces finite clauses. (i) Nit ta cay thi: ca2 pen khru: intend comp will be teacher Nit intends to be a teacher. (ii) Nit ta cay wa: ca2 pen khru: intend comp will be teacher Nit; intends that (shej) will be a teacher. As for relative clauses, they can be introduced by ("i: but not by wa:. (ii) a. n rjsi: (lem) thi: Nit ya:kday pha?: ma.k book (cl) comp want expensive very The book that Nit wanted to buy was very expensive, b. * n rjsi: (lem) wa: Nit ya:kday ph«: ma:k book (cl) comp want expensive very The book that Nit wanted to buy was very expensive. 103 B2: # Nik pen ?aca.n be teacher Nick is a university lecturer. In (15B1), a specific definite interpretation correlates with the presence of r*i:, this suggests that thi: should be treated as a definite marker. If thl: were acting as a complementizer in a relative clause, we would expect the sentence to be infelicitous because it is a sentence fragment, illustrated in the context below. A: khun p"u.1 tht.T) [Nik nay] you talk about VARIABLE Which Nick are you talking about? (lit. You are talking about Nick which?) Bl: # Nik thi: S5.TI maha.withaya;lay def teach university The Nick that teaches at a university lecturer. B2: p"u.-t t"i.Tj Nik th i: S5.TI maha.wit ayaday Talk about def teach university (I) am talking about the Nick that teaches at a university lecturer. In this context, the definite marker thi: is obligatory. I claim that it is the morpheme thi: that encodes a uniqueness presupposition. We saw in (15) the presence of the overt pen copula in nominal predicates. In Thai, predicational/characterizatinal contexts are distinct from equative/identificational contexts: the former requires the pen, the latter requires the copula /ft: (Kuno and Wongkhomthong 1981). This is illustrated in (16). (16) a. c''an pen khru: I be teacher I am a teacher. 104 b. * chan kht: I be I am a teacher According to Kuno and Wongkhomthong (1981), the contrast between (16a) and (16b) is attributable to the nature of characterizational sentences such that 'being a teacher' is one of the characteristics that the subject possesses. (16b) is ungrammatical due to the mismatch between a characterizational sentence and the use of the identificational copula ti: in the sentence. The above example shows the context where pen can occur and ti: cannot. The next example shows the opposite, namely a context where ti: can occur and pen cannot. (17) a. Nik kht: khon ti: chan rak be cl comp I love Nick is the one that I love. b. * Nik pen khon thi: chan rak be cl comp I love Nick is the one that I love. (17a) is an identificational sentence such that the speaker identifies the person who she loves. This is compatible with the identificational nature of ti:. (17b), on the other hand, produces a conflict by the use of characterizational copula pen in an identificational context, and hence the sentence is ungrammatical. Contrastive clefts are only compatible with the characterizational copula pen, as shown in (18). When the identificational copula ti: is used, the sentence is ill-formed and does not gave a contrastive focus interpretation. We see that there is a correlation between the presence of the pen copula and the availability of a contrastive focus reading. (18) CONTEXT: A mother is furious to know that her expensive china was broken into pieces. She looks fiercely at Nick, Nat and New. Nat suddenly bursts out and utters (18a) that Nick is the one who did it, NOT him. khru: teacher 105 a. pen AS A CHARACTERIZATION AL COPULA Nik thi: pen khon tham def be nom do Nick was the one who did it. b. kf'i: AS AN IDENTIFICATIONAL COPULA *Nik thi: khi: khon t"am def be nom do [Nick was the one who did it.] To summarize, we have seen that contrastive pen clefts have the following three semantic properties: an existential presupposition (due to the open variable in the cleft clause), a uniqueness presupposition (due to the definite marker thi:), and a contrastive focus (due to the characterizational copula pen). I now consider the syntactic properties of contrastive clefts. 4.3 The syntax of contrastive clefts with overt pen Contrastive pen clefts have the following syntactic properties: (i) The cleftee is in the subject position of the pen copula. (ii) The cleft is a nominalized clause with the nominalizer if on. (iii) The morpheme if'on is a subject oriented nominalizer. The example in (19) illustrates a contrastive cleft where the cleftee is introduced by the definite marker fi:. The cleft is a nominalized clause headed by the nominalizer if on. This nominalized clause is coindexed with its cleftee linked by the characterizational copula pen. (19) [Nik thi:] pen. [NornP khon tham can ta*:k] def be nom cause plate break Nick was the one that broke a plate. The structure is shown in (20). The characterizational pen copula takes a nominal predicate (here a nominalized clause) which in turn combines with a DP subject. The cleftee is a DP 106 marked by the definite morpheme thi:. The cleft part is a nominalized clause headed by the nominalizer ton. Assuming the nP as a nominalized phrase, the referential argument in the specifier of nP is co-indexed with a pro subject. (20) DP Nik fl eam V plate ta?:k break 4.3.1 ton as a subject oriented nominalizer In this section, I argue that the morpheme ton acts as a subject-oriented nominalizer in contrastive pen clefts. The nominalized clause behaves like a normal noun. To test whether ton is a real nominalizer, one can introduce the future maker ca? in the sentence. We expect that when the future marker is introduced, the sentence will be ill-formed because the nominalizer and the verb cannot be separated. This is confirmed by the data in (21). (21) a. khon cam nom employ an employer b. * k'on ca? cam nom fut employ [a future employer] 107 In additon to occurring pre-verbally, if on also occurs post-verbally, as in (22a), in which case it functions as an indefinite pronoun rather than a nominalizer. The introduction of a future marker in such contexts is well-formed, (22b). (22) a. pro ca:rj khon employ one & (i) an employee = (ii) (pro) employs someone. b. pro ca? ca:rj kh6n fut employ one (pro) will employ someone. We predict that if a nominalized clause like if on cd:rj in (23a) indeed behaves like a noun phrase, it should have the same distribution as ordinary nominal expressions. For example, it can be modified (23b), quantified (23c) or pluralized by adding a numeral and classifier (23d). (23) a. c'an pen [Von ca.-n] I be nom employ I am an employer. b. [khon ca.Tj] cay ray nom employ heart cruel The employer is cruel. c. [Von ca.-n] ba.Tj k'on cay ra.y nom employ some cl heart cruel Some employers are cruel. d. rja.n ni: mi: [khon ca.-n] prama.-n ha: khon job dem exist mon employ about five cl In this project, there are about 5 employers. 108 A nominalizer the morpheme ton is subject-orientd29. When we try to nominalize the object with the nominalizer ton, the sentence becomes ungrammatical, as in (24a). An object can only be nominalized if it is first promoted to subject vial passivization, as in (24b). (24) a. khon ca.-rj khun nom employ you ca.Tj employ ca:n employ In a contrastive cleft, the cleft clause is nominalized by the nominalizer ton. In so much as it appears that the cleft clause in contrastive pen clefts is nominalized, only the subject argument is available. Morevoer, the nominalizer ton can only nominalize the subject of a transitive verb, as shown in (25-28). The (b) examples of (26-28) show that the nominalizer nak is used for intransitive. (25) a. khon kep nan nom collect money a bill collector/ a waiter b. khon taz.Tj ba.n nom decoratehouse an interior designer an employer b. * khon khun nom you [an employee] b. khon thu:k nom pass an employee However, we will see in the next chapter that it can also appear as a classifier, in which case it is not subject restriction interpretations. That is, when there is no subject restriction, we have a relative clause. 109 khon tham su:an nom do garden a gardener khon pu.ay nom sick a sicker (26) a. * k on wirj nom run [a runner] b nak wirj nom run a runner (27) a. * khon r5.Tj nom sing [a singer] b. nak r5:rj nom sing [a singer] (28) a. * khon da.Wa.Tj nom travel [a traveler] nak da.-ntVrj nom travel a traveler 110 Consistent with its status as a nominalized clause, we predict that no future marking is allowed in the cleft clause. The prediction is borne out. In (29a), when the future tense marker is introduced before contrastive pen copula, the copula is now associated with the future tense meaning 'will be'. The example in (29b), as expected, is ill-formed when the nominalizer and its verbs get separated by the future tense marker. (29) a. khaw t"i: ca? pen khon hen phT: he del fut be nom see ghost He will be the one who sees the ghost. b. * khaw thi: pen khon ca? hen phT: he det be nom fut ghost see [He is the one who will see the ghost.] To summarize, there is an asymmetry between subject and object when the kfon nominalizer is introduced; only subjects of transitive verbs can be nominalized. The next two sections explore the semantic and syntactic properties of contrastive wh-clefts and argue that the apparent cases of partial and long-distance wh-movement are in fact instances of contrastive wh-clefts. 4.4 The semantics of reduced contrastive wh-clefts This section shows that the semantics of reduced wh-clefts have the same properties as contrastive wh-clefts. I compare contrastive wh-clefts, the apparent cases of "partial" and "long-distance" wh-movement and wh in-situ. I show that there are semantic parallels between contrastive wh-clefts and the apparent cases of wh-movement, while wh in-situ lacks such properties. I propose that apparent cases of wh-movement involve no movement but are in fact reduced wh-clefts. In Thai, contrastive wh-clefts have the following semantic properties: (i) The cleft clause is associated with an existential presupposition. (ii) The cleftee is associated with a uniqueness presupposition. (iii) The cleftee is associated with contrastive focus. Ill Consider how these properties are exemplified in (30). (30) [Dpkhray ti:] pen [NomP khon tham cam ta»:k] VARIABLE.+HUMAN def be nom cause plate break Who was the one that broke a plate? As we saw the cleft clause is in fact a nominalized clause. This cleft clause has an existential presupposition in the sense that it presupposes the existence of someone who broke the plate but we do not know who that person was. In (30), the cleftee is a DP, here tray thi:. This DP has uniqueness presupposition: in this example 'who' seeks the unique individual that satisfies the existential presupposition introduced by the cleft clause. As before, the source of the uniqueness presupposition is the definite marker t''i:. The cleftee, in addition to satisfying a uniqueness presupposition, is also contrastively focused (due to the pen copula), and so the wh-expression is here contrasted with any other accessible discourse referent. Thus, (30) can be uttered in a situation where there is a presupposition that the proposition holds true of no other individual that the set referred to in the focus position. 4.4.1 Reduced contrastive wh-clefts have an existential presupposition I argue that apparent cases of wh-movement are reduced wh-clefts. I use a question-answer pair diagnostic to show that apparent cases of wh-movement are associated with an existential presupposition parallel to contrastive wh-clefts. Reduced wh-clefts require an existential presupposition of someone as an answer, the same way that contrastive wh-clefts do. Wh in-situ, on the other hand, is not associated with an existential presupposition. To illustrate this, consider a context where the answer to the wh-question is the empty set 'no one'. Such an answer is infelicitious with wh-clefts (31a-b), but felicitious with wh in-situ (31c). CONTRASTIVE WH-CLEFTS (31) a. Q: [khray] ti: pen khon tham cam ta?:k VARIABLE. -/-HUMAN def be nom cause plate break Who was the one that broke a plate? 112 A:#may mi: ni neg have No one REDUCED WH-CLEFTS b. Q: [k"ray] thi: tham ca.-n ta?:k VARIABLE.+HUMAN def cause plate brake Who was the one that broke a plate? A:#may mi: ni neg have No one WH IN-SITU c. Q: [khray] VARIABLE. +HUMAN Who broke a plate? A: may mi: ni neg have No one 'No one' is infelicitous as an answer in (31a-b) because the existence presupposition requires that the set of propositions in the domain is not empty. In contrast to this, 'no one' is a felicitous answer to wh in-situ because an existential presupposition is not expected in (31c). 4.4.2 Reduced contrastive wh-clefts have a uniqueness presupposition Using the same question-answer test, one can show that reduced wh-clefts have the same semantic property as contrastive wh-clefts in that a speaker expects a unique entity to be singled out. Thus, while contrastive wh-clefts and their reduced counterpart do not permit more than one person to be singled out (32a-b), wh in-situ questions do (32c). tham ca.n ta?:k cause plate brake 113 CONTRASTIVE WH-CLEFTS (32) a. Q: [khray] t'1: pen khon tham ca.n ta?:k VARIABLE.+HUMAN def be nom cause plate break Who was the one that broke a plate? A: #Nik Nat Nick, Nat REDUCED WH-CLEFTS b. Q: [khray] thi: tham ca.n ta?:k VARIABLE. +HUMAN def cause plate brake Who was it that broke a plate? A:# Nik Nat Nick, Nat WH IN-SITU c. Q: [khray] tham ca.n ta*:k VARIABLE.+HUMAN cause plate brake Who broke a plate? A: Nik Nat Nick, Nat In this section, I look at another semantic property that contrastive wh-clefts share with their reduced counterparts: both are associated with a contrastive focus. 4.4.3 Reduced wh-clefts have a contrastive focus One of the clearest ways of showing the presence of a contrastive focus reading is to create a context where such a contrastive reading is explicitly accepted. In such contexts, we expect that answers that encode a contrastive focus reading are felicitous with contrastive wh-clefts and reduced wh-clefts, but are infelicitous with wh in-situ. The data below turn out just as expected. 114 We see that the apparent cases of wh-movement have the same inherent contrastive focus reading found in contrastive wh-clefts (33a-b). CONTRASTIVE WH-CLEFTS (33) a. Q: [khray] thi: pen khon tham cam tx:k VARIABLE.+HUMAN det be nom cause plate break Who was the one that broke a plate? A: Nik pen khon tham may chay Nat be nom cause neg yes Nick was the one who did it and Nat wasn't the one who did it. REDUCED WH-CLEFTS b. Q: [khray] thi: tham cam ta»:k VARIABLE. +HUMAN det cause plate brake Who was the one that broke a plate? A: Nik pen khon t''am may chay Nat be nom cause neg yes Nick was the one who did it and Nat wasn't the one who did it. WH IN-SITU c. Q: [k'Vay] tham cam ta?:k VARIABLE. +HUMAN cause plate brake Who broke a plate? A: # Nik pen khon tham may chay Nat be nom cause neg yes Nick was the one who did it and Nat wasn't the one who did it. We thus conclude that Thai does not exhibit "partial" and "long-distance" wh-movement, though superficially it looks as if it does. The apparent instances of wh-movement are attributable to contrastive wh-clefts. The section shows that reducted wh-clefts have the same semantic properties as contrastive wh-clefts, as summarized (34). 115 (34) Semantic properties Contrastive wh-clefts Reduced wh-clefts i. Existential presuppostition V V ii. Uniqueness presupposition V V iii. Contrastive focus V This section has focused on the semantic properties of contrastive wh-clefts. I now consider the syntactic properties of wh-clefts and show that reduced wh-clefts have the same properties as wh-clefts. 4.5 The syntax of reduced contrastive wh-clefts The syntactic structure of a contrastive wh-cleft is shown in (35). The cleftee is a wh-expression marked by the definite marker fi:. The cleft is a nominalized clause headed by the nominalizer if on. This nominalized clause is coindexed with its cleft linked by the characterizational copula pen. The wh-expression matches in feature with the c-commanding probe Q[wh] base-generated in C position. Hence, wh-expression in subject position of pen copula is construed as wh-interrogative satisfying probe-goal relation. 116 (35) CP be n vP plate tee:k break Contrastive wh-clefts have the following syntactic properties: a nominalized clause that gives rise to an existential presupposition, thl: as a definite marker that gives rise to a uniqueness presupposition and a characterizational pen copula that gives rise to contrastive focus. Their reduced counterparts, however, lack an overt copula, and lack an overt nominalizer. Only the definite marker thi: is present in reduced wh-clefts. Reduced contrastive wh-clefts have syntactic properties that parallel those of contrastive clefts: (i) The wh-expression as the cleftee is in the subject position of an abstract predicational copula (the covert counterpart of pen). (ii) The cleft is a nominalized clause with an abstract nominalizer (the covert counterpart of ton). (iii) The nominalizer is a subject-oriented nominalizer. The question that naturally arises is 'how are reduced wh-clefts associated with an existential presupposition and a contrastive focus reading when the nominalizer ton and pen copula are absent?' I repeat the examples below. 117 CONTRASTIVE WH-CLEFTS (36) a. [khray] t"i: pen khon tham ca.n ta?:k VARIABLE.+HUMAN def be nom cause plate break Who was the one that broke a plate? REDUCED WH-CLEFTS .h i .hj. Ji b. [knray] fi: fam ca.n ta>:k VARIABLE. +HUMAN def cause plate break Who was the one that broke a plate? Even though the definite marker t''i: induces a uniqueness presupposition, the uniqueness presupposition implies existential presupposition in the sense that the existence of someone must be presupposed before a unique individual is picked out. We can say that the presence of thi: in reduced wh-clefts is the source for both existential and uniqueness presupposition. For contrastive focus, I mentioned earlier that a notable property of reduced wh-clefts is that they can only allow 'bare' wh-expressions as opposed to 'D-linked' wh-expressions. We will see in chapter 5 that D-linked wh-expressions involve reduced identificational wh-clefts. This suggests that there might be a correlation between contrastive focus and 'bare' wh-expressions. The structure of the reduced wh-cleft is illustrated below. (37) CP cam I plate tseik break 118 To summarize, I have argued that reduced wh-clefts have the same semantic and syntactic properties as contrastive wh-clefts. If they have the same structure, then the wh-expression is base-generated at the left edge of the clause where it matches in feature with the probe Q[wh] in C. 4.6 Consequences of the contrastive wh-cleft analysis Considered together, the semantic and syntactic evidence reviewed above suggests that the apparent cases of wh-movement are derived from contrastive wh-clefts. I now consider three consequences of analyzing them as reduced wh-clefts: (i) There will be an asymmetry between the subject and the object. (ii) Wh-objects are only promoted to wh-subjects via passivization. (iii) Only [+human] subjects are allowed. 4.6.1 An asymmetry between wh-subjects and wh-objects In contrastive wh-clefts, only wh-subjects can occur as a cleftee. As discussed in section 4.3.1, the nominalizer ton is subject oriented, as shown by the contrast between (38a) and (38b). If the apparent cases of wh-movement derive from contrastive wh-clefts, we predict the presence of a subject/object asymmetry in that only wh-subjects can occur as cleftees. CONTRASTIVE WH-CLEFTS (38) a. [khray] thi: pen khon ca.Tj Britney VARIABLE.+HUMAN def be nom employ Who was the one that hired Britney? b. * [khray] t'l: pen khon Britney ca.T) VARIABLE. + HUMAN def be nom employ [Who was the one that Britney hired?] The prediction is confirmed by the data in (39). As expected, there is a subject/object asymmetry in reduced wh-clefts. The reduced counterpart is grammatical in (39a) when the copula and nominalizer are absent corresponding to the full form of the contrastive wh-cleft in 119 (38a). In contrast to this, (39b) is ill-formed, the ungrammatically arises due to the subject restriction on the nominalizer ifon. This is consistent with the claim that apparent instances of wh-movement are derived from reduced contrastive wh-clefts. REDUCED WH-CLEFTS ; (39) a. [khray thi;] ca.Tj Britney VARIABLE. +HUMAN def employ Who was the one that hired Britney? b. * [khray thi:] Britney ca.Tj VARIABLE. +HUMAN def employ [Who was the one that Britney hired?] 4.6.2 Wh-object can occur as a cleftee with the presence of passive markers The contrastive wh-cleft analysis correctly predicts that in reduced wh-clefts, wh-objects can occur as a cleftee when the object in the cleft clause is passivized before being nominalized. We see in the data below that wh-objects can occur as the cleftee if they are promoted to subject by one of the passive makers do:n, day or thu:k. (40) a. [khray] t'l: do.-n thamth6.t m?:awa.Tini: VARIABLE. +HUMAN def pass punish yesterday Who was the one that was punished yesterday?-b. [khray] t'l: da.y rap khat 11:ak pay kha?:nkhan ?o:limpik VARIABLE.+HUMAN def pass receive choose choose go contest Olympic Who was the one that was chosen to go to the Olympic? c. [khray] thi: thu:k l?:ak hay rap f\m ka.n siksa: VARIABLE.+HUMAN def pass choose give receive scholarship nom study Who was the one that was chosen for a scholarship? 120 4.6.3 Only [+human] subjects are allowed Another consequence of analyzing apparent cases of wh-movement as reduced wh-clefts is that [-human] wh-expressions are expected to not occur as cleftees in reduced wh-clefts. As we saw earlier, the nominalizer t on is subject oriented, and more specifically is restricted to [+human] entities. Hence, wh-expressions with a [-human] feature equivalent to English 'what' are not compatible with the nominalizer t on. This explains why the full form on wh-clefts in (41a) is ungrammatical. By hypothesis, (41b) is a reduced wh-clefts of (41a), so we predict that the [+human] will hold, and it does. CONTRASTIVE WH-CLEFTS (41) a. * [?aray] thi: pen khon tham bam thalom VARIABLE.-HUMAN def be nom cause house collapse [What was the thing that caused the house to collapse?] REDUCED WH-CLEFTS b. * [?aray] thi: tham bam thalom VARIABLE.-HUMAN def cause house collapse [What was the thing that caused the house to collapse?] We have seen the following three restrictions hold of in reduced wh-clefts: only wh-subjects can occur as a cleftee (due to the inability of the nominalizer ton to nominalize the object), wh-objects can occur as a cleftee only when passive markers are present (the object can only nominalize through passivization), only [+human] subject can occur as cleftees (due to the incompatibility of the nominalizer ton with [-human] arguments). These restrictions fall out naturally from the contrastive wh-cleft analysis: the apparent cases of wh-movement are reduced wh-clefts and so they have the same structure as contrastive wh-clefts. In this chapter, I argued that despite superficial appearances, apparent cases of wh-movement are best analyzed as a kind of contrastive wh-cleft. I have shown that the semantic and syntactic properties also hold of reduced wh-clefts. 121 Chapter 5 Wh-expressions at the left edge of the clause: identificational wh-clefts 5.1 Introduction There are other apparent cases of partial and long-distance wh-movement that crucially allow only discourse-linked wh-expressions (equivalent to English 'which N') to occur at the left edge. I call these D-linked wh-expressions. With this in mind, consider the ungrammatical example in (la), in which the wh-expression appears to have undergone movement from its base-generated position. The ungrammaticality is attributable to the absence of the morpheme thi:, which is required to co-occur with the wh-expression. In contrast, (lb) is a grammatical example, in which thi: is present following the wh-expression in an intermediate position. A further grammatical example is given in (lc), with thi: present in a wh-expression in clause-initial position. THAI WH IN-SITU (1) a. * khun khit wa: [phu:cha.y khon nay] Nit ?a>.p ch3.p you think comp man cl VARIABLE hide like [Which man do you think Nit secretly has a crush on?] THAI APPARENT PARTIAL WH-MOVEMENT b. k"un khit wa: [phu:cha.y khon nay] t"l: Nit ?a?.p you think comp man cl VARIABLE comp hide Which man do you think Nit secretly has a crush on? THAI APPARENT LONG-DISTANCE WH-MOVEMENT c. [phu:cha.y khon nay] thi: khun khit wa: Nit ?aep man cl VARIABLE comp you think comp hide like Which man do you think Nit secretly has a crush on? Despite the surface appearance of "partial" and "long-distance" wh-movement, I argue that /z: clauses with D-linked wh-expressions such as those in (lb) and (lc) are identificational wh-122 clefts, and as such are the reduced counterparts of identificational ifi: clefts. Identificational ifi:. clefts have a head noun or a classifier, an overt copula (ifi:.), and an overt complementizer (thi:): the clefted element may be a nominal expression, as in (2a), or a wh-expression, as in (2b). The reduced counterparts of identificational clefts appear in (3). I adopt the same convention as in chapter 4 of referring to reduced identificational clefts such as (3a) as "reduced clefts", and reduced identificational wh-clefts such as (3b) as "reduced wh-clefts". IDENTIFICATIONAL CLEFT (2) a. Ron khi; khon thi: Nit ?a>.p ch3.p be cl comp hide like Ron is the one that Nit secretly has a crush on. IDENTIFICATIONAL WH-CLEFT b. [phu:cha.y k"on nay] khi: khon thi: Nit ?a?.p ch3.p man cl VARIABLE be cl comp hide like Which man is the one that Nit secretly has a crush on? REDUCED (IDENTIFICATIONAL) CLEFT (3) a. Ron t"i: Nit ?a?.p ch3.p comp hide like Ron is the one that Nit secretly has a crush on lit = (It is) Ron that Nit secretly has a crush on... REDUCED (IDENTIFICATIONAL) WH-CLEFT b. [phu;cha.y k"on nay] thi; Nit ?a>.p ch3.p man cl VARIABLE comp hide like Which man is the one that Nit secretly has a crush on? Note that (3a) is a relative clause: the Ron under the discussion is restricted to present in the scene. (3b) contains the discoursed-linked wh-expression, which contains a noun (phu:chay), a classifier (kl'on), and a variable expression (nay). This phrase is equivalent to the English 'which NP'. Discoursed-linked wh-expressions (called "D-linked") are questions that require answers restricted to the set of entities common to both the speaker and hearer (Pesetsky 1987). 123 That is, D-linking is a relation between a referent and a discourse. However, a which-NP expression in Thai is even more contextually restricted than its English counterpart. Thai D-linked wh-expressions require that the answer select from a deictically prominent set of entities in the context. With the deictic interpretation of D-linked wh-expressions in mind, I argue that (2) and (3) have the same structure. On this view, the reduced counterparts of (full) identificational wh-clefts lack an overt copula, and lack an overt classifier in the cleft clause. The remainder of this chapter shows that thi: clauses with D-linked wh-expressions are reduced wh-clefts. Section 5.2 discusses the semantics of identificational clefts with the overt copula ti:. Section 5.3 looks at the syntax of identificational clefts and shows that the cleft clause is a relative clause. In section 5.4 and 5.5,1 show that the semantics and syntax of reduced wh-clefts have the same properties as identificational wh-clefts. Recall that in the previous chapter, we saw that one of the properties that contrastive wh-clefts possess is a contrastive focus reading. I show, in this chapter, that identificational clefts do not have such a property. Identificational wh-clefts are compared to contrastive wh-clefts, and are shown to be associated with existential and uniqueness presuppositions, but not contrastive focus. The chapter ends by considering two predictions that fall out from the identificational wh-cleft analysis. Thai wh-intervention effects are in fact reduced wh-clefts. That is, wh-intervention effects derive from identificational wh-clefts. This is to satisfy the probe-goal relation which determines that variable expressions must be at the left edge of the clause in order to match with the Q[wh] probe, and hence are interpreted as wh-expressions. Another consequence of analyzing D-linked thi: clauses as reduced wh-clefts is that it reflects the morphological composition of Thai wh-argument locatives. D-linked wh-locatives in Thai are not bare wh-expressions akin to 'where'; instead, they are morphologicaly complex forms equivalent to 'which place'. 5.2 The semantics of identificational clefts with the overt copula ti:. Like the contrastive clefts discussed in the previous chapter, identificational clefts are associated with an existential presupposition and a uniqueness presupposition. As we shall see, the existential force correlates with the presence of the identificational ti: copula, while the uniqueness presupposition arises from the definite description in the cleft relative clause. Consider how these properties are exemplified in (4). 124 (4) [NP Ron] khi: [Xp khon thi: Nit ?a>.p ch3.p ] be cl comp hide like Ron is the one that Nit secretly has a crush on. Here, the cleft clause corresponds to XP, which as we shall see below is in fact a relative clause. This cleft clause contains the definite description that gives rises to the uniqueness presupposition. In (4), the cleftee is an NP, Ron. This NP correlates with the uniqueness presupposition: in this example Ron is the unique individual that satisfies the existential presupposition introduced by the identificational kfi: copula. As we shall see below, the identificational kfi: copula is the source of an existential presupposition in the sense that it presupposes the existence of someone who secretly has a crush on Nit. 5.2.1 Identificational clefts have an existential presupposition One of the semantic properties of identificational clefts in Thai is that the identificational ifi: copula is associated with an existential presupposition. To illustrate this, the example (4) is repeated below. (5) [NP Ron] khi:. [Xp khon thi: Nit ?a>.p ch3.p ] be cl comp hide like Ron is the one that Nit secretly has a crush on x. —» Existential presupposition: there is x such that Nit secretly has a crush on. I use the same test illustrated in the previous chapter to show the presence of an existential presupposition by creating a context where such a presupposition is explicitly denied. In such contexts, we expect that identificational clefts will be infelicitous. The following context is provided to exemplify the presence of an existential presupposition in Thai clefts (6b), which is clearly lacking in the non-clefted focus construction. (6) SI Nit may day ?a?.p ch3.p khray neg hide like VARIABLE.+HUMAN Since Nit did not secretly have a crush on anyone,... 125 a. nan ma.ykhwa.mwa: [Ron may chay khon thi: Nit ?a?.p ch3.p ] that mean neg yes cl comp hide like ... it follows that, it is not Ron that Nit secretly has a crush on. b. # nan ma.yk'Va.-mwa: may chay [ Ron kht: khon thi: Nit ?a».p ch3p] that mean neg yes be cl comp hide like ...it follows that Ron was not the one that Nit secretly has a crush on. (6a) is felicitous because an existential presupposition is not expected in a non-cleft clause. In contrast to this, (6b) is infelicitous. This is because the identificational ti: copula is the source of an existential presupposition. We see that such a presupposition is denied in (6b). The contradiction between the absence of the existential presupposition and the presence of the identificational ti: copula results in infelicity. 5.2.2 Identificational clefts have a uniqueness presupposition Another semantic property of identificational clefts in Thai is that a uniqueness presupposition arises from the definite description in the cleft relative clause. In a cleft focus, only the unique value of the item substituted for the focus value can be asserted. To illustrate, consider a situation containing a contextually given set of individuals, a unique individual from which has to be picked out, as in (7a). When the speaker tries to add an additive particle such as 'also' to assert one more value, the sentence becomes ungrammatical, as in (7b). - This ungrammaticality is predicted because the result contradicts with the uniqueness presupposition property contained in the cleft clause construction. INDIVIDUALS IN DOMAIN: {Ron, James, Ian} (7) a. CONTEXT: Nit secretly has a crush on Ron; Nit does not secretly have a crush on James; Nit does not secretly have a crush on Ian. [NPRon] k'i: [Xp k"on thi: Nit ?ahp ch3.p ] be cl comp hide like Ron is the one that Nit secretly has a crush on. 126 Uniqueness presupposition: only x such that Nit secretly has a crush on. CONTEXT: Nit secretly has a crush on Ron and James; Nit does not secretly have a crush on Ian. [Np Ron] khi: [XP khon thi: Nit ?a*.p ch3.p] be cl comp hide like k"i: [XP khon thi: Nit ?a».p cb3.p ] nurankan be cl comp hide //Ae same-each James k3; also Ron is the one that Nit secretly has a crush on and James is also the one that Nit secretly has a crush on. (7b) confirms the claim that only a unique individual can be asserted to be the focus value of open proposition in identificational clefts. 5.2.3 Identificational clefts do not have a contrastive focus (8) illustrates that identifictional clefts do not have a contrastive focus. (8) a. [NP Ron] khi: [Xp khon thi: Nit ?a».p ch3.p ] be cl comp hide like Ron is the one that Nit secretly has a crush on. —» Existential presupposition: there is x such that Nit secretly has a crush on. —» Uniqueness presupposition: only x such that Nit secretly has a crush on. b. * [Ron] k"i: [Xpkhon t'l: Nit ?a».p ch3.p ] may chay Bill be cl comp hide like neg yes Ron is the one that Nit secretly has a crush on, not Bill. In (8b), we see that there is no contrastive focus between 'Ron' and someone else. This illustrates that a cleftee may be unique without being contrastive. This is attributable to the presence of identificational ifi: copula Recall from chapter 4 that is the characterizational pen copula is the source of contrastive focus. Identificational ifi: copula is, however, associated 127 with an "identificational interpretation" which in turn correlates with an existential and uniqueness presupposition only. I repeat the contrastive cleft example from chapter 4 below. (9) [DpNikthi:] pen [xpkhon tham ca.n ta?:k] det be nom cause plate break Nit was the one that broke a plate. —> Existential presupposition: x broke plate —> Uniqueness presupposition: x = only Nick -> Contrastive focus: x = Nit and no one else To summarize, we have seen that identificational clefts have the following two semantic properties: an existential presupposition (due to the identificational ti: copula), a uniqueness presupposition (due to the definite description in the cleft relative clause). However, identificational clefts lack a contrastive focus reading. Therefore, these two types of clefts must be distinguished in Thai. The table in (10) summarizes the differences between these two clefts. (10) Semantic properties Contrastive clefts Identificational clefts (i) Existential presuppostition V (ii) Uniqueness presupposition V V (iii) Contrastive focus A/ X 5.3 The syntax of identificational clefts with overt ti:. I now consider the syntax of identificational clefts that contain an overt copula ti:. Identificational ti: clefts have the following syntactic properties: (i) The cleftee is in the subject position of the identificational ti: copula (ii) The cleft clause is a relative clause with a head noun and/or a classifier followed by the complementizer thi: The example in (11) illustrates an identificational cleft which involves a relative clause. The cleftee is the subject of the ti: copula. The cleft is a relative clause that contains a head noun and/or a classifier ton followed by the complementizer thi:. 128 (11) [Ron] khi: [DP pro [CP khon thi: Nit ?a*.p ch3.p ] ] be cl comp hide like Ron is the one that secretly has a crush on Nit. The structure of (11) is shown in (12). The head noun can be optional when the classifier is present. Assume that pro is a head noun in the cleft clause. The classifier agrees in [+human] feature with (pro) that is co-indexed with the cleftee (Ron). And the cleftee is linked by the identificational k!'i: copula. The head noun is in turn related to a null operator in Spec CP. The null operator does not involve A' movement. Cheng and Sybesma (1999) argue that classifiers have an individualizing-singularizing function similar to that of determiners in other languages, which enables them to pick out a single individual from what is described by an NP. I follow their agument and assume that the kl'on classifier in Thai is functioning like a determiner. ?a?.p ch3.p e( hide like The next subsections discuss the identificational lfi: copula, followed by discussion of the internal structure of the relative clauses. I show how Thai allows the head noun in relative clauses to be dropped, and also focus on the status of the classifier, and the status of fi: as a complementizer30. 30 In contrastive clefts thi: is analyzed as a definite marker, but in identificational clefts fi: is analyzed as a complementizer. 129 5.3.1 Identificational khi: copula In addition to the characterizational pen copula that is used in contrastive clefts, Thai also has an identificational /fi: copula. Kuno and Wongkhomthong (1981) argue for a semantic distinction between these two copulas: while pen is used for characterizational sentences , Jfi: is used for identificational sentences. One of the syntactic properties that the ifi: copula has is that the subject and the predicate can be switched. This is unsurprising, given that the subject and the predicate are identified as the same person, as in (13). (13a) and (13b) show no semantic distinction. (13) a. f'aksin khi: na.yok khon patcuban be prime minister cl present Thaksin is the present Prime Minister. b. na.yok khon patcuban khi: khun thaksin prime minister cl present be The present Prime Minister is Thaksin. This predicts that in identificational clefts, the cleft clause should be able to precede the cleftee (due to the nature of identificational tf'i: copula). This prediction is bome out, as in (14), in which Ron and the person that Nit secretly has a crush on are interpreted as one and the same person. This shows that an identificational cleft sentence makes use of identificational ifi: to assert that there is a particular entity that is identified by the definite description contained in the relative clause. (14) a. [Ron] khi: [Xp khon thi: Nit ?«fcp ch3.p ] be cl comp hide like Ron is the one that Nit secretly has a crush on. b. [xpkhon t'l: Nit ?«e.p ch3.p ] khi: [Ron] cl comp hide like be The one that Nit secretly has a crush on is Ron. 130 5.3.2 Cleftee is a relative clause In this section, I argue that in an identificational ti: cleft, the cleft is a relative clause. This differs from the contrastive pen cleft discussed in chapter 4 where it was shown that the cleft is a nominalized clause. I show that identificational clefts are associated with the following syntactic properties: (i) The cleft is in the subject position of the identificational ti: copula. (ii) In a cleft clause, a head noun is dropped. (iii) The morpheme ton is a classifier and a relative operator in a relative clause. (iv) The morpheme thi: is a complementizer in a relative clause. 5.3.2.1 pro as a head noun in relative clauses In the present analysis, the cleftee is in a relation with a relative clause which takes a pro head. In Thai relative clauses, the head noun and the classifier usually co-occur, as in (15a). When the head noun is overt, the classifier can be optional, as in (15b). The classifier, on the other hand, must be obligatory when the head noun is covert to allow recoverability of the head noun, as in (15c). (15d), on the other hand, is ungrammatical because both head noun and its classifier are not allowed to be dropped simultaneously in a relative clause. RELATIVE CLAUSE (15) [N head noun [ classifier ^ a. N cl b. N 0 c. 0 cl d. * 0 0 The [noun classifier] patterning in (15) is illustrated in the example (16). In (16a), the nominal 'dog' is the head noun. The classifier tua, when present, must agree with the head noun ma: that it occurs with. Here, the relevant feature is [-human]. In (16b), the classifier is optional when the head noun is present. In (16c), the classifier is present without the head noun. This can be generated in a discourse context in which the referent of the classifier is discourse 131 familiar or is established deictically. (16d) shows that Thai does not allow a head noun and classifier to drop simultaneously in a relative clause, even though it is a pro-drop language. (16) a. ma: tua fl: dog cl comp The dog that Nit bought... Nit si: buy ma:.. come ma: 0 ti: dog comp The dog that Nit bought Nit st: buy ma:.. come 0 tua fl: cl comp The dog that Nit bought Nit si: buy ma: . come d. * 0 0 ti: Nit si: ma: ... comp buy come [That Nit bought... ] Classifiers must agree with the head noun. The example (17) illustrates how the classifier agrees in features with the noun that it occurs with. In (17a), ton is a classifier for nouns that are [+human], and agrees in [+human] feature on the head noun 'man'. On the other hand, tua in (17b) is a classifier for nouns that are [-human], and it produces a conflict in feature agreement between the classifier and the head noun 'men'. Resulting in ungrammaticality. RELATIVE CLAUSE (17) a. p"u:cha.y khon ti: Nit ?a>p ch3.p man cl comp hide like The man that Nit secretly has a crush on... 132 b. * phu:cha.y tua thi: Nit ?a>.p ch3.p man cl comp hide like The man that Nit secretly has a crush on... Examples (18-19) illustrate how different nouns that share the same feature take the same classifier. The classifier if on in (18) agrees in [+human] feature with each of the following head nouns, while the classifier tua in (18) agrees in [-human] feature with the head nouns. 'a man' 'a teacher' 'a thief a dog' 'a doll' 'a robot' Recall that in the previous chapter, I analysed the morpheme if on as a nominalizer, and showed that it was subject-oriented. However, k!'on can also appear as a classifier, in which case it is not restricted to a subject interpretations. This means that when there is no subject restriction, we have a relative clause, which gives us an identificational cleft; however, when there is a subject restriction, we have a nominalized clause, which is used to construct contrastive clefts. 5.3.2.2 The morpheme fi: is a complementizer in a relative clause In Thai, there is never a overt wh-operator in relative clauses; however, a head noun and/or a classifier is obligatorily followed by fi:. Cross-linguistically, relative clauses are embedded inside a nominal expression which it modifies (Alexiadou et al, 2000). Examples of English relative clauses are given below. (18) k"on [+humanl a. pNi.'cVy^human] b. khrU.'[+lHiman] C. kham0.y[+human] (19) tua [-human] a. ma;[_hurnan] b. tukkata:r -human] C. hunyon[-human] 133 ENGLISH RELATIVE CLAUSE (20) a. * [The man [who that Hike]] b. [ The man [who 0 I like ]] c. [The man [0 that Hike]] d. [The man [0 0 Hike]] In (20a), the relative pronoun and complementizer are both overt, and hence the ungrammaticality of the sentence arises. This can be explained by the Doubly-Filled Comp filter (Sportiche 1992). This filter requires either the specifier or the head to be lexically filled in a given projection but not at the same time. In (20b), the relative pronoun 'who' is overt, and the complementizer is covert. In contrast to this, in (20c) the relative pronoun is covert, and the complementizer is overt. In (20d), we see that both relative pronoun and the complementizer can be absent. The restrictions on relative operator and complementizer patterning are illustrated in (21) for Thai relative clauses. Taken together, the examples in (21a-d) show that tHi: is an overt complementizer and it must be obligatorily present in a relative clause. (21) a. phu:cha.y khon 'thi: Nit ch3.p man cl comp like The man that Nit likes... * phu:cha.y khon 0 Nit ch3.p man cl like The man that Nit likes... p"u:cha.y 0 ti: Nit ch3.p man comp like The man that Nit likes... * p"u:cha.y 0 0 Nit ch3.p man like The man that Nit likes... 134 Thus, in an identificational cleft, the cleftee is the subject of an identificational ti: copula and the cleft is formed by a relative clause. 5.4 The semantics of reduced identificational wh-clefts This section shows that the semantics of reduced wh-clefts have the same properties as identificational wh-clefts, while wh in-situ does not. I argue that there are semantic parallels between identificational wh-clefts and thi: clauses with D-linked wh-expressions. Wh in-situ constructions, however, lack these properties. With this in mind, consider the example in (22). IDENTIFICATIONAL WH-CLEFT (22) [phu:cha.y k"on nay] khi: khon thi; Nit ?a>.p ch3.p man cl VARIABLE be cl comp hide like Which man is the one that Nit secretly has a crush on? As we have seen, the cleft clause is in fact a relative clause. It is the definite description in the cleft relative clause that gives rise to a uniqueness presupposition. The cleftee phu:chay ton nay 'which man' and the cleft clause are connected by the identificational ti: copula. The identificational ti: copula, in turn, is the source of an existential presupposition. However, identificational wh-clefts do not have a contrastive focus reading. The cleftee is a unique individual but need not be contrasted with any other accessible discourse referent. 5.4.1 Reduced identificational wh-clefts have an existential presupposition I argue that thi: clause with D-linked wh-expressions are reduced wh-clefts that lack an overt ti: copula. I use the same diagnostic as discussed in the previous chapter to show that thi: clauses with D-linked wh-expressions are associated with an existential presupposition parallel to identificational wh-clefts. Reduced wh-clefts require the existence of someone in the answer. In contrast, an answer to wh in-situ question is not restricted to the presence of an existential presupposition. 135 In identificational wh-clefts, the speaker presupposes the existence of the cleft that Nit likes x and asks the question to find out who is x such that Nit has a crush on x. When the answer is given as 'Nit has a crush on no one', the presupposition is explicitly denied. This results in an anomaly if 'no one' is given as an aswer to the identificational wh-clefts, as in (23a). The fact that the answer of 'no one' answer is infelicitious in (23b) entails that (23b) behaves parallel to the identificational wh-cleft in (23a). This follows if (23b) is a reduced form of (23a). In contrast to wh-clefts, wh in-situ is not associated with an existential presupposition (23c). IDENTIFICATIONAL WH-CLEFT (23) a. Q: [phu:cha.y khon nay] khi: khon thi: Nit ?a»p ch3.p man cl VARIABLE be cl comp hide like Which man is the one that Nit secretly has a crush on? A:#may mi; ni Neg have No one REDUCED WH-CLEFT b. Q: [phu:cha.-y khon nay thi:] Nit ?a>.p ch3.p man cl VARIABLE comp hide like Which man is the one that Nit secretly has a crush on? A:#may mi: ni Neg have No one WH IN-SITU c. Q: Nit ?a?.p ch3.p [phu:cha.y khon nay] hide like man cl which Which man does Nit secretly have a crush on? A: may mi: ni Neg have No one 136 5.4.2 Reduced identificational wh-clefts have a uniqueness presupposition Consider the data in (24). We found the answer to (24a) infelicitous when the two members are picked out from the answer set. One of the semantic properties that identificational clefts have is that they require an answer to be uniquely singled out. The cleftee must contain one and the only one member to satisfy a uniqueness presupposition. The peculiarity of the answer given in (24a) is due to a violation of the uniqueness requirement on the cleft-focus construction. Reduced wh-cleft behave the same way (24b). On the other hand, with wh in-situ questions, the answer is not required to be a unique individual. The question can be uttered in a situation in which the speaker does not expect Nit to like only one person, as in (24c). IDENTIFICATIONAL WH-CLEFT (24) a. Q: [p"u:cha.y khon nay] kh*: khon thi: Nit ?a?.p ch3.p man cl which be cl comp hide like Which man is the one that Nit secretly has a crush on? A: #ch3.p sa.Tj k''on Sdmpo.Tj lae? SdmcVy like two cl and She likes both: Sompong and Somchaay. REDUCED WH-CLEFT b. Q: [(phu:cha.y) khon nay] t'l: Nit ?a*.p ch3.p man cl VARIABLE comp hide like Which man is the one that Nit secretly has a crush on? A:#c''3.p S3.TJ khon Sompo.Tj lae? SdmcVy like two cl and She likes both: Sompong and Somchay. WH IN-SITU c. Q: Nit ?a?.p ch3.p [phu:cha.y khon nay] hide like man cl which Which man does Nit secretly have a crush on? 137 A: ch3p tharj SJ.TJ khon Somp^.Tj like two cl She likes both: Sompong and Somchay. lae? SdmcVy and So far, we have seen that identificational wh-clefts have the same semantic properities as contrastive wh-clefts, as discussed in chapter 4. The next section provides the same test (a question-answer pair) to exemplify the absence of a contrastive focus in identificational wh-clefts. Although contrastive focus is absent in identificational wh-clefts, it is clearly present in contrastive wh-clefts. 5.4.3 Reduced identificational wh-clefts lack a contrastive focus This section shows that reduced identificational wh-clefts lack contrastive focus. We find that the contrastive focus answer given to identificational wh-clefts, as in (25a-b) is infelicitious because the answer is more informative than what the question requires. The identificational wh-cleft is asking the hearer to pick out a single and unique member from the answer set, rather than asking for a person who is being contrasted with any other accessible discourse referent. Notice that the same answer is infelicitous with the wh in-situ construction as well. IDENTIFICATIONAL WH-CLEFT (25) a. Q: [phu:c"ay khon nay] khi: khon thi: Nit ?a>.p ch3.p man cl VARIABLE be cl comp hide like Which man is the one that Nit secretly has a crush on? A:#S6mpa;n may chay SdmcVy neg yes Sompong, not Somchaay. REDUCED WH-CLEFT b. Q: [p"u:c"a.y khon nay thi:] Nit ?a>.p man cl VARIABLE comp hide Which man is the one that Nit secretly has a crush on? ch3p like 138 A:#Srjmp3.Tj may chay SdmcVy neg yes Sompong, not Somchaay. WH IN-SITU c. Q: Nit ?a?.p ch3.p phu:cha.y khon nay hide like man cl VARIABLE Which man does Nit secretly have a crush on? A: #Sdmpa.Tj may chay SdmcVy neg yes Sompong, not Somchay. We see that the reduced wh-clefts have the same semantic properties as identificational wh-clefts: an existential presupposition (due to the infelicity of a negative answer to a wh-cleft), a uniqueness presupposition (due to the infelicity of the answer denoting more than one member in the answer to a wh-cleft), and the lack of a contrastive focus (due to the infelicity of a contrastive focus answer). Note that it is existential and uniqueness presuppositions (not a contrastive focus in this case) which distinguish wh-cleft constructions from wh in-situ. This is illustrated in a table below. (26) Semantic properties Identificational wh-clefts Reduced wh-clefts i. Existential presuppostition V V ii. Uniqueness presupposition V iii. Contrastive focus X X The next section considers the syntactic properties of identificational wh-clefts and shows that the reduced wh-clefts have the same syntactic properties as wh-clefts. 139 5.5 The syntax of reduced identificational wh-clefts I established above that identificational wh-clefts have the following syntactic properties: (i) The cleftee as a D-linked wh-expression is in the subject position of the identificational ti: copula (ii) The cleft clause is a relative clause with a head noun and/or a classifier followed by the complementizer t''i: IDENTIFICATIONAL WH-CLEFT (27) [p"u:c"a.y khon nay] khi: [pro [khon thi: Nit ?a?.p ch3.p ] ] man cl VARIABLE be cl comp hide like Which one is the one that Nit secretly has a crush on? The structure of (27) is illustrated in (28). The cleftee is in the subject position (Spec IP) of the ti: copula. Notice that the NP 'man' is moved to appear in the specifier position of DP preceding the classifier in D position. In the cleft clause, the head noun is covert as pro, while the classifier is overt. The classifier agrees in [+human] feature with the pro co-indexed with the cleftee (Ron), and is linked by the identificational ti: copula. The head noun is in turn related to a null operator in Spec CP. The null operator does not involve A' movement. 140 (28) CP c rp cl cp NP D C IP ?a?.p ch3.p ej hide like We see that the D-linked expression gets a wh-construal. As an underspecified goal, the [wh] feature on the qualified probe Q is copied onto it, yielding a wh-construal. The question that arises with respect to the cleft clause of (28) is 'why is it possible in reduced wh-clefts to drop both the pro subject and the classifier at the same time?' As discussed in the previous sections, identificational wh-clefts have the following properties: an existential force correlates with the presence of the identificational ifi: copula, while a uniqueness presupposition arises from the definite description in the cleft relative clause. In their reduced counterparts, we see that only the complementizer thi: is overt, as illustrated in (29). 141 (29) CP c rp cl cp NP D c IP ?33.p ch3.p ej hide like As discussed in section 5.3.2.2, a head noun and its classifier cannot be dropped in a normal relative clause. The structure in (29), however, indicates that the head noun and its classifier in the cleft relative clause are both dropped. What is happening is that they do spell out but in the cleftee via the D-linked wh-expression. And hence, there is no need to spell out twice. Recall that one of the syntactic properties that identificational wh-clefts have is that the cleft and the cleftee can be reversed. In reduced wh-clefts, the cleft clause 'that Nit secretly has a crush on' is not allowed to be in the subject position. The D-linked wh-expression must occur at the left edge. Why is this so? Note that this language allows a long-distance probe-goal relation. The goal, in a reversed position, can still match with the probe in C position because there is no other closer probe. The reason why the D-linked wh-expression has to be at the left edge may be because the left edge position is a structurally dedicated focus position in this language. 142 5.5.1 Evidence for a non-movement analysis : identificational wh-clefts/relative clause parallel This section provides more evidence that thl: clauses with D-linked wh-expressions are in fact reduced wh-clefts. Hoonchamlong (1991) and Wannapaiboon (2001) argue that Thai relative clauses do not involve A' movement (due to an absence of island effects). That is, the head nouns are base-generated outside the relative clause. In the relative clause, an empty operator is an A' binder, binding the null resumptive pronoun via A' binding. If the apparent cases of wh-movement with D-linked wh-expressions were to derive from identificational wh-clefts, we predict that they should show no island effects. (30a) is an example of a relative clause where the subject DP inside a complex NP island is being relativized. The head noun 'woman' and the optional resumptive pronoun inside the island are linked. If there were movement, an empty operator from the island would be moved out of the island. Since such movement is not allowed, we expect (30a) to be ungrammatical. However, (30a) is a well-formed sentence. Therefore, we can deduce that no movement of the empty operator occurs. A D-linked wh-expression parallel to (30a) is illustrated in (30b). We find no island effect in reduced wh-clefts, as expected. If there were movement of the D-linked wh-expression in (30b), we would expect ungrammaticality because no movement can occur across the island. However, (30b) is grammatical. Hence, there is no wh-movement. COMPLEX NP ISLAND (30) a. khaw ch3.p phu.yirj khon thi: khaw dayyin kha.wli; ma: wa: (tha:) kamlarj ca? ta?.Tjna:n he like woman cl comp he hear rumor come comp(she) prog will marry He likes the woman that he heard the rumor that (she) is going to get married. COMPLEX NP ISLAND b. phu.yin khon nay thi: khaw dayyin kha.wh: ma: wa: (tha:) kamlarj ca? ta*.Tjrja:n woman cl which comp he hear rumor come comp (she) prog will marry lit = Which woman is the one that he heard the rumor that (she) is going to get married? Consider now instances with adjunct islands. If the D-linked wh-expression were to extract from the adjunct island in (31), we would expect the sentences to be ungrammatical. However, 143 since (31b) is well-formed, we deduce that no wh-movement occurs. ADJUNCT ISLAND (31) a. khaw kamlarj n?kthin pNa.virj khon thi: famhay khaw ru:cak Nit phr5? tha: c\i:ay nas?nam he prog think of woman cl comp cause he know because she help introduce He is thinking of the woman who made him know Nit because she introduced (him to Nit). ADJUNCT ISLAND b. phu.yirj khon nay thi; thamhay khaw ruxak Nit phr5? tha: chu:ay na??nam woman cl which comp cause he know because she help introduce Which woman is the one that made him know Nit because she introduced him to Nit? To conclude, thi: clauses with D-linked wh-expressions are reduced identificational wh-clefts. They do not involve wh-movement; rather D-linked wh-expressions are base-generated in the subject position agreeing in [wh] feature with the Q[wh] probe. 5.6 Consequences of the identificational wh-cleft analysis This section considers two major consequences of analyzing thi: clauses with D-linked wh-expressions as reduced wh-clefts. First, there should be no asymmetry between wh-subjects and wh-objects. They are both predicted to occur in a cleftee position. Second, there should be no featural restrictions on D-linked wh-expressions. In particular, both [+human] and [-human] are predicted to occur in a cleftee position. 5.6.1 The lack of asymmetry between wh-subjects and wh-objects We see that in identificational wh-clefts, the cleft clause is formed by a relative clause. In the relative clause, the head nouns in subject or complement position can be relativized. Analyzing D-linked wh-expressions as reduced wh-clefts, we expect the D-linked wh-expressions in the cleftee position can be associated with a gap in either the subject or the object position of a relative clause. 144 With this in mind, consider (32a), where the D-linked wh-expression is construed as a wh-subject in a cleftee position, while in (32b), the D-linked wh-expression is construed as a wh-object. The cleftees are co-indexed with the relativized head nouns in both positions. The relativized head nouns, in turn, are co-indexed with the (optional) resumptive pronoun in subject and object position inside the relative clause, as in (33a-b). The reduced identificational wh-clefts with D-linked wh-expressions show the lack of a subject-object asymmetry. In (34a), the reduced D-linked wh-expression is construed as a wh-subject, and in (34b) it is construed as a wh-object. IDENTIFICATIONAL WH-CLEFT: WH-SUBJECT (32) a. [phu:chay khon nay] kl: khon thi: ?a?p ch3.p Nit man cl VARIABLE be cl comp hide like Which man is the one that secretly has a crush on Nit? IDENTIFICATIONAL WH-CLEFT: WH-OBJECT b. [phu:cha.y khon nay] k"i: khon fl: Nit ?a>p ch3p man cl VARIABLE be cl comp hide like Which man is the one that Nit secretly has a crush on? RELATIVE CLAUSE: RELATIVIZED SUBJECT (33) a. [p'Wa.y k"on ti: ?a>.p ch3p Nit]... man cl comp hide like The man that secretly has a crush on Nit... RELATIVE CLAUSE: RELATIVEZED OBJECT b. phu:cha.y khon ti: Nit ?a?p ch3.p ]... man cl comp hide like The man that Nit secretly has a crush on... REDUCED WH-CLEFT: WH-SUBJECT (34) a. [phu:chay khon nay] ti: ?a*p ch3p Nit man cl VARIABLE.+HVMAN comp hide like Which man is the one that secretly has a crush on Nit? 145 REDUCED WH-CLEFT: WH-OBJECT b. [phu:cha.-y khon nay] tl: Nit ?a»p ch3.p man cl VARIABLE+HUMAN comp hide like Which man is the one that Nit secretly has a crush on? These above examples in (31-33) suggest that reduced D-linked wh-expressions are indeed derived from identificational wh-clefts because wh-subject and wh-object are able to occur in a cleftee position, as expected if relativization is involved. 5.6.2 Both [+human] and [-human] are allowed in a cleftee position In reduced wh-clefts with D-linked wh-expressions, we see a [+human] feature 'which man' in a cleftee position. We expect that the cleftee should not be restricted to only [+human]. The cleft clause is formed by a relative clause, in which case any [±human] head noun featured can be relativized. Reduced wh-clefts are given a parallel analysis to identificational wh-clefts. Hence, we should expect to see [-human] occur in a cleftee position. This prediction is in fact borne out, as shown in (35). REDUCED WH-CLEFT: [-HUMAN] WH-SUBJECT (35) a. ma: tua nay thi: kat Nit dog cl VARIABLE comp bite Which dog was the one that bit Nit? REDUCED WH-CLEFT: [-HUMAN] WH-OBJECT b. ma: tua nay thi; Nit si: ma: dog cl VARIABLE comp buy come Which dog was the one that Nit bought? The next section ends the chapter by discussing the two predictions that fall out from the identificational wh-cleft analysis. 146 5.7 Predictions of the identificational wh-cleft analysis This section discusses the two predictions of treating thl: clauses with D-linked wh-expressions as identificational wh-clefts. These predictions help shed some light on the nature of wh-intervention effects and wh-argument locatives in Thai. Both wh-intervention effects and wh-argument locatives derive from reduced (identificational) wh-clefts. 5.7.1 Identificational wh-clefts rather than wh-intervention effects Beck (1996) and Beck & Kim (1997) investigate wh-intervention effects cross-linguistically and argue that wh-expressions cannot move across certain interveners (i.e. negation quantifiers, focus phrases) to take scope over the clauses at LF. However, these interveners do not block overt wh-movement. The following data from Korean and Mandarin illustrates their generalization31 (Beck & Kim 1997, Kim 2000). In (36a), 'anyone' is an intervener when the wh-expression 'who' covertly moves across the intervener at LF. This intervention effect results in ungrammaticality of the sentence. In contrast, in (36b), we see that it is grammatical when the wh-expression overtly moves across the intervener 'anyone'. WH-INTERVENTION EFFECTS: KOREAN (36) a. * amuto nuku-hil po-chi anh-ass-ni anyone who-acc see-chi not do-past-Q Whom did no one see? b. nuku-lulj amuto tj po-chi anh-ass-ni who-acc anyone see-chi not do-past-Q Whom did no one see? In (37a), 'anyone' is an intervener and it blocks the LF movement of the D-linked wh-expression 'which book'. Hence, ungrammaticality of the sentence arises. In (37b), the 31 Beck and Kim (1997) provide an array of data illustrating intervention effects for wh-licensing and NPI licensing. Interveners include negation, quantifiers and focus phrases (i.e. 'only', 'even' and 'also'). 147 intervener 'anyone' does not block the overt wh-movement of the D-linked wh-expression, resulting in grammaticality of the sentence. WH-INTERVENTION EFFECTS: CHINESE (37) a. * [shei ye] kan bu dong [na-ben shu] who also (=anyone) read not understand which-cl book Which book could no one understand? b. [na-ben shu] [shei ye] kan bu dong which-cl book who also (=anyone) read not understand Which book could no one understand? At first glance, it seems that there are wh-intervention effects in Thai. This is illustrated in the examples below. The goal (as a variable expression) does not have a wh-construal when the intervener maym:ikhray 'no one' appears between the in-situ variable and the clause over which the variable takes scope. The goal can only have an NPI-construal, as in (38a). However, in (38b) when the goal is at the left edge, it only gets a wh-construal. * WH-CONSTRUAL V NPI-CONSTRUAL (38) a. maymi:khray ch3.p [khray] no-have- VARIABLE. +HUMAN like VARIABLE.* HUMAN * (i) Who does nobody like? = (ii) Nobody likes anyone. V WH-CONSTRUAL * NPI-CONSTRUAL b. [k"ray] thi: maymi:khray ch3.p VARIABLE+HUMAN comp no-have- VARIABLE. +HUMAN like = (i) Who is the one that no one likes? ?- (ii) Nobody likes anyone. 148 In (39a), the D-linked wh-expression does not have an NPI construal, in contrast to the previous example given in (38a). The example (39a) is instead ungrammatical. But when the D-linked wh-expression is at the left edge, the sentence becomes grammatical. It can have a wh-construal. * WH-CONSTRUAL * NPI-CONSTRUAL (39) a. * maymi:khray khawcay no-have- VARIABLE. +HUMAN understand Which book did no one understand? WH-CONSTRUAL NPI-CONSTRUAL [nans?: lem nay] thi: maymi:khray khawcay book cl VARIABLE.±HUMANcomp no-have-VARIABLE.+HUMAN understand Which book was the one that no one understood? Despite the surface parallels between the Thai example in (38b-39b), and the corresponding examples of wh-intervention effects in Korean and Chinese in the (36b-37b), I will argue that (38b-39b) are a kind of reduced (identificational) wh-clefts. Let us first consider the example (38a), where the goal (as a variable expression) receives only an NPI construal. This can be straightforwardly accounted for by the probe-goal analysis, as schematically illustrated in (40a). The feature of the closest probe is the one that is copied and filled in for the underspecified goal. Here it is a [Neg] feature. Hence, Match is satisfied. However, we would expect the D-linked wh-expression in (39a) to be construed as an NPI, rather the ungrammaticality of the sentence. We saw earlier in chapter 3 that the interpretation assigned to a D-linked wh-expression is predictably determined by the probe-goal relation that holds between the operator (as probe) and the D-linked wh-expression (as goal), the same way as bare wh-expressions do. This raises an interesting question regarding why the negative quantifer blocks wh-construal for the D-linked wh-expression in (39a), but does not license NPI construal either, rather it forces the overt movement? Intuitively, what is happening is that the NPI reading is ruled out for pragmatic reasons. The D-linked wh-expression is only chosen to have a wh-construal in this context. This results in ungrammaticality, as in (40c). In the probe-goal analysis, the goal looks for the [nans?: lem nay] book cl VARIABLE.-HUMAN 149 closest probe to match in feature. The goal cannot look past the closest c-commanded probe (due to a locality condition). . , (40) a. [CP Q[+wh] [IP Subject [NP Neg[+neg] [vp Verb cp M ] ] ] ] b. c. [CP Q[+wh] [IP Subject [NpNeg[+neg] [VP Verb cp [+neg ]]]] [CP Q[+wh|. [IP Subject [NpNeg[+neg] [VP Verb cp t+wh] ]]]] In order for a wh-construal to be available for (37a-38a), the wh-expressions must be at the left edge to match with the [wh] feature on the Q[wh] probe via a reduced wh-cleft construction, as argued in this chapter. The structure of reduced identificational wh-clefts is illustrated in (41). Notice that in reduced wh-clefts, the Q[wh] probe is now the closest c-commanding probe, satisfying the probe-goal relation as in (41b). The negative quantifier is generated lower than the wh-expression, and hence is not qualified to be the probe for the wh-expression. (41) a. [Cp Q[+wh ] [IP [DP 9 n ] [i 0 [DP 0 [CP pro [c thi: [rp NPI-Subject [+neg] Verb ] ] ] ] ] b. [CP Q[+wh | [IP [DP 9 l+wh ] ] [i 0 [DP 0 [CP pro [c thi: [ff NPI-Subject [+„eg] Verb ] ] ] The examples (38b) and (39b) are repeated in (42). Rather than treating (42) as a wh-intervention effect, I argue that the examples in (42) are reduced wh-clefts. I will show that the semantics and syntax of (42) have the same properties as identificational wh-clefts. Recall that there are two notable properties about the clefted element in reduced identificational wh-clefts. First, the clefted element may only be a D-linked wh-expresssion. Secondly, it may have a wh-object construal that is linked to the gap inside the relative clause. We see that (42a) contradicts the first property. It is the bare wh-expression that occupies the clefted element position. However, this bare wh-expression is equivalent to the D-linked wh-expression 'which N\ This question requires an answer restricted to the set of individuals common to both the speaker and hearer. Crucially, there is a presupposed set of individuals from which the answer is selected. We see that the wh-expression 'who' in (42a) has a wh-object construal linked to the gap inside the relative clause 'who/which one; is the onej that nobody likes e; ?' Recall that the wh-object construal in the clefted element is not possible for a reduced (contrastive) wh-cleft (due to the internal structure of a nominalized clause). Despite its appearance, I argue that (42a) is a kind of reduced wh-cleft. In addition to their syntactic parallels to wh-clefts, the examples in (42) have the same semantic properties as identificational wh-clefts. 150 V WH-CONSTRUAL * NPI-CONSTRUAL (42) a. [khray] thi: maymi:khray ch3.p VARIABLE.+HUMAN comp no-have- VARIABLE. +HUMAN like = (i) Who is the one that no one likes? (lit = which one is the one that no one likes?) 5* (ii) Nobody likes anyone. V WH-CONSTRUAL * NPI-CONSTRUAL b. [nans?: lem nay] thi: maymi:khray khawcay book cl VARIABLE.-HUMAN comp no-have-VARIABLE.+HUMAN understand Which book was the one that no one understood? One of the clearest ways of showing that (42) has the same semantic properties as identificational wh-clefts is to create a question-answer pair where a presupposition is explicitly denied in the answer. In such an answer to the wh-cleft. question, we expect that the answer will be infelicitious. As predicted, the answer (i) to the question (43) repeated from (43) is pragmatically odd. This supports the argument that wh-clefts have an existential presupposition. Another semantic property discussed in this chapter is that wh-clefts have a uniqueness presupposition. Only a single and unique individual can be identified in order to satisfy the uniqueness presupposition. The answer (ii) to the (43) question turns out to be infelicitious, just as predicted there should be when there are two individuals identified in the answer. 151 These semantic and syntactic effects are attributable to a cleft structure . That is, wh-intervention effects derive from the identificational wh-cleft analysis. REDUCED WH-CLEFT (43) Q: [khray thi:] maymi:khray VARIABLE.+HUMAN comp no-have- VARIABLE. +HUMAN like Who/which one is the one that no one likes? Ai: # may mi: ni neg have no one Aii: # Paris, Nicole 5.7.2 Wh-argument locatives as D-linked wh-expressions This section discusses thi: clauses with wh-argument locatives. I first argue that thi: clauses with wh-argument locatives are instances of reduced wh-clefts. Then, I consider the consequence of analyzing thi: clauses with wh-argument locatives as reduced identificational wh-clefts. It is possible that thi: clauses with wh-argument locatives in Thai are a D-linked wh-expression equivalent to English 'which place', rather than a bare wh-expression 'where'. I provide another example in (i) where a bare wh-expression 'what' is clefted. As argued, 'what' has the status of being a D-linked wh-expression, and it is a reduced 'identificational' wh-cleft. Context: In a class, a teacher is giving a lecture and she sees a student frowning, (i) Q: [?aray thi:] khun may khawcay varaiable. -human comp you neg understand What/which part was the part that you did not understand? A: trorj thi: ?aca.-n bo:k wa:... part comp teacher say comp The part that you said that... A':#may mi: kha? khawcay fNikya.Tj neg have a female polite marker understand everything Nothing A": #mi; s5;rj cut trorj thi;... la=? k3: have two point part that... and also I have two points that I did not understand, the part that.. .and trorj thi.\.. part that also the part that... 152 I argue that the examples in (44-45) are instances of reduced wh-clefts. Consider the wh-question in (44), where the wh-argument locative occurs with a fi: clause. As a clefted element, it is linked to the gap in the relative clause in the subject position. The wh-argument locative, in (45) is related to the gap in object position. The fi: clauses with wh-argument locatives33 appear to have the same properties as identificational wh-clefts. REDUCED WH-CLEFT: LOCATIVE SUBJECT (44) Q: [thi: nay] t'l: thamhay k\m prafapcay place VARIABLE comp make you impress Which place impressed you? A: Vancouver REDUCED WH-CLEFT: LOCATIVE OBJECT (45) Q: [t"i: nay] tl: Nit ch3.p pay nan place VARIABLE comp like go sit Which place does Nit hang out? A: ra.n klay ?empo.Ti:am store near Emporium The club near Emporium. Recall that one of the notable properties of identificational wh-clefts is that the clefted element may only be a D-linked wh-expression. By this claim, there is a possibility that Thai wh-argument locatives may have the status of D-linked wh-expressions, rather than bare wh-expressions. There is also independent evidence to treat wh-argument locatives as D-linked wh-expressions. The evidence comes from how wh-argument locatives are morphologically 33 However, wh-adjunct locatives do not have such properties. They are ungrammatical with the fi: clause, as shown in the examples in (i) and (ii) mamu.arj mango si: buy (i) *[thi: nay t"i:] Nit place variable comp Where did Nit buy mango? (ii) *[t"T: nay t"i:] Nit tas.rjrja.-n place variable comp many Where did Nit get married? This probably suggests that such properties are only restricted to wh-arguments. 153 composed, as in (46b). Compare the examples of Thai D-linked wh-expressions in (46a) and (46b). In (46a), the NP 'man' is optional and is accompanied by its counterpart classifier. Notice that in (46b), the noun and its classifier are homophonous. This is not accidental because some nouns can serve as a classifier on their own; in this case'place' serves as its own classifier34. nay VARIABLE b. (t'l;) t'l: nay place cl VARIABLE which place? The examples below illustrate that locative thi: is in fact functioning as both a noun 'place' and its counterpart classifier. In the Thai 'how many' question construction, 'how many' only takes classifiers as its noun. In (47a), it is ungrammatical when 'how many' is followed by the noun 'book'. But when it is followed by its classifier (lem)', as in (47b), the sentence becomes grammatical. In order to show that thi: has a dual function 'place' and a classifier for 'place', we expect thi: to be grammatical when we introduce it in 'how many' type question. The example in (47c) turns out just as predicted. HOW MANY QUESTION (47) a. * khun si: ma; ki; nans?: you go come how many book How many books did you buy? b. k\in si: ma: ki: lem you go come how many cl How many books did you buy? 34 ("v. 'place', on the other hand, is accidently homophonous with the complementizer H:. D-LINKED WH-EXPRESSION (46) a. (pVcV-y) k"on man cl which man? 154 c. khun pay du: ma: ki: thi: you go look come how many cl How many places did you take a look? The fact that thi: can function as a classifier on its own suggests that thi: nay '(place) cl which' is a D-linked wh-expression. This concludes my discussion that treating a wh-argument locative as a reduced wh-cleft reveals its internal structure as a D-linked wh-expression. To summarize, I have argued, in this chapter, that thi: clauses with D-linked wh-expressions in Thai take the form of an identificational wh-cleft type structure35. I have provided several arguments to show that these t'"i: clauses with D-linked wh-expressions have the same predictably semantic and syntactic properties as identificational wh-clefts. Reduced identificational wh-clefts have two semantic properties: an existential presupposition and a uniqueness presupposition. I have argued that the existential presupposition correlates with the presence of the identificational ll'i: copula, while the uniqueness presupposition arises from the definite description in the cleft relative clause. In reduced identificational wh-clefts, the clefted element may be associated with the gap in subject and object position in the relative clause. As such, the cleft element may have a wh-object construal. This is not possible in reduced contrastive wh-clefts. I finished this chapter by discussing the two major consequences of analyzing thi: clauses with D-linked wh-expressions as reduced wh-clefts. First, there are no wh-intervention effects in Thai, despite appearances. They are in fact reduced wh-clefts. I showed that they have the same semantics and syntax as identificational wh-clefts. Second, analyzing thi: clauses with wh-argument locatives as reduced wh-clefts leads to the conclusion that Thai wh-argument locatives are in fact D-linked wh-expressions. 35 I reject a focus phrase analysis because it cannot account for the full range of facts associated with the semantic and syntactic properties of the thi: clause in Thai. 155 Chapter 6 Conclusions 6.1 Summary of major findings This dissertation has presented an in-depth examination of Thai wh-expression as variables. I have claimed that wh-expressions are variables with no inherent interrogative force. As variables, they acquire an interpretation by virtue of being in the scope of an operator. I propose that the syntactic relation between the operator and the variable is best analyzed in terms of a probe-goal relation. The probe-goal relation requires that the features of probe and goal Match. The analysis of Thai wh-expressions that I propose claims that Match can be satisfied via feature copying. In particular, I argue that, in Thai, the operator-variable relation, as a probe-goal relation, satisfies Match through feature copying. Thai wh-expressions are variables; as such they are underspecified goals whose featural content needs to be filled in. Chapter 2 argued that the probe is a covert Q[Wi,], and the goal, on the other hand, is underspecified for a feature, copying the [wh] feature of the probe onto the goal. I claimed that the probe-goal relation in Thai is established via Match without Move. Much of the chapter was spent discussing the two major consequences of my claim that the covert Q[Wh] probe is base-generated in C position from where it assigns wh-scope. First, the proposed analysis correctly predicts the distribution of Thai wh in-situ. More specifically, it derives the absence of an asymmetry between wh-subjects and wh-objects, as well as the absence of an asymmetry between wh in-situ arguments and wh-adjuncts with respect to island effects. Second, the probe-goal analysis accounts for restrictions on interpretation. In particular, it derives the absence of pair-list readings in multiple wh-questions, as well as the absence of list-answer readings in wh-constructions that contain a quantifier. Chapter 3 explores the role of negation as a probe, and asserts that negation and yes-no question markers are allomorphs. In the context of negation, the goal (as a variable expression) matches with the [Neg] feature on the Neg probe, hence yielding a Negative Polarity Item (NPI). In a yes-no construction, the goal matches with the [+polarity] feature on the Q[poiarity] probe, yielding an Existential Polarity Item (EPI). I further argued that the interpretation assigned to a goal is predictably determined by the c-command relation that holds between the probe and the goal. For matrix clauses, the proposed analysis correctly predicts the presence of a subject/object 156 asymmetry, and the presence of a complement/adjunct asymmetry. Moreover, the analysis captures the fact that both subject/object asymmetries and complement/adjunct asymmetries with respect to NPI- and EPI-construals only hold in matrix clauses. Thus, while NPI- and EPI-construals are unavailable with subjects and adjuncts in matrix clauses, they are available in embedded clauses. This is because a matrix negative probe or matrix yes-no question probe is available for the embedded goal. The availability of NPI- and EPI-construals in embedded clauses is a side-effect of a locality condition which requires that the closest c-commanding probe is the one that enters into the probe-goal relation. Chapter 4 is mainly concerned with reduced contrastive wh-clefts. Crucially, I argued that the probe-goal relation in Thai does not induce Move, contrary to superficial appearances. In fact, cases of apparent wh-movement of bare wh-expressions are instances of reduced contrastive wh-clefts. Specifically, this chapter deals with the semantic and syntactic properties of contrastive wh-clefts with the overt copula pen. I show that the semantics and syntax of reduced contrastive wh-clefts have the same properties as contrastive wh-clefts. I close the chapter by considering the three major consequences of analyzing apparent cases of wh-movement as reduced wh-clefts. First, only wh-subjects can occur as a cleftee (due to the inability of the nominalizer ton to nominalize the object). Second, wh-objects can occur as a cleftee only when passive markers are present (the object can only nominalize through passivization). Last, only 'who' can occur as a cleftee (due to [+human] restriction imposed by the nominalizer ton). These restrictions fall out naturally from the contrastive wh-cleft analysis. Chapter 5 is devoted to another type of cleft, reduced identificational wh-clefts. In particular, they involve clefting of D-linked wh-expressions. I discuss the semantics and syntax of identificational wh-clefts with the overt identificational ti: copula. I show that the semantics and syntax of reduced identificational clefts have the same properties as identificational wh-clefts. The chapter ends by considering two predictions that fall out from the identificational wh-cleft analysis. Thai wh-intervention effects are in fact reduced wh-clefts. This satisfies the probe-goal relation that goals must be at the left edge of the clause in order to match with [wh] feature of the Q probe, yielding a wh-construal. Another consequence of analyzing thi: clauses with D-linked wh-expressions as reduced wh-clefts is that it reveals the internal structure of wh-argument locatives as D-linked wh-expressions. 157 6.2 A note on wh-adjuncts and Move In this dissertation, I claim in chapter 4 and 5 that wh-argument focus is attributable to a cleft clause structure. A question that needs to be addressed for future research is 'how do Thai wh-adjuncts mark focus?' I showed, in chapter 3, that adjuncts can have NPI- and EPI- construals when they are in an embedded clause (by matching with a matrix negative probe). This suggests that adjuncts have the same status of variable expressions as wh-arguments. However, I observe that Thai wh-adjuncts—in particular, wh-temporal and wh-rationale undergo A' movement with respect to focus. This is because a [+foc] feature of the probe is copied onto the underspecified goal. The [+foc] feature of a head has an EPP feature, and it triggers the goal to move to the probe and form a specifier of the probe. Notice that the goal and the probe do not form a constituent due to an intervening subject. This particular issue is beyond the scope of this thesis. I leave this open for future research. WH-TEMPORAL (1) a. [rm;aray] Lek ca? ri:ancop VARIABLE fut graduate When is it exactly that Lek will graduate? WH-RATIONALE b. [thammay] Nit thtn ya: VARIABLE even divorce Why was it exactly that Nit divorced? I have shown that wh-arguments are at the left edge in a cleft focus structure, satisfying the probe-goal relation. I will now address some final thoughts about wh-adjuncts, tying in with the proposed analysis. Consider the wh-question in (2a), in which the wh-temporal appears in-situ, and on the other hand, in (3a), in which the wh-temporal has undergone movement. WH IN-SITU (2) a. Lek ca? ri.ancop [m?:aray] fut graduate VARIABLE When will Lek graduate? WH-MOVEMENT 158 b. [rm:aray] Lek ca? ri:ancop VARIABLE fut graduate When is it exactly that Lek will graduate? (2a) and (2b) involve a contextual presupposition that Lek has been studying for her Ph.D in Canada for more than six years. The speaker wants to know when Lek will graduate, as in (2a). On the other hand, (2b) is emphatic in that the speaker expects a specific answer as to a particular time of Lek's graduation. In (3a), we see that wh-rationale appears in-situ. Without the movement, the sentence becomes ungrammatical. The wh-rationale in (3b), on the other hand, moves to a clause-initial position, and grammaticality of the sentence obtains. Notice that in (3b), there is a presence of a focus morpheme ftrj in a preverbal position. It is this morpheme that forces the wh-expression to undergo A' movement. This explains why (3a) is ungrammatical. WH IN-SITU (3) a. * Nit th?rj ya: [thammay] even divorce VARIABLE Why was it exactly that Nit divorced? WH-MOVEMENT b. [thammay] Nit thtn ya: VARIABLE even divorce Why was it exactly that Nit divorced? The examples in (4) show that the movement is unbounded, similar to that of wh-movement. The wh-expressions undergo movement cyclically from their base-generated position, to an intermediate position, and then to a clause-initial position. (4) a. mlaray k],un khit wa: Nit ca? ri:ancop VARIABLE you think comp fut graduate When do you think that Nit will graduate? b. thammay khun khit wa: Ra.n nixak Nit 159 VARIABLE you think comp Why do you think that Ron knew Nit? know Wh-focus obeys the same island conditions. In (5a), when the wh-rationale 'why' is inside the relative clause island, it is ungrammatical, unlike in-situ wh-arguments discussed in chapter 2. The wh-rationale cannot occur freely in the island, as in (5a) nor can it be extracted out of the island. This indicates that overt wh-focus shows the usual island properties associated with wh-movement. When the wh-expression is moved out of the island, as in (5b), an island violation is found. Notice that the wh-expression in (5c) can only be interpreted as questioning the matrix clause. The wh-expression cannot be construed as questioning the embedded clause. This shows that the wh-expression is not able to associated with the embedded verb across the relative clause island. RELATIVE CLAUSE ISLAND (5) a. * khun ch3.p [NP nans?: [Cp thi: Nit khia.n t'ammay ] ] you like book comp write VARIABLE [Why did you like the book that Nit wrote?] b. * khun ch3.p thammayi [Npnarjs?: [cpthi: Nit khia.-n tj ] ] you like VARIABLE book comp write [Why did you like the book that Nit wrote?] c. thammayj khun ch3.p t( [NP nans?: [cpthi: Nit khia.n tj]] VARIABLE you like book comp write = (i) What was the reason that you like the book? ^ (ii) Why did you like the book that Nit wrote? With respect to the sensitivity of island effects, the above examples confirm that wh-focus undergoes A' movement to the left edge position (Spec CP). The wh-expression matches with [wh] feature with Q[wh] probe via Move. 160 6.2.1 Wh-focus adjuncts as topicalization? This section considers alternative analyses of topicalization and wh-clefts. I will argue that wh-focus is a kind of A' movement rather than either of these alternatives. When wh-adjuncts (i.e. temporal and rationale) are moved to a sentence-initial position, they do not involve the presence of the complementizer thi:. At first glance, this can be viewd as evidence for a topicalization analysis. However, Thai topicalized arguments are not subject to island constraints (Hoonchamlong 1991). This is illustrated in (6b). The topic element (Ben) is co-indexed with the optional resumptive pronoun (he) in the complex NP island. If there were movement, (6b) should be ungrammatical. However, no island violation is found. (6) a. chan dayyin kha.w wa: Ben ta?.Tjrjan I hear news comp marry I heard the rumor that Ben married. b. Beni chan dayyin kha.w wa: (khawj) ta?.-nnan I hear news comp he marry Ben], I heard the rumor that he] married. Temporal adjuncts in (7b), on the other hand, are constrained by the island. In (7a), the adjunct is inside the complex NP island, and it has an embedded scope interpretation. But when it is moved out of the island, the embedded scope interpretation is no longer available, as in (7b). The temporal adjunct can only modify the matrix verb. This shows that the movement is subject to island effects. If no movement takes place, it should be able to associated with the gap in the island. (7) a. chan dayyin k''a.w wa: Su.mi ta».Tjrjan ?a.thitthid^.-w I hear news comp marry week-last I heard the news that Soomii married last week. ?a.thitthi;la?.Wi chan dayyin kha.-w wa: Su.mi tae.Tjrjan tj week-last I hear news comp marry Last week, I heard the news that Soomii married. 161 As mentioned earlier, the fact that wh-focus adjuncts are sensitive to island constraints suggests that they cannot be topicahzation. Hence, topicalization is not an appropriate analysis for wh-adjunct movement. 6.2.2 Wh-focus adjuncts as a wh-cleft? Can Thai wh-focus adjuncts be analyzed as wh-clefts? We see that, in reduced wh-clefts, it is obligatory that the morpheme thi: is present immediately following the wh-arguments. Wh-focus adjuncts, on the other hand, do not require the presence of the morpheme thi:. I have discussed that the morpheme thi: must be present because presuppositions found in clefts arise from the morpheme This is clearly lacking in the non-cleft focus construction. The example in (8) illustrates that when temporal and rationale adjuncts are clefted, they require the presence of the complementizer Hence, the absence of complementizer thi: of wh-focus adjuncts does not follow from a cleft analysis. (8) a. wanphut khi: wan *(thi:) Nik pay tala.-t day-Wed be day comp go market Wednesday is the day that Nick goes to the market. b. k''wa.Tnbi: khi: sa.tie.t *(thi:) tham hay kha.w la:k ri:an nom- bore be cause comp make give him quit study Boredom was a cause that made him quit his study. / 162 Bibliography Aoun, Joseph and Audrey Li. 1993. Wh-elements in situ: syntax or LF? Linguistics Inquiry 24, 199-238. Baker, C. Lee. 1970. Notes on the description of English questions. Foundation of Language 6, 197-219. Beck, Sigrid and Shin-Sook Kim. 1997. On Wh-and operator scope in Korean. Journal of East Asian Linguistics 6, 339-384. Blain, Eleanor. 1997. Wh-constructions in Nehiyawewin. Doctoral dissertation, UBC. Chang, Lisa. 1995. Wh-in-situ Phenomena in French. Master Thesis, UBC. Cheng, Lisa. 1991. On the typology of Wh-questions. Doctoral dissertation, MIT. Cheng, Lisa and Sybesma R. 1999. Bare and not-so-bare nouns and the structure of NP. Linguistic Inquiry 30 (4), 509-542. Chomsky, Noam. 1977. On Wh-movement. In P. Culicover, T. Wasow, and A. Akmajian Formal Syntax, 71-132. Chomsky, Noam. 2000. Minimalist inquiries: the framework. In step by step: Essays on Minimalist syntax in honor of Howard Lasnik, ed. R. Martin, D. Michaels, and J. Uriagereka, 89-155. Cambridge, MA: MIT Press. Cole, Peter and Cher Leng Lee. 1997. Yes-no questions in the Teochew of Singapore. Journal of East Asian Linguistics 6 (2), 1997. Cole, Peter and Gabriella Hermon. 1994. Is there LF Wh-movement? Linguistics Inquiry 25, 239-262. Cole, Peter and Gabriella Hermon. 1998. The typology of Wh-movement: Wh questions in Malay. Syntax 1, 221-258. Cole, Peter, Gabriella Hermon and Norhaida Bt. Aman. 2000. Clefted questions in Malay? In David Gil and James Collins (eds.), To appear in Malay/Indonesian Linguistics. London: Curzon Press. Cole, Peter and Gabriella Hermon. 2000. Partial Wh movement: evidence from Malay. In Lutz. U, Mueller. G, and A. Von Stechow (eds.), To appear in The Syntax and Semantics of Partial Movement. Cambridge: Cambridge University Press. Cook, Clare. 2004. Adjunct extraction in Nehiyawewin and Yoruba: neither adjunction nor extraction. Syntax generals paper, University of British Columbia. Cook, Clare. 2004. Argument-adjunct asymmetries in wh-construction. CLA presentation, Manitoba, May 2004. 163 Dechaine, Rose-Marie. 2002. Decomposing focus: evidence from Yoruba. Triggers for Movement workshop, Tilburg University, 24-26 October, 2002. Dechaine, Rose-Marie and Martina Wiltschko. 2002. Decomposing pronouns. Linguistic Inquiry 33, 409-442. Den Dikken, Marcel and Pornsin Singhapreecha. 2004. Complex noun phrases and linkers. Syntax 7, 1-33. Denham, Kristin. 2000. Optional Wh-movement in Babine-Witsuwit'En. Natural Language & Linguistic Theory 18 (2), 199-251. E. Kiss, Katalin. 1988. Identificational focus versus information focus. Language 74 (2), 245-73. Ekniyom, P. 1982. A study of information structuring in Thai sentences. Doctoral dissertation, Univeristy of Hawaii. Guglielmo, Cinque. 1999. Adverbs and functional heads: a cross-linguistic perspective. Oxford: Oxford University Press. Hagstrom, Paul. 1998. Decomposing questions. Doctoral dissertation, MIT. Haspelmath, Martin. 1997. Indefinite pronouns. Oxford: Clarendon Press. Hedburg, Nancy. 2000. The referential status of clefts. Language 76 (4), 891-920. Hoonchamlong, Yuphaphann. 1991. Some Issues in Thai Anaphora: a Government and Binding Approach. Doctoral Dissertation, University of Wisconsin-Madison. Huang, James. 1982. Logical relations in Chinese and the theory of grammar. Doctoral dissertation, MIT. Huang, C.T. James. 1982. Move Wh in a language without Wh movement. Linguistic Review 1, 369-416. Hundius, H. and U. Kolver. 1981. Syntax and semantics of numeral classifiers in Thai. Studies in Language 7, 165-214. Jackendoff, Ray. 1972. Semantic interpretation in Generative Grammar. Cambridge, MA: MIT Press Jelinek, Eloise. 1998. Wh-clefts in Lummi. In D. Bates (ed.), ICSNL 33 Proceedings. University of Washington, Seattle. Jones, Susie. 2004. Syntax of Yoruba questions. ACAL 35 Presentation, Harvard, April 2004. Kim, Shin-Sook. 2002. Focus matters: two types of intervention effect. WCCFL XXI, UC Santa Cruz, 5-7 April, 2002. Kobsinphat, W. 1988. Empty categories in Thai. Doctoral dissertation, University of Washington. 164 Koopman, Hilda. 2000. The Spec Head configuration. In the syntax of specifiers and heads: Collected essays of Hilda J. Koopman. Routledge. Kuno, Susumu and Preya Wongkhomthong. 1981. Characterizational and identificational i sentences in Thai. Studies in language 5, 65-109. Li, Yen-hui Audrey. 1992a. Indefinite Wh in Mandarin Chinese. Journal of East Asian Linguistic 1, 125-155. Lee, Felicia. 2001. Wh- and focus are not the same projection. In K. Megerdoomian and L.A. Bar-el (eds.) WCCFL 20 Proceedings. Somerville, MA: Cascadilla Press. McCawley, James. 1994. Remarks on the syntax of Mandarin yes-no questions. Journal of East Asian Languages 3 (2), 179-194. Miiller, Gereon and Wolfgang Sternefeld. 1993. Improper movement and unambiguous binding. Linguistic Inquiry 24, 461-507. Nishigauchi, Taisuke. 1990. Quantification in the theory of grammar, Kluwer academic publishers, Dordrecht. Percus, Onn.1997. Prying open the cleft. In. K. Kusumoto (ed.) NELS27, 337-351. Pesetsky, David. 1987. Wh-in-situ: movement and unselective binding. The Representation of (In)defmiteness. Eric J. Reuland and Alice G.B. ter Meulen eds. Cambridge, Massachusettes: MIT Press. 98-129. Phanthumetha, Nawawan. 1991. Thai linguistics for teachers (In Thai). Sukhothaithammathirat University. Bangkok: Phapphim. Phillips, Colin. 1996. Order and structure. Doctoral dissertation, MIT. Phillips, Colin. 2003. Linear order and constituency. Linguistic Inquiry 34, 37-90. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.) Elements of Grammar: Handbook of Generative Syntax, 281-337. Dordrecht: Kluwer Ross, John R. 1967. Constraints on variables in syntax. Doctoral dissertation, MIT. Rochemont, Michael. 1986. Focus in generative grammar. Amsterdam/Philadelphia: John Benjamins publishing company. Saddy, Douglas. 1991. Wh scope mechanisms in Bahasa Indonesia. MIT Working Papers in Linguistics 15, 183-218. Saito, Mamoru. 1985. Some asymmetries in Japanese and their theoretical implications. Doctoral dissertation, MIT. Sawai, Naomi. 2002. The syntax of wh-questions in Nuu-chah-nulth. Syntax generals paper, University of British Columbia. Shank, Scott. 2004. Domain widening. Doctoral dissertation, UBC. 165 Shiobara, Kayono. 2004. Linearization: a derivational approach to the syntax-prosody interface. Doctoral dissertation, UBC. Sportiche, Dominique. 1998. Partitions and atoms of clause structure: subjects, agreement, case and clitics. New York: Routledge. Sornlertlamvanich, Charoenporn, and Isahara. 1997. Orchid corpus: Thai part of speech tagged corpus. In the Proceedings of NLPRS 1997. Stein, Mark Jeffrey. 1981. Quantification in Thai. Doctoral dissertation, University of Massachusetts at Amherst. Thomas Ernst. 2002. The syntax of adjuncts. Cambridge: Cambridge University Press. Visonyanggoon, Saisunee. 2000. Parallelism between Noun Phrases and Clauses in Thai. Doctoral dissertation, MSU. Wannapaiboon, Naruemon. 2001. Thai Relative Clauses: evidence for (non)movement analysis. Master Thesis, UCSC. Watanabe, Akira. 1992. Subjacency and S-structure Movement of Wh in-situ. Journal of East Asian Linguistics 1, 255-291. Williams, Edwin. 2003. Representation theory. Massachusetts: The MIT Press. 166 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0099823/manifest

Comment

Related Items