Thesis - Archive ouverte UNIGE
Transcription
Thesis - Archive ouverte UNIGE
Thesis The syntactic structures of Russian wh-questions ROJINA, Nina Abstract This dissertation addresses a number of issues involving the syntactic structure of Russian wh-questions. This work proposes a different perspective on the syntactic structure of Russian wh-questions. The dissertation offers a comparative analysis of Russian wh-questions with other languages involving multiple wh-movement (a.o. Bulgarian and Serbo-Croatian), and argues against the common perception of Russian as a language where a wh-phrase does not undergo wh-movement or show any Superiority effects. The data and analyses presented in this work argue for the opposite. The dissertation adopts the Cartographic approach to the syntactic analysis of wh-phrases in Russian. Reference ROJINA, Nina. The syntactic structures of Russian wh-questions. Thèse de doctorat : Univ. Genève, 2011, no. L. 724 URN : urn:nbn:ch:unige-174157 Available at: http://archive-ouverte.unige.ch/unige:17415 Disclaimer: layout of this document may differ from the published version. [ Downloaded 15/10/2016 at 17:07:45 ] Département de Linguistique Faculté des Lettres THE SYNTACTIC STRUCTURES OF RUSSIAN WH-QUESTIONS Thése de doctorat Présentée à la Faculté des Lettres de l‘Université de Genève pour obtenir le grade de docteur ès lettres per NINA ROJINA 2011 Jury de thèse: Prof. Luigi Rizzi (Université de Genève/Sienne) – Directeur M.e.r. Christopher Laenzlinger (Université de Genève) – Directeur Prof. Jacques Moeschler (Université de Genève) – Président Prof. Iliana Krapova (Université de Venise) Prof. Guglielmo Cinque (Université de Venise) i ii ACKNOWLEDGEMENTS First of all I would like to say a tremendous ‗Thank you!‘ to Luigi Rizzi and Christopher Laenzlinger, my supervisors, who supported me and my research during one of the toughest periods of my work on the dissertation and without whom I would never be able to finish it. Thank you for your support, patience and encouragement. I am fortunate to work with you. My gratitude also goes to Jacques Moeschler, president of the board, for his generosity, his advice, and feedback on my work. Also, I would like to thank the members of my examination committee for criticism and feedback on my research. Special thanks to Iliana Krapova and Guglielmo Cinque for their interest in my work and feedback on my early research. Many thanks also go to my colleagues; especially I would like to mention Gabriella Soare, Goljihan Kachaeva, Greg Ellison, Stephanie Dürrleman. I want to thank you guys for discussions, support, and the occasional beer. My gratitude also goes to everybody in the Department of Linguistics at the University of Geneva, especially Genoveva Puskas, Eric Haeberli, Eric Wehrli, Eva Capitao, Ur Shlonsky. I have received important feedback when presenting my work at Slavic Linguistic Society meetings in Indiana, Berlin and Ohio. I would like to thank Steven Franks, Natalia Kondrashova, and Catherina Rudin for their interest in my presentations and discussions after the conferences. I am immensely grateful to the people who took time to answer my sometimes weird questions about Russian and Bulgarian. Special thank you goes to Marina Zyryanova, Polya Vitkova, Velentina Dolgova, Elena Djakonova, and of course to my parents. Thank you for being patient with me and answering my questions at any time of day and night. iii Of course, I would never be here if it was not for my first supervisor Elena Marinova from Kemerovo State University in Russia. She was the one who inspired me and encouraged to continue the research in the sphere of Linguistics. I would like to specially thank Kamila and Thomas, who became my family in Geneva and who are always happy to let me stay at their place during my visits. My gratitude also goes to John Egil Turbekkmo, who coped with all my travelling to Geneva and to the conferences and taking days off for writing my dissertation. I would like to give another ‗Thank you!‘ to my ‗editors‘ Glyn Hicks, James Pennington, and Greg Ellison for reading my papers during all these years and inserting the articles where needed. A huge ‗Thanks!‘ is in order for my ‗Swiss‘ family Jacques and Veronique, for basically becoming my parents in Geneva and being a great support during my stay here. And of course enormous thank you to my family in Siberia for their love, encouragement, belief in me, even when I thought I did not have strength to do it. I did it because of you! Finally, I would like to thank Mikael for appearing in my life and encouraging me to finish the Dissertation. I dedicate this dissertation to my friends and family. iv Table of Contents ACKNOWLEDGEMENTS .................................................................................................................... III CHAPTER 1. INTRODUCTION .............................................................................................................. 1 1.1. 1.2. 1.3. 1.4. CARTOGRAPHIC APPROACH TO SYNTACTIC STRUCTURES ........................................................... 2 RELATIVIZED MINIMALITY ........................................................................................................ 3 CRITERIAL FREEZING ................................................................................................................. 5 THE ORGANIZATION OF THE DISSERTATION ............................................................................... 7 CHAPTER 2. MULTIPLE WH-FRONTING LANGUAGES AND SUPERIORITY EFFECTS ...... 9 2.1. INTRODUCTION .......................................................................................................................... 9 2.2. DEFINING SUPERIORITY EFFECT ............................................................................................... 12 2.2.1. D-linking ............................................................................................................................. 15 2.2.2. More than two-wh-phrases ................................................................................................. 16 2.3. SOME PROPERTIES OF BULGARIAN, SERBO-CROATIAN AND RUSSIAN WH-QUESTIONS ............. 18 2.3.1. Constituency of wh-phrases ................................................................................................ 18 2.3.2. Wh-islands .......................................................................................................................... 21 2.3.3. Superiority effects in Bulgarian .......................................................................................... 23 2.3.4. Superiority effects in Serbo-Croatian ................................................................................. 26 2.3.5. Why is focus movement of wh-phrases not subject to the Superiority effect? ..................... 28 2.3.6. Problems for Bošković‟s analysis ....................................................................................... 31 2.3.7. Superiority effects in Russian ............................................................................................. 34 CHAPTER 3. TOWARDS AN ANALYSIS OF RUSSIAN MULTIPLE WH-ELEMENTS ............. 41 3.1. BULGARIAN ORDER OF WH-PHRASES ........................................................................................ 41 3.1.1. The order of wh-adjuncts .................................................................................................... 42 3.1.2. The order of wh-objects w.r.t wh-adjuncts ......................................................................... 43 3.1.3. The order of wh-subjects w.r.t wh-adjuncts ........................................................................ 46 3.1.4. The order of D-linked and non-D-linked wh-phrases ......................................................... 48 3.1.5. The order of clitic resumed and non-clitic resumed wh-phrases ........................................ 49 3.1.6. Some conclusions ................................................................................................................ 51 3.2. RUSSIAN ORDER OF WH-PHRASES ............................................................................................. 54 3.2.1. Meyer‟s analysis ................................................................................................................. 54 3.2.2. Analysis of the order of Russian wh-matrix questions ........................................................ 56 3.2.3. Analysis of the order of Russian wh-embedded questions .................................................. 66 3.2.4. A comparison with Meyer‟s results .................................................................................... 75 3.3. PREVIOUS APPROACHES TO MULTIPLE WH-QUESTIONS ............................................................. 76 3.3.1. The adjunction analysis ...................................................................................................... 76 3.3.2. The wh-cluster analysis ...................................................................................................... 77 3.3.3. The multiple [Spec, CP] analysis ....................................................................................... 78 3.3.4. The split CP analysis .......................................................................................................... 80 3.4. ANALYSIS OF RUSSIAN MULTIPLE WH-QUESTIONS ................................................................... 83 3.4.1. Hierarchy prior to movement ............................................................................................. 83 3.4.2. Cartographic approach to Russian multiple wh-questions................................................ 85 CHAPTER 4. WHY-WH-ELEMENTS AND THEIR VARIETY ....................................................... 95 4.1. POČEMU VS. ZAČEM ................................................................................................................ 97 4.1.1. Long-distance dependencies ............................................................................................. 103 4.1.2. The occurrence of “why” phrases in multiple wh-questions ............................................ 105 4.1.3. Position of počemu and začem ......................................................................................... 107 4.2. AGGRESSIVELY NON-D-LINKED AND NOMINAL WH-ELEMENTS .............................................. 113 4.2.1. The case form of „kakogo čërta‟ ....................................................................................... 115 4.2.2. “What” is “why”: differences between čego and čto. ...................................................... 120 4.2.3 Why does „kakogo čërta‟ mean only „why‟? .................................................................... 123 4.2.4. Syntactic properties of aggressively non-D-linked wh-phrases ........................................ 124 4.2.5. Kakogo čërta, čego, čë and multiple wh-questions........................................................... 131 4.3. COORDINATION OF WH-ELEMENTS ......................................................................................... 137 4.3.1. Coordination prior to the wh-movement? ........................................................................ 140 4.3.2. Coordination of two CPs followed by sluicing? ............................................................... 141 v 4.3.3. 4.3.4. 4.3.5. Coordination in CP? ........................................................................................................ 146 Spurious coordination ...................................................................................................... 152 The analysis ...................................................................................................................... 157 CHAPTER 5. MYSTERY OF THAT-TRACE EFFECT ................................................................... 161 5.1. INDICATIVE COMPLEMENTIZER ‗ČTO‘..................................................................................... 164 5.1.1. An NIC approach .................................................................................................................. 164 5.1.2. An ECP account (Rizzi, 1990) ......................................................................................... 165 5.1.3. Richards‟ account (1999) ................................................................................................. 169 5.1.4. Szczegielniak (1999) ......................................................................................................... 171 5.1.5. Rizzi and Shlonsky (2007) ................................................................................................. 174 5.2. SUBJUNCTIVE COMPLEMENTIZER ‗ČTOBY‘ ............................................................................ 185 5.3. ANALYSIS OF RUSSIAN COMPLEMENTIZERS ‗ČTO‘ AND ‗ČTOBY‘ .......................................... 198 CHAPTER 6. WH-SCOPE MARKING CONSTRUCTIONS ........................................................... 211 6.1. SOME PROPERTIES OF WH-SCOPE MARKER IN RUSSIAN .......................................................... 214 6.2. PREVIOUS ANALYSIS OF WH-SCOPE MARKING CONSTRUCTIONS ............................................. 218 6.2.1. The Direct Dependency Approach ........................................................................................ 219 6.2.2. The Indirect Dependency approach ...................................................................................... 222 6.2.3. The Mixed Dependency Approach ........................................................................................ 228 6.3. ANALYSIS OF RUSSIAN WH-SCOPE MARKING CONSTRUCTIONS .............................................. 233 6.3.1. The status of the Slavic wh-scope marker ............................................................................. 234 6.3.2. Syntactic structure of wh-scope marking constructions ........................................................ 238 CHAPTER 7. CONCLUDIING REMARKS ...................................................................................... 247 BIBLIOGRAPHY .................................................................................................................................. 253 vi CHAPTER 1. INTRODUCTION The goal of this dissertation is to address a number of issues involving the syntactic structure of Russian wh-questions. The dissertation proposes a comparative analysis of wh-questions in Slavic languages along the lines of current Minimalist guidelines. This work offers a different perspective on the syntactic structure of Russian whquestions. Russian, being a multiple wh-fronting language, has been considered by a number of linguists (among others Bošković, 1997, 1998a; Stepanov, 1997) as a language without an overt wh-movement to the left periphery and a language not exhibiting Superiority effects in all contexts. The comparative analysis offered in this work suggests the opposite. The data presented in this dissertation clearly demonstrate overt movement of a wh-phrase to the [Spec;CP], as well as the presence of Superiority effects in Russian. The dissertation adopts the Cartographic approach to the syntactic analysis of wh-phrases in Russian; the underlying idea of which is to draw maps as precise and detailed as possible of syntactic configurations (Rizzi, 1997, 2001, 2002). The theoretical implications and predictions related to wh-movement in multiple wh-fronting languages are outlined and analysed in this dissertation, leading towards an analysis of multiple wh-questions in Russian. Before going into the analysis of wh-questions, it is necessary to outline the ideas that underlie the present dissertation which studies the syntax of Russian whquestions on a comparative basis with other multiple wh-fronting languages, with Bulgarian essentially. 1 1.1. Cartographic approach to syntactic structures The Cartographic approach has been developed by a number of linguists, among others Rizzi (1997, 2001, 2002), Cinque (1999, 2006), and Belletti (2004), in order to propose a detailed map of the structure of the clause, assuming that all languages share the same principles of structure formation and the same functional projections. The fact that the complementizer system can be conceived as a structural zone consisting of distinct functional heads and their projections brings Rizzi (1997) to postulate the Force-Finiteness system which expresses relations between a C system and the immediate IP and the Topic-Focus system, which is present in the structure only when needed, i.e. when a constituent bears Topic or Focus features that need to be accommodated in the left periphery of the clause. Rizzi (1997) concludes that the C system has the following structure: FORCE (TOP*) FOC (TOP*) FIN IP Rizzi suggests that there is only one structural Focus position and an indefinite number of Topics. The relative operators occupy [Spec;ForceP], while question operators (whphrases) occupy [Spec;FocP]. If Rizzi suggests a split CP analysis, Cinque (1999) proposes a fine-grained analysis of the adverbs and adverbial position within IP, where each functional head expresses different types of mood, tense, aspect and mode. Moreover, Cinque (1999) suggests that the ordering of the adverbs in a clause is determined by a universal hierarchy, which is determined by phrase structure, with the adverbs occupying unique specifies of functional projections. Krapova and Cinque (2005) employ the cartographic approach to multiple whfronting in Bulgarian, suggesting that each class of wh-phrases targets a specific position in the left periphery. The important aspect of their analysis is that the order of wh-elements is dependent upon the order prior to their wh-movement, i.e. the order realised in the IP or Mittelfeld. Following their lead I analyse Russian multiple whquestions and come to the same conclusion, i.e. in Russian wh-questions the order is 2 pre-defined in the Mittelfeld, which is later realized in CP. This will be the focus of Chapter 3, where Russian data are compared to Bulgarian and the syntactic analysis, on the basis of the Cartographic approach, is offered. 1.2. Relativized Minimality Krapova and Cinque‘s (2005) suggestion that the order of wh-elements is dependent on the order prior to their movement leads them to develop a principle which will ensure that the order is preserved under movement. In this view, Relativized Minimality (Rizzi, 2001) comes into the picture. Relativized Minimality captures the intuition that a local structural relation is one that must be satisfied in the smallest possible environment in which it can be satisfied. Rizzi (2001) proposes a theory of chains, which state that a chain cannot be built between X and Y in configuration …X…Z…Y… 1. Y is in a Minimal Configuration with X iff there is no Z such that (i) Z is of the same structural type as X, and (ii) Z intervenes between X and Y Y and X are in a minimal configuration when Y is a trace and X is a target and there is no intervener Z between them, which c-commands Y and does not c-command X and which is specified with the same features as the target. ……Xt………….Z………….Y(t)… [F] [F] [N] c-command 3 Rizzi (2004) suggests the definition of ‗chain‘ given in (2): 2. (A1,....An) is a chain iff, for 1 < i < n (i) Ai = Ai+1 (ii) Ai c-commands Ai+1 (iii) Ai+1 is in a MC with Ai According to Rizzi (2004), the chain is defined by the following syntactic properties: a. Identity, i.e. each position is identical to any other position in internal structure. b. Prominence, defined by c-command. c. Locality, determined by the notion ‗Minimal Configuration,‘ which is defined in (1). Relativized Minimality explains the ungrammaticality of example in (3), chain formation fails between the target and the trace since they are not in a Minimal configuration, because who is the intervener between how and the trace and who is the same structural type as the target. 3. *How do you wonder who could solve this problem <how>? Rizzi (2004) replaces the structural-type based definition of Relativized Minimality by the feature-type (4): 4. ―same structural type‖ = Spec licensed by features of the same class in (5) 5. a. Argumental: person, number, gender, case b. Quantificational: Wh, Neg, measure, focus.. c. Modifier: evaluative, epistemic, Neg, frequentative, celerative, measure, manner... d. Topic Under the feature-type definition Relativized Minimality effects are expected to arise within the same feature class, but not across classes, i.e. a quantificational specifier acts as an intervener on a quantificational chain, but a pure modificational specifier does not, etc. (Rizzi, 2004). Thus, Relativized Minimality (2004) involves a more fine-grained 4 typology of intervening A‘ specifiers, which trigger Minimality effects on A‘ chains; this accounts for the very selective locality patterns. Krapova and Cinque (2005) employ Rizzi‘s Relativized Minimality in their study of Bulgarian multiple wh-questions. They suggest that within a system in which Superiority is subsumed under Relativized Minimality the preservation of the order of the wh-phrases can be ensured through the requirement where only a whole chain, not just a link of a chain, counts as an intervener. This requirement is a modification of Chomsky‘s principle which states that only the head of an A-chain (equivalently, the whole chain) blocks matching under the Minimal Link Condition. (Chomsky, 2001:17). Thus, Krapova and Cinque (2005) modify Rizzi‘s (2001) definition given in (1) as ‗no chain can intervene between the trace and the target, if this chain bears the same features as the target.‘ 1.3. Criterial freezing Rizzi (2007) provides empirical evidence supporting the view that A‘ chains are delimited by s-selectional and scope-discourse (criterial) positions which are of immediate relevance for the interpretive systems: 6. ....____Xcrit ...... ____Xs-sel ...... Rizzi (2006) argues that the two positions in (6) have two critical properties (Rizzi, 2007:146): a. They are unique (i.e. there is only one position of each kind per chain) b. They delimit the chain (i.e. the chain includes no position lower than the Sselectional position and no position higher than the Criterial position) The connection between the s-selectional position and the criterial position is ensured by successive movement. Lasnik & Saito (1984; 1992), Rizzi (2006), Rizzi & Shlonsky 5 (2007) argue that there is a principle of Criterial Freezing, which has an effect of terminating the chain as soon as the first Criterial position is reached. 7. Criterial Freezing: A phrase meeting a Criterion is frozen in place. Rizzi (1996, 1997) interprets the Criterion as the requirement on the creation of a local Spec-head configuration which is then passed on to the interface system where a relevant interpretive instruction is triggered (Rizzi, 2007:146). This principle makes an expression available to the interface as soon as the expression reaches a Criterial position. One of the examples of Criterial freezing is provided in (8), where a wh-phrase satisfying the Q Criterion in the embedded clause cannot continue movement to the matrix CP (Rizzi, 2007: 147). 8. a. Bill wonders [which candidate Q [ you voted for t ]] b.* Which candidate does Bill wonder [ t‘ Q [ you voted for t ]] Later, Rizzi (2007) reconsiders the definition of Criterial Freezing in (7) and concludes that Criterial Freezing does not freeze the whole phrase, therefore satisfying the criterion, but instead, it should be restricted to the carrier of the relevant criterial feature. Adopting Chomsky‘s (2000) terminology, Rizzi (2007) makes use of terms Criterial Probe (a left-peripheral head endowed with critical features) and Criterial Goal (an element carrying the same features and attracted to the Spec of Criterial Probe) and proposes a new definition of Criterial Freezing (Rizzi, 2007:147). 9. Criterial Freezing: In a criterial configuration, the Criterial Goal is frozen in place. Thus, according to this new definition, in the criterial configuration, only the element carrying the crucial feature is frozen in place, while other elements of the phrase are available for further movement. 6 1.4. The organization of the Dissertation The thesis is organized as follows. In Chapter 2, I examine the properties of whquestions in Slavic languages, such as Bulgarian, Serbo-Croatian and Russian. I discuss the phenomena of Superiority, D-linking and wh-islands in the above-mentioned languages from a comparative perspective. This chapter outlines Bošković‘s (2002) proposal regarding Superiority effects in Slavic languages and provides empirical evidence against the common assumption that Russian is a wh-in-situ language, which does not have wh-movement and thus does not show Superiority effects. Chapter 3 starts with an outline of the structure of the Bulgarian left periphery within the Cartographic framework. Krapova and Cinque (2005) offer a detailed analysis of the Bulgarian data and provide evidence suggesting that the order of whphrases in the left periphery reflects the order prior to wh-movement, which is captured by the principle of Relativized Minimality. After introducing the Bulgarian data, I follow Krapova and Cinque‘s lead and apply their analysis to the Russian data, providing a comparative analysis of wh-order in Russian and Bulgarian. I demonstrate that both languages make use of residual hierarchy of animacy, i.e. [+human] wh elements will prevail other [-human]. Once the ordering is distinguished and before suggesting a syntactic analysis of Russian wh-questions, the Chapter offers a discussion of the four main approaches (the adjunction analysis, the wh-cluster analysis, the multiple [Spec;CP] analysis, and the split CP analysis) offered in the literature in order to capture the syntactic peculiarities of multiple wh-questions. This Chapter concludes with a proposal for the syntactic analysis of Russian wh-questions by adopting the Cartographic approach proposed by Rizzi (1997). Chapter 4 is devoted to the so-called why-wh-elements, i.e. wh-adjuncts whose interpretation is akin to the English why. First, I study the neutral why-wh-elements začem and počemu, which distinguish purpose from reason varieties of why. After having offered a syntactic analysis for the two whys, I discuss aggressively non-Dlinked (kakogo čërta, kakogo figa – „what the hell‘) and nominal (čto, čë and čego) wh- 7 elements meaning why. The final section of the Chapter offers a discussion and an analysis of the coordination of wh-elements. Chapter 5 studies the phenomenon of that-trace effect. Russian, unlike other well studied languages, exhibit the absence of subject-object asymmetry with respect to thattrace effect. For example, in English only extraction of a wh-object is possible over the complementizer that. However, in Russian, neither object nor subject can be extracted over the indicative complementizer čto. Nonetheless, it is still possible to extract both a wh-subject and a wh-object over a subjunctive complementizer čtoby. This Chapter presents previous approaches discussed in the literature to account for that-trace effect and offers an analysis of indicative and subjunctive clauses in Russian within the Cartographic framework. Chapter 6 focuses on the syntactic properties of yet another type of interrogatives in Russian, known in the literature as wh-scope marker questions or questions with partial wh-movement. The Chapter offers discussions of the Direct Dependency Approach (van Riemsdijk, 1982, McDaniel, 1989), The Indirect Dependency approach (Dayal, 1994, 2000), and the Mixed Dependency Approach (Mahajan, 1996, 2000, Stepanov, 2000, Herburger, 1994). I investigate the status of the Russian wh-scope marker and come to the conclusion that kak is a wh-expletive, endowed with a [+wh] feature. Finally the Chapter offers a syntactic analysis of the Russian wh-scope marking construction within the Cartographic framework. Chapter 7 provides concluding remarks and the Bibliography concludes the dissertation. 8 CHAPTER 2. MULTIPLE WH-FRONTING LANGUAGES AND SUPERIORITY EFFECTS 2.1. Introduction Slavic multiple wh-fronting languages (henceforth MWF languages) and Superiority effects in these languages have attracted much attention from linguists. Multiple whmovement in Slavic languages has been thoroughly studied in the literature by a number of linguists, e.g. Bošković (1997, 1998, 2001, 2002), Stjepanoviç (1995) for SerboCroatian; Alboiu (2002), Laenzlinger and Soare (2005) for Romanian; Stepanov (1997), Kazenin (2002), Grebenyova (2004), Bošković (2002) for Russian; Pesetsky (2000), Rudin (1988), Krapova and Cinque (2005) for Bulgarian. Traditionally MWF languages are divided into two (Rudin, 1988) or three (Bošković, 2002) types with respect to the Superiority effect. Rudin differentiates two types of MWF languages. The languages of the first type are represented by Bulgarian and Romanian, which display strict ordering of whquestions, where all fronted wh-elements are located in [Spec, CP]. In languages of the second type, such as Serbo-Croatian, Czech, Polish, Russian and Ukrainian, the Superiority effect appears in only some cases and only the first element is moved to [Spec, CP]. Bošković (2002) offers a different approach to MWF languages. He divides them into three types: Superiority effect in all contexts (Bulgarian), Superiority effect in some contexts (Serbo-Croatian), and absence of Superiority effect in all contexts (Russian). Russian is classified differently in the two approaches mentioned above. This makes it interesting to study the language in detail with respect to both Bulgarian and Serbo-Croatian, and to determine which class the language actually belongs to. 9 It is known from the literature that Russian is a language with fronted whphrases and is argued by Bošković (2002) and Stepanov (1997) to be a language which does not exhibit Superiority effects in any context, including embedded clauses and long distance constructions, contrary to Serbo-Croatian. Moreover, they claim that Russian shows a free order of subject and adjuncts and subject and objects wh-phrases.1 10. a. Kto kak postaraetsja, u togo who how will-try tak i polučitsja. that-one that-way and will-come out ‗The way whoever tries, that way it will come out.‘ b. Kak kto postaraetsja, tak how who will-try u togo i polučitsja. that-one that-way and will-come out ‗The way whoever tries, that way it will come out.‘ 11. a. A ètomu čeloveku kto kogo predstavil? and that man who whom introduced ‗And to that man, who introduced whom?‘ b. A ètomu čeloveku kogo kto and that man predstavil? whom who introduced ‗And to that man, who introduced whom?‘ Interestingly, Stepanov (1997) cites two instances from Rudin (1996) where she argues that examples repeated here as (12 and 13) show a preferred word-order. 12. a. Kto kogo videl? who whom saw ‗Who saw whom?‘ b. ??Kogo kto videl? whom who saw ‗Who saw whom?‘ 13. a. Kto čto videl? who what saw ‗Who saw what?‘ 1 Examples are taken from Bošković 2002 where he refers to Stepanov‘s judgements, which myself and other native speakers strongly disagree with, i.e. (10b) and (11b) are ungrammatical. 10 b.*Čto kto videl? what who saw ‗Who saw what?‘ Stepanov (1997) assumes that this asymmetry is not characteristic of Russian syntax, but instead is caused by non-syntactic reasons. Stepanov regards it as some low-level phonological constraint excluding sequences of phonologically similar or identical whconstraints; and therefore, he disregards these contrasts. Stepanov claims that the order of wh-elements in wh-matrix and embedded clauses is free, which he straightforwardly interprets as the absence of Superiority effect. In terms of an economy-based approach, this conclusion suggests that no Attract wh takes place and consequently wh-phrases do not raise to [Spec, CP] to check overtly the interrogative Q feature. Furthermore, Stepanov (1997) argues that Russian interrogative C has a weak [+wh] feature, and therefore it does not trigger overt whmovement even though it is inserted overtly. Bošković (2002) points out that Superiority serves as an indicator for when MWF languages involve wh-movement, and according to the data presented by Stepanov (1997). Russian does not have any overt movement to [Spec, CP] and the [+wh] feature of C is weak. Ignoring the fact that wh-elements in Russian are fronted, Stepanov assumes that Russian patterns with wh-in-situ languages (like Chinese or Japanese) in the absence of an overt wh-movement, and therefore he suggests treating Russian as a wh-in-situ language. The reason for fronting wh-elements in Russian, according to Stepanov, is focus checking. He assumes that Russian wh-phrases are inherently focused, and hence fronting occurs for checking their focus feature, not a wh-feature. In this dissertation, I argue against the approach proposed by Stepanov and Bošković and offer another analysis of Russian wh-questions. I claim that wh-elements in Russian are not focus-driven, but have a strong wh-feature in C, which triggers movement. I show that Russian exhibits Superiority effects and more so patterns with Bulgarian and Serbo-Croatian than with Chinese or Japanese in this respect. In order to be able to judge the claims made by Bošković and Stepanov I consider it to be important to refer to the notion of Superiority effect, and only then study this phenomenon in Serbo-Croatian and Bulgarian, given that Russian might reflect the syntactic properties of wh-elements of one of these languages. 11 2.2. Defining Superiority effect The Superiority effect occurs in multiple wh-questions where more than one whelement is relevant to the answering patterns for the question (Pesetsky, 2000). In this case the syntax needs to decide which pattern of wh-movement will be utilized within the question and the Superiority effect is considered to be a restriction on this decision. The Superiority effect behaves differently in different languages. Thus, one of the examples of this phenomenon is observed in English, where only one wh-element moves overtly to a higher position. In (14a) the leftmost wh-element moves to the higher position, and in (14b) the lower wh-element what crosses who and raises above it, leading to ungrammaticality. 14. a. Who ____ said what? b. *What did who say ____? This phenomenon was first observed by Kuno and Robinson (1972), and it is stated as a property of a wh-movement that does not allow a wh-element to be crossed by another wh-word. Kuno and Robinson‘s constraint is presented in Pesetsky (2000:15) as: A wh-word cannot be preposed crossing over another wh. Chomsky (1973) generalizes Kuno and Robinson‘s constraint and accounts for it in terms of a condition on rule application that forbids applying a given rule R to an expression Y if there is a superior expression Z to which it equally applies (Hornstein, 1995). No rule can involve X, Y in the structure…X…[α…Z…-WYV…] where the rule applies ambiguously to Z and Y and Z is superior to [m-commands] Y. The category A is superior to the category B if every major category dominating A dominates B as well but not conversely. 12 Later, Chomsky (1995) reduces this condition and refers to it as an Attract Closest (henceforth AC) rule, where ‗movement is viewed as triggered by particular features of a ‗target‘ head K‘ (Pesetsky, 2000:15). α can raise to target K only if there is no legitimate operation Move β targeting K, where β is closer to K. (Chomsky, 1995:296) This condition explains the contrast between (15a) and (15b), where who is superior to what, and according to the Attract Closest rule, who moves to C given that it is the closest element. 15. a. Whoi did John tell ti that he should buy what? (Bošković, 1998:5) b. *Whati did John tell who that he should buy ti? Hornstein (1995) reviews approaches to Superiority and reduces the contrast in (15) to some version of Empty Category Principle2 (henceforth ECP). He suggests that Superiority effects are manifestations of Weak Cross Over3 (henceforth WCO) and that the unacceptability of *What did who say? reflects an illicit pronoun-building configuration at LF (Hornstein, 1995:124). Hornstein adopts the WCO account of quantifier/wh interactions. He suggests that configurations that display Superiority effects are parallel to those that forbid pairlist readings. Thus (16a) has a pair-list reading which implies an answer listing the buyer and buyee pairs, i.e. Mary bought a car, John bought a house etc. 16. 2 a. Who bought what? Empty Category Principle (introduced by Chomsky, 1981) is a principle which requires that empty categories be properly governed, where proper government is defined as follows: 3 A properly governs B iff A theta-governs B or A antecedent-governs B. A theta-governs B iff A governs B and A theta-marks B. A antecedent-governs B iff A governs B and is coindexed with B. WCO follows from the leftness condition: A wh-trace cannot be coindexed with a pronoun to its left (Haegeman, 1994:417). 13 b. *What did who buy? He assumes that a pair-list reading in multiple wh-questions is generated in the same way as in (17) involving a quantifier: 17. a. What did everyone say? b. Who said everything? 18. [CP Whati [IP everyonej [IP tj say [proj ti]]]] In (18), pro is bound by a quantifier everyone, and this bound structure is a generator, since it can generate a list. Thus a pair-list reading is a function of two properties: 1) certain whs have a functional reading that involves bound pronouns, 2) pronouns bound by quantifiers that are generators can produce a pair-list reading (Hornstein, 1995:125). Based on these assumptions Hornstein assumes that the wh-element who (in 16a) occupies [Spec, CP] and functions like everyone on the pair-list reading, i.e. who is a quantifier that generates the list and wh-in-situ what is interpreted functionally. 19. [Whoi [ti bought [proi N]]] 20. *[Whati [[proi N] bought ti]] (16b) at the same time will get the structure in (20), where proi is co-indexed with a variable ti on its right, which results in WCO violation. Hornstein‘s (1995:128) account ties the properties of three different construction types together: WCO structure, pair-list readings of wh/quantifier constructions, and multiple wh-questions. He expects wh-questions to be well-formed in the cases where pronoun binding is licit and a pair-list interpretation is available. The Superiority effect is most apparent in multiple wh-fronting languages, where all wh-phrases have to move to a higher position. I will come back to the MWF languages below, but first I will mention the cases where the Superiority effect is restricted because of a number of properties in the syntactic structure. 14 2.2.1. D-linking It has been observed in the literature that D-linking (after Pesetsky 1987) produces exceptions to the Superiority effect. According to Pesetsky (2000:16), D-linking is a phenomenon ―when a wh-question asks for answers in which the individuals that replace the wh-phrases are drawn from a set that is presumed to be salient to both speaker and hearer, the multiple questions can appear to violate AC.‖ The answers to that kind of question are supposed to be drawn from a set of individuals previously mentioned in the discourse or when this set is clear for both speaker and hearer (Pesetsky, 2000). According to Pesetsky (2000) and Krapova and Cinque (2005), the inherently D-liked wh-phrases are ‗which‘- type phrases. Consider the examples in (21) where the Superiority effect disappears and both wh-phrases can move to a higher position: the crossover of the higher wh-phrase by the lower one does not lead to ungrammaticality. The context is obviously known for speaker and hearer and thus it is silent in the question. 21. a. Whichi person ti bought which book? b. Whichi book did which person buy ti? The elimination of the Superiority effect with D-linking is observed in many other languages. For example, consider German in (22) or Bulgarian in (23) which are highly rigid languages. 22. a. Welcher Student hat welches Buch gelesen? which student had which book read ‗Which student read which book?‘ b. Welches Buch hat welcher Student gelesen? which book had which student read ‗Which book did which student read?‘ 23. a. Koja kartina na koj prijatel si (ja) posvetil? which painting to which friend have-you it dedicated 'Which painting have you dedicated to which friend?' 15 b. Na koj prijatel koja kartina si (ja) posvetil? to which friend which painting have-you it dedicated 'Which painting have you dedicated to which friend?' In (22) and (23) the Superiority effect disappears and both wh-elements can freely move to a higher position. 2.2.2. More than two-wh-phrases Kayne (1984) first observed that the ECP violation and Superiority effect are mitigated with the addition of an extra wh-element. He offers examples repeated here in (24) where (24a) is ungrammatical due to ECP violation, but the addition of a third wh-element neutralizes the violation (24b) (Kayne, 1984:176). 24. a. *We are trying to find out which man said that which woman was in love with him. (Vs. ?We are trying to find out which man said that he was in love with which woman.) b. ?We are trying to find out which man said that which woman was in love with which boy. Improvement is also observed in the domain of the Superiority violation: 25. a.*I‘d like to know where who hid it. (vs. I‘d like to know who hid it where.) b. ? I‘d like to know where who hid what. Kayne suggests that the example in (25a) could require the wh-phrase in A-position to be linked to the one in Comp, thus subject to the Connectedness Condition for Empty Category (Kayne, 1984:175). 16 Hornstein (1995) analyses the lessening of the Superiority effect with the addition of en extra WH from the point of view of Weak Crossover (WCO)4. 26. a.*What did who buy there? (Hornstein, 1995:143) b. ?What did who buy where? He offers a structure for (26a) as in (27) where WCO is violated, i.e. what is moved across (who=) [proj person] and a wh-trace of what is co-indexed with (who=) [proj person], which yields a WCO violation. 27. Whatj [(who=) [proj person] bought tj there] According to Hornstein (1995:144), a wh in CP acts as a generator for the pair-list reading and whs-in-situ are interpreted functionally, i.e. they involve implicit pronouns linked to the generator‘s variable. Now consider (26b), which has a structure in (28). Hornstein (1995:144) argues that with adopting the linking version of WCO,5 the presence of a second functionally interpreted wh-element will lessen the effect of WCO. 28. Whatj [(who=) proj person] bought tj [(where=) proj place] Thus in (28), both who and where are interpreted functionally; who – as a person and where – as a place and a pronoun is not linked to a variable on its right, hence no violation of WCO (in Hornstein‘s terminology). Pesetsky (2000) studies this phenomenon in Bulgarian and offers an analysis, which I discuss in Section 2.3.3. 4 Weak crossover (WCO) appears when a wh-phrase moves to the front of the sentence and "crosses over" a coindexed pronoun. 5 Hornstein (1995:100) revises the notion of WCO and replaces indexation with linking: A pronoun cannot be linked to a variable on its right 17 To summarize, the Superiority effect occurs in languages where more than one whelement is relevant to answering the question, forcing the syntax to move one of the whphrases to a higher position. According to the Superiority effect and the Attract closest rule (which subsumes the Superiority condition) the left-most wh-phrase will raise to [Spec, CP] and the lower wh-element will stay in-situ (for English) or move to a lower position from the leftmost wh-phrase (for MWF). The Superiority effect can be mitigated if a multiple wh-question contains either D-linked wh-elements or more than two wh-phrases. In the next section I introduce some properties of Bulgarian, Serbo-Croatian, and Russian wh-elements and mention the accounts proposed in the literature with respect to the Superiority effect in these three languages. 2.3. Some properties of Bulgarian, Serbo-Croatian and Russian wh-questions 2.3.1. Constituency of wh-phrases Bošković (2002) follows Rudin (1988) in observing the difference in constituency of wh-phrases in Bulgarian and Serbo-Croatian. According to their observations, SerboCroatian, contrary to Bulgarian, allows the intervention of clitics, adverbs, and parentheticals between fronted wh-phrases (examples from Bošković, 2002:1-2). 29. a. Ko mu je šta dao? (Serbo-Croatian) who him is what given ‗Who gave him what?‘ b. Ko je prvi koga udario? who is first whom hit ‗Who hit whom first?‘ c. Ko, po tebi, šta pije? who, according to you what drinks 18 ‗Who, according to you, drinks what?‘ 30. a. *Koj ti e kakvo kazal? (Bulgarian) who you is what told ‗Who told you what?‘ b. *Zavisi ot tova, koj prâv kogo e udaril. depends on it who first whom is hit ‗It depends on who hit whom first.‘ c. ?*Koj, spored tebe, kakvo e kazal? who according to you what is said ‗Who, according to you, said what?‘ Rudin (1988) argues that impenetrability of the domain of wh-phrases in Bulgarian is evidence that they form a constituent which does not allow any intervention of other elements. This is contrary to Serbo-Croatian, which allows insertion of lexical material between the wh-elements, indicating that they do not form a constituent. It is interesting to note that Cinque and Krapova in their paper (2005:8) provide Bulgarian data where the intervention of parentheticals is possible. Consider the examples in (31) and (32): 31. a. Koja kartina, spored tebe, kâde ja e risuval tozi xuţdonik? which painting, according to you, where it is painted this artist ‗According to you, which painting did this artist paint where?‘ b. Koja kniga, spored tebe, na koj prijatel da ja dam? which book, according to you, to which friend should it give-I ‗According to you, which book should I give to which friend?‘ 32. a. Na kogo za Boga kakvi kartini pak šte davaš? to whom for God‟s sake what paintings again will give-you ‗What paintings will you again be giving to whom?‘ b. *Kakvi kartini na kogo za Boga pak šte davaš? what paintings to whom for God‟s sake again will give-you c. ?/*Na kogo kakvi kartini za Boga pak šte davaš? to whom what paintings for God‟s sake again will give-you 19 Examples (31 and 32) show that Bulgarian allows an intervention of parentheticals between wh-phrases. Moreover, examples (32b and 32c) demonstrate that the parenthetical za Boga cannot occur lower than kakvi kartini and should follow na kogo, which argues for the non-constituency of the wh-phrase. Furthermore, the example in (30c) presented by Rudin as unacceptable is judged by native speakers as a grammatical sentence; the ungrammaticality of (30a) can be explained by the property of clitic clusters, which should be in the verbal domain and (30b) can be explained by the property of the ordinal numbers (p.c. Iliana Krapova) which should follow the verb. This property is also observed in Russian: 33. Zavisit ot togo kto (*pervym) kogo udarit pervym. depends from that who (first) whom hits first ‗It depends on who hit whom first.‘ Bošković (2002) also mentions a fact regarding the contrast between (29a) and (30a) where he points out different properties of Serbo-Croatian and Bulgarian clitics. SerboCroatian clitics can occupy only the second position of the sentence and impose no requirements on their host, while Bulgarian clitics are verbal clitics. This feature cannot be considered sufficient either. Thus, the distinction between Serbo-Croatian and Bulgarian based on the impenetrability of wh-phrases is not strong enough. Russian, like Serbo-Croatian and Bulgarian, allows the intervention of lexical materials, such as nouns, pronouns, and parentheticals between fronted wh-phrases. Consider below the Russian examples, where (34) shows insertion of a pronoun, (35) insertion of a parenthetical, and (36) of a noun. 34. Kto tebe čto skazal? who you what said ‗Who said to you what?‘ 35. Kakaja kartina, po tvoemu mneniju, gde byla narisovana. which picture to your opinion where was painted ‗According to you, which painting was painted where?‘ 36. Komu Maša kogda zvonila? whom Masha when called ‗Whom Maša called when?‘ 20 Insertion of a lexical material between the wh-elements indicates that wh-elements do not form a constituent in Russian, Bulgarian, and Serbo-Croatian. 2.3.2. Wh-islands Rudin (1988) distinguishes Serbo-Croatian and Bulgarian by extraction out of whislands, indicating that Bulgarian allows more than one wh-phrase in [Spec, CP], while Serbo-Croatian does not. 37. a. Vidjah edna kniga, kojatoi se čudja saw-1s a koj znae koj prodava ti. (Bulgarian) book which wonder-1s who knows who sells ‗I saw a book which I wonder who knows who sells.‘ b.*Vidio sam knjigu kojui se pitam ko zna ko prodaje ti.(Serbo-Croatian) seen am book which wonder-1s who knows who sells ‗I saw a book which I wonder who knows who sells.‘ Rudin claims that in the Bulgarian example, kojato can escape the Wh-island constraint by moving through the embedded [Spec, CP], occupied by koj. Serbo-Croatian, on the other hand, does not allow more than one wh-element to be located in [Spec, CP], which is why movement of koju does not avoid the Wh-island constraint (Bošković, 2002:3). It should be mentioned that the literature generally focuses on argument extractions, ignoring adjunct ones. The example in (38), where wh-adjunct extractions out of wh-islands leads to ungrammaticality, contradicts the claim made by Rudin that Bulgarian is not sensitive to the wh-island constraint (Bošković, 2002:3). In (35) dali forms a strong island constraint. 38. *priãinata, poradi kojatoi [Ivan znae dali the-reason for which Boris e zaminal ti] Ivan knows whether Boris is left ‗the reason for which Ivan knows whether Boris left.‘ 21 Bošković (2002) also notices that Bulgarian exhibits wh-island effects with non-Dlinked wh-questions, as compared to the D-linked wh-extractions in (37a) (the example is taken from Rudin, 1988): 39. a.*Kakvoi se čudiš koj znae koj prodava ti? what wonder-2s who knows who sells ‗What do you wonder who knows who sells?‘ b.*Kakvoi se čudiš koj prodava ti? (p.c. Polya Vitkova) what wonder-2s who knows Russian, like Serbo-Croatian and Bulgarian, does not allow extraction out of whislands, making the languages similar. Although, unlike Bulgarian, Russian does not allow extraction of D-Linked wh-elements out of wh-islands (40b), it is mitigated when both wh-elements are fronted (40b‘). 40. a.*Čtoi ty dumaeš‘ kto znaet kto prodaet ti? what you think who knows who sells ‗What do you wonder who knows who sells?‘ a.‘*Chtoi ty dumaeš‘ kto prodaet ti? what you think who sells b.*Kakuju knigui tebe posovetovali kakomu rebenkuj podarit‘ ti tj? which book you were advised to which child to present ‘ b . Kakuju knigui kakomu rebenkuj tebe posovetovali podarit‘ ti tj? which book to which child you were advised to present To summarize thus far, data concerning the penetrability of wh-phrases and wh-islands make the languages, to some extent, look similar. It is essential to look at the behaviour of the languages with respect to Superiority effects, which might allow us to propose an analysis that will account for the Superiority effect in all three languages. 22 2.3.3. Superiority effects in Bulgarian Bulgarian is well known for its obligatoriness of movement of all wh-phrases to a higher position. The highest wh-element has to move to [Spec, CP] first. Consider the examples in (41), where only the highest wh-phrase is allowed to move first (41a). Movement of kakvo over koj in (41b) leads to ungrammaticality. 41. a. Koj kakvo vizda? who what sees ‗Who sees what?‘ b.*Kakvo koj vizda? what who sees ‗What does who see?‘ According to Pesetsky (2000), in English the Superiority effect follows from Attract Closest, and in Bulgarian, the left-most (i.e. the closest) wh-element will be also the first to move. Superiority effect in Bulgarian (Pesetsky, 2000:22) The leftmost wh-phrase in a Bulgarian multiple question is the wh-phrase that moves overtly in the corresponding English multiple question. Since Bulgarian is a MWF language, all wh-phrases will move overtly to the left periphery of the sentence, indicating that after the left-most element is moved, the other wh-phrases in the sentence ―must have ‗tucked‘ in underneath the first phrase, forming a lower specifier‖ (Pesetsky, 2000:22). 42. a. Koj kǔde C ____ udaril Ivana___? 1 who where 2 hit Ivan ‗Who hit Ivan where?‘ 23 b. *Kǔde koj C____ udaril Ivana___? 2 where who 1 hit Ivan ‗Who hit Ivan where?‘ In (42a), the left-most koj moves first, and then kǔde is tucked in underneath it; in (42b), on the other hand, the lower wh-element moves first and thus crosses over the higher wh-element koj, which leads to violation of the Superiority effect, rendering the sentence ungrammatical. Richards (1997), studying multiple wh-movement in Bulgarian, points out that second instances of movement should ―tuck in,‖ and in addition to the Attract Closest rule, the Shortest Move condition should be obeyed as well. This is essential, inasmuch as the first element will be moved to C by Attract Closest and the second element must be raised higher than the first one if there is no condition such as Shortest Move. Obeying this condition, the second element will create a specifier lower than the first one. ―In this way, the interaction of Attract Closest and Shortest Move causes the command relations between two wh-phrases after movement to mirror their order before movement‖ (Pesetsky, 2000:23). So far I have discussed only multiple questions with two overtly moved whphrases. It would be interesting to see if the same rules apply for wh-questions with three wh-elements. What is expected is the same mirror order of wh-phrases before movement. Nonetheless, Pesetsky (2000) and Bošković (2002) notice, this is not the case. They report the possibility of the following orders: wh1 wh2 wh3 and wh1 wh3 wh2. 43. a. Koj1 kogo2 kak3 udaril? who whom how hit b. Koj1 kak3 kogo2 udaril? who how whom hit (43) shows that in wh-questions with three wh-phrases the first wh-element is attracted to a higher position triggered by Attract Closest and thus pays the ‗Attract Closest tax,‘ while the other two elements do not have to obey this condition anymore. Richards (1997) develops this constraint into the Principle of Minimal Compliance (PMC), which 24 implies that the ‗attractor-oriented‘ constraint on movement applies only to the first movement operation (Pesetsky, 2000:25). Principle of Minimal Compliance (henceforth PMC) (Pesetsky, 2000:36): For any dependency D that obeys constraint C, any elements that are relevant for determining whether D obeys C can be ignored for the rest of the derivation for purposes of determining whether any other dependency D obeys C. An element X is relevant to determining whether a dependency D with head A and tail B obeys constraint C iff a. X is along the path of D (that is, X=A, X=B, or A c-commands X and X c-commands B), and b. X is a member of the class of elements to which C makes reference. PMC for our purposes (Pesetsky, 2000:36): Once an instance of movement to α has obeyed a constraint on the distance between source and target, other instances of movement to α need not obey these constraints. Thus PMC explains the acceptability of the order of wh-elements in example (43), where the first wh-phrase obeys the AC constraint and the other wh-phrases do not need to obey it anymore. It should be noted that although AC is subject to PMC, Shortest Move, on the other hand, is not. This constraint does not govern the distance, but instead it chooses the exact position which is near a given attractor where the movement takes place (Pesetsky, 2000:27). To summarize, Bulgarian shows a strict ordering of wh-phrases (unless three whphrases or D-linked wh-elements occur) and is sensitive to the Superiority effect. 25 The Superiority effect in Bulgarian requires the movement of the leftmost whphrase in a multiple wh-question to a higher position and the rest of the wh-phrases are ‗tucked in‘ below the first moved wh-element. Wh-phrases in Bulgarian obey the Attract Closest and Shortest Move conditions. Attract Closest is subject to the PMC, according to which the first wh-element in a question with three wh-phrases satisfies the Attract Closest condition and the rest do not have to obey it. Shortest Move does not govern the distance (and thus is not subject to PMC), but rather chooses the exact position for the movement, which is near a given attractor. Now I turn to the Superiority effect in Serbo-Croatian to see if this language is subject to the same constraints as Bulgarian. 2.3.4. Superiority effects in Serbo-Croatian The second type of MWF language discussed by Bošković (2002) is exemplified by Serbo-Croatian, which exhibits the Superiority effect in some contexts: long distance questions and embedded questions (Bošković, 2002:8). 44. a. Ko si koga tvrdio da je istukao? who are whom claimed that is beaten ‗Who did you claim beat whom?‘ b. *Koga si ko tvrdio da je istukao? whom are who claimed that is beaten ‗Who did you claim beat whom?‘ Bošković (2002) claims that examples in (44) exhibit a Superiority effect because whphrases have to move in these kinds of sentences; while in short-distance matrix questions (45) a wh-phrase does not undergo overt wh-movement to CP, although it still must be fronted for independent reasons, therefore demonstrating the lack of the Superiority effect in this context. 26 45. a. Ko koga voli? who whom loves ‗Who loves whom?‘ b. Koga ko voli? whom who loves ‗Who loves whom?‘ Bošković (1998:9) comes to a simple conclusion: ―Whenever we have wh-movement in Serbo-Croatian Superiority is operative.‖ Since all wh-elements must be fronted, as in (46), Bošković (2002) claims that they are fronted for independent reasons. This movement is not driven by the strong [+wh] feature of C, given that all wh-elements must move in languages like SerboCroatian, and only movement of one wh-phrase should be sufficient for checking the strong [+wh] feature. 46. a. Ko šta kupuje? who what buys ‗Who buys what?‘ b.?/*Ko kupuje šta? who buys what ‗Who buys what?‘ According to Bošković (2002), there is no reason to differentiate the movement of ko from šta by requiring ko to be moved to [Spec, CP]. He suggests that both wh-phrases undergo a movement which is different from a wh-movement. Bošković (2002) argues that in matrix short-distance null C questions (like in (46)), the merger of C happens only at LF, and this allows Serbo-Croatian to avoid overt wh-movement to [Spec, CP]. He attempts to explain that non-wh-fronting of wh-phrases (i.e. movement that is not motivated by checking a strong [+wh-feature] on C) is not sensitive to the Superiority Condition. Bošković follows Stjepanović (1995), who argues that the reason for non-wh-fronting of wh-phrases is focus. She argues that Serbo-Croatian wh-phrases are inherently contrastively focused and therefore must undergo focus movement (Bošković, 2002:11). So, according to Bošković, this non-whmovement is focus movement. Similar claims have been made for a number of 27 languages, such as Aghem, Basque, Hungarian, and Korean (Horvath 1986, Rochemont 1986, Kim 1997). Horvath (1986) argues that wh-phrases move to the position for contrastively focused phrases if such a position is present in the language. According to Bošković (2002:11), this suggestion seems plausible, ―given the similarity in the interpretation of wh-phrases and contrastively focused phrases.‖ Bošković (2002) makes an attempt to argue for non-wh-fronting of wh-phrases for Bulgarian, suggesting it to be an instance of focus-movement. In order to support his assumption, he provides an example in a footnote (page 12), arguing for the unacceptability of leaving the wh-phrase in-situ in an echo-question. My informants find this question with an echo reading absolutely acceptable, allowing the wh-phrase to stay in-situ; hence, this data works against Bošković‘s proposal. 47. Ivan e popravil kakvo? ivan is fixed what He suggests that while in Bulgarian wh-movement affects only one wh-phrase and the others are focused-moved, only the first wh-moved wh-element will be subject to the Superiority condition, which implies that the highest wh-phrase should move first and the movement of others should not matter6. 2.3.5. Why is focus movement of wh-phrases not subject to the Superiority effect? According to Bošković‘s (1998) proposal, the first element is moved to [Spec, CP], and thus it is the only one which is subject to Superiority effect, while other elements are focus triggered and insensitive to Superiority. The questions that arise are why there should not be any Superiority effect in focus-moved wh-questions and what the difference between wh- and focus-movement is. Bošković (1998a) argues that wh-movement differs from focus-movement with respect to where the attractor lies. It is assumed that with the former the attractor lies in 6 Krapova and Cinque (2005) argue for a fixed order of wh-phrases. Their analysis is discussed in 3.1. 28 the target, which explains fronting of only one wh-phrase overtly to [Spec, CP]. The first element moves to [Spec, CP] in order to check strong [+wh] feature, and there is no need for other wh-elements to undergo wh-movement. Regarding the remaining wh-elements, after the first one is moved to check the strong wh-feature, they must undergo focus-movement, which indicates that the feature driving movement lies in the wh-phrases, but not in the target of the movement. He suggests that if the strong feature was in the target then it would be sufficient to move only one wh-element and leave the rest in-situ; but this is not the case. 48. a. Ko šta gdje kupuje? who what where buys ‗Who buys what where?‘ b.*Ko kupuje šta gdje? who buys what where c.*Ko šta kupuje gdje? who what buys where d.*Ko gdje kupuje šta? who where buys what Therefore, the main difference between the two movements is where the strong features driving the movements reside, i.e. in the target with wh-movement; and in the elements undergoing movement with focus-movement. Bošković (1998a) argues that this difference is responsible for the different behaviour of the movements with respect to the Superiority effect. Bošković (1998:15) suggests the following configuration: 49. Wh-movement F Wh-phrase1 Wh-phrase2 Wh-phrase3 +wh +wh +wh +wh strong weak weak weak 29 50. Focus-movement F Wh-phrase1 Wh-phrase2 Wh-phrase3 +focus +focus +focus +focus weak strong strong strong Subsequently, what he suggests is that in (49) the functional head F has a strong [+wh] feature which is checked through the shortest movement. As a result, the wh-phrase1 will have to move to F in order to check a strong [+wh] feature in the F; the movement of wh-phrase2 or wh-phrase3 to check a strong [+wh] feature will lead to the Superiority effect. Now, in (50), the strong features are in wh-phrases, and the relevant feature must be checked via the shortest movement, rendering the order of movement to FP irrelevant. Bošković claims that the derivation in which wh-phrase1 checks its focus feature before wh-phrase2, and the derivation in which wh-phrase2 checks its focus feature before wh-phrase1 are equally economical. The same nodes are crossed to check the strong focus feature, and that is why the Superiority effect does not manifest here. Bošković (1998:15) suggests that in Bulgarian, wh-phrases have a strong focus feature and C has a strong [+wh] feature, and none of these features can be checked before the interrogative C is introduced into the structure7. The difficulty here is understanding the order in which the features will be checked. Bošković argues that the order features checking in wh-phrases does not matter. Whether or not a strong focus feature of koj - ‗who‘ in (51) is checked first or last, the same number of maximal projections will be crossed. As for the strong feature of C, it should be checked by the highest wh-element. 51. a. Koj kogo kak e tselunal? who whom how is kissed ‗Who kissed whom how?‘ 7 Bošković assumes that wh1 will move to [Spec, CP], while the rest will be attracted to C (multiple [Spec, CP] structure). 30 b. Koj kak kogo e tselunal? who how whom is kissed ‗Who kissed whom how?‘ To summarize, Bošković proposes that in (51) koj moves first to [Spec, CP] in order to check a strong [+wh] feature in the target, and then kak and kogo can move to C in any order. The attractor of wh-movement in languages like English, where only one whelement is moved to C and the rest are left in-situ, is clearly an Attract one-F head. Moreover, in languages with more than one potential attractee, an Attract one-F head will always attract the highest attractee and an Attract all-F element, i.e. the focus attractor, will attract all focus feature-bearing elements. Therefore, the Superiority effect is expected with Attract one-F head but not with Attract all-F elements. 2.3.6. Problems for Bošković‟s analysis First of all, as Bošković (1998:15) himself admits, his account goes against the original formulations of Superiority made by Chomsky (1973) as well as other accounts (Pesetsky 1982, Lasnik and Saito 1992, etc.) No rule can involve X, Y in the structure…X…[α…Z…-WYV…] where the rule applies ambiguously to Z and Y and Z is superior to [mcommands] Y. The category A is superior to the category B if every major category dominating A dominates B as well but not conversely. This account would predict a Superiority effect in both wh- and focus-movement. Bošković (1998:15) mentions the problem with the accounts proposed before by suggesting that ―it is simply not possible to make the information concerning where the feature driving the movement lies, which determines whether a question will exhibit a Superiority effect, relevant to Superiority in a principled way.‖ 31 Another issue that would be interesting to refer to is the order of feature checking. In which order does wh1 check its features? If we assume that wh-elements have a weak wh-feature but a strong focus feature, does it mean that a wh-element will first move to check its strong focus feature, and only after that it will be triggered to [Spec, CP] to check its weak wh-feature against the strong wh-feature in the trigger? Or is it triggered straight to [Spec, CP], not allowing it to check its focus feature; and if so, what happens to this feature then? There are also problems arising with data. Recall again the main assumption made by Bošković for Bulgarian: the first wh-element undergoes wh-movement and is thus subject to the Superiority effect, and the rest simply move to check focus features. 52. a. Koj kogo kak e tselunal? who whom how is kissed ‗Who kissed whom how?‘ b. Koj kak kogo e tselunal? who how whom is kissed ‗Who kissed whom how?‘ The examples he offers in (52) can be alternatively analyzed as instances of the Principle of Minimal Compliance (discussed in section 2.3.3), which implies that in whquestions with three wh-elements the first wh-phrase obeys the AC constraint and the other wh-phrases do not need to obey it. Now consider the example with four whphrases (53)8, which one would expect to behave the same way as (52), following from Bošković‘s analysis, i.e. the highest wh-phrase should move first and the movement of others should not matter. 53. a. Koj kogo kade kak e tselunal? wh1 wh2 wh3 (Bulgarian) wh4 who whom where how is kissed ‗Who kissed whom how?‘ b.??/*Koj kade kak kogo e tselunal? who where how whom is kissed 8 I would like to thank Polya Vitkova for the Bulgarian data. 32 c.*Koj kak kade kogo e tselunal? who how where whom is kissed d.?Koj kade kogo kak e tselunal? who where whom how is kissed e.??/*Koj kogo kak kade e tselunal? who whom how where is kissed Apparently, the order of other wh-elements, after the first one is moved, does actually matter, as the examples in (53b-e) demonstrate. These data strongly contradict the assumptions made by Bošković. According to his theory, wh2 wh3 wh4 are focus moved elements and thus should be insensitive to any Superiority effect. Furthermore, Bošković‘s analysis does not account for D-linked wh-phrases (54), where changing of the order of wh-phrases does not lead to ungrammaticality; his analysis requires additional explanations. 54. a. Koi kartini koi hudojnik e narisuval? which pictures which artist (Bulgarian) is drawn b. Koi hudojnik koi kartini e narisuval? Another question that arises is how to treat wh-questions with wh-in-situ in echo questions (55), since, in Bulgarian, the second element kakvo has a strong focus feature which should trigger the wh-element to check it in C (55a). Note echo questions are easily found in other languages as well, for example Russian (55b), English (55c), and French (55d). 55. a.Ivan e popravil kakvo? ivan is fixed (Bulgarian) what b. Ty byl gde? (Russian) you were where c. You went where? (English) d. Il lit quoi? (French) he reads what 'He reads what?' 33 It is problematic for Bošković‘s analysis that a) the order of wh-phrases is rather strict in wh-questions with more than three wh-elements (which means that not only the highest wh-word is subject to the Superiority effect), b) examples with D-linked whphrases do not exhibit the Superiority effect and c) wh-in-situ is permitted in echo questions. 2.3.7. Superiority effects in Russian It is claimed by many linguists (Kazenin, 2002; Bošković, 1998; Zavitnevich, 2001; Rudin, 1996 etc.), that Russian does not exhibit the Superiority effect in multiple whquestions. Consider the examples in (56) where both variants are possible. 56. a. Gde kogo ty poslednij raz videl? where whom you last time saw ‗Whom did you see last time and where?‘ b. Kogo gde ty poslednij raz videl? whom where you last time saw ‗Whom did you see last time and where?‘ At the same time there are cases in Russian where obvious preference of a particular order of wh-words is observed. Stepanov (1997) observes cases in which the Superiority effect emerges, as in (57), where the wh-subject and the wh-inanimate object seem to have a fixed order. 57. a. Kto čto videl? who what saw ‗Who bought what?‘ b.*Čto kto videl? what who saw ‗Who bought what?‘ 34 Rudin (1996) evaluates (58b) as grammatically degraded, i.e. showing some Superiority effect. 58. a. Kto kogo videl? who whom saw ‗Who saw whom?‘ b. ??Kogo kto videl? whom who saw ‗Who saw whom?‘ Another observation is made by Zavitnevich (2001), who claims that začem – ‗why‘ occurring in multiple questions cannot appear in front of another wh-word, rather only following it. 59. a. Kuda začem on hodil? where why he went ‗Where did he go and why?‘ b. *Začem kuda on hodil? why where he went ‗Where did he go and why?‘ Stepanov (1997) argues for an absence of any constraints in embedded and long distance constructions. The data I have collected suggests the opposite: 60. a. Kak ty predpolagaeš kto komu podskaţal pravilnyj otvet? how you assume who whom offered correct answer ‗What do you assume, who gave a correct answer to whom?‘ b.*Kak komui ty predpolagaeš kto ti podskaţal pravilnyj otvet. how whom you assume who offered correct answer c.*Kak ktoi ty predpolagaeš ti komu podskaţal pravilnyj otvet. how who you assume whom offered correct answer d.*Ktoi kak ty predpolagaeš ti komu podskaţal pravilnyj otvet. who how you assume whom offered correct 35 answer Despite the data introduced in (57, 58, 59) above, it has been proposed that Russian is a language which does not exhibit the Superiority effect, i.e. it does not have any overt movement to [Spec, CP] and that the [+wh feature] of C is weak. Stepanov suggests treating Russian as a wh-in-situ language, drawing this conclusion based on interpretative aspects of multiple questions, i.e. single or multiple pair list answers. He presents an English example in (61) where who raises overtly to [Spec, CP], while what stays in-situ. 61. Who bought what? It has been noticed that questions as in (61) allow only a multiple pair list answer, like Bob bought a car, Mary bought a flower etc. A single pair answer, on the other hand, is not allowed. He suggests that the reason for this is associated with at least one of the wh-words being in [Spec, CP] in overt syntax (Stepanov, 1997). Bošković (1998) notices that if a language (e.g. Chinese and Japanese) allows both types of readings, then this language does not have ‗true‘ wh-movement, and the wh-phrase is not in [Spec, CP]. Taking this into consideration, he views Russian as allowing both multiple and single pair answers. This makes it similar to Chinese and Japanese, and thus a language where the wh-element does not undergo real wh-movement, but stays in situ. 62. Kto čto kupil? who what bought ‗Who bought what?‘ I should comment that neither my informants nor I perceive a single pair answer in this sentence, but instead a multiple pair list reading. The answer for multiple pair list readings involves listing propositions involving ordered pairs as in (63), whereas a single pair reading involves a single proposition (64) (Grebenyova 2004:11). 36 63. Who invited who to the dinner? (John is at a formal dinner where there are diplomats and journalists. Each journalist was invited by a different diplomat. John wants to find out all the details, so he asks the host) Answer: Mr. Smith invited Mr. Jones, Ms. Black invited Mr. Green, etc. 64. Which diplomat invited which journalist to the dinner? (John knows that a very important diplomat invited a very important journalist to a private dinner. John wants to find out all the details, so he asks the caterer) Answer: Ms. Black invited Mr. Smith. Note that English lacks single pair readings in non-D-linked multiple questions, and the same is true for Bulgarian. Conversely, it is available in Serbo-Croatian and Japanese. It was noticed by Hagstrom (1998) that in languages allowing for single pair readings, fronting of the lower wh-phrase over the higher one forces the single pair reading. Bošković refers to this phenomenon as Interpretive Superiority, meaning that instead of unacceptability of any reading, one of two potential readings gets lost (Grebenyova, 2004:12). 65. Šta je ko kupio? Serbo-Croatian (*PL/SP) what is who bought ‗Who bought what?‘ Therefore the absence of a single pair reading in a language would suggest the unacceptability of fronting the lower wh-element over the higher one. It was observed by Bošković (2001) that single pair readings are unavailable in multiple interrogatives where overt syntactic wh-movement takes place, in languages such as English, German, or Bulgarian. Bošković refers to Russian as a wh-in-situ language, which does not exhibit any wh-movement and allows single pair readings. As I mentioned above, Stepanov (1997) reports an example (repeated here as (66)) where he claims that Russian has a single pair reading and thus does not allow movement of wh-phrases to [Spec, CP] in overt syntax. 37 66. Kto čto kupil? who what bought ‗Who bought what?‘ Grebenyova (2004) argues against the judgments reported in Stepanov (1997) and claims that the above sentence does not get a single pair reading, according to the judgments she collected (I agree that only multiple pair list reading is available in this sentence). Moreover, it has already been noticed in the literature that Russian shows a Superiority effect which emerges when čto is moved over kto (67). 67. *Čto kto sprosil? what who asked ‗Who asked what?‘ According to Bošković‘s claim, Russian never shows the Superiority effect in any context and thus does not have syntactic wh-movement. The example in (67) proves the opposite, and as a single pair reading can occur only in languages which do not have overt syntactic wh-movement, it does not look like Russian qualifies. Grebenyova (2004) provides some other evidence for the lack of single pair reading in Russian. Particularly, she observes that fronting of an object wh-phrase over a subject wh-phrase does not allow a single pair reading. According to observations made for other languages this type of construction should force this reading. Consider the example in (68)9 68. ??Kogo kto priglasil na uţin? whom who invited (*SP) to dinner ‗Who invited who to the dinner?‘ She also observes contexts with multiple sluicing, which depend crucially on the interpretation of multiple interrogatives in Russian. Thus, she provides an example, which I repeat here as (69), where the antecedent clause forces a single pair reading of 9 I provide data below which show that fronting of object over wh-subject is quite degraded as well. 38 the embedded wh-elements, although this reading is unavailable in Russian (Grebenyova, 2004:20). 69. *Kto-to priglasil kogo-to na tanec, no ja ne znaju kto kogo someone invited s omeone to dance but I not know who whom ‗Someone invited someone to a dance but I don‘t know who invited whom‘ Grebenyova shows that the example in (69) can be improved as in (70). However, in this case the sentence does not get a single pair reading, but the so-called order-reading, where the question is about the direction of the events, rather than the identity of the agents. 70. Maša i Ivan pošli na večerinku. Kto-to iz nix priglasil drugogo na tanec, no ja Maša and Ivan went to party. One of them invited the-other to dance but I ne znaju kto kogo not know who whom ‗Maša and Ivan went to a party. One of them invited the other to a dance, but I don‘t know who invited who.‘ The observation is supported by the ungrammaticality of the sentence with sluiced whadjuncts. It supports the idea that a single pair reading is unavailable in Russian, and adjuncts do not allow order-readings in general; this is a property of arguments. While arguments can be switched around, an argument and an adjunct cannot (Grebenyova, 2004:21). 71. *Kto-to sprjatal gde-to someone hid zdes‘ klad, no ja ne znaju kto gde. somewhere here treasure but I not know who where ‗Someone hid the treasure somewhere here but I don‘t know who hid it and where.‘ As was shown in (69), while a single pair reading does not allow sluicing, the multiple pair list reading, on the other hand, permits it (72). 39 72. Kaţdyj priglasil kogo-to na tanec, no ja ne znaju kto kogo. everyone invited someone to dance but I not know who whom ‗Everyone invited someone to a dance but I don‘t know who invited who.‘ To summarize, the data above showed that Russian does not allow single pair readings and, moreover, exhibits some cases of the Superiority effect, bringing the language closer to Bulgarian. Therefore, based on the fact that Russian employs overt syntactic wh-movement, it cannot be analyzed as a wh-in-situ language10. Based on the evidence introduced above, I will provide Russian data from multiple wh-questions to see if Russian shows any matrix/embedded asymmetry. 10 Contrary to Bošković (1998, 2001, etc.), who claims that Russian is an in-situ languages, which does not have overt movement of wh-element to [Spec, CP]. 40 CHAPTER 3. TOWARDS AN ANALYSIS OF RUSSIAN MULTIPLE WH-ELEMENTS In this chapter I employ the approach offered by Krapova and Cinque (2005) for the analysis of the order of wh-phrases in Bulgarian multiple wh-fronting questions11 to analyze the sequence of fronted wh-elements in Russian matrix and embedded questions. Analyzing Russian data, I focus mainly on wh-subject and wh-object and time, place and manner wh-adjuncts. I do not discuss wh-adjuncts such as začem, and počemu12 here; I will study these in detail in Chapter 4. 3.1. Bulgarian order of wh-phrases As mentioned before, Bulgarian is known to display a rather rigid ordering of whphrases in multiple wh-fronting. At the same time the literature reports cases of rather free ordering of wh-questions in Bulgarian, and it appears that some of them violate the Superiority effect. 73. a. Kakvo kâde raste? (Billings and Rudin 1996, 42, and fn.10) what where grows ‗What grows where?‘ 11 12 Krapova and Cinque (2005) study only matrix questions, they do not provide embedded questions data. Chapter 4 deals in detail with wh-adjuncts like začem – ‗what for‘, počemu – ‗why‘, kakogo čerta - ‗what the hell,‘ etc. I refer to these kind of wh-adjuncts as why-wh-adjuncts. 41 b. Kâde kakvo raste? (Billings and Rudin 1996, 42, and fn.10) where what grows ‗What grows where?‘ Krapova and Cinque argue in their paper that in Bulgarian wh-phrases occupy different positions as a consequence of their internal makeup and interpretation (Krapova and Cinque, 2005:2). Thus, in (73a) it is more important to know what grows and in (73b) where it grows. That is why the order of wh-phrases in (73) varies depending on interpretation, and thus no violation of the Superiority effect occurs. Krapova and Cinque study the order of wh-adjuncts with respect to wh-objects and wh-subjects and attempt to define the order of wh-phrases. 3.1.1. The order of wh-adjuncts Krapova and Cinque show that the order of wh-adjuncts appears to be very strict: koga – ‗when‘ must precede kâde – ‗where‘; kak – ‗how‘ must follow both koga and kâde13: 74. a. Koga kâde šte hodiš tova ljato? when where will go-you this summer ‗When will you go where, this summer?‘ b. *Kâde koga šte hodiš tova ljato? where when will go-you this summer 75. a. Kâde kak si se dârţal? where how are-you behaved ‗Where did you behave how?‘ b. *Kak kâde si se dârţal. how where are-you behaved 13 In this section all examples are from Krapova and Cinque (2005). 42 76. a. Koga kak si se dârţal? when how are-you behaved ‗When did you behave how?‘ b *Kak koga si se dârţal? how when are-you behaved 3.1.2. The order of wh-objects w.r.t wh-adjuncts Wh-arguments bearing certain grammatical relations do not occupy the same position with respect to wh-adjuncts, rather they occupy a position depending on their internal makeup (features like [human] and [D-linking]). Prepositional indirect object The examples (77)-(81) show that na kogo – ‗to whom‘ must precede all adjuncts, and na kolko - ‗to how many N‘ will occupy a lower position, following koga – ‗when‘ and kâde – ‗where‘ but preceding kak – ‗how‘. 77. a. Na kogo kak šte prepodadeš tozi urok? to whom how will teach-you this lesson ‗To whom will you teach this lesson how?‘ b. *Kak na kogo šte prepodadeš uroka? how to whom will teach-you this lesson 78. a. Na kogo kâde si daval podarâci? to whom where are-you given presents ‗To whom did you give presents where?‘ b. ???Kâde na kogo si daval podarâci? where to whom are-you given presents 43 79. a. Na kogo koga šte se obadiš? to whom when will call-you ‗Who will you call when?‘ b.??Koga na kogo šte se obadiš? when to whom will 80. call-you a. Koga/kâde na kolko xora si pomagal? when/where to how many people are-you helped ‗How many people did you help when/where?‘ b. *?Na kolko xora koga/kâde si pomagal? to how many people when/where are-you helped 81. a. Na kolko xora kak moţeš da pomogneš? to how many people how can-you to help ‗How many people can you help how?‘ b.*Kak na kolko xora moţeš da pomogneš? how to how many people can-you to help Krapova and Cinque argue that the different distribution of na kogo –‗to whom‘ and na kolko – ‗to how many N‘ is explained by a [human] feature. Na kogo [+human], na kolko [underspecified] since N can be both human and non-human. Multiple questions containing two [+human] wh-objects show strict ordering as well: direct objects must precede indirect ones. Therefore, the order so far is the following: Kogo > na kogo > koga > kâde > na kolko > kak (whom) (to whom) (when) (where) (to how many) (how) Direct [-human] wh-objects Looking at the distribution of direct [-human] or underspecified wh-objects shows that the distribution is different from [+human] wh-objects: 82. a. Koga/kâde kakvo kupuvaš? when/where what buy-you 44 ‗When are you buying what?‘ b.?(?)Kakvo koga/kâde kupuvaš? what when/where buy-you 83. a. Koga/kâde kolko (pari) si poxarčil? when/where how much (money) are-you spent ‗How much (money) did you spend when/where?‘ b. *Kolko (pari) koga/kâde si poxarčil? how much (money) when/where are-you spent 84. a. Kakvo kak šte napraviš? what how will do-you ‗What will you do how?‘ b. *Kak kakvo šte napraviš? how what will do-you The examples above show that kakvo – ‗what‘ and kolko – ‗how much N‘ will follow koga – ‗when‘ and kâde - ‗where‘ and precede kak – ‗how‘. Based on the data discussed above the following tendency is distinguished: 1. All [+human] wh-objects precede wh-adjuncts koga –‗when‘, kâde – ‗where‘ and kak – ‗how.‘ 2. All [-human] or underspecified wh-objects follow koga – ‗when‘ and kâde – ‗where‘ but precede kak – ‗how‘. 3. All [+human] wh-objects precede [-human] or underspecified wh-objects. 4. All direct [+human] wh-objects precede indirect [+human] wh-objects. Table 1. Direct [+h] Indirect [+h] Adjunct [-h] or unspecified object (place & time) Adjunct (manner) kogo na kogo koga kâde na kolko N kak (whom) (to whom) (when) (where) (to how many) (how) kolko N (how many) kakvo (what) 45 3.1.3. The order of wh-subjects w.r.t wh-adjuncts Krapova and Cinque (2005) notice that wh-subjects occupy different positions depending on their makeup: A [+human] wh-subject koj – ‗who‘ behaves the same way as a [+human] wh-object, i.e. it precedes all adjuncts (85-86). 85. a. Koj kâde šte spi? (Billings and Rudin 1996, 41) who where will sleeps ‗Who will sleep where?‘ b. * Kâde koj šte spi? where who will sleeps 86. a. Koj koga pristiga? who where arrives ‗Who will arrive when?‘ b. *Koga koj pristiga? where who arrives The examples below show that [-human] wh-subjects kakvo – ‗what‘ and underspecified kolko N – ‗how much N‘ pattern together with [-human] wh-objects, i.e. they follow koga – ‗when‘ and kâde – ‗where‘ but precede kak – ‗how‘ (87-90). 87. a. Koga kakvo te pravi štastliv? when what you-acc makes happy ‗What makes you happy when?‘ b.??Kakvo koga te what 88. pravi štastliv? when you-acc makes happy a. Kâde kakvo stava sega po sveta? where what happens now in world-the ‗What is happening where around the world?‘ b. *Kakvo kâde stava sega po sveta? what where happens now in world-the 46 89. a. Kâde/koga kolko se investira v častnija sektor? where/when how much refl.cl. invests in private-the sector ‗Where/when how much is invested in the private business?‘ b. *Kolko kâde/ koga se investira v častnija how much where/when refl.cl. invests 90. a. Kakvo kak ti sektor? in private-the sector haresva? (p.c. Polya Votkova) what how to you like ‗What how do you like?‘ b. *Kak kakvo ti how what 91. a. Kakvo haresva? to you like prichiniava kakvo? what-subj causes (p.c. Polya Votkova) what-obj ‗What causes what?‘ b. *Kakvo prichiniava kakvo? what-obj causes what-subj It is difficult to test the word order of [-human] or underspecified subject and [-human] or underspecified object, given that there is no subject/object distinction in human/underspecified wh-phrases, that is, unless we are using verbs like cause, which chooses a wh-object. Questions with the verb cause clearly distinguish precedence of kakvo – subject over kakvo – object in (91). To depict this I use a dotted line to show that there is no clear order between the whelements. Thus, the data discussed so far can be summarized as follows: Table 2. [+human] Adjunct [-human] or underspecified (place & time) subject object Adjunct (manner) koj kogo na kogo koga kâde kakvo kakvo kak (who) (whom) (to whom) (when) (where) (what) (what) (how) kolko N na kolko N (how many) (to how many) kolko N (how many) 47 It is noteworthy to mention that for some speakers, kakvo, kâde, and koga can precede kogo as well as na kogo, which are obvious violations of the orders presented in the Table 2. Krapova and Cinque (2005:6) suggest that ―the problem posed by these marked orders of kakvo and kâde/koga can be made sense of if they are taken to access (more markedly) a higher position, the one reserved for D-linked phrases.‖ In the next subsections I present D-linked and clitic resumed wh-phrases, inasmuch as the empirical evidence shows that they are located higher than the other wh-elements. 3.1.4. The order of D-linked and non-D-linked wh-phrases Phrases in which koj functions as a specifier, i.e. koj/koja/koe/koi (N) - ‗which‘ phrases are inherently D-linked and must precede all non-D-linked wh-phrases, as shown in (92) (Krapova and Cinque, 2005). The only exception to the order D-linked > non-D-linked seems to be the fronting of a D-linked wh-phrase over a koj-subject (93). According to Krapova and Cinque, this is explained by the fact that koj – ‗who‘ belongs to the same paradigm as koj (N) – ‗which‘, where the head N is interpreted as [+human]. If this was true then there would be no Superiority effect between koja studentka and koj. Example (93) shows the opposite. Therefore, I would suggest the order between these two elements, i.e. a [+ human] wh-subject must precede a D-linked wh-element. 92. a. Koi kartini na kogo za Boga iskaš pak da podarjavaš? which paintings to whom for God‟s sake want-you again to donate-you ‗Who on earth do you want to donate which pictures to again?‘ b. *Na kogo za Boga koi kartini pak iskaš da podarjavaš? to whom for God‟s sake which paintings again want-you to donate-you 93. a. Koj koja studentka šte izpita? who which student will examine ‗Who will examine which student?‘ 48 b. *Koja studentka koj šte izpita? which student who will examine ‗Who will examine which student?‘ I would suggest the following order of wh-elements in Bulgarian, which differs from Krapova and Cinque‘s order in placing of koj higher than D-linked [-h] elements. Table 3. D-linked wh-14 Subj Non-D-linked wh-phrases [+h] koj koj/koja/koe/koi (N) kogo na kogo koga kâde kakvoSubj kakvoObj kak (who) (which) (whom) (to whom) (when) (where) (what) (what) (how) D-linked (kogo) kolkoSubj N (na) kolkoObjN subj (na kogo) (how many) (tohow many) (marked)kakvoSub/Obj (marked)kâde/koga 3.1.5. The order of clitic resumed and non-clitic resumed wh-phrases In this section I discuss another type of multiple wh-construction, which contains an inherently D-linked phrase resumed by a clitic (94). 94. Koja kartina na kogo/ na koj prijatel si (ja) posvetil? which painting to whom/to which friend have-you it dedicated ‗Which painting have you dedicated to whom/to which friend?‘ Krapova and Cinque (2005:185) claim that ―clitic resumption is not available for the second of two wh-phrases when the first is not itself resumed by a clitic,‖ as in (95), but 14 In the D-linked column Krapova and Cinque indicate the possibility of D-linking kogo and na kogo, although they do not give evidence for this in the paper. 49 if the leftmost D-linked wh-phrase is separated by a parenthetical from the second whphrase then clitic resumption becomes obligatorily, as in (96). 95. a. Na kogo koja kartina si (*ja) posvetil? to whom which painting have-you it dedicated ‗Which painting did you dedicate to whom?‘ b. Na koj prijatel koja kartina si mu ja posvetil? to which friend which painting have-you it dedicated 96. Koja kartina, spored tebe, kâde *(ja) e risuval tozi xuţdonik? which painting, according to you, where it is painted this artist ‗According to you, which painting did this artist paint where?‘ It is evident that the material which precedes the parenthesis requires a clitic and the material that follows the parenthetical requires the absence of the clitic. Accordingly, Krapova and Cinque (2005) suggest that (94) represents two different structures. One of them involves a position (some XP) which can be targeted only by clitic resumed (inherently) D-linked wh-phrases; the other involves a position (some YP) which can be targeted only by non-clitic resumed (inherently) D-linked wh-phrases. 97. [XP Cl-D-linked wh[parenthetical [YP non-Cl-D-linked wh[ZP non-D-linked wh [IP...cl... The parenthetical can also occupy a sentence initial position, as shown in (98). 98. Vpročem, koja kartina, spored tebe, koj *(ja) e narisuval? by-the-way which painting according-to-you who it is painted ‗By the way, according to you, who painted which painting?‘ parenthetical [XP parenthetical [YP [ZP [IP clXP Cl-D-linked wh- D-linked wh- non-D-linked wh- 50 Within a CP structure XP can be identified with Topic position (TopP), YP with a position specialized for D-linked phrases (D-LP); and ZP with a position specialized for non-D-linked phrases, the traditional CP. From the data above the following order is revealed: clitic resumed Topic whelements precede D-linked wh-elements, which in turn precede non-D-linked whelements. 3.1.6. Some conclusions Krapova and Cinque (2005) point out that there is evidence which suggests that the order of wh-phrases reflects the order prior to wh-movement. The argument for this idea is based on the order of the adjuncts koga – ‗when‘ > kâde – ‗where‘ > kak - ‗how‘, confirming the UG hierarchy of adjuncts, where Temporal precedes Locative, which precedes Manner. It has been also noticed that the order of Bulgarian [+human] wh-phrases (koj – ‗who‘, kogo – ‗whom‘, and na kogo – ‗to whom) with respect to wh-adjuncts, and to wh-phrases underspecified (or negatively specified) for the feature [human] (kakvo – ‗what‘ and kolko/kakâv N – ‗how many N‘) appears to reflect their relative order prior to wh-movement (Krapova and Cinque, 2005:190). The fact that the order of wh-phrases reflects the order prior to wh-movement suggests some principle which ―will have to ensure that this order is preserved under wh-movement‖ (Krapova and Cinque, 2005:191). For these purposes they refer to Rizzi‘s (1990, 2001a) Relativized Minimality. To recall the idea of Relativized Minimality: In order to explain the impossibility of wh-movement in (99b) Rizzi (1990) suggests a theory of chains, which state that a chain cannot be built between X and Y in configuration …X…Z…Y… 99. a. I wonder who could solve the problem in this way? b. *How do you wonder who could solve this problem <how>? 51 In (99b), the wh-chain starting from an adverb position fails across a subject who (Rizzi, 2001a). Thus, according to Rizzi‘s assumptions, in (99b) the chain cannot connect how and its trace, given that another wh-element who intervenes between them. Rizzi refers to this principle as to Relativized Minimality. 100. Y is in a Minimal Configuration with X iff there is no Z such that (i) Z is of the same structural type as X, and (ii) Z intervenes between X and Y Y and X are in a minimal configuration when Y is a trace and X is a target and there is no intervener Z between them, which c-commands Y and does not c-command X and which is specified with the same features as the target. ……Xt………….Z………….Y(t)… [F] [F] [N] c-command So, going back to (99b), chain formation fails between the target and the trace since they are not in a Minimal configuration, because who is the intervener between how and the trace and who is the same structural type as the target. Krapova and Cinque (2005:192) suggest that ―within a system in which Superiority is subsumed under Relativized Minimality, the preservation of the pre-whmovement order of the wh-phrases in the case of multiple movements can be ensured through the requirement‖ where - only a whole chain, not just a link of a chain, counts as an ‗intervener.‘ This requirement is a modification of Chomsky‘s principle which states that only the head of an A-chain (equivalently, the whole chain) blocks matching under the Minimal Link Condition. (Chomsky, 2001:17). Thus, a modified (100) would be ‗no chain can intervene between the trace and the target, if this chain bears the same features as the target.‘ (Krapova and Cinque, 2005:193) Based on the assumptions made above, Krapova and Cinque (2005:193) explain the ungrammaticality of (101b) by arguing that the entire chain kakvo intervenes between the trace of kak and its target, whereas in (101a) only a link of a chain (not an entire chain) intervenes between the trace and the target. 52 101. a. [CP kakvo [CP kak [IP šte napraviš t t ]]] [+wh] [+wh] ① ② [+wh] b. *[CP kak [CP kakvo [IP šte napraviš t [+wh] [+wh] t ]]] [+wh] As for D-linked wh-phrases, Krapova and Cinque claim that these wh-phrases have special features which match the corresponding feature in the target (say a [+D-L] feature) and this allows the intervener between the target and the trace. Consider the example in (102) which is slightly marginal but still acceptable. 102. [YP kakvo [CP kâde [IP raste t [+D-L wh] [+wh] t ]]] [+wh] In (102), the intervener kâde bears distinct features from kakvo, and its trace does not violate Relativized Minimality and is thus acceptable. To summarize: Krapova and Cinque (2005) come to the conclusions that the order of wh-phrases reflects the order prior to wh-movement within the Mittelfeild, and the cases which seem to violate the Superiority effect involve selective movements forced by a feature not present in the intervener. The Superiority effect is captured under Relativized Minimality, where links of a chain do not count as interveners, following Chomsky. In the next section, I attempt to employ the analysis proposed by Krapova and Cinque for Russian and see if there is any wh-phrase order observed. 53 3.2. Russian order of wh-phrases In this section, I follow the structure of the analysis offered by Krapova and Cinque. I employ their approach for Russian matrix and embedded questions. First, I introduce an analysis offered by Meyer (2004), who studies Russian wh-elements in detail and distinguishes the Superiority effect in Russian. 3.2.1. Meyer‟s analysis Meyer (2004), researching Superiority effects in Russian, Czech and Polish based on evidence from a series of controlled acceptability studies and text corpora, came to a conclusion that all three languages exhibit the Superiority effect in one or another way15. Meyer (2004a) looks for the Superiority effect by conducting three different studies. The first study is devoted to the relative order of wh-subjects and wh-objects. The results of this indicate a general preference16 for wh-subjects to precede wh-objects, which, according to Meyer, could be interpreted as a Superiority effect. The second study focuses on the relative order of wh-arguments and whadjuncts. Meyer defines a preference of kto – ‗who‘ over kak – ‗how‘ order, subjects over adjuncts, and a weak preference of wh-objects to precede wh-adjuncts; he found instances of both orders. In the third study, Meyer looks at the alleviation of wh-order effects via Dlinked wh-elements, but he does not consider Russian for the third study, inasmuch as he claims that Russian D-linked wh-phrases behave differently than in other languages, i.e. they do not move high enough. I disagree with Meyer‘s claim concerning D-linked wh-elements and propose my own analysis of the elements in the next section. 15 I will not discuss Czech and Polish here, see Meyer (2004, 2004a). 16 Referring to Meyer‘s analysis, I use his vocabulary, i.e. for instances I claim to be a clear example of Superiority effect, Meyer uses term « preference » etc. 54 During the experiment Meyer comes to an interesting conclusion that Russian belongs to a group of languages in which Superiority affects both matrix and embedded wh-questions; the same is argued for Bulgarian. Superiority-like preferences distinguished by Meyer are listed in the table below: Table 4. Order Example Order Example [+h]Subject/[+h]dir Object kto(who)>kogo(whom) > kogo(whom)>kto (who) [+h]Subject/[+h]indir Object kto(who)>komu (whom) > komu(whom)>kto(who) [+h]Subject/[+h]prep Object kto(who)>PP kto(who) > PPkto(who)>kto(who) [+h]Subject/[-h]dir Object kto(who) > čto Acc (what) > čto Acc (what)> kto(who) [+h]Object/[-h]Object komu(whom) > čto (what) > čto (what) > komu(whom) [+h]Subject/D-linked Object kto(who) > kakoj (which) > kakoja(which) > kto(who) [+h]Subject/Ajunct kto(who) >kogda(when), > kogda(when), gde(where), gde(where), kak(how) [-h]Object / Ajunct kak (how) > kto (who) čto (what) > kogda(when), < kogda(when), gde(where), gde (where), kak (how) > kak (how) > čto (Acc) In the next section, I propose an analysis of Russian wh-questions following the approach introduced by Krapova and Cinque for Bulgarian. I want to see if Meyer‘s findings correspond to mine and if Russian behaves the same way as Bulgarian, i.e. I will try to define the preferred wh-orders for Russian wh-questions. The data were collected from 20 native speakers regularly residing in Russia (South-Western Siberian region), who do not have any interactions with other languages. The collection of the data was made in 2 stages: during the first stage only wh-matrix questions were asked for evaluation, and during the second one only whembedded questions. In this way subjects judging wh-embedded questions do not refer to the answers they gave during the first section. There was one-week break between the two stages. 55 3.2.2. Analysis of the order of Russian wh-matrix questions The order of wh-adjuncts The data introduced in (103-105) show that the language does not require strict order of kogda – ‗when‘ and gde – ‗where‘, but both of them should precede kak – ‗how‘. 103. a. Gde kogda ty ego videl ? where when you him saw ‗Where did you see him and when?‘ b. Kogda gde ty ego videl? when where you him saw ‗When did you see him and where?‘ 104. a. Gde kak ty spal? where how you slept ‗Where did you sleep and how?‘ b. *Kak gde ty spal? how where you slept ‗How did you sleep and where?‘ 105. a. Kogda kak ty ušel? when how you went away ‗When did you leave and how?‘ b. *Kak kogda ty ušel? how when you went away ‗How did you leave and when?‘ The order of an object with respect to wh-adjuncts In Bulgarian, according to Krapova and Cinque (2005), wh-objects do not occupy the same position with respect to wh-adjuncts: the position will depend on the internal makeup of the wh-element, i.e. [+human] will be placed higher than [-human]. I want to 56 introduce the judgments of native speakers on this issue and determine whether it is true for Russian as well. According to the judgments there is a clear preference of ordering [+human] object over kak (106), but there is no strong preference of the object placed above whadjuncts kogda and gde as in (107-108). 106. a.*Kak komu ty pomog? how whom you helped ‗Whom did you help and how?‘ b. Komu kak ty pomog? whom how you helped 107. a. ?Gde kogo ty poslednij raz videl? where whom you last time saw ‗Who did you see last time and where?‘ b. Kogo gde ty poslednij raz videl? whom where you last 108. a. Kogo kogda ty time saw priglasil? whom when you invited ‗Whom did you invite and when?‘ b. ?Kogda kogo ty priglasil? when whom you invited According to my native speakers, the ordering of wh-elements in (107 and 108) depends on which information is more prominent, i.e. place or object as in (107) or time or object as in (108); however, there is still a slight preference of an object over adjunct. Note that Meyer (2004) came to the exact conclusion. In the next set of examples, I test the theory mentioned in Krapova and Cinque, who assume for Bulgarian that cumulative objects or skolkoN – ‗how many N‘ being underspecified for [human] feature should be placed lower than wh-objects bearing a [+human] feature. Since we do not have a very clear picture of the object with respect to the kogda – ‗when‘ and gde – ‗where‘ order I expect that we will not be able to get a clear ordering in this case either. 57 109. a. Skol‘kim druz‘jam gde ty podaril to how many friends knigi? where you gave books ‗To how many friends did you give books and where?‘ b. Gde skol‘kim druz‘jam ty podaril knigi? where to how many friends 110. a. Skol‘kim you gave books druz‘jam kogda ty podaril knigi? to how many friends when you gave books ‗To how many friends did you give books and when?‘ b. Kogda skol‘kim druz‘jam ty podaril knigi? when to how many friends 111. you gave books a. So skol‘kimi ljud‘mi kak ty razgovarival? with how many people how you spoke ‗With how many people did you talk and in which way?‘ b. *Kak so skol‘kimi ljud‘mi ty razgovarival? how with how many people you spoke 112. a. Skol‘hih ljudej gde ty vstretil? how many people where you met ‗How many people did you meet and where?‘ b. Gde skol‘kih ljudej ty vstretil? where how many people you met 113. a. Skol‘kih ljudej kogda ty vstretil? how many people when you met ‗How many people did you meet and when?‘ b. Kogda skol‘kih ljudej ty vstretil? when how many people you met 114. a. Skol‘kim ljudjam kak ty pomog? how many people how you helped ‗How many people did you help and how?‘ b. *Kak skol‘kim ljudjam ty pomog? how how many people you helped As the above examples show, skolko N – ‗how many N‘ independently of being direct or indirect does not exhibit any order with respect to wh-adjuncts, except with kak – ‗how‘ which should follow skolko N. 58 At this point, the order for wh-objects with respect to wh-adjuncts is not clear and appears to be no higher than kogda – ‗when‘ and gde – ‗where‘, yet it precedes kak‗how‘. The order of a wh-subject with respect to a wh-object As shown in (115-117), the order of wh-subjects with respect to wh-objects appears to be quite strict: wh-subjects both [+human] and [-human] must precede wh-objects. This is an example of a clear Superiority effect. 115. a.Kto komu pokazal film? who whom showed movie ‗Who showed movie to whom?‘ b.*Komu kto pokazal film? whom who showed movie 116. a. Čto komu pokazalos‘? what whom appeared ‗What appeared to whom?‘ b.*Komu čto pokazalos‘? whom what appeared 117. a. Kto kogo uvidel? who whom saw ‗Who saw whom?‘ b.*Kogo kto uvidel? whom who saw? The order of a wh-[+human] object with respect to wh-adjuncts For Russian speakers, the only strong preference is a wh-[+human] object preceding kak. 59 118. a. Komu kogda ty rasskaţeš‘ novosti? whom when you will tell news ‗Whom will you tell the news and when?‘ b. ?Kogda komu ty rasskaţeš‘ novosti? when whom you will tell 119. a. Kogo gde news ty predloţiš‘ vstretit‘? whom where you will offer to meet ‗Whom will you offer to meet and where?‘ b. ?Gde kogo ty predloţiš‘ vstretit‘? 120. where whom you will offer to meet a. Komu kak ty peredaš‘ eti knigi? whom how you will give these books ‗Whom will you offer these books and how?‘ b.*Kak komu ty peredaš eti knigi? how whom you will give these books There is a slight preference of a [+human] object over time and place adjuncts in whmatrix questions. I will not record this as an evidence of Superiority effect in Matrix questions, although it is very clear in embedded questions, which will be described in the next section. The order of a wh-[-human] object with respect to wh-adjuncts No order is observed, only the preference of čto over kak (123), as for [+human] objects 121. a. Gde čto ty videl? where what you saw ‗What did you see and where?‘ b. Čto gde ty videl? what where you saw 122. a. Kogda čto ty videl? when what you saw ‗What did you see and when?‘ 60 b. Čto kogda ty videl? what when you saw 123. a.*Kak čto ty sdelal? how what you did ‗What did you do and how?‘ b. Čto kak ty sdelal? what how you did The order of wh-[+human] with respect to wh-[-human] As shown in examples (124-127), the distribution of wh-[+human] and wh-[-human] depends not only on the internal marking, i.e. wh-[+human] is predicted to be placed higher, but apparently also on the function in the sentence. Wh-subjects will be placed higher than wh-objects disregarding the internal makeup. Consider the examples in (124-125), where wh-[+human] precedes [-human] or underspecified wh-phrase skolko in (125). Examples (126-127) illustrate the case when wh-[-human] čto is the subject of the sentence, forcing it to precede wh-[+human] object. 124. a. Kto čto tebja sprosil? who what you asked ‗Who asked you and what?‘ b. *Čto kto tebja sprosil? what who you asked 125. a. Komu skol‘ko ty dal? whom how much you gave ‗Whom did you give how much (money)?‘ b.?/*Skol‘ko komu ty dal? how much whom you gave 126. a. ??Kogo čto sdelaet sčastlivym? whom what will make happy ‗What will make who happy?‘ b. Čto kogo sdelaet sčastlivim? what whom will make happy 61 127. a.*S kem čto slučilos? with whom what happened ‗What happened with whom?‘ b. Čto s kem slučilos? what with whom happened The examples above show that the subject will always precede the object and [+human] objects precede [-human] or underspecified objects. The order of wh-subjects with respect to wh-adjuncts The examples in (128-130) show that subjects both [-human] and [+human] precede any kind of adjuncts. 128. a. Kto gde budet spat‘? who where will sleep ‗Who will sleep where?‘ b.*Gde kto budet spat‘? where who will sleep 129. a.Čto kogda proizajdët? what when will happen ‗What will happen when?* b.*Kogda čto proizajdët? when what will happen 130. a. Kto kak rešit etu zadaču? who how will sove this problem ‗Who will solve this problem and how?‘ b.*Kak kto rešit etu zadaču? how who will solve this problem To summarize the order we have thus far, we get the following picture: 62 Table 5 Subject [+human] Object [-human] Object Adjunct Adjunct (place&time) (Manner) kak (how) kto (who) komu (whom) kogda (when) čto (what) kogo (whom) gde (where) čto (what) skol‘ko N (how much N) The table above shows that subjects precede objects and adjuncts, [+human] objects precede [-human] objects and manner adjuncts but there is no clear preference over place and time adjuncts (an interrupted line shows that there is no strong word order). Manner adjuncts occupy the lowest position. D-linked wh-questions The last case left to investigate is D-linked wh-questions in Russian matrix clauses. In Bulgarian, D-linked wh-phrases always have to be placed higher than non-D-linked, and koj – ‗who‘ is situated higher than any wh-elements. D-linked elements in Russian matrix clauses exhibit some features similar to Bulgarian D-linked elements. Kto – ‗who‘ is higher than D-linked elements as in (134)17, and [-human] D-linked wh-elements do not show order preference over [+human] wh-objects (131-132), with the exception of a prepositional [+human] whobject located lower that D-linked wh-elements (133). D-linked wh-elements are higher than wh-adjuncts, as in (135-136), and a D-linked wh-subject precedes [+human] whobject, as in (138). 17 Recall that the same judgement we see in Bulgarian example (93). Romanian, being another multiple wh-fronting language does not seem to show the same order. Consider the examples below: a. Pe care student cine il va examina? (Gabriella Soare, p.c.) which student who will examine ‗Who will examine which student?‘ b. ?Cine pe care student il va examina? who which student will examine According to Soare (p.c.), cine –‗who‘ here is not D-linked and the order in (b) gets degraded. 63 131. a. Kakuju kartinu komu on podaril? which picture whom he presented ‗Which picture did he give who?‘ b. Komu kakuju kartinu on podaril? whom which picture he presented 132. a. Kakoj roman kogo zaintrigoval? which novel whom interested ‗Which novel interested who?‘ b. Kogo kakoj roman zaintrigoval? whom which novel interested 133. a. Kakoj podozrevaemyj s which suspected kem razgovarival? with whom talked ‗Which suspected talked with whom?‘ b.*S kem kakoj podozrevaemyj razgovarival? with whom which suspected 134. talked a. Kto kakuju studentku budet ekzamenovat‘? who which student will examine ‗Who will examine which student?‘ b.*Kakuju studentku kto budet ekzamenovat‘? which student 135. who will examine a. Kakoj professor kogda prišël? which professor when came ‗Which professor came when?‘ b. *Kogda kakoj professor prišël? when which professor came 136. a. Kakomu rebenoku kak ty pomog? which child how you helped ‗Which child did you help and how?‘ b.*Kak kakomu rebenoku ty pomog? how which 137. child you helped a. Kakoj student čto sprosil? which student what asked ‘Which student asked what?‘ 64 b.*Čto kakoj student sprosil? what which student asked 138. a. Kakoj malčik kogo uvidel? which boy whom saw ‗Which boy saw who?‘ b.*Kogo kakoj malčik uvidel? whom which boy saw To summarize the order we have thus far, we get the following picture: Table 6. Subject [+human] Object [-human] Object Adjunct D-linked Object Adjunct (place&time) (Manner) kto (who) komu (whom) kogda (when) kak (how) čto (what) kogo (whom) gde (where) kakoj/kakaja N (which N) kakoj/kakaja N čto (what) (which N) skol‘ko N (how much N) PP object The table shows that subjects precede all wh-elements, [+human] objects, and D-linked objects do not show word order. D-linked elements precede all adjuncts, [-human] whelements and prepositional object. Note that [+human] objects show only slight preference over adjuncts of time and place, although they do show a clear order with respect to [-human] objects. Now I will look at word order in embedded questions, which will show a more detailed order of wh-elements. 65 3.2.3. Analysis of the order of Russian wh-embedded questions As was seen from the data above, Russian demonstrates some Superiority-like effects in a few cases in matrix clauses. Now I turn to embedded clauses. I will show that embedded clauses allow a clearer distinction of the order of the wh-elements. These data yield a main/embedded clause asymmetry. I follow Krapova and Cinque‘s approach for this analysis. The order of adjuncts As shown in (139-141), there is no preference in the order of gde and kogda, although both should precede kak, the same Superiority effect we see in matrix questions. 139. a. Ona interesovalas‘, gde kogda ja ego videl. she wondered where when I him saw ‘She was wondering where I saw him and when.‘ b. Ona interesovalas‘, kogda gde ja ego videl. she wondered 140. when where I him saw a. Ona interesovalas‘, gde i she wondered kak ty spal18. where and how you slept ‗She was wondering where i slept and how.‘ b.*Ona interesovalas‘, kak i she wondered 141. gde ty spal. how and where you slept a. Ona interesovalas‘, kogda i kak ty dobralsya do doma. she wondered when and how you reached till home ‗She was wondering when I got home and how.‘ b.*Ona interesovalas‘, kak i kogda ty dobralsya do doma. she wondered 18 how and when you reached till home I will discuss the nature of conjunction ‗i‘ in Chapter 4. 66 The order of a wh-object with respect to wh-adjuncts The examples in (142-145) show that wh-objects must precede all wh-adjuncts. Remember that in matrix questions there is no clear ordering between objects and adjuncts except manner adjuncts, which have to follow objects. 142. a.*Ona interesovalas‘, kak komu ty she wondered pomog. how whom you helped ‗She was wondering whom i helped and how.‘ b. Ona interesovalas‘, komu kak ty she wondered 143. pomog. whom how you helped a.*Ona interesovalas‘, gde komu ty she wondered pomog. where whom you helped ‗She was wondering whom you helped and where.‘ b. Ona interesovalas‘, komu gde ty she wondered 144. pomog. whom where you helped a.*Ona interesovalas‘, gde kogo ty poslednij raz videl. she wondered where whom you last time saw ‗She was wondering whom you saw last time and where.‘ b. Ona interesovalas‘, kogo she wondered 145. gde ty poslednij raz videl. whom where you last time saw a.*Ona interesovalas‘, kogda kogo ty pozval. she wondered when whom you called up ‗She was wondering whom you called and when.‘ b. Ona interesovalas‘, kogo kogda ty she wondered pozval. whom when you called up The examples in (146-149) show that a cumulative object precedes all adjuncts. In matrix questions, on the other hand, there was no word order between them except manner adjuncts, which are found lower in a structure in respect to any wh-element. 146. a. Ona sprosila, skol‘kix she asked, ljudej i gde on ubil. how many people and where he killed ‗She asked how many people he killed and where.‘ 67 b. *Ona sprosila, gde she asked, 147. i skol‘kix ljudej on where and how many people he ubil. killed a. Ona pointeresovalas‘, so skol‘kimi ljud‘mi kak často Peter obschalsja. she wondered, with how many people how often Peter was in touch ‗She was wondering with how many people he was in touch and how often.‘ b.*Ona pointeresovalas‘, kak často so skol‘kimi she wondered, 148. how often with how many people Peter was in touch a. Ona sprosila, skol‘hih she asked, ljud‘mi Peter obschalsja. ljudej kogda on vstretil. how many people when he met ‗She asked how many people he met and when.‘ b. *Ona sprosila, kogda skol‘hih she asked, 149. when how many people he met a.Ona sprosila, skol‘kih she asked, ljudej on vstretil. ljudej kak on nazval. how many people how he named ‗She asked how many people he named and how.‘ b.*Ona sprosila, kak skol‘kih she asked, ljudej on nazval. how how many people he named Multiple questions containing [+human] objects are ordered according to a direct/indirect distinction, i.e. direct objects always precede indirect objects (150-151). The same was noted for Bulgarian by Krapova and Cinque (2005) and Billings and Rudin (1996): 150. a. Oni interesovalis‘, kogo komu ona predstavila pervym. they were wondering, who to whom she introduced first ‗They was wondering who she introduced first and to who.‘ b.??Oni interesovalis‘, komu kogo ona predstavila pervym. they were wondering, to whom who she introduced first 151. a. Ona sprosila, kogo s she asked kem ja sputal. whom with whom i mixed ‗She asked who I confused with who.‘ b.*Ona sprosila, s she asked kem kogo ja sputal. with whom whom i mixed 68 To summarize the order we distinguished so far we see that the direct object precedes indirect, all objects precede adjuncts, and there is no order distinguished between kogda – ‗when‘ and gde – ‗where‘, but they must precede kak – ‗how‘. The order of wh-subject with respect to wh-object The examples in (152-153) show that the wh-subject always precedes the wh-object; the same Superiority effect is visible in matrix clauses. 152. a. Ona sprosila, kto komu pokazal film. she asked, who whom showed movie ‗She asked who showed movie whom.‘ b.*Ona sprosila, komu kto pokazal film. she asked, whom who showed movie 153. a. Ona interesovalas‘, kto o she wondered , čëm menja sprašival. who about what me asked ‗She was wondering who asked me about what.‘ b.*Ona sprosila, o čëm kto menja sprašival. she wondered , about what who me asked The order of wh-[-human] with respect to wh-adjuncts In the order of wh[-human], with respect to wh-adjuncts, informants give a clear preference to čto – ‗what‘ over all adjuncts. In this case, čto behaves the same way as a cumulative object. 154. a. *Ona sprosila, gde she asked, čto ty videl. where what you saw ‗She asked what you saw and where.‘ b. Ona sprosila, čto gde she asked, ty videl. what where you saw 69 155. a.*Ona sprosila, gde she asked, skol‘ko deneg ja potratil. where how much money I spent ‗She asked how much money I spent and where.‘ b. Ona sprosila, skol‘ko she asked, deneg gde ja potratil. how much money where I spent To summarize the order we have distinguished thus far, we get the following picture: Table 7. Subject kto (who) Dir object Indir object [-human] Adjunct Adjunct [+human] [+human] object (place&time) (manner) kogo (who) komu (to who) skol‘ko N gde (where) (how much N) kogda (when) kak (how) čto (what) The order of wh-[+human] with respect to wh-[-human] The data show that wh-[+human] and wh-[-human] occupy different positions depending on their role in the sentence, i.e. if they are a subject or an object, and on their internal makeup, i.e. [+/-human]. In (156), kto – ‗who‘ is a subject and [+human] precedes čto – ‗what‘, which is an object and [-human]. In (157 and 158), komu – ‗whom‘, which is specified as a [+human] object, precedes čto – ‗what‘, a [-human] object and skolkoN – ‗how many N‘, an underspecified human object. In (159 and 160), čto – ‗what‘, serving as a subject, will precede the objects kogo – ‗whom‘ and s kem – ‗with whom‘, respectively. 156. a. Ona sprosila, kto o she asked, čëm menja sprašival. who about what me asked ‗She asked who asked me and about what.‘ b.*Ona sprosila, o čëm kto menja sprašival. she asked, about what who me asked 157. a. Ona sprosila, komu čto ja pokazal. she asked, whom what i showed 70 ‗She asked whom I showed what.‘ b. *Ona sprosila, čto komu she asked, 158. ty pokazal. what whom you showed a.Ona sprosila, komu skol‘ko she asked, let ty dal. whom how many years you gave ‗She asked who I gave what age.‘ b.*Ona sprosila, skol‘ko she asked, 159. let komu ty dal. how many years whom you gave a. ??/*Ona sprosila, kogo čto sdelaet she asked, sčastlivym. whom what will make happy ‗She asked what would make whom happy.‘ b. Ona sprosila, čto kogo sdelaet sčastlivim. she asked, what whom will make happy 160. a. *Ona sprosila, s she asked, kem čto slučilos. with whom what happened ‗ She asked what happened with who.‘ b. Ona sprosila, čto she asked, s kem slučilos. what with whom happened The order of a wh-subject with respect to wh-adjuncts Wh-subjects in Russian always occupy a higher position than wh-adjuncts, and the internal makeup of the subject does not play any role, contrary to Bulgarian matrix questions, where a [+human] subject will precede adjuncts and a [-human] subject will follow where and when but precede how. 161. a. Ona sprosila, kto gde she asked, budet spat‘. who where will sleep ‗She asked who would sleep and where.‘ b. *Ona sprosila, gde kto budet spat‘. she asked, 162. where who will sleep a. Ona sprosila, čto kogda projasnitsja. she asked, what when will be clear 71 ‗She asked what would become clear and when.‘ b. *Ona sprosila, kogda čto projasnitsja. she asked, when what will be clear The order of a wh-object with respect to a prepositional wh-object It has been mentioned above that direct objects precede indirect objects (see examples (150-151)). The examples below show that there is also a clear order between prepositional objects and non-prepositional ones. 163. a. Ona menja sprosila, čto she me asked s kem ja obsudil. what with whom I discussed ‗She asked me what I discussed with whom.‘ b. *Ona menja sprosila, s she me 164. asked čto ja obsudil. with whom what I discussed a. Ona sprosila, skol‘kih she asked kem ljudej o čëm ja sprašival. how many people about what I was asking ‗She asked how many people I was asking and about what.‘ b.*Ona sprosila, o she asked čëm skol‘kih l judej ja sprašival. about what how many people I was asking Based on the data above, the following hierarchy appears: Table 8. Subject Dir object Indir obj Object [+human] [+human] [-human] kto (who) kogo komu skolko N čto (what) (who) (to who) PP Adjunct Adjunct (place&time) (manner) o čëm gde (where) kak (how much N) (about what) kogda (when) (how) čto (what) s kem (with whom) 72 The order of D-linked and non-D-linked wh-phrases Russian D-linked wh-phrases kakoj/kakaja/kakoe – ‗which‘ are used as specifiers to nouns. Which-kind wh-phrases are inherently D-linked in Bulgarian, and according to Krapova and Cinque (2005) must precede all non-D-linked wh-phrases. Krapova and Cinque (2005) also mention an exceptional case where non-D-linked koj – ‗who‘ in Bulgarian precedes D-linked wh-phrases, which is explained by koj belonging to the same paradigm as kojN – ‗whichN‘. Koj has a necessary interpretation as [+human] while kojN is [human] underspecified. It is interesting to note that Russian native speakers require a strict order in embedded questions, where a D-linked kakoj-type precedes non-D-linked wh-phrases (remember that in matrix questions the kakoj-type precedes adjuncts, [-human], and underspecified objects, while there is no clear order between D-linked elements and [+human] objects). Russian exhibits the same restrictions as Bulgarian, where kto – ‗who‘, bearing [+human] features and belonging to the same paradigm as which-whphrases, has to precede [human] non-specified D-linked phrases and D-linked objects as in (169). Examples (166-168) show that a D-linked subject precedes wh-objects; (165) exemplifies the precedence of a D-linked object over a non-D-linked one, which is not the case in matrix questions. 165. a. Mne interesno kakuju kartinu komu on podaril. to me interesting which picture whom he presented ‘I am interested in which picture he presented to whom.‘ b.??Mne interesno komu kakuju kartinu on podaril. to me interesting whom which picture he presented 166. a. Byl zadan vopros, kakaja devuška kogo zainteresovala. was asked question which girl whom made interested ‗A question was raised which girl interested who.‘ b.??Byl zadan vopros, kogo kakaja devuška zainteresovala. was asked question whom which girl 167. a. Vsem stalo made interested interesno, kakoj roman kogo zaintregoval. to all became interesting which novel whom intrigued ‗Everybody became interested in which novel intrigued whom.‘ b. ??Vsem stalo interesno kogo kakoj roman zaintregoval. 73 to all became interesting whom which novel intrigued 168. a. Oni interesovalis, kakoj podozrevaemyj s they were interested,which suspect kem razgovarival. with whom talked ‗They were interested in which suspect talked with whom.‘ b. *Oni interesovalis, s kem kakoj podozrevaemyj razgovarival. they were interested, with whom which suspect 169. talked a. Bylo rešeno, kto kakuju studentku budet ekzamenovat‘. was decided who which student will examine ‘It has been decided who will examine which student.‘ b. *Bylo rešeno, kakuju studentku kto budet ekzamenovat‘. was decided which student who will examine Subsequently, including D-linked wh-phrases, the order will be the following: Table 9. Subject D-link object Dir obj Indir obj [+h] kto kakojN kogo (who) (whichN) (who) [+h] Object Prep Adjunct Adjunct [-human] object (place&time) (manner) komu skolko N o čëm gde (where) kak (to who) (how muchN) (about what) kogda (when) (how) čto (what) s kem čto (what) (with whom) kakojN (whichN) Table 9 shows a clear word order where subjects precede D-linked objects, which in turn precede non-D-linked elements. Non-D-linked elements occupy a position according to their internal make-up, i.e. [+human] precedes [-human] or underspecified human, direct objects are higher than indirect ones, and prepositional objects follow all non-prepositional ones. Adjuncts of place and time are located lower than other whelements but precede manner adjuncts. 74 3.2.4. A comparison with Meyer‟s results In this section I compare the data I presented above with Meyer‘s results, mentioned in section 3.2.1. Remember that Meyer conducted two sub-studies for Russian, which led him to the conclusion that Russian does exhibit the Superiority effect in some contexts. He revealed the following patterns of word-order preference: - a wh-subject precedes a wh-object - a wh-subject precedes a wh-adjunct - [+human] precedes [-human] - a slight word order preference of a wh-object over a wh-adjunct Meyer also claims that there is no matrix/embedded clause distinction in Russian multiple wh-questions. His analysis supports a few conclusions I drew out of my data: wh-subjects do indeed precede wh-objects and wh-adjuncts, kto – ‗who‘ precedes a D-linked object, a [human] object follows [+human] objects and there is no clear picture in wh-object/whadjunct ordering, except that kak – ‗how‘ must follow all wh-elements (the latter fact is not described in Meyer‘s analysis). However, my data contradict Meyer‘s claim that the matrix/embedded clause distinction is irrelevant. The data show that the order of whelements in embedded clauses is much stricter and more differentiated. In general, the results mentioned by Meyer (2004) correspond to the results I came up for matrix clauses, i.e. in both analyses, wh-subjects precede wh-objects and wh-adjuncts, and there is no evidence of a fixed linear order for wh-arguments. It appears, there are several drawbacks in Meyer‘s analysis. Firstly, Meyer in his study did not pay attention to the wh-adjunct/wh-adjunct ordering. Secondly, studying wh-subject/wh-object and wh-subject/wh-adjunct, he looked only at [+human] whsubject, ignoring [-human] wh-subject. Thirdly, Meyer did not look at order of whobject vs. wh-object, meaning direct/indirect/prepositional objects. I presume that if he had looked at the points mentioned above he would have found a matrix/embedded asymmetry and a fixed linear order for wh-arguments in embedded clauses, which is apparent from the data presented in this work. 75 3.3. Previous approaches to multiple wh-questions There are four main analyses of multiple wh-questions mentioned in the literature: the adjunction analysis, the wh-cluster analysis, the multiple [Spec, CP] analysis, and the split CP analysis. I will discuss each of them one by one, making reference to Russian. 3.3.1. The adjunction analysis Rudin (1988) argues for splitting multiple wh-fronting languages into two groups. The first group displays strict ordering of wh-questions, and is represented by Bulgarian and Romanian. The second group, on the other hand, includes such languages as SerboCroatian, Czech, Polish and Russian, where according to Rudin only the first element is moved to [Spec, CP] and the Superiority effect appears only in some cases. Accordingly, Rudin proposes two different structures for two groups. Consequently, the first group employs a right-adjunction structure as in (170), where the order of whelements is obtained by successive right-adjunction to the [Spec, CP]. 170. CP (Bulgarian) IP Spec dade wh Spec (give) kakvo wh (what) kogo (whom) wh koj (who) The languages in the second group are claimed to utilize IP-adjunction, i.e. the first element is moved to [Spec, CP] and the rest are left-adjoined to IP (171). 76 CP 171. Spec C´ kdo (who) (Czech) C (clitic+ parenth) IP IP Wh kdy (when) Wh VP koho (whom) V pozval (invited) The adjunction analysis predicts the Superiority effect for the languages of both groups, i.e. according to Rudin, Russian exhibits the Superiority effect, as well as Bulgarian, Romanian, Serbo-Croatian, Czech, and Polish. 3.3.2. The wh-cluster analysis Grewendorf (2001) argues for wh-elements forming a cluster by adjunction of one to the other(s) for feature checking purposes, which follows from the Cluster Hypothesis (Grewendorf, 2001:94): A particular feature of a wh-element acts as a checker for other wh-elements Based on this hypothesis then, in Bulgarian, the direct object kakvo right adjoins to kogo together forming a cluster, which further right-adjoins to the subject koj in [Spec, IP], and then the entire cluster moves together to [Spec, CP] (Laenzlinger and Soare, 2005:4). 172. Kojj kogoi kakvok [tj e pital ti tk]? who whom what is asked 77 CP Spec C´ C IP Q I´ Wh koj koj kogo VP I kogoi ti tk kakvok Grewendorf (2001) suggests that because of the uninterpretable feature [Q] in kogo, kakvo first overtly moves and adjoins to kogo, matching the [Q] features. This leads to deletion of the uninterpretable [Q] feature in the head of kogo. Next, the uninterpretable feature [Q] of wh-subject koj attracts the wh-cluster formed from kogo and kakvo, matching the interpretable [Q] features of kogo and deleting the uninterpretable feature [Q] of koj. Finally the uninterpretable feature [Q] of C matches with the interpretable feature [Q] of the wh-subject and attracts the wh-cluster consisting of three wh-phrases to [Spec, CP]. 3.3.3. The multiple [Spec, CP] analysis The multiple [Spec, CP] analysis has been proposed by a number of linguists: Bošković (1998, 2000), Richards (1997, 2001), Pesetsky (2000), etc. Bošković (1998:12) assumes that all multiple wh-fronting languages have obligatory non-wh-fronting of wh-phrases (which is argued to be an instance of focus movement)19. Under this analysis the first wh-element is moved to [Spec, CP] in order to check the strong [+wh] feature of C, 19 For analysis of the focus movement approach, please refer to Chapter 2. 78 whereas movement of other wh-elements is an instance of pure focus-movement. He suggests that in Bulgarian C has two attracting features: Attract one-F [+wh]-feature and Attract all-F [+focus]-feature. The most economical way would require moving the highest wh-phrase first to [Spec, CP] and then the remaining ones will be moved to C in any order to check their strong focus features. Pesetsky (2000) and Richards (1997)20 suggest that in a MWF language, all whphrases will move overtly (for checking wh-features) to the left periphery of the sentence, which means that after the left-most element has been moved, the other whphrases in the sentence ‗must have ―tucked in‖ underneath the first phrase, forming a lower specifier‘ (Pesetsky, 2000:22). Richards (1997) also points out that second instance of movement should ―tuck in,‖ and in addition to the Attract Closest rule, the Shortest Move condition should be obeyed as well. Obeying this condition, the second element will create a specifier lower than the first one. Therefore, the structure for multiple [Spec, CP] analysis will be the following: 173. (Bulgarian) CP C wh[+wh] koj (who) wh[+F] kogo C (whom) wh [+F] kakvo (what) IP dade (give) Pesetsky (2000) adopts overt wh-phrase movement for Bulgarian and structurally similar languages. He builds his analysis on the presence of a certain type of a complementizer which depends on the lexicon of the language, i.e. a C0-spec for wh-insitu languages, a C1-spec for English and a Cm-spec for Bulgarian. His proposal distinguishes multiple vs. non-multiple wh-questions, but fails to account for the order of wh-elements in a multiple wh-fronting language. 20 For more discussion see section 2.3.3. 79 3.3.4. The split CP analysis The Split CP analysis has been offered by a number of linguists for languages with a strict word order such as German and Bulgarian. The split CP analysis is also referred to as a Cartographic Approach (which presents a more detailed ‗map‘ of the structure). Thus, Grohmann (1998, 2001), discussing the phenomenon of multiple whquestions in German, comes to a conclusion that the German left-periphery contains a FocP hosting one wh-phrase and a lower projection FP hosting the second wh-phrase. A reason for offering this approach is given by a possibility of a topicalized element occupying a position between two wh-elements as in (174). 174. a.[FocP Wer hat [TopP alle/viele/die meisten Bücher [FP wo gekauft ]]]? who had all/many/the most books where bought ‗Who bought all/many /most books where?‘ b.*Wer hat kein Buch/wenige/höchstens drei Bücher wo who had no book /few gekauft? /at most three books where bought ‘Who bought no book/few/at most three books where?‘ In (174a), quantificational elements can be placed between wh-elements, while in (174b) negative quantificational phrases kein – ‗no N‘, wenige N – ‗few N‘ cannot, which strongly suggests that the position between the wh-elements is in fact a Topic position. Grohmann shows that both wh-elements move overtly to FocP and FP, and that anything in between fills in Rizzi‘s (1997) TopP. Laenzlinger and Soare (2005) employ the Cartographic approach to multiple wh-fronting languages, which fits in with the Kayne‘s Antisymmetry (Kayne 1994)21. 21 Kayne‘s Antisymmetry denies the assumption that Universal Grammar (UG) allows a given hierarchical representation to be associated with more than one linear order. He argues that phrase structure always determines linear order, so that if two phrases differ in linear order, they must also differ in hierarchical structure. (Kayne, 1994 :3). Kayne shows that asymmetric c-command invariably maps into linear precedence, from which it follows that complement positions must always follow their associated head, and that specifiers and adjoined elements must always precede their sister phrase. Thus he argues for a uniquely imposed specifier-head-complement order. As a natural consequence to such 80 The Cartographic approach is employed by Belletti (2004), Cinque (1999, 2002), and Rizzi (1997, 2004), where CP (or the Vorfeld) and IP (or the Mittelfeld) are areas rich in functional projections, roughly represented in (175). 175. [ForceP [TopP* [FocP [TopP* [FinP … [MoodP* [ModeP* [TP* [AspP* [VoiceP …]]]]]]]]]] The crux of this approach is to create a detailed map of the structure coming from the assumption that all languages share the same principles of structure formation and the same functional projections and vary in types of movements and overt realization of functional projections. The example in (174) illustrates that both CP and IP fields are split in several projections, which are hierarchically ordered. Krapova and Cinque (2005) employ the Cartographic approach and Relativized Minimality in their paper on multiple wh-fronting in Bulgarian. In their analysis, each class of wh-phrases targets a specific position in the left periphery (see the discussion above of their analysis in chapter 2). The important aspect of their analysis is that the order of wh-elements is dependent upon the order prior to their wh-movement. Krapova and Cinque (2005:192) suggest that ―within a system in which Superiority is subsumed under Relativized Minimality, the preservation of the pre-wh-movement order of the wh-phrases in the case of multiple movements can be ensured through the requirement‖ where only a whole chain, not just a link of a chain, counts as an intervener, i.e. ‗no chain can intervene between the trace and the target, if this chain bears the same features as the target.‘ Laenzlinger and Soare (2005), analyzing Krapova and Cinque‘s approach, suggest that they propose as many Foc positions (in the terms of Rizzi, 1997) as fronted wh-elements in Bulgarian. Laenzlinger and Soare (2005), using the same approach for Romanian, follow Rizzi‘s idea of the existence of only one FocP at the left periphery. They argue for a structure where multiple wh-fronting relies on the mechanism of ‗heavy pied-piping.‘ They offer the following structure: order, movements are also highly restricted to given positions in the role of suitable landing sites for moving constituents. 81 176. Cine ce a citit? who what read ‗Who read wahat?‘ ForceP FocP Foc Foc unFF MP Spec SubjP =AgrsP V a citit (read) cine +wh (who) strong FF (wh) ObjP =AgroP ce (what) strong FF (wh) vP Following Krapova and Cinque‘s assumptions, Laenzlinger and Soare suggest that whelements leave the VP-shell in order to reach a position in the Mittelfeld (IP), where they can check their A-features (case, phi). These positions are labelled as [Spec, SubjP] for wh-subjects and [Spec, ObjP] for wh-objects (Laenzlinger and Soare, 2005:29). FocP has an uninterpretable wh-feature, which needs eliminating by agreement with a wh-element that it c-commands, and EPP/OCC features as well. Both wh-elements move overtly to [Spec, FocP] by pied-piping. Following Krapova and Cinque (2005) and Laenzlinger and Soare (2005), I employ the Cartographic approach for Russian wh-questions based on the data introduced in Chapter 3. Before offering an analysis I introduce examples which show that the prior order reflects the order after the movement. 82 3.4. Analysis of Russian multiple wh-questions 3.4.1. Hierarchy prior to movement Russian is an SVO language that is very often considered to be a free-order language. The following order is observed in the language: 177. Ivan pročёl stat‘ju o beţencax. Ivan read article about refugees ‗Ivan read an article about refugees.‘ 178. Stat‘ju o beţencax Ivan pročёl. article about refugees Ivan read 179. ?Stat‘ju o beţencax pročёl Ivan. article about refugees read Ivan 180. ??Ivan stat‘ju o beţencax pročёl. Ivan article about refugees read Example (177) is a default order which is used in everyday life and written language. In (178) we put a stress on statja o bežencax – ‗article about refugees‘, here it is important for a speaker to stress that Ivan read an article, not a book about refugees. The order in (179 and 180) is possible, although quite marginal. The principal order in Russian is SVO. Furthermore, I consider the order of adjuncts in IP structure comparing with the order observed in wh-questions. Recall the order of the adjuncts in Russian multiple embedded wh-questions,22 where gde and kogda precede kak. In the IP structure, the adjuncts behave in the following way: no preference of Time over Place (181), and Manner somehow precedes all adjuncts (182-183). 22 Here I will refer to embedded wh-elements, given that the order there is more distinguished. 83 181. Time Ja ego videl včera Place Time v universitete (včera). I him saw yesterday in university (yesterday) ‗I saw him yesterday at the university.‘ 182. Manner Manner Place Manner On gromko razgovarival (gromko) v sosednej komnate (*gromko). he loudly 183. spoke Manner (loudly) in next Manner On bystro vse room Time (bystro) zakončil večerom he quickly everything (quickly) finished (loudly) Manner (*bytro). in the evening (quickly) Based on the examples above, it looks like Manner adjuncts are situated higher than Time and Place, but I will follow Nilsen (2000) and Cinque (2002a), who in their studies prove the position of the Manner adjunct to be lower than Time and Place. The adverb of Manner was observed to occupy a higher position in a few languages, which was explained by some movements in the structure. I will not look at this issue here; I will just adopt the position of those linguists who claim Manner adjuncts to be situated lower than Time and Place. Therefore, in Russian, the order of adjuncts kogda, gde and kak reflects the order prior to their movement, where kogda and gde precede kak, and are not ordered between each other. Similarly, the order of [+human] wh-phrases with respect to adjuncts appears to reflect their relative order prior to wh-movement, i.e. objects and subjects precede adjuncts, as in (184). 184. Subj Obj Place Time Manner včera/ v papyhah. On vstretil eje v sadu/ he met in the garden/yesterday/in a hurry her A Subject [-human] always precedes a [-/+ human] object, as in wh-questions; this is a neutral order. 84 185. Subj [-h] Object Eto udivilo Petju This impressed Peter The examples above serve as an argument that the order in the IP structure (Mittelfeld) reflects the order after the movement. Subject precedes Object, following from default SVO order. Adjuncts of Time and Place do not reveal any strict ordering but both of them must precede Manner adjunct. This serves as one more piece of evidence for analyzing wh-questions according to the Cartographic approach. 3.4.2. Cartographic approach to Russian multiple wh-questions Rizzi (1997) proposes an approach whereby the C system expresses at least two kinds of information, one facing outside and the other facing inside. He suggests treating the first kind of information (information about the sentence type, i.e. a question, a declarative, a relative etc.) as the specification of Force and the second one as finiteness (since according to him C expresses distinctions related to tense), which in turn selects IP. He also suggests distinguishing Focus/Topic within the C system, which will be occupied by question operators, while Force is occupied by relative operators. Force . . . (Topic) . . . (Focus) . . . Fin IP Rizzi (1997) argues for one structural Focus position and an indefinite number of Topics per clause. 186. C Credo Top Foc Top IP che a Gianni, QUESTO, domani, gli I.belive that to Gianni this dovremmo dire. tomorrow to.him we.should say ‗I believe that to Gianni, THIS, tomorrow we should say.‘ 85 Rizzi suggests that question operators (wh-phrases in our case) are in Spec of FocP, since in Italian focalized constituents and question operators compete for the same position, and thus cannot co-occur. 187. *IL PREMIO NOBEL a chi THE NOBEL PRIZE deverebbero dare? (Italian) to whom should they give It should be pointed out that Rizzi (1997:330) comes across examples where the Wh element is marginally compatible with an embedded focalized element in embedded questions as in (188a), contrary to main questions (188b). 188. a.? Mi domando A GIANNI che cosa abbiano ditto (, non a Piero) ‗I wonder TO GIANNI what they said (, not to Piero)‘ b. *A GIANNI che cosa hanno ditto (, non a Piero) ‗TO GIANNI what they said (, not tp Piero)‘ Rizzi suggests that the Wh element occupies an independent position distinct from [Spec;FocP] in Italian embedded questions. Russian, on the other hand allows the co-occurrence of a focalized element and a wh-element in the structure, i.e. a focalized element and a wh-element in Russian do not compete for the same position (189) and thus, following Rizzi‘s suggestion, we will offer an extra projection for Russian wh-phrases, distinct from [Spec;FocP] and to which we refer to as WhP. 189. (situation: two people talking about the presents for children. One of them is asking: ‗What did you buy for the kids, especially for Masha?‘ The other replies that he got a car for Peter, a doll-house for Kate etc, not mentioning Masha first. The first speaker is impatient and asking:) Da MAŠE čto ty kupil, a ne drugim? to MASHA what you bought, but not to others In (189), Maša is in a Focus position which co-occurs with čto – ‗what‘, which seems to occupy a different position. 86 The question is where this WhP will be situated. With respect to the Focus position, it can appear higher or lower than FocP (190). I follow Rizzi, who argues for one structural focus position and the non-recursion of Focus in the structure. 190. MAŠE čto MAŠE ty kupil, a to MASHA what ne drugim? you bought, but not to others With respect to the Topic position, WhP can be situated higher and lower than Topic in questions with multiple wh-elements as in (191a). Topic can never precede wh-elements (191c, d), but must follow it (191e). 191. a. Kogda Maša komu zvonila? when Masha whom called b. *Kogda kazhdyj rebenok komu zvonil? when every child whom called c. *Maša kogda komu zvonila? Masha when whom called d. ??Maša kogda zvonila? Masha when called e. Kogda Maša zvonila? when Masha called I assume that Maša is in a Topic position since this position cannot be filled by quantificational elements as in (191b). The fact that wh-elements in Russian can be placed higher than Topic strongly suggests the occurrence of wh-movement in Russian. Based on the examples above, I come to a conclusion that Russian WhP, which is used as a landing position for wh-elements in the structure, can be situated higher or lower than FocP and TopP can intervene between WhPs, but cannot precede whelement as in (191c and d) Force . . . (WhP) . . . (Topic) . . . (WhP) . . . (Focus). . . (WhP) . . . Fin IP The data above clearly demonstrates that Russian wh-elements do not compete for FocP, inasmuch as they allow a focalized element to occur in the same structure, 87 therefore a WhP is used as a landing site for wh-elements and not FocP as in Italian following Rizzi. Furthermore, it was shown that wh-elements in CP reflect the order in the Mittelfeld, the order prior to wh-movement. Moreover, wh-elements allow the intervention of lexical material, which can occupy TopP. This strongly supports the movement of wh-elements, contrary to Bošković and Stepanov, and additionally it serves as an argument countering the idea of pied-piping movement of wh-elements to [Spec, FocP], as proposed by Laenzlinger and Soare (2005). Following from the discussion above, I propose the following structure for Russian multiple wh-questions: 192. Ona interesovalas‘, komu She wondered ForceP Pavel kak pomog. Obj [+h] manner adj [strongWh] [strongWh] whom Pavel how helped FocP WhP TopP komu WhP (whom) Pavel (Pavel) . . . IP kak . . . ObjP (how) MannerP tk tj I assume that wh-elements which have strong [wh] features move to a particular position in the C system to check wh-features. The hierarchy realized in the Mittelfeld is retrievable through chain crossing at the left periphery. I would suggest the following hierarchy for wh-elements: [ForceP[FocP[SubjP[d.obj+h[ind.obj+h[obj- h [PPobj [adjunct adv [manner adv]]]]]]]]]]] 88 Notice that in the presented hierarchy above, adjuncts of Time and Place are not distinguished, and adjunct of manner is situated lower. [+human] objects precede [human] objects as well as prepositional ones. Now a question arises when we try to account for the cases where no order is distinguished as such, as in the case with wh-adjuncts kogda and gde, where no preferred order is distinguished, i.e. kogda can precede gde and vice versa. I suggest that in this case the order in the C system depends on the order realized in Mittelfeld. Thus, if in the Mittelfeld gde precedes kogda, the order is preserved prior to the movement, and the same order will be visible on the surface after the movement. Preference of one wh-element over another one depends on the prominence of the information. Thus in (193), gde – ‗where‘ is more important for the speaker than kogda – ‗when‘ and (194) shows prominence of kogda – ‗when‘ information over gde – ‗where‘. 193. Ona interesovalas‘, gde kogda ja ego videl. she was wondering where when I him saw 194. Ona interesovalas‘, kogda gde ja ego videl. she was wondering where when I him saw Structure for (193) Structure for (194) ForceP ForceP FocP FocP WhP WhP WhP WhP gdej kogdaj . . .IP (where) kogdak (when) gdek . . .LocP (when) . . .TimeP LocP (where) TimeP tj . . .IP tj tk tk In the Mittelfeld of (193), the locative adjunct gde – ‗where‘ precedes the time adjunct kogda – ‗when‘, and in (194) gde follows kogda. The order in the Mittelfeld is preserved and reflected in CP. 89 This approach accounts for the order of Russian wh-embedded questions. The question that arises is how to treat the data in Russian wh-matrix clauses. The order of wh-elements in matrix questions which appears as non-defined can be argued to be ―ruled‖ by the internal make-up of wh-elements and prominence of information. Due to an absence of a controller which is present in embedded questions, the order is based on internal make-up or importance of information, while in embedded questions it is ―chosen‖ by a licenser. Recall the order distinguished for matrix questions: Table 10 Subject [+human] Object [-human] Object Adjunct Adjunct (place&time) (Manner) kak (how) kto (who) komu (whom) kogda (when) čto (what) kogo (whom) gde (where) kakoj/kakaja N (which N) kakoj/kakaja N čto (what) (which N) skolko N (how much N) PP object Subjects precede all wh-elements, [+human] objects and D-linked objects do not show differentiated word order. D-linked elements precede all adjuncts, [-human] whelements and prepositional object. Note that [+human] objects show only a slight preference over adjuncts of time and place, though a clear order with respect to [human] objects. Based on the data above I assume a structure for Matrix questions in (195) where the subject is located higher than the [+human] object which in turn is located higher than [-human] objects and adjuncts, which precede Manner adjuncts. 195. [ForceP[FocP[SubjP[[+hum]objP [objP, adjunctP[ manner adjunctP]]]]]] Following Rizzi and Shlonsky (2007), I assume that a wh-subject moves to the leftmost wh-projection (WhSubjP) without checking any EPP features in [Spec, TP]. I suggest that in matrix questions the order is ―ruled‖ by the internal make-up of wh-elements, prominence of information and function of the elements. For example, 90 [+human] elements always precede [-human] elements as in (196), more important information will be placed higher than less important (197), and the subject is functionally placed higher than other elements (198). 196. a. Komu čto on pokazal? whom what he showed ‗Whom did he show what?‘ b.*Čto komu on pokazal? what whom he showed 197. Gde kogda on ego uvidel. where when he him saw ‗Where did he see him and where?‘ 198. a. Kto kogo uvidel? who whom saw ‗Who saw whom?‘ b.*Kogo kto uvidel? whom who saw The order is pre-defined in the Mittelfeld, which is later realized in CP. A structure for (196) is presented as (199), where, in the Mittelfeld, komu – ‗whom‘, due to internal make-up and function, is located higher than čto – ‗what‘, and the same order is realized later in CP. This also fulfils Rizzi‘s Relativized Minimality in view of Krapova and Cinque, where no chain can intervene between the trace and the target if the chain bears the same features as the target. In (196), only a part of chain is crossed, which is allowed; yet notice that having the order where čto precedes komu will lead to crossing the whole chain, i.e. a violation of Relativized Minimality, which renders the sentence ungrammatical, as in (196b). 91 199. ForceP FocP WhP WhP komuj (where) . . . IP čtok [+human] Obj (what) [-human] Obj tj tk To summarize, it is evident that the Cartographic approach works for accounting for the syntactic structure of Russian wh-elements, taking into account the strict order revealed in embedded questions. Following Rizzi‘s (1997) approach, I propose to extend the C system and distinguish additional projections, i.e. ForceP, FocP, TopP. Notice that I propose extra projections WhP for wh-elements (following Rizzi‘s (1997) suggestion for wh-movement in embedded clauses), which serve as landing sites for wh-phrases (contrary to Rizzi, who suggests FocP as a landing projection for wh-elements in matrix clauses). Following Krapova and Cinque (2005), I assume that the order of wh-elements is realized in the Mittelfeld prior to the movement and is preserved under the notion of Relativized Minimality. In order to account for a not-really-free order of wh-elements in wh-matrix questions, I propose that the order is ―ruled‖ by internal make-up, syntactic function of wh-elements and prominence of information, based on which order is realized in Mittelfeld, preserved under the notion of Relativized Minimality and emerges in CP. This chapter has offered a preliminary analysis of wh-elements in Russian. Covering Subjects, Objects, and Adjunct wh-elements, I ignored for now a type of whelements which should be classified as ―adjuncts‖. These adjuncts can be referred to as ―why-wh-elements‖ based on their main meaning of questioning purpose and reason. I devote the next chapter of my dissertation to ―why-wh-elements‖, which comprise at least 6 subgroups: (1) počemu – ‗why‘, asking for a reason as in Why are you here?; (2) začem – ‗why/what for‘, asking for a purpose as in Why did you come?; (3) aggressively non-D-linked kakogo cherta, kakogo figa, kakogo lešego etc – ‗why the hell‘ as in Why 92 the hell did not you tell me about this?!; (4) nominal čto/čë/čego – equivalent to English ‗what‘ as in What are you doing here? In a meaning ‗why are you here?‘; (5) prepositional na čto tebe, na koj tebe, za kakim xrenom – ‗what for‘, and (6) why with a particle xrana li/figa li/čego že – ‗why‘. 93 94 CHAPTER 4. WHY-WH-ELEMENTS AND THEIR VARIETY Different languages employ different ways for asking for a reason, one of which is the use of wh-adjuncts whose interpretation is akin to English ‗why.‘23 English itself exhibits several different adjuncts for asking for reasons (why, how come, why the hell) or purpose (what for) (200). 200. a. Why did John come? b. How come you are here? c. Why the hell did you never tell me about it? d. What are you asking this for? In French one can identify at least three such elements pourquoi, pour quelle raison, pour quoi diable, in Italian three – perchè, per quale ragione, come mai, in German five – warum, wieso, wofür, wass für, was, etc. In this Chapter, I address six types of Russian ‗why‘wh-elements which demonstrate some common properties: (1) počemu– ‗why, for what reason‘ as in (201a), (2) začem–‗why/what for/for which purpose‘ as in (201b), (3) aggressively non-Dlinked kakogo cherta, kakogo figa, kakogo lešego etc as– ‗why the hell‘ as in (201c), (4) nominal čto/čë/čego as in (201d,e,f), (5) prepositional na čto tebe, na koj tebe, za kakim xrenom – ‗what for‘ as in (201g,h) and (6) why with a particle xrana li/figa li/čego že – ‗why the hell‘ as in (201i). 201. a. Počemu ty zadal etot vopros? why you asked this question ‗Why did you ask this question?‘ 23 I refer to wh-adjuncts with interpretation of asking for reason as ‗why‘wh-elements. 95 b. Začem ty podnjal ruku? why you raised hand ‗Why did you raise your hand?‘ c. Kakogo čërta ty prišël? which devil you came ‗Why the hell did you come?‘ d. Čto ty prišël? why you came ‗Why did you come?‘ e. Čё ty molčiš‘? why you keep quiet ‗Why do you keep quiet?‘ f. Čego ty plačeš‘? why you crying ‗Why are you crying?‘ g. Na koj tebe eto nado? on which you this need ‗Why the hell do you need this?‘ h. Za kakim xrenom ty pripёrsya? for which horseradish you came ‗Why the hell did you show up?‘ i. Tak čego ţe ty prišël? so what prt you came ‗So why did you come?‘ All the wh-elements listed above can express either a reason or purpose and show some similarities in their distribution. It should be specially pointed out that ‗why‘-elements from (201c) to (201i) have an aggressive connotation (i.e. the speaker is irritated or annoyed); while počemu and začem can be characterised as neutral. From the first glance at ‗why‘-elements, začem and počemu stand out as more standard elements for asking for a reason, therefore I split above mentioned phrases into two categories: standard/neutral and aggressive. Later in the research, these two groups will be shown to correspond to different syntactic positions. First I focus on začem and počemu, then I address other types of why in detail. 96 4.1. Počemu vs. začem The wh-adjuncts počemu and začem, being equivalent to English ‗why‘, demonstrate a similarity in their distribution but exhibit some semantic differences. Thus, Tolkovyj Slovar‘ Russkogo Jazyka (1992) gives the following definitions: Počemu – reason ‘why’ Začem – purpose ‘why’ po kakoj pričine – ‗because of which reason‘ s kakoj celjuj – ‗with what purpose‘ vsledstvii čego – ‗because of what‘ dlja čego – ‗for what‘ It is clear from above that the two non-homonymous lexical items počemu and začem distinguish purpose from reason varieties of why. Počemu triggers reason as in (202) while začem triggers purpose as in (203) 202. Q: Počemu on zlitsja? why he angry ‗Why is he angry?‘ A: Potomu čto ego komanda proigrala because his team lost ‗Because his team lost.‘ 203. Q: Začem on zapel? why he started singing ‗For what purpose did he start singing?‘ A: Čtoby privleč‘ vnimanie in order to get attention ‗In order to get attention.‘ Stepanov and Tsai (2007), studying the syntactic distribution of Slavic whys, underline that the distribution of počemu and začem is more or less the same for simple matrix questions as in (202 and 203). They point out, however, that the behaviour of the two items changes with respect to negation, where negative počemu questions are perfectly grammatical as in (204), while negative začem questions are ill-formed as in (205). 97 204. Počemu ty ne otvečaeš na moi zvonki? why you not answering on my calls ‗Why are you not answering on my phone calls?‘ 205. *Začem ty ne napomnil mne ob etom? why you not reminded me about this ‗Why did not you remind me about this?‘ Stepanov and Tsai explain the distinction between the two elements based on syntactic properties of the two. They claim that začem is base-generated in the vP domain, below NegP. The ungrammatically is due to the intervention effect caused by začem crossing the negative operator, which creates a weak island as is shown in (206). 206. ForceP WhP Wh` FocP začemk (why) NegP tyj (you) [neg] Neg` TP ne (not) T` DP tj VP VP V` napomnili (reminded) k DP tj V whP ti tk Počemu, on the other hand, seems to be base-generated in CP, which is above NegP24as in (207). 24 The idea of why being base-generated in CP was first discussed by Rizzi (2001) for Italian and later adopted by Shlonsky and Soare (2009) for Romanian. I will outline their analyses in 4.1.3. 98 207. 204 ForceP WhP Wh` FocP počemu (why) NegP tyj (you) Neg` [neg] TP ne (not) T` DP tj VP VP otvečaeši k (answering) V` DP tj V ti The idea of different placement of the two elements is supported by a particular morphological paradigm concerning the formation of negative wh-words (Stepanov and Tsai, 2007:18). Almost all wh-phrases can form negative counterparts except počemu – ‗why‘ as in (208f) and kak – ‗how‘ as in (208g): 208. a. Ne-komu/ ne-kogo not-whom-dat/ not-whom-acc b. ne-čemu/ ne-čego not-what-dat/ nor-what-acc c. ne-gde not-where d.ne-kogda not-when e. ne-začem not-why f.*ne-počemu not-why g.*ne-kak not-how 99 209. a. Mne nekomu zvonit‘. to me not-whom-dat call ‗I have none to call.‘ b. Mne nezačem s toboj razgovarivat‘. to me no-why with you talk ‗I have no purpose in talking to you.‘ If we assume that this is correct and začem cannot form a negative question as in (205) based on its position below NegP, the question that arises is why other wh-questions such as kogo – ‗whom‘, kogda – ‗when‘, gde – ‗where‘, et., can form negative questions, if they are base-generated lower than NegP like začem and will have to cross a negative operator on their way to CP. 210. Kogo ty ne oprosil? whom you not questioned ‗Whom did not you question?‘ (Situation: Peter gave a lecture and was planning to check that each student understood. At the end of class, and he had not manage to ask a couple of students) 211. Kuda ty ne zašel včera? where you not drop by yesterday ‗Where did not you drop by yesterday?‘ (Situation: a child was supposed to drop by a bakery, a pharmacy and a sport shop, but met a friend and did not manage to fulfil his chores) The possibility of forming negative wh-phrases in (210 and 211) is due to the D-Linked nature of the elements. Since an answer to the questions above is drawn from a set which is clear for both the speaker and hearer. It gets more intriguing when we try to form negative questions with kak – ‗how‘. Clearly, negative kak-question is grammatically ill-formed (212) like začem in (205). Interestingly, the presence of particle že improves the question and makes it the perfectly grammatical (213), though the question is no longer a ‗manner‘-question, but a ‗why‘-question. I assume that particle že makes kak in (213) D-linked, forcing a context known for both the speaker and listener. 100 212. *Kak ty eto ne sdelal? how you this not did ‗*How did not you do it?‘ 213. Kak ţe ty ne posetil etot muzej? how prt you not visited that museum ‗How come you did not visit this museum, I cannot believe it!‘ (Situation: Stuart was on a trip to London and he visited numerous museums but forgot to visit the National Gallery. The speaker is upset that Stuart neglected to take his advice to visit it). From the data presented above it is obvious that only one bare wh-element počemu can be used with negative questions, which suggests a rather high base-generated position for this wh-element. Other wh-elements can appear in negative questions only when the wh-phrase is D-linked as I showed in examples above. Začem, on the other hand, cannot be D-linked and cannot be used in negative questions. Stepanov and Tsai (2007:25) address a question of agentivity and claim that unlike počemu, začem cannot appear in a priori non-agentive contexts. Examples below are taken from their research (where whyR stand for reason-why and whyp for purposewhy): 214. Počemu/?*začem Ivan byl arestovan? whyR [passive] whyp Ivan was arrested ‗Why was Ivan arrested?‘ 215. Počemu/*začem eta kniga upala na pol? whyR whyp this book fell [unaccusative] on floor ‗Whys did this book fall on the floor?‘ 216. Počemu/*začem Ivan ljubit cvety? whyR [experiencer/transitive sentient verbs] whyp Ivan loves flowers ‗Why does Ivan love flowers?‘ 217. Počemu/*začem Ivan zaplakal? whyR [unergative sentient verbs] whyp Ivan started to-cry ‗Why did Ivan start to cry?‘ 218. Počemu/*začem smerkaetsja? R why p why [weather verbs] getting-dark 101 ‗Why is it getting dark?‘ I disagree with their data and claim that both počemu and začem can be used in a priori non-agentive contexts, though it should be pointed out that due to semantic differences between the two elements not every verb can be used with začem. Below I provide data showing the occurrence of začem in the above-mentioned contexts. 219. Začem Ivan byl priglašen na uţin, whyp Ivan was invited čtoby ego isportit‘? [passive] on dinner, in order to it spoil? ‗Why was Ivan invited for dinner?‘ Answer: ‗in order to destroy your plans!‘ 220. Začem ty prygala tak vysoko, znala ţe čto upadeš‘? whyp you jumped so high, [unaccusative] knew prt that will fall ‗Why did you jump so high, you knew you would fall.‘ Answer: ‘I thought I could fly if I jumped high enough!‘ 221. Začem ja ego poljubila? p why [experiencer/transitive sentient verbs] I him fell in love ‗Why did I fall in love with him?‘ Answer: ‗because he is the only man left in a city.‘ 222. Začem Ivan zaplakal? [unergative sentient verbs] whyp Ivan started cry ‗Why did Ivan start crying?‘ Answer: ‗he wanted his mother to buy him a toy-car‘ 223. Nu začem xolodaet?! [weather verbs] prt whyp getting-cold ‗Why is it getting cold?‘ (rhetorical question) The examples above illustrate the use of začem in a prior non-agentive contexts, contra Stepanov and Tsai. Examples (219-221) have a reading of reproach or criticism. Stepanov and Tsai (2007:25) suggest that a characteristic feature of the purpose why is its volitional character: in order to ask for the purpose for doing something, one must imply a volitional agent to whom this doing is ascribed. Furthermore, they propose, that since agentivity is associated with the syntactic level of vP, purpose why should be regarded as a vP adverbial. Reason why, on the other hand, takes the 102 underlying event as its internal argument and it enters the structure as an operator with sentential scope. There seem to be other syntactic similarities and differences between the two elements, among others, long-distance dependencies and occurrence in multiple wh-questions. 4.1.1. Long-distance dependencies In English, two wh-elements for asking reasons why and how come show differences in terms of whether they allow long-distance dependencies. As has been pointed out by Collins (1990), why allows long-distance dependencies, provided that there is no island, while how come allows only local dependency. 224. a. Why did John say Mary left? (ambiguous) b. How come John said Mary left? (only matrix) Strict locality of how come follows from the strict locality of the head movement constraint, which is assumed to be clause-bound and, following Collins, how come is analyzed as an interrogative C head. Notice that English why, on the other hand, can modify both embedded and matrix clauses (224a). Russian data show that počemu and začem can modify both clauses, like ‗other‘ wh-elements: the matrix clause is modified whenever a complementizer čto – ‗that‘ is present and the embedded clause is modified in its absence25. Examples (225b, 226b and 227b) show long-distance dependencies and modify embedded clauses, while examples in (225a, 226a and 227a) show local dependency and modify matrix ones. 225. a. Kogda on skazal, čto Alex pridet? when he said that Alex will come 25 I discuss čto -‗that‘ complementizer in detail in Chapter 5. 103 (only matrix) ‗When did he say that Alex was coming?‘ b. Kogda ty dumaeš‘, John vstretilsja s when you think John met Mašej? (only embedded) with Masha ‗When do you think John met with Masha?‘ 226. a. Počemu ty skazal, čto on uvolnjaetsja? why you said (only matrix) that he resigns ‗Why did you say that he was resigning?‘ b. Počemu ty skazal, on uvolnjaetsja? why (only embedded) you said he resigns ‗Why did you say he was resigning? 227. a. Začem ty skazal, čto on hočet why you said so mnoj vstretitsja? that he wants with me (only matrix) meet ‗Why did you say that he wanted to meet me?‘ b. Začem ty skazal, on hočet why you said so mnoj vstretitsja? he wants with me (only embedded) meet ‗Why did you say he wanted to see me?‘ Examples from (225) through (227) show that the presence of čto - ‗that‘ changes the meaning of the sentence and does not allow ambiguous readings, while the absence of the complementizer forces the embedded clause reading for ‗other‘ and ‗why‘ whelements. According to the judgments of native speakers, in (226a and 227a) the speaker is asking a reason/purpose for saying that; in (226b) – the reason for resigning and (227b) – the purpose for meeting. Examples with začem and počemu in (226b and 227b) can get some sort of echo reading in a particular context as below: A: Ja slyšal čto Peter xočet vstretitsja s I heard that Peter wants to meet toboj čtoby obsudit problemy v Izraile. with you in order to discuss problems in Israel ‗I heard that Peter wants to meet with you to discuss the problems in Israel.‖ B: Začem ty skazal, on xočet why you said so mnoj vstretitsja? he wants with me to meet ‗WHY did you say he wants to meet me?‖ (B speaker is either surprised or thinking that he did not quite catch the purpose of the meeting) 104 It follows from the above data that both elements allow long-distance dependencies, though it depends on the presence of the that-complementizer. 4.1.2. The occurrence of “why” phrases in multiple wh-questions If we continue to look at English why and how come, trying to distinguish similarities and differences with Russian whys, we come to a conclusion, following Ochi‘s analysis (2004:2), that the English counterparts differ in their occurrence in multiple whquestions. In this respect, why can easily co-occur in multiple wh-questions (228a) while how come cannot (228b). 228. a. Why did John eat what? b.*How come John ate what? In Russian, on the other hand, začem can easily occur in multiple wh-questions (229) while počemu may occur only with arguments (230). 229. a. Kto začem prixodil? who why came ‗Who came and why?‘ b. Čto začem oni prinesli? what why they brought ‗What did they bring and why?‘ c. Kogda začem on prixodil? when why he came ‗When did he come and why?‘ 230. a. Kto počemu sejčas ne spit? who why now not sleep ‗Who does not sleep now and why?‘ b. Kogo počemu vybrala Rossija? who why chose Russia 105 ‗Who did Russia choose and why?‘ c. S kem počemu ona razošlas‘? with whom why she left ‗Whom did she leave and why?‘ d.*Kogda počemu oni smejalis‘? when why they laughed ‗When did they laugh and why?‘ e.*Kak počemu oni vstretilis‘?26 how why they met ‗How did they meet and why?‘ Following Rizzi‘s Relativized Minimality and the idea that počemu is base-generated in CP, it does not come as a surprise that only argument wh-elements can move over the element. According to Relativized Minimality, Y is in a Minimal Configuration with X iff there is no Z such that (i) Z is of the same structural type as X, and (ii) Z intervenes between X and Y So, Y and X are in a minimal configuration when Y is a trace and X is a target and there is no intervener Z between them, which c-commands Y and does not c-command X and which is specified with the same features as the target. ……Xt………….Z………….Y(t)… [F] [F] [N] c-command In (230 a, b, c) počemu and the arguments bear different features and therefore počemu is not an intervener (231), whereas in (230 d, e) the elements have the same features (all of them are adjuncts) which makes počemu an intervener (232). 26 It is worth pointing out that in Bulgarian how and why cannot co-occur either (p.c. Iliana Krapova) *Kak i zašto e otgovoril ? how and why he-answered ‗How did he answer and why ?‘ 106 231. ……ktot…………. počemu ………….(t)… [argument] 232. [adjunct] *…kogdat…………. počemu ………….(t)… [adjunct] [adjunct] In the next section I study other types of ‗whys‘ found in Russian, I will refer to them as aggressively non-D-linked like kakogo cherta – why the hell and nominal adjunct whelements čto, čë and čego. The questions is: Do they manifest the same restrictions as počemu and začem or do they have some features which bring them closer to English how come? 4.1.3. Position of počemu and začem There is numerous literature devoted to the analysis of the position of why in the syntactic structure (Hornstein 1995, Ko 2005, Rizzi 2001, Shlonsky & Soare 2009, Stepanov & Tsai 2007 a.o.). Rizzi (2001) suggests that Italian why – perché – is base generated in [Spec; IntP] which is situated higher than [Spec; FocP], a position of wh-elements. Shlonsky & Soare (2009) argue that why is externally merged as the specifier of a dedicated functional projection which they label as CauseP which is configured above NegP, adverbials and most probably above the canonical subject position. They propose that the ambiguity of (233) is a result of why positioning. 233. Why did you say that John left? (1st reading) the speaker is questioning the reason for saying (2nd reading) the speaker is questioning the reason for John‘s leaving In order to obtain the first reading, why raises from [Spec;Cause] to [Spec;Int] which is the criteria position (in Rizzi‘s terms); the second reading is obtained due to why‘s movement from its base position in the embedded [Spec;Cause] to a criterial position in the matrix left periphery. 107 ForceP IntP TopP whyi FocP WhP CauseP FinP ti In order to support their argument, Shlonsky & Soare (2009:2) analyze infinitival questions headed by wh-elements. 234. I asked Bill a. whether to serve spiced aubergines for dinner b. who to serve c. what to serve the guests d. when to serve pickled aubergines e. how to serve pickled aubergines f. where to serve pickled aubergines g. ?? why to serve pickled aubergines The examples above show that not all wh-elements can occupy the left periphery of infinitivals. The same goes for Topic as in (235). 235. *John decided [his mother] not to invite for his wedding. Shlonsky & Soare (2009) propose that the examples above argue for the splicing of infinitival clauses at WhP: 236. Force P IntP Topic P Focus P WhP FinP As a result, a WhP is not an attractor for why, otherwise (234g) would pattern with other wh-infinitivals, and a proposed position for why [Spec;IntP] is a part of the left 108 periphery which seems to be truncated in infinitivals, as shown above. The ungrammaticality of (234g) according to Shlonsky & Soare (2009) follows from the absence of an appropriate landing site for why in the embedded infinitival left periphery. It should be pointed out that the same sequence is observed in Russian, i.e. whyR(eason) cannot head an infinitival. 237. Ja sprosil Bila a. komu podavat‘ pirogi. I asked Bill whom to serve pies b. čto podavat‘ gostjam what to serve to.guests c. kogda podavat‘ pirogi when to serve pies d. gde podavat‘ pirogi where to serve pies e. *počemu podavat‘ pirogi whyR to serve pies f. začem podavat‘ pirogi sejčas whyP to serve pies now (237e) demonstrates the ungrammaticality of using počemu as the head of an infinitival, while začem behaves as other wh-operators. This suggests one more time a different syntactic position for the two elements. Since počemu and začem occupy different positions in the structure and začem shows a lot of similarities with other wh-elements, I put začem aside for now and focus on počemu in order to distinguish a base position for the element in the structure. Another empirical argument against the external merge of why in [Spec;IntP] is provided by Shlonsky & Soare (2009:6) based on Romanian data which is a rigid multiple fronting language. The data show that Romanian de ce – ‗why‘ can occur with another wh-constituent and de ce has to follow it, as in (238). 238. a. Cine de ce a plecat? who why has left ‗Who has left and why?‘ 109 b.*De ce cine a plecat? why who has left The same is true for Russian as in (230), repeated here as (239). 239. a. Kto počemu sejčas ne spit? who why now not sleep ‗Who does not sleep now and why?‘ a`.*Počemu kto sejčas ne spit? why who now not sleep b. Kogo počemu vybrala Rossija? who why chose Russia ‗Who did Russia choose and why?‘ b`. *Počemu kogo vybrala Rossija? why c. S who chose Russia kem počemu ona razošlas? with whom why she left ‗Whom did she leave and why?‘ c`.*Počemu s why kem ona razošlas? with whom she left This serves as an argument for a position which is referred to by Shlonsky & Soare (2009) as [Spec;Cause]. According to them de ce is externally merged in [Spec;Cause] And other wh-elements are raised over de ce to [Spec;Wh] and de ce in its turn remains in-situ. I will adopt their suggestion for the landing site [Spec;Cause] for Russian as well, but I am reluctant to assume that počemu stays in-situ 110 ForceP IntP TopP FocP WhP CauseP kogo (whom) FinP počemu (why) twh Interestingly, for Romanian, a topic should precede all wh-elements, including multiple questions with de ce as in (240). 240. Carte, cui de ce ai oferit-o? book who why the (you) have offered-it ‗The book, to whom did you offer and why?‘ In Russian, on the other hand, it is not possible to have Topic or Focus higher or between the elements when počemu occurs. Furthermore, nothing can intervene between the two wh-elements. 241. a. Kogo počemu vybrala Rossija? who why chose Russia ‗Who did Russia choose and why?‘ b. *Kogo Rossija počemu vybrala? who Russia why chose c. *Rossija kogo počemu vybrala? Russia who why chose d.*Kogo po tvoemu mneniju počemu vybrala Rossija? who in your opinion why chose Russia 111 I suggest that when počemu occurs with another wh-element, they together form a cluster and move further to IntP; that is why topicalized or focalized elements can appear lower than počemu. ForceP IntP TopP FocP WhP CauseP kogo FinP (whom) počemu (why) twh Another interesting observation is ungrammaticality forced by the co-occurrence of počemu with another adjunct which I explained in 4.1.2 by means of Relativized Minimality. Notice that the presence of a coordinator i - ‗and‘ makes the question grammatical, I revert to the phenomenon of wh-phrases coordination in 4.3. 242. a. *Kogda počemu oni smejalis‘? when why they laughed ‗When did they laugh and why?‘ b. *Počemu kogda oni smejalis‘? why when they laughed c. Kogda i počemu oni smejalis‘? when and why they laughed There is a lot of discussion about short moved why vs. long moved why as in example (233) repeated here as (243), where two readings are available, dependent on the base position of why. 243. Why did you say that John left? (1st reading) the speaker is questioning the reason for saying (2nd reading) the speaker is questioning the reason for John‘s leaving 112 In Russian, on the other hand the reading depends on the presence of complementizer čto (that), as has been discussed in 4.1.1. 244. a.Počemu ty skazal, čto on uvolnjaetsja? why you said (only matrix) that he resigns ‗Why did you say that he was resigning?‘ b.Počemu ty skazal, on uvolnjaetsja? why (only embedded) you said he resigns ‗Why did you say he was resigning? To sum up, I have argued following Shonsky and Soare‘s (2009) analysis, that in Russian počemu is externally –merged in the functional projection CauseP, which is located higher than FinP, but lower than IntP. Počemu moves to [Spec;Int], a criterial wh position, which is different from WhP. If and when some other wh-elements move to the left periphery, they form a cluster together with počemu and move together to [Spec;Int] for checking features. Začem on the other hand is base generated in VP, like other wh-elements, and moves to the left periphery to satisfy wh-features. 4.2. Aggressively non-D-linked and nominal wh-elements As noted in Franks and Rojina (2008), semantically, kakogo čërta, kakogo figa, čego, etc. are not unlike ordinary wh-questions: 245. a. Počemu ty prišël? why you came ‗Why did you come?‘ b. Začem ty prišël? what for you came ‗What did you come for?‘ 113 c. Čto ty prišël? what you came ‗Why did you come? d. Čego/Čë ty prišël?27 what you came ‗Why did you come? e. Kakogo figa/čërta ty prišël? which figs/devil you came ‗Why the hell did you come?‘ Examples (245) present information questions which imply that somebody came and the speaker wants to know the reason/purpose for coming. Questions in (245a-b) are neutral, i.e. emotionally indifferent, whereas examples in (245c-e) have aggressive, even negative connotation, implying that the speaker did not expect the person to come and is annoyed that he did. It thus connotes a negative attitude on the part of the speaker. This negative attitude is part of kakogo čërta‘s more general status as a Polarity Item (PI), as described by den Dikken and Giannakidou (2002) for English wh- the hell/on earth/the devil/the fuck phrases. I will use ―the hell‖ to stand in for all of such collocations, which can in general be called ―aggressively non-D(iscourse)-linked‖ expressions. They are completely promiscuous in English, since, as shown in (246), they can be attached to any wh-element: 246. a. What the hell did you buy? b. When the hell is he coming? c. Who the hell cares what you think? d. How the hell should I know? e. Where the hell did I put my keys? In English, by adding ‗the hell‘ to different wh-elements we get different readings, for example: time as in ‗when the hell‘ (246b), subject as in ‗who the hell‘ (246c), place as 27 čë is considered to be a phonologically reduced form of čego. 114 in ‗where the hell‘ (246e), etc. In Russian, on the other hand, we can add ‗the hell‘ part only to kakogo – ‗which‘, and kakogo čërta has only the ‗why‘ function. Obenauer and Poletto (2000) refer to the mentioned above types of questions as rhetorical questions. As per Quirk‘s et alii (1985) definition, a rhetorical question is interrogative in structure, but has the force of a strong assertion and it generally does not expect an answer. Obenauer and Poletto (2000:122) argue that German questions like (247) can express surprise without having the force of negative assertion, i.e. it does not necessarily deny that there might be a reason (acceptable for the speaker) for laughing. 247. Was lachst du denn so? what laugh you „denn‟ thus ‗Why are you laughing like this?‘ According to Obenauer and Poletto (2000), the question in (247) does not require an answer and fits the definition of rhetorical questions. They use term ‗rhetorical questions‘ when discussing this type of questions. 4.2.1. The case form of „kakogo čërta‟ Looking at kakogo čërta phrase type, it is important to raise a question of the case form of noun čërta. At first sight it seems to be either accusative or genitive cases. Possibility of kakogo čërta being accusative is driven by the fact that čërt ‗devil‘ is animate, as in (248): 248. Ivan videl čërta na kaţdom šagu. Ivan saw devil on each step ‗Ivan saw the devil everywhere.‘ On this reasoning, (249a) would be a variant of (249b), in which the direct object of delaeš‟ is animate kakogo čërta instead of inanimate čto. 115 249. a. Kakogo čërta ty tut delaeš‘? which devil you here doing ‗What the hell are you doing here?‘ = ‗Why the hell are you here?‘ b. Čto ty tut delaeš‘? what you here doing ‗What are you doing here?‘ = ‗Why are you here?‘ Let‘s call this the ―direct object accusative analysis‖. However, the fact that the direct object can have a ―why‖ meaning in (249b) seems to us to be a special case, since even in the English translation of (249b) what can have the meaning of why. That is, (249b) can be a literal question, i.e., ‗What is the thing that you are doing here?‘ or it can mean ‗Why are you here?‘. In English, this is probably because the verb do—and its Russian counterpart delat‟- are semantically bleached. One cannot after all use expression in (250) to mean ‗Why are you here?‘: 250. What are you fixing/buying/making here? The reason one might want to treat kakogo čërta and čto in (249) as direct objects is because delat‟ indeed requires an accusative object. When that object is missing, the sentence can only be elliptical or defective. Now recall that genitive čego, as in (245d) repeated here as (251a), can have a similar ‗why‘ function. Interestingly, (251a) with genitive čego is degraded, just like (251b) with true lexical ‗why‘ počemu: 251. a. ?Čego ty tut delaeš‘? what you here doing ‗Why the hell are you doing here?‘ b. *Počemu ty why tut delaeš‘? you here doing ‗Why the hell are you doing here?‘ In (251a) čego is genitive and therefore does not satisfy case properties required by the verb ‗do‘- in (251b) the verb delaeš‟ requires an object in accusative case. 116 A problem with the ‗direct object accusative analysis‘ arises in examples like (252), where kakogo čërta and čto are used to mean ‗why‘, even when there is something we readily identify as the direct object: 252. a. Kakogo čërta ty tut čitaeš‘ gazetu? which devil you here reading newspaper ‗Why the hell are you reading the paper here?‘ b. Čto ty tut čitaeš‘ gazetu? what you here reading newspaper ‗Why are you reading the paper here?‘ We conclude in Franks and Rojina (2008) that, although kakogo čërta in (249a) may indeed be the object of delaeš‟, hence technically accusative, this approach will not generalize to all instances. The ‗direct object accusative analysis‘ is therefore irrelevant and we put it aside. A second irrelevant possibility that treats kakogo čërta as the animate accusative parallel to čto might be constructed based on examples like (253): 253. Čto ty dumaeš‘, počemu ona plačet? what you think why she cries ‗Why do you think she is crying?‘ Notice however that this čto simply serves to indicate that the embedded interrogative počemu must be interpreted with matrix scope. Čto here is a wh-scope marker; it does not mean ‗why‘ per se.28 Hence (253) is comparable to (254): 254. ?Čto ty dumaeš‘, kuda ona idet? what you think, where she is going ‗Where do think she is going?‘ 28 This phenomenon is known in literature as partial wh-movement or wh-scope marking which is thoroughly discussed by Grewendorf, Fanselow, Stepanov etc. I will devote Chapter 6 to discussion of partial wh-movement in Russian. 117 In Franks and Rojina (2008), we call this analysis the ‗wh-scope marking accusative‘ and similarly put it aside. There are, nonetheless, true instances of čto where it just seems to mean ‗why‘. (245c) was one, which is repeated here as (255a): 255. a. Čto ty prišël? what you came ‗Why did you come? b. Čto ty plačeš‘? what you crying ‗Why are you crying? As pointed out in Franks and Rojina (2008), this is neither a conceivable direct object nor a scope marker. Accusative čto also appears meaning ‗why‘ with verbs that take oblique objects as in (256): 256. a. Čto ty nam nikogda ne zvoniš‘? what you us never not call ‗Why don‘t you ever call us? b.Čto ty tak boiš‘sja sobak? what you so afraid of dogs ‗Why are you so afraid of dogs? It can be referred to as an ‗intrinsic accusative čto.‘ The fact that accusative čto occurs in contexts in which there already is an object (252), there can be no object (255), or any credible object is not accusative (256), does not necessarily mean it is not accusative. After all, non-argument time and distance phrases are accusative in exactly the same contexts, such as in (257): 257. a. Ona plakala vsju nedelju. she cried whole week ‗She cried all week.‘ b.On vsju ţizn‘ bojalsja he all life sobak. was afraid of dogs 118 ‗He was afraid of dogs all his life.‘ Fowler and Yadroff (1993) however argue that these are quasi-arguments, bearing thetaroles but being non-referential. As suggested in Franks and Rojina (2008), čto in the meaning ‗why‘ does not have such a thematic status. Fowler and Yadroff (1993) consider two accounts of the source of accusative that would divorce case from verbal government: (i) it is intrinsically accusative or (ii) there is a hidden P assigning accusative. Dong (1995) in his reply to Fowler and Yadroff (1993), argues that the latter approach cannot be correct. Franks and Dziwirek (1993) counter with one traditional reason for considering accusative time and distance phrases as quasi-arguments essentially marked as accusative, as if they were direct objects. This is the claim that genitive time and distance phrases under negation, as in (258), are genitive-marked direct objects under negation. 258. Ivan ne spal ni odnoj minuty. Ivan not slept no one minute ‗Ivan did not sleep even (for) one minute.‘ They argue that this is best understood as partitive, not genitive. It is at least conceivable that čto and kakogo čërta are marked accusative in some idiosyncratic way. The other possibility, which is far more plausible, is that kakogo čërta is genitive. The main reason why this seems an inevitable conclusion is that genitive čego exists in the same ‗why‘ function, as in (245d), repeated in (259a): 259. a. Čego/Čë ty prišël? what Gen you came ‗Why did you come?‘ b. Čego/Čë on plačet? what Gen you crying ‗Why is he crying?‘ The examples in (259), with genitive čego, imply that there must be some way of marking wh-phrase as genitive in the meaning ‗why‘. 119 Moreover, if we take into account inanimate constructions similar to kakogo čërta, namely kakogo figa/xrena etc, the picture becomes clearer – it is Genitive case. 260. a. Kakogo xrena (Gen) ty pripersja? which horse-radish you came ‗Why the hell did you come?‘ b.*Kakoj xren (Acc) ty pripersja. which horse-radish you came We came to conclusion in Franks and Rojina (2008) that since xren is necessarily inanimate and the potential accusative variant xren in (260b) is impossible, it must be genitive in the meaning ‗why‘. Therefore kakogo čërta is really genitive too. As for the accusative marking of kakogo čërta, this mechanism could presumably be either the intrinsic one or the hidden P one; I take no stand on how to deal with obtaining the genitive formally. As it was claimed in Franks and Rojina (2008), it is simply that kakogo čërta is genitive for the same reason čego is, when it has the meaning ‗why‘. Since Russian uses both čto and čego to mean ‗why‘, Franks and Rojina (2008) propose to unify these: čto can mean ‗why‘ and it can be marked accusative or genitive. As for the account of kakogo čërta, based on inanimate constructions of the same kind, I would claim that kakogo čërta is genitive. 4.2.2. “What” is “why”: differences between čego and čto. It is time to look at differences between čto and čego when they have the ‗why‘ meaning. Franks and Rojina (2008) point out that interpretively it is a matter of style and pragmatics where (261b) has more of a meaning of surprise and/or annoyance than does (261a): 261. a. Čto ty plačeš‘? what you crying 120 ‗Why are you crying? b. Čego ty plačeš‘? what you crying ‗Why are you crying Another difference is attributed to parsing, but may be more fundamental. For example, the complementizer function of čto trumps the interrogative function in (262a), though of course when čego is used, as in (262b), only ‗why‘ makes sense. 262. a. Ja znaju, čto ty prišla. I know what you came ‗I know that you came.‘ (*‗I know why you came.‘) b. Ja znaju, čego ty prišla. I know what you came ‗I know why you came.‘ (*‗I know that you came.‘) Presumably, to the extent Russian requires there to be a complementizer in an embedded statement and znat‟ takes embedded statements, (262a) must be as indicated. Notice, however, that the possibility of having ‗why‘ meaning in (262a) improves somewhat if we use negative verb ne znat‟ – ‗not know‘, which can also take an embedded question (263a): 263. a. ?Ja ne znaju, čto ty prišla. I not know what you came ‗I don‘t know that you came.‘ (?‗I don‘t know why you came.‘) b. Ja ne znaju, čego ty prišla. I not know what you came ‗I don‘t know why you came.‘ (*‗I don‘t know that you came.‘) 121 In the right context and with the right intonation on čto, (263a) can be interpreted as ‗why‘. (264a) is even better, since sprosit‟ ‗to ask‘ only takes questions: 264. a.?Ja sprosil ego, čto on prišël. I asked him what he came ‗I asked him why he came.‘ b. Ja sprosil ego, čego on prišël. I asked him what he came ‗I asked him why he came.‘ Why should (264a) be degraded at all? Franks and Rojina (2008) assume that a reason for this is that čto can be construed as a complementizer but more importantly also because sprosit‟ can take an accusative direct object. That is, čto can cause garden-path problems, i.e. be interpreted as a direct object. The same is shown in (265), where the initial accusative čto in (265a) cannot have the ‗why‘ interpretation but čego and počemu in (265b, c), respectively, are fine. The reason is the garden path problem, since rešit‟ takes an accusative direct object. 265. a. *Čto ty rešila, čto on boitsja sobak? what you decided that he afraid dogs ‗Why did you decide that he is afraid of dogs?‘ b. Čego ty rešila, čto on boitsja sobak? what you decided that he afraid dogs ‗Why did you decide that he is afraid of dogs?‘ c. Počemu ty rešila, why čto on boitsja sobak? you decided that he afraid dogs ‗Why did you decide that he is afraid of dogs?‘ In sum, ‗what‘ can mean ‗why‘ but it can be either genitive or accusative: čego and čto are stylistic and pragmatic variants. Kakogo čërta is however morphologically invariant. 122 4.2.3 Why does „kakogo čërta‟ mean only „why‟? Recall from English (246), repeated here as (266), that the hell expressions are completely promiscuous, attaching to any wh-word. 266. a. What the hell did you buy? b. When the hell is he coming? c. Who the hell cares what you think? d. How the hell should I know? e. Where the hell did I put my keys? So why, one might ask, does Russian only have the one expression kakogo čërta in the meaning of ‗why‘? The answer is that this is a lexical property, and not paradigmatic. English the hell attaches to various wh-questions, but Russian kakogo čërta is a fixed phrase. Note that some South Slavic languages have a paradigmatic way of making a wh-phrase aggressively non-D-linked, namely by adding li, as in Bulgarian (267); see Boškovic (2001) or Franks (2006) for discussion: 267. Kakvo li šte what mi podariš?! li what me will give ‗Whatever are you going to give me?!‘ Bulgarian, Macedonian li more generally has this meaning as part of its focus marker function. Franks and Rojina (2008) note, however, that Russian particle že can play a comparable role, as in (268): 268. Čto ţe ty mne podariš‘? what že you to-me will give ‗Whatever are you going to give me?!‘ It can even be used to intensify kakogo čërta, as in (269): 123 269. Kakogo ţe čërta ty prišël? which prt devil you came ‗Why in damnation did you come?!‘ Putting že aside29 and returning to our question of why kakogo čërta is special, we reduce this question to that of why čego is special. That is, for some reason čego can mean ‗why‘ and kakogo čërta (similarly kakogo xrena, kakogo figa, kakogo lešego etc) is just an ‗epithet‘ which functions as a more colourful version of čego. Thus it is like the idiot in (270), which in its anaphoric epithet function is a more colourful version of he (with additional R-expression-like properties): 270. a. George W. Bush‘s father sent the idiot to Yale (because he couldn‘t get into OSU). b. Donal Rumsfeld‘s father was surprised that Princeton accepted the creep. Interestingly, words that can serve as anaphoric epithets have negative connotations, just like kakogo čërta. In short, kakogo čërta is to čego as the idiot is to him: in both, a fixed noun phrase substitutes for a pronoun. 4.2.4. Syntactic properties of aggressively non-D-linked wh-phrases The meanings of kakogo čërta and čego are however similar, since both are aggressively non-D-linked. In Pesetsky‘s sense (1987), aggressively non-D-linked whelements are phrases that, because of particular lexical choices, are incompatible with a D-linked interpretation. Therefore, expressions such as what on earth, why the hell, etc. are used in order to express surprise and to imply that the speaker and hearer share no preconceptions about the range of possible answers. In English, then, the aggressively 29 I will return to the že issue in section 4.2.5 124 D-linked phrase which conflicts with aggressively non-D-linked the hell, as shown in (271)30, from Pesetsky (1987): 271. a. What the hell book did you read that in? b. *Which the hell book did you read that in? It has been noticed in the literature (Obenauer and Poletto (2000), Obenauer (2004), Rojina (2007)) that aggressively non-D-linked elements exhibit a variety of syntactic peculiarities which differentiate them from other wh-elements. One of the features that can be pointed out is their impossibility to remain in situ but rather the requirement to move to SpecCP, as shown by (272): 272. a. *Ty prišël kakogo čërta? you came which devil ‗*You came why the hell?‘ b. *Ty plačeš‘ čego? you crying what ‗*You are crying why?‘ For the same reason, they never occur as the in situ phrase in multiple wh-questions, as in English (273): 273. a. *How did you fix what the hell? b. *Who bought what the hell? Ordinary whys začem and počemu, on the other hand, can occur in situ (albeit with an echo reading) as in (274). 274. a. Ty prišël začem? you came what for 30 Note that although in Russian, čërta - ‘devil‘ technically combines with kakogo ‗which‘, this is a fixed expression and the meaning yielded is not the D-linked ‗which‘, but epithetical ‗why‘ (see section 4.2.3 above) 125 ‗You came why?‘ b. Ty plačeš‘ počemu? you are crying why ‗You are crying why?‘ Furthermore, kakogo čërta differs from začem and počemu in its ability to occur with veridical predicates such as znat‟ - ‗know‘, while kakogo čërta cannot, the same is true for French (275c). Compare (275a) with (275b): 275. a. Ja znaju počemu pticy letjat zimoj na jug. I know why birds fly winter to south ‗I know why birds fly south in the wintertime.‘ b.*Ja znaju kakogo čërta on sjuda prišël. I know what devil he here came ‗*I know why the hell he came here.‘ c. *Je sais pour quoi diable il est venu. I know for which devil he came However, negating the matrix clause in (275b) renders kakogo čërta acceptable in Russian, English and French, as shown in (276). 276. Ja ne znaju kakogo čërta on sjuda prišël. I not know what devil he here came ‗I don‘t know why the hell he came here.‘ ‗Je ne sais pour quoi diable il est venu.‘ I not know for which devil he came Negation scoping over kakogo čërta appears to have an ameliorating effect. Following den Dikken and Giannakidou (2002), these data support the idea of treating kakogo čërta as a Polarity Item (PI). Veridical verbs do not license PI in their scope, but insertion of negation ne makes the sentence grammatical by providing a non-veridical ccommanding element. These data also support Obenauer and Poletto‘s (2000) suggestion to treat this kind of questions as rhetorical questions, thus distinguishing them from true wh-questions. 126 Moreover, kakogo čërta can be found in other PI environments: it appears in the complement of matrix interrogative verbs like sprosit‟ - ‗ask‘ as in (277a) as compared with (277b); after verbs with ‗negative‘ force like otkazat‟sja - ‗refuse‘ as in (278a) as compared to (278b); and in the scope of focus marking tol‟ko - ‗only‘ as in (279). 277. a. Ja ego sprosil, kakogo čërta on zadal ètot vopros. I him asked which devil he asked this question ‗I asked him why the hell he asked this question.‘ b. On zadal kakie-nibud‘ voprosy? he asked any kind questions ‗Did he ask any questions?‘ 278. a. On otkazalsja mne otvetit‘ he refused kakogo čërta on pripersja. me to answer which devil he came ‗He refused to answer me why the hell he came.‘ b. On otkazalsja mne čto-libo he refused skazat‘. me what ever tell ‗He refused to tell me anything‘ 279. Tol‘ko ja znaju kakogo čërta mne zdes‘ nado! only I know which devil me here need ‗Only I know why the hell I need to be here.‘ Obenauer and Poletto (2000) argue that wh-elements in rhetorical questions (in their terms) move higher than they do in true wh-questions. They draw this conclusion based on Italian data. The first argument comes from the fact that in Italian a subject can only occur at the left edge of the whole interrogative structure (280a) or at its right edge (280b), leading to a structure for true wh-questions in (281) (Obenauer and Poletto, 2000). 280. a. Gianni, cosa ha fatto? John what has done ‗What has John done?‘ b. Cosa ha fatto, Gianni? what has done John 127 281. [SubjP DP [FocP true WH...]] Obenauer and Poletto (2000:127) come across an example repeated here in (282) where the subject is located between the auxiliary avrebbe and the past participle potuto, which is a case of Germanic inversion. 282. a. Cosa mai avrebbe Gianni potuto fare, in quell frangente? what ever had+conditional John b. *Cosa avrebbe could do in that situation Gianni fatto se sua madre gli avesse dato l‘eredità? what had+conditional John done if his mother had given him the heritage In (282a), both the wh-element and the inflected verb occupy position higher than the subject. According to Obenauer and Poletto (2000), Germanic inversion of the DP is allowed only in one type of interrogative questions, namely rhetorical questions, while it is ungrammatical in all other interrogative structures (282b). This leads them to assume that a wh-element can occur in more than one position and each position corresponds to a distinct interpretation of the question. 283. [XP rhetorical WH [SubjP DP [FocP true WH...]]]31 The second argument in favour of the idea that wh-elements move higher in rhetorical questions than in true wh-questions is provided by the position of Left Dislocated elements32 (Obenauer and Poletto, 2000:129). They notice that in true-wh-questions the only position for a Left Dislocated element is at the left of the wh-element as in (284a), leading to a structure in (285), where the Topic position containing a Left Dislocated item is higher than the position of a moved wh-element. 284. a. A Gianni, cosa gli to John 31 hai dato? what to-him have you-given Obenauer and Poletto (2000) follow Rizzi (1997) suggesting that wh-items occur in [Spec;FocP] position in true questions. 32 Obenauer and Poletto (2000) distinguish Left Dislocation from Hanging Topic based on the fact that Hanging Topics can be only DPs and not PPs. In order to disambiguate the two constructions they use an indirect object with a preposition. 128 b.*Cosa, a Gianni, gli hai dato? what to John to-him have you-given 285. [TopP LD [FocP true WH...]] The situation is however different with respect to rhetorical questions (Obenauer and Poletto, 2000:129): 286. ?Cosa mai, a Gianni, avresti potuto dirgli che lo tirasse su in un momento simile what ever to John had+conditional could tell-him that cheered him up in such a moment. The example in (286) clearly contrasts with example in (284b). Based on these data, Obenauer and Poletto (2000:131) assume that the position of a wh-element in rhetorical questions in higher than in true questions, thus they propose a structure of the left periphery of this type of questions illustrated in (287): 287. [XP rhetorical WH [TopP LD [FocP true WH...]]] Russian data, on the other hand, behave differently. Consider examples in (288) and (289): 288. a. *K Maše kto prixodil? to Maša who came ‗Who visited Masha?‘ b. Kto k Maše prixodil? who to Maša came 289. a. *K Maše kakogo čërta on prixodil? to Maša what devil he came? b. ?Kakogo čërta k Maše on prixodil? what devil to Maša he came? 129 In wh-questions as in (288) a Left Dislocated element k Maše must be located to the right of a wh-element, the same is true for the questions with aggressively non-D-linked wh-elements (289). Although (289b) is marginal, there is still a clear contrast with (289a). Furthermore, a subject in Russian wh-questions (which occupies a Topic position in the left periphery) must be located lower than a wh-element in both types of questions. 290. a.??/*Oleg komu zvonil? Oleg whom called ‗Whom did Oleg call? b. Komu Oleg zvonil? whom Oleg called 291. a. *Oleg kakogo čërta zvonil? Oleg what devil called ‗Why the hell did Oleg call? b. Kakogo čërta Oleg zvonil? what devil Oleg called This suggests that in Russian both ‗true‘ and aggressively non-D-linked wh-elements occupy a position higher than Topic, therefore, we assume the following structure: 292. [WhP ‗true‘ Wh/rhetorical Wh [TopP DP]] Thus, Obenauer and Poletto‘s (2000) claim that the position of a wh-element in rhetorical wh-questions is higher than in true wh-questions is not supported by the Russian data. Furthermore, the data presented in next section will reveal that aggressively non-D-linked wh-elements will follow ‗true‘ wh-elements in coordinated multiple wh-questions. 130 4.2.5. Kakogo čërta, čego, čë and multiple wh-questions I now turn to a major domain of this chapter, that of how kakogo čërta, čego, čë behave in multiple wh-questions. The special restrictions on aggressively non-D-linked expressions in multiple wh-questions are so significant that they essentially serve as a diagnostic for such status. First of all, kakogo čërta as well as čego (meaning ‗why‘) cannot be used in a multiple wh-question. Consider Russian (293)33: 293. a.*Kto kakogo čërta pripersja?! who what devil came ‗*Who came why the hell?‘ b.?*Kto kakogo čërta mnoj interesovalsja? Who what devil me was interested ‗*Who was asking about me why the hell?‘ c.*Kto čego boitsja sobak? who what fears dogs ‗*Who is afraid of dogs why the hell?‘ d.*Kto kakogo čërta ţalovalsja? who which devil complained ‗*Who complained why the hell? Notice that in the English translations, why the hell is similarly bad. Interestingly, this can be saved by using coordination, as in English (294)34: 294. a. Who came and why the hell did he?‘ b. Who was asking about me and why the hell was he?‘ 33 There is a parenthetical exclamatory function of kakogo čerta which in some contexts can emerge in (293b) 34 There is a growing recent literature on the phenomenon of coordinated multiple wh-questions in Russian and other languages, including Kazenin (2002), Merchant (2007), Citko (2008), Gribanova (2009), I will discuss this issue in section 4.3. 131 c. Who is afraid of dogs and why the hell is he?‘ Note that, while VP-ellipsis is good in (294), IP-ellipsis (Sluicing) is not, consider (295): 295. a. Who came and why (*the hell)?‘ b. Who was asking about me and why (*the hell)?‘ c. Who is afraid of dogs and why (*the hell)?‘ Some similar effects arise in Russian. Consider (296), the coordinated wh-question version of (293) where addition of a coordination saves the question. Note that in (296) we get only single pair reading, i.e. there is a particular time when the speaker got irritated by somebody coming to visit him. 296. Kto i kakogo čërta pripersja?! who and what devil came ‗Who came and why the hell did he?‘ (*PL/SP) The same is true for čego and čë (297) where presence of the coordination makes it grammatically acceptable. Notice that čto in a meaning of ‗why‘ cannot be used in multiple wh-questions as shown by the ungrammaticality of (298). 297. a. Kto *( i) who and čego prišël? what came ‗Who came and why the hell did he?‘ b. Kogda *(i) čë on zvonil? when and what he called ‗Who called and why the hell did he?‘ 298. *Kto i čto prišël? who and WHAT came ‗Who came and why? It is however unclear if this restriction derives from the aggressively non-D-linked status of kakogo čërta, since multiple wh-fronting of ‗why‘ in Russian is generally 132 severely degraded without coordination, see (299). That is, počemu and začem, like kakogo čërta require coordination in between them as in (300), the same is true for English and Bulgarian (301) (p.c. Iliana Krapova). 299. Kto ??(i) počemu ţalovalsja? who (and) why complained ‗Who complained *(and) why?‘ 300. Čto *(i) začem on kupil? what (and) what for he bought ‗What did he buy *(and) why?‘ 301. Koga i zašto e idval? when and why came ‘When did he come and why?‘ Notice in (302) a Superiority Effect takes place when we try to put any why-phrase in an initial position, even with coordination. Interestingly, the same data were revealed in Bulgarian, consider examples in (303) (p.c. Iliana Krapova) 302. a.*Začem i čto on kupil? what for and what he bought ‗*Why and what did he buy?‘ b.*Kakogo čërta i which čto on kupil? devil and what he bought ‗*Why the hell and what did he buy?‘ c.*Počemu i why kto ţalovalsja? and who complained ‗*Why and who complained?‘ 303. a.*Zašto i koj e idval? why and who came ‗Who came and why?‘ b. Koj i zašto e idval? who and why c. Zašto i came a kogo e dal tova? why and to whom gave that 133 It is however worth remarking that in the acceptable corresponding English sentences, the ‗and wh-phrase‘ is always after the wh-question, e.g.: ‗Who complained and why?‘ rather than ‗*Who and why complained?‘ which sounds odd. This is because English does not have multiple wh-fronting in general. Russian of course has the ‗stranded‘ wh construction as well as English, and here kakogo čërta behaves just like other adjuncts, as in (304): 304. a. Kto pripersja i kakogo čërta?! who came and which devil ‗Who came and why the hell did he? b. Kto pripersja i who came začem? and what for ‗Who came and why? Kazenin (2002) however convincing argues that this is an elliptical construction and should be differentiated from the truly coordinated multiple wh-construction which, as it is generally argued (cf. e.g. Gribanova, 2009), is a possibility only in multiple whfronting languages. In short, aggressively non-D-linked expressions such as kakogo čërta are only acceptable in coordinated multiple wh–questions. However, even in this context there are important limitations: as indicated in (296) the pair-list reading is suppressed in favor of a single-pair interpretation. That is, unlike ordinary wh-elements, kakogo čërta does not exhibit a pair-list reading. This, according to den Dikken and Giannakidou (2002), follows from the PI status of aggressively non-D-linked expressions such as the hell or kakogo čërta. The reason then that the hell cannot be included in (295) is because this would force a single-pair reading onto the multiple wh–question, but English does not tolerate single-pair readings in general. This implies that Sluicing has the interpretation of a multiple wh–question but VP-ellipsis would just amount to two independent questions connected through the discourse: Who came? and Why the hell did he come? That is how the single pair reading is expressed in English. Returning to Russian, the coordinated questions are to be compared with ordinary multiple wh-fronting questions (which of course English lacks). 134 In Russian, multiple wh-questions generally only allow the pair-list reading. This judgment is shown in (305): 305. Kto čto prines? who what brought (PL/*SP) ‗Who brought what?‘ (305) can only be answered by a list of pairs of ‗bringers‘ and ‗things brought‘; it is not an appropriate way of asking for a single pair, in a context where the interlocutor knows that some person brought some thing, but not who it was that did the bringing and what it was that that person brought. Let us look now at the particle že in the light of Superiority effect which was mentioned in (268) and (269), repeated here as (306): 306. a. Čto ţe ty mne podariš‘? what prt you me give ‗What are you going to give me?! b. Kakogo ţe čërta ty prišël? which prt devil you came ‗Why in damnation did you come?!‘ Interestingly že can follow only an initial wh-phrase, which suggests that this particle occupies a Focus position between wh-phrases and has a meaning as part of focus marker function. 307. a. *Kto čto ţe kupil? who what prt bought ‗*Who bought what on earth? b. Kto ţe čto kupil? who prt what bought ‗Who on earth bought what?‘ 135 ForceP WhP Wh` FocP kto Foc` (who) WhP ţe Wh TP čto (what) kupil (bought) But if Superiority is violated že can never be used, as shown in (308): 308. *Čto ţe kto kupil? what part who bought *‗What on earth did who buy?‘ Now, how does kakogo čërta behave with respect to long distance dependencies? As it has been pointed out in Rojina (2007) kakogo čërta manifests a common set relating to English how come, and an identical set to English why the hell: it is impossible to get a reading in which kakogo čërta modifies embedded clauses. 309. a. Kakogo čërta ty skazal, čto Peter ne pridet? which devil you said (only matrix) that Peter not come-FUT ‗Why the hell did you say that Peter would not come?‘ (only matrix) (possible answer: because I did not want to upset you) b.*Kakogo čërta ty skazal, __ Peter ne pridet? which devil you said Peter not come-FUT The examples in (309) show that the expression kakogo čërta can modify only the matrix clause as in (309a); moreover it requires the use of čto in the embedded clause (309b). This suggests that kakogo čërta is based generated in the matrix clause. This suggests that the presence of either a wh-element trace or a complementizer in a CP 136 domain is crucial in Russian embedded clauses, since kakogo čërta cannot be base generated in an embedded clause and thus cannot leave a trace in [Spec, CP]. Čto, čë, čego on the other hand cannot modify embedded clauses when complementizer-čto occupies C in an embedded clause (310a); the interpretation is strictly clause-bound. Although they can easily manifest it when there is no čto (310b): 310. a. Čë/čego ty dumaeš‘ čto on ego boitsja? WHAT you think that he him fears ‗Why do you think that he is afraid of him?‘ b. Čë/čego ty dumaeš‘ on ego boitsja? WHAT you think he him fears ‗Why do you think he is afraid of him?‘ 4.3. Coordination of wh-elements This section deals with a coordination phenomenon of wh-elements in multiple whquestions in Russian. Although, as it was shown above, Russian multiple wh-questions are well studied and much literature is devoted to the subject, coordination of these elements surprisingly has been studied only by a few linguists, among them Kazenin (2002), Gribanova (2009), Haida and Repp (to appear). Previous sections discussed examples where the use of coordination i - ‗and‘ is crucial for the grammaticality of multiple wh-questions as in (311). 311. a. Kto *(i) kakogo čërta pripersja?! who and what devil came ‗Who came and why the hell did he?‘ b. Kto *(i) čego prišël? who and what came ‗Who came and why?‘ c. Kogda *(i) when čë on zvonil? and what he called 137 ‗When did he call and why?‘ d. Kto *(i) počemu ţalovalsja? who (and) why complained ‗Who complained and why?‘ Furthemore, we come across example like (312) where, a la Gribanova (2009), substantial differences, interpretive and structural, are revealed between the two types. We see below that (312a) yields a single-pair interpretation, while in (312b) only pairlist reading is available. 312. a. Kto i kakoj gorod zaxvatil? who and which city conquered ‗Who conquered which city?‘ The Germans conquered Paris. (SP) ?? The Germans conquered Paris, the Russians conquered Berlin,.... (PL) b. Kto kakoj gorod zaxvatil? who which city conquered ‗Who conquered which city?‘ ??The Germans conquered Paris. (SP) The Germans conquered Paris, the Russians conquered Berlin,.... (PL) Before addressing theoretical issues I would like to point out some properties of coordinated wh-elements in Russian. First, wh-phrases in Russian multiple questions can appear adjacent to a coordinator either preverbally or postverbally, as noted in Gribanova (2009). 313. a. Čto i gde oni našli? what and where they found ‗What did they find and where?‘ b. Čto oni našli i gde? what they found and where ‗What did they find and where?‘ 138 Secondly, when a question consists of more than two wh-elements, the conjunction ‗i‘ must occupy the position preceding the last wh-element. 314. Čto (*i) gde i kogda slučilos‘? what (and) where and when happened? ‗What happened when and where?‘ Thirdly, any sort of wh-items can be coordinated, as pointed out in Gribanova (2009); (315a) shows a coordination of two arguments, (291b) presents a coordination of two adjuncts, (315c) shows a coordination of an argument and an adjunct. 315. a. Kto i kakoj gorod zaxvatil? who and which city conquered ‗Who conquered which city?‘ b. Gde i kogda ego našli? where and when him found ‗Where was he found and when?‘ c. Kto i gde budet spat‘? who and where will sleep ‗Who will sleep and where?‘ The last property that should be mentioned is that ordering restriction appears between some coordinated wh-elements (contra Gribanova 2009, Kazenin 2002). 316. a. Kto i čego prišël? who and why came ‗Who came and why?‘ b.*Čego i kto prišël? why and who came ‗Who came and why?‘ The question that arises is how to treat these types of constructions. Several analyses for coordinated wh-elements have been proposed, the most important of which I account for below. 139 4.3.1. Coordination prior to the wh-movement? Grimshaw (1978) argues that conjoined wh-phrases in English are coordinated prior to the wh-movement. This analysis is inapplicable to Russian, since under her analysis whcoordination is possible only for phrases of the same categories and grammatical relation as in (317a). (317b) is ungrammatical because who and what cannot be both subject and object of bought, and (317c) is ungrammatical because put subcategorizes for a locative adverbial, a category to which when does not belong (Grimshaw, 1978). 317. a. Which book and which pencil did John buy? b. *John asked who and what bought. c. *Where and when did Bill put the book? In Russian, on the other hand, coordination of wh-elements does not require the identity of categories as in (318). 318. a. Čto i gde oni našli? what and where they found ‗What did they find and where?‘ b. Kto *(i) čego prišël? who and why came ‗Who came and why?‘ c. Kto i počemu mnoj interesovalsja? who and why by me interested ‗Who was interested in me and why?‘ Some informants report coordination of [+human] wh-arguments to be marginal, even impossible for some of them, as shown in (319). 319. a. Kto (??i) s kem vstretilsja? who (and) with whom met ‗Who met whom?‘ 140 b. Kogo (??i) komu ty predstavil? who (and) to whom you introduced ‗Who did you introduced to whom?‘ Based on the data above, I conclude that wh-arguments cannot be coordinated prior to their movement to the left periphery due to the fact that they are not categorically identical. Putting Grimshaw‘s analysis aside and thus declining possibility of whcoordination prior to the movement I refer to other possible analyses. One of them was proposed by Kazenin (2002) and Camacho (2003), who offer treating this kind of coordination as instances of two CPs coordination followed by sluicing within the left CP. 4.3.2. Coordination of two CPs followed by sluicing? One of the suggestions, that looks at first mostly plausible, is to treat coordinated multiple wh-constructions as instances of coordinated sentences accompanied by sluicing. Thus, the sentence in (320) gives us the structure in (321) (taken from Kazenin 2002). 320. Kuda i začem on pošёl? where and what for he went ‗Where did he go and what for?‘ 321. [Kudai [on pošёl ti]] i [ začemj [on pošёl tj]]? where he went and what for he went 141 CP C‘ C IP e According to the sluicing analysis, the elided phrase is IP and the empty category in the position of IP is governed by C´ with a [+Q] feature. Kazenin also points out that sluicing can happen in the second conjunct as in (322). 322. Kuda on pošёl i začem? where he went and what for ‗Where did he go and why?‘ [Kudai [on pošёl ti]] i [ začemj [on pošёl tj]]? where he went and what for he went Kazenin (2002) points out in his analysis two main problems for sluicing. The first problem occurs when accounting for examples like in (323 and 324) which, according to the sluicing analysis, would be impossible due to the fact that the second of the conjoined wh-phrases is an argument (examples taken from Kazenin, 2002). 323. Kto i kakoj gorod zaxvatil? who and which city conquered ‗Who conquered which city?‘ 324. Začem i komu eto nuzhno? why and whom this need ‗Who needs this and why?‘ Following Kazenin (2002), example in (323) would yield a structure as in (325) where we are forced to postulate the presence of an empty pronominal element in the object position of the first IP, as it is required by the verb‘s thematic requirement, i.e. zaxvatil requires an object in Accusative case to follow. 142 325. a. Kto i kakoj gorod zaxvatil? who and which city conquered [CP Ktoi [IP ti zaxvatil prot]] i [CP [kakoj gorod]t [IP proi zaxvatil tt]? b.*Kto zaxvatil i kakoj gorod zaxvatil? who conquered and which city conquered Kazenin argues that this structure results in a cataphoric dependency between the whphrase kakoj gorod and the pronominal in the first IP, which is restricted in Russian, especially when the antecedent of a pronominal is a wh-phrase. The second problem with the sluicing analysis, according to Kazenin (2002), concerns the possibility of ‗canonical‘ sluicing constructions parallel to whconstructions. The examples in (326) and (327) show the possibility of such constructions in Russian. 326. Kuda on pošёl i začem? where he went and why ‗Where did he go and why?‘ 327. Kuda i začem on pošёl? where and why he went ‗Where did he go and why?‘ Analyzing wh-construction using the sluicing approach will lead us to a conclusion that Russian permits two directions of deletion in sluicing constructions; however there are restrictions on sluicing with respect to directionality (Kazenin, 2002). Sluicing with a ‗stranded‘ wh-phrase is impossible when the stranded wh-phrase is an argument (328329). When the second wh-phrase is an adjunct, sluicing is able to operate in both directions as is shown in (326-327). 328. Kto prixodil i začem? who came and what for ‗Who came and why?‘ Ktoi [ti prixodil] i začemj [proi prihodil tj] 329. Začem prixodil i why came kto? and who 143 Začemj [proi prihodil tj] i ktoi [ti prihodil]? The impossibility of stranding a wh-argument is explained under the sluicing analysis by creating a cataphoric dependency between a wh-phrase and an empty pronoun in the first IP. (329) is ungrammatical because of a cataphoric dependency, kto is the antecedent of pro, unlike in (328) where no cataphoric dependency arises. The sluicing approach fails to explain the ungrammaticality of examples with unergative verbs as in (331) compared with (330) or with a non-wh-subject (333) as compared to (332); there is no cataphoric dependency, though sentences are still ungrammatical. 330. Komu on plakalsja i to whom he cried začem? and why Komui [on plakalsja ti] i 331. *Začem on plakalsya i why he cried začemj [on plakalsja tj]? komu? and to whom Začemj [on plakalsja tj] i komui [on plakalsja ti] 332. S kem on razgovarival i with whom he spoke 333. and why *Pochemu on razgovarival i why he talked pochemu? s kem? and with whom Pochemuj [on razgovarival tj] i s kemi [on razgovarival ti]? The last evidence against the sluicing analysis, proposed by Kazenin (2002), is the fact that when two wh-phrases are coordinated without ellipsis, the wh-phrase of the first conjunct can correspond either to an empty or to an overt pronominal in the second conjunct (Kazenin, 2002). 334. [Kogoi Petja izbil] i [za čto Petja egoi/??proi izbil]? whom Peter beat and for what Peter him beat ‗Whom did Peter beat and what for did Peter beat him?‘ 335. [Ktoi videl Petju] i [kogda oni/??proi videl Petju]? who saw Peter.ACC and when he saw Peter.ACC Who saw Peter and when did he see him?‘ 144 According to Kazenin, we would expect to find an overt pronoun corresponding a whphrase in the first conjunct if it was a result of a sluicing operation. Thus, sluicing of (334) is presented in (336), which is ungrammatical. 336. *Kogoi i za čto Petja egoi izbil? whom and for what Peter him beat Kogoi [Petja izbil ti] i za čto [Petja egoi/??proi izbil tj]? Merchant (2008) mentions a few more problems with the sluicing analysis. One of the examples comes from Hungarian where, according to Lipták (2003), definiteness (object agreement on the verb) should be obligatorily, but here it is disallowed. 337. a. Érdekel (hogy) mit csinálsz és hogyan csinál-od /*-sz. interest.3S (that) what ACC do.2s INDEF and how do2s DEF/*INDEF ‗I care about what you do and how.‘ b. Érdekel (hogy) mit és hogyan csinál- *od /-sz. interest.3S (that) what ACC and how do2s *DEF/ INDEF ‗I care about what you do and how.‘ Another problem arises in Vlach, where the otherwise obligatory clitic in the second conjunct cannot appear. 338. a. Acari s who kundu (*łu) ai vijutu? and when (him) have 2s seen ‗Who did you see, and when?‘ b. Acari ai vijutu s kundu *( łu) ai vijutu? who have 2s seen and when (him) have 2s seen ‗Who did you see, and when?‘ Taking aside the sluicing option, I will refer to another possible analysis – coordination of wh-phrases in CP. 145 4.3.3. Coordination in CP? Gribanova (2009) assumes that all wh-elements are generated in their normal base position and later join to IP (she adopts Rudin‘s (1988) analysis for Russian multiple wh-elements) and the internal structure of the coordinated constituent has the structure formulated in Zoerner (1995), where coordinated elements appear in the Spec and complement of the head &. &P DP DP & wh1 wh2 Taking Zoerner‘s proposal into account, Gribanova assumes the following structure for coordinated multiple wh-constructions in IP-fronting languages. CP IP C IP &P t....t wh & wh Gribanova proposes the same structure for CP-fronting languages with the &P constituent in the specifier position of the CP. 146 CP C‘ &P IP wh & wh t....t Going back to a phenomenon observed in the beginning of section 4.3 (repeated in 339), Gribanova (2009) argues that a different structural pattern is responsible for the distribution of pair-list and single-pair readings. A coordinated wh-construction yields a single-pair interpretation, while a non-coordinated wh-construction has only a pair-list reading (Gribanova 2009:141). 339. a. Kto i kakoj gorod zaxvatil? who and which city conquered ‗Who conquered which city?‘ The Germans conquered Paris. (SP) ??The Germans conquered Paris, the Russians conquered Berlin,... (PL) b. Kto kakoj gorod zaxvatil? who which city conquered ‗Who conquered which city?‘ ??The Germans conquered Paris. (SP) The Germans conquered Paris, the Russians conquered Berlin,.... (PL) To explain the phenomenon above, Gribanova assumes that a certain type of structural configuration at LF is the precondition for a pair-list reading to result, as in (339b). She argues that the syntax of coordinated wh-constructions does not meet this precondition and therefore the pair-list reading is unavailable. The structural precondition originates from work on quantifier absorption (Higginbotham and May (1981), Barss (2000)). 147 Quantifier absorption can map two or more structurally adjacent quantifiers at LF into one binary quantifier, which is a wh-item in this discussion.35The binary quantifier binds all relevant variables simultaneously, giving rise to a bijective interpretation for a multiple question (Gribanova 2002:145). 340. Which man admires which woman? [WHx: x a man][Why: y a woman] x admires y [WH21,2 x,y: x a man & y a woman] x admires y Quantifier absorption is defined so that for every x there is a unique y, and for every y there is a unique x. There is a presupposition that a complete answer to such a question requires an exhaustive listing of pairs which results in a pair-list reading. Structural adjacency for quantifiers at LF is a key to deriving the right distribution of pair-list and single-pair interpretations. Therefore, in order to undergo quantifier absorption, quantifiers must be in a particular structural configuration at LF – they must be structurally adjacent as per the definition in (341) (Gribanova 2002:146). 341. α and β are structurally adjacent iff a. α c-commands β, and b. α c-commands no head that c-command β. (341b) requires that the c-command is immediate, i.e. no head may intervene between the two quantifiers.36 Based on the requirement above, quantifier absorption is blocked in the coordinated wh-constructions, i.e. there is an intervening head & which is ccommanded by wh1 and which itself c-commands wh2. 35 Gribanova (2009) points out that in Russian the syntactic ordering and structure of wh-elements can be assumed to be identical to their LF ordering and structure. 36 Gribanova‘s proposal differs from previous proposals in defining structural precondition. Before the precondition was formulated as a requirement of either α to immediately c-command β or that α and β ccommand each other (Gribanova 2002:146) 148 &P & The head & blocks structural adjacency DP DP & wh1 wh2 Therefore, the structural adjacency precondition for quantifier absorption is not met and if we take adjacency as a relevant precondition for deriving the pair-list reading, then coordinated wh-constructions are predicted never to result in pair-list interpretations (Gribanova 2002). This prediction is confirmed by data from Russian as well as other languages, i.e. Czech (342), Romanian (343), Hungarian (344), etc. 342. Kdo a kdy to našel? (Sp/*PL) who and when it found-3SG (Skrabalova,2006) ‗Who found it and where?‘ 343. Cine si pe cine a intalnit? who and whom (Sp/?PL) met (Soare, p.c.) ‗Who met and whom?‘ 344. Mikor és mit adott János Marinak a múzeumban? when and what gave John Mary to the museum in (Sp/*PL) (Gazdik, 2010) ‗What sis John give to Mary in the museum and when?‘ Therefore, adopting structural adjacency as a required configuration for obtaining pair-list interpretation has produced the correct result, i.e. it prevents a coordinated wh-construction from being assigned a pair-list reading. Non-coordinated wh-constructions, on the other hand, meet the requirement for structural adjacency as shown below, i.e. wh1 c-commands wh2 and wh1 c-commands no head that c-command wh2. 149 CP DP C‘ IP wh1 wh2 C t....t Gribanova (2009) correctly predicts that other wh-constructions, in which two wh-items are not structurally adjacent, do not allow pair-list interpretation either. She refers to Serbo-Croatian where an overt complementizer li appears between the two wh-elements forcing single-pair reading (example taken from Grebenyova 2004:181). 345. Ko li koga pozva na večeru? who comp whom invited to dinner ‗Who (on earth) invited whom to dinner?‘ (SP/??PL) Kazenin (2002) proposes another solution for accounting single-pair vs. pair-list interpretations. According to Kazenin a pair-list interpretation is derived when the first wh-element moves into the position designated for universal quantifiers at LF. Singlepair reading follows from the fact that moving of one of the coordinated wh-items in a quantifier position would violate the Coordinate Structure Constraint37, which is assumed to be operative at the LF. Since no wh-element can move out of the coordinate structure to LF, no pair-list interpretation is obtained. Kazenin‘s analysis fails to explain the lack of a pair-list interpretation in Serbo-Croatian examples with clitic placement. It seems that the quantifier absorption analysis proposed by Gribanova (2009) covers a broader range of cases. Although Gribanova‘s analysis looks most plausible, there are a few cases an application of her analysis would serve inadequate. First, her analysis does not explain why Superiority effect persists as in (346 349). 37 Ross (1967:98-99) introduced the Coordinate Structure Constraint which implies that in a coordinate structure, no conjunct may be moved, nor may any element contained in a conjunct be moved out of that conjunct. 150 346. a. Ki és kiről beszélt? (Hungarian) who and who.ABOUT talked (Lipták, 2003) ‗Who talked and about whom?‘ b. ???Kiről és ki beszélt? who. ABOUT and who talked ‗Who talked and about whom?‘ 347. a. Kiedy i ile Jan zjadł pączków? (Polish) when and how-many Jan ate doughnuts (Citko, 2008) ‗When and how many doughnuts did Jan eat?‘ b. *Ile i kiedy Jan zjadł pączków? how-many and when Jan ate doughnuts Intended: ‗How many doughnuts did Jan eat and when was it?‘ 348. a. Koj i koga ste si hodi v Bulgaria? who and when will REFL go in Bulgaria (Bulgarian) (Gracanin-Yuksek, 2010) ‗Who is going to Bulgaria and when?‘ b.*Koga i koj ste si hodi v Bulgaria? when and who will REFL go in Bulgaria ‗Who is going to Bulgaria and when?‘ 349. a. Kto i kakogo čërta pripersja?! who and what (Russian) devil came ‗Who came and why the hell did he?‘ b.*Kakogo čërta i what kto pripersja?! devil and who came ‗Who came and why the hell did he?‘ Secondly, coordination of more than two wh-elements, allows pair-list reading 350. Kto kogo i komu predstavil? (PL/*SP) who whom and who introduced ‗Who introduced who to whom?‘ 351. Kto kogo i gde vstretil? (PL/*SP) who whom and where met ‗Who met whom and where?‘ 151 Finally, coordinator ‗i‘ must be placed before the last wh-element when there are more than two wh-elements. 352. Čto (*i) gde i kogda slučilos? what (and) where and when happened? ‗What happened when and where?‘ 4.3.4. Spurious coordination Based on the above discussion, it seems that the conjunction ‗i‘ which is used in coordinated wh-constructions is not the ‗ordinary‘ conjunction that we see in sentences like (353). 353. Vladimir i Alena vstretilis v restorane. Vladimir and Alena met in restaurant ‗Vladimir and Alena met at a restaurant.‘ In Russian, coordination is generally possible only for phrases of identical categories which occupy identical syntactic positions, as we see in (353), where two arguments in Nominative cases are coordinated. It should be noted that Russian exhibits examples like (354), where NPs are not expected to be conjoinable due to different categories and thematic roles. 354. Vse i vsё znajut. Chaves and Paperno (2007) everyonenom and everythingacc knows ‗Everyone knows everything.‘ This phenomenon was first noted by Sannikov (1989) and later referred to as hybrid coordination by Chaves and Paperno (2007). In their discussion of hybrid coordination, Chaves and Paperno point out that the presence of the conjunction does not alter the meaning of the sentence, i.e. (355a) has basically the same meaning as (355b). 152 355. a. Nikto i nikogo ne pobedit. nobodynom and nobodyacc not will win ‗Nobody will beat anyone.‘ b. Nikto nikogo ne pobedit. nobodynom nobodyacc not will win ‗Nobody will beat anyone.‘ It can be argued that the particle ‗i‘ is just homophonous with the conjunction lexeme ‗i‘ and there is no coordination as such. Moreover, there are instances when the particle ‗i‘ is realized as a focus particle with the meaning ‗even‘ as in (356) (Chaves and Paperno, 2007:3). 356. Petya i Vanju pobedit. Petja and (=even) Vanja will win ‗Petja will beat even Vania.‘ Clearly ‗i‘ does not function in (356) as a coordination conjunction, but as a focus particle adjacent to Vanja which can be interpreted as ‗Vanja is one of the strongest participants but Petja has shown that he is able to beat even Vanja‘. Putting the focus particle ‗i‘ seen in (356) aside, Chaves and Paperno (2007) come to a conclusion that the element ‗i‘ that we see in (354) and (355) is nothing else but a coordination marker in hybrid coordination, because it does not explain the absence of a focus reading seen in (356), or the fact that co-arguments cannot appear discontinuously. Moreover, they argue that the entire sequence of the conjoined elements behaves like a syntactic block in the conjunction. Chaves and Paperno (2007) point out the same peculiarities that we came across earlier in the discussion of coordinated wh-constructions. Hybrid coordination allows for the ‗coordination of unlikes‘ phenomena (noticed first in Gazdar et al., 1985) (examples from Chaves and Paperno (2007)) 357. Vsem vezde i vsё do lampochki everyonedat everywhere and everythingnom don‟t care ‗Nobody cares about anything anywhere‘ 153 358. Ne sposoben nikto not able i ni s kem pomenyat‘sya mestami nobody and no with body change places ‗Nobody is able to change places with anyone‘ In (357) the conjuncts include adverbials, while in (358) – a prepositional phrase. Note that the same is observed for coordinated wh-elements. 359. Čto i gde oni našli? what and where they found ‗What did they found and where?‘ Chaves and Paperno (2007) note that hybrid coordination shows some restrictions that are different from non-coordinate counterparts. a) none of the conjuncts can contain modifier phrases, for example an adjective, as in (360) and b) conjuncts are required to be of the same semantic type as in (361). (examples from Chaves and Paperno (2007)) 360. a. Vse lysye vsё znayut. everyone bold everything knows b. *Vse lysye i vsё znayut. everyone bold and everything knows 361. *vse i chto-to vidyat. everybody and something see It is clear from Chaves and Paperno‘s (2007) analysis that Russian has a coordination construction with conjuncts that can have different grammatical roles. These structures are non-canonical but have the same meaning as their non-coordinated counterparts. Theses conjuncts also show a number of constraints that are not exhibited by standard coordinated phrases, such as a) none of the conjuncts can contain modifier phrases, for example an adjective and b) conjuncts are required to be of the same semantic type. These constraints provide evidence that this is a special kind of coordination. Comparing the data presented by Chaves and Paperno (2007) and data presented in this section, we come to a conclusion that coordination in both instances is the case of hybrid coordination, in Chaves and Paperno terms. 154 Now, returning to coordinated wh-constructions. Merchant (2008) analyzing coordination in Vlach multiple fronting, encounters the same properties of conjunction as discussed above and proposes to treat this conjunction as ‗spurious‘, which can be p used as a discourse marker and not meaning q[p^q]. Merchant proposes the structure in (362) where the coordinator is spurious, i.e. it does not coordinate the wh-phrases themselves, but rather occurs as a clausal left-edge conjunctive particle (related to free focus uses of these elements). 362. CP WH1 & C‘ WH2 C IP t1....t2 Merchant accounts for the same properties of coordinated questions in Vlach as we have already noticed for Russian Vlach coordinated wh-questions only allow single pair answers 363. Acari (s) kundu ari vatimatə muma-ts? who (and) when has killed (*PL/SP) mother-your ‗Who killed your mother and when?‘ Merchant notes that without the conjunction s ‗and‘ in Vlach, the question becomes infelicitous due to the one-time-only nature of the predicate. Interestingly, in Russian, some coordinated constructions exhibit the same property. Thus, example (364) becomes infelicitous without the conjunction. 364. Kto i kakogo čërta pripersja?! who and what devil came ‗Who came and why the hell did he?‘ 155 (*PL/SP) Multiple wh-fronting languages which exhibit Superiority effect show that the Superiority effect persists in spurious coordinated questions. 365. a. Ki és kiről beszélt? (Hungarian) who and who.ABOUT talked (Lipták, 2003) ‗Who talked and about whom?‘ b. ???Kiről és ki beszélt? whoABOUT and who talked ‗Who talked and about whom?‘ 366. a. Kto i kakogo čërta pripersja?! who and what (Russian) devil came ‗Who came and why the hell did he?‘ b.*Kakogo čërta i what kto pripersja?! devil and who came ‗Who came and why the hell did he?‘ The coordinated wh-elements cannot occur in ‗balanced‘ coordination, i.e. coordination where conjuncts have the same grammatical, morphological, thematical features (for morphologically rich languages) as in Russian (367), contra ‗canonical‘ coordination, which must be balanced as in (368). 367. O kom i s kem ty razgovarival? about whomPREP (and) withINSTR whom you talked ‗Who did you talk about and with whom?‘ 368. Ja ne ljublju smotret‘ filmy o I not like watch politike i o movies about politicsPREP and about warPREP ‗I do not like watching movies about politics and war.‘ The same has been noted for Vlach by Merchant (2008) 369. a. *S acari s vojne či ari vijtu? and who and what has seen ‗Who saw something and what did they see?‘ 156 b. S fičorlu s fiata anu vijutu mama-ts and the boy and the girl have seen mother-your ‗Both the boy and the girl saw your mother.‘ In order to account for the lack of pair-list interpretation, Merchant agrees with the Quantifier Absorption analysis offered by Gribanova (2009). He also attempts to offer another possibility where the conjunction itself imposes the presupposition for a singlepair answer. This suggestion is ruled out by the data in (370), where only a pair-list interpretation is available, though conjoined by a spurious &. 370. Kto kogo i gde vstretil? (PL/*SP) who whom and where met ‗Who met whom and where?‘ Based on the discussion of the examples above, I will follow Chaves and Paperno (2007) and Merchant (2008) in their proposal to treat conjunction in wh-coordination differently from a canonical conjunction; I will refer to this phenomenon as ‗spurious‘ coordination following Merchant. Following Merchant, I assume that the coordinator ‗i‘ is spurious in that it does not coordinate the wh-phrases themselves, but rather occurs as a clausal left-edge conjunctive particle (related to free focus uses of these elements). 4.3.5. The analysis Before proposing a structure that can capture spurious wh-coordination, I outline a few more peculiarities that will allow me to offer a structure in the end. First, I would like to address puzzling examples in (371 and 372), where the pair-list interpretation arises, though it should not, according to quantifier absorption analysis. 157 371. Kto kogo i gde vstretil? (PL/*SP) who whom and where met ‗Who met whom and where?‘ 372. Kto kogo i komu predstavil? (PL/*SP) who whom and who introduced ‗Who introduced who to whom?‘ This also rules out Kazenin‘s proposal which suggests that a single-pair reading follows from the fact that the moving of one of the coordinated wh-item in a quantifier position would violate the Coordinate Structure Constraint. Secondly, it is important to point out that coordinated wh-elements form a cluster, and any insertion of any kind of elements is impossible. 373. *Kto, v 1975, i kakoj gorod zaxvatil? who in 1975, and which city 374. *Kto, Maša interesovalas, conquered kogo i komu predstavil? who Masha was interesred, whom and who introduced 375. O kom Misha i s kem razgovarival, a ne Vanja? about whom Misha and with whom talked, a not Vanja Taking the various presented above into consideration, I partially adopt structure proposed by Zoerner (1995) and Gribanova (2009) and suggest the following structure for coordinated wh-construction. 376. Kto i kogo videl? who and whom saw ‗Who saw somebody and who was it?‘ 158 ForceP FocP FinP Spec IP ktoi kogoj (who) & (whom) IP ti....tj I assume that wh-elements are coordinated above the IP after they move to a [Spec;FocP] where spurious ‗i‘ realises as a discourse marker bearing some focus features. Wh elements move directly to FocP and not to whP based on the data in (Error! Reference source not found.), which suggests that coordinated wh-elements and a focused element cannot co-exist, i.e. they are competing for the same position.Quantifier Absorption takes place at LF level and the question gets a singlepair interpretation. Let‘s consider now a wh-question with three wh-elements two of which are coordinated. 377. Kto kogo i komu predstavil? (PL/*SP) who whom and whom introduced ‗Who introduced who to whom?‘ ForceP FocP ktoi WhP (who) WhP Spec FinP ti kogoj (who) & IP komuk (who) ti..tj...tk 159 I assume that in the case of three wh-elements, the highest wh-element kto moves to [ForceP] where it checks [Q] feature. The two lower ones move to a [Spec;FocP] where they form a cluster. When the cluster is formed, coordinated wh-elements stay in [FocP] since kto has already checked the [Q] feature. Kto + cluster [kogo&komu] form a union which satisfies requirements of the quantifier absorption, i.e. kto + cluster are structurally adjacent and the c-command is immediate, i.e. no head may intervene between the two quantifiers a la Gribanova (2002:146). 378. α and β are structurally adjacent iff a. α c-commands β, and b. α c-commands no head that c-command β. Kto + cluster is a non-coordinated wh-construction, which meets the requirement for structural adjacency, i.e. wh1 c-commands wh2 and wh1 c-commands no head that ccommand wh2 Therefore, the construction in (377) yields a pair-list interpretation. To summarize, coordinator ‗i‟ found between wh-phrases in Russian is a spurious coordinator, in Merchant‘s terms, which is different from a canonical coordinator. It can be characterized as a discourse marker bearing some focus features. Multiple whfronting languages which exhibit Superiority effect show that the Superiority effect persists in spurious coordinated questions. Conjunction ‗i‘ seems to be mandatory in some examples Kto i kakogo čërta pripersja?! - ‗Who came and why the hell did he?‘ furthermore, presence of spurious ‗i‘ forces single-pair reading, unless a wh-question has three wh-elements. Putting this issue aside, let us address now phenomenon of complementizer čto –‗that‘ briefly mentioned in section 4.1.1. I discuss this subject thoroughly in Chapter 5. 160 CHAPTER 5. MYSTERY OF THAT-TRACE EFFECT In section 4.1.1, we came across examples where presence of a complementizer čto (that) can alter the meaning of the sentence. Recall examples from previous chapter repeated below in (379) which show that the matrix clause is modified whenever a complementizer čto (that) is present and the embedded clause is modified in its absence. The example in (379a) shows local dependency and a modified matrix clause, while example in (379b) shows long-distance dependencies and a modified embedded clause. 379. a. Kogda on skazal, čto Alex pridet? when (only matrix) he said that Alex will come ‗When did he say that Alex was coming?‘ Answer: He told me yesterday. b. Kogda on skazal Alex pridet? when (only embedded) he said Alex will come ‗When did he say that Alex was coming?‘ Answer: Alex is coming on Tuesday. The above examples show that the presence of čto (that) changes the meaning of the sentence and does not allow ambiguous readings, while the absence of the complementizer forces the embedded clause reading. I refer to this phenomenon as thattrace effect which has been well addressed by many linguists among which Cowart (1997), Beletti & Rizzi (1988), Pesetsky (1982), Bayer (1984), Haider (1983), Grewendorf (1988), Szczegielniak (1999), Featherston (2005), etc.; however this phenomenon has not received enough attention from linguists working in the Russian language. 161 It has been observed for English that the presence of that forces a reading on the matrix clause, whereas its absence makes the reading ambiguous as in (380)38. 380. a. Why did you say that he was coming? (only matrix) b. Why did you say he was coming? (ambiguous) English reveals a clear subject-object asymmetry; this effect has been extensively studied by Cowart (1997). In English, extraction of a wh-subject out of a clause with that is not allowed, while an object can be easily extracted (381). 381. a. *Whoi do you think [ ti´ that [ti will be coming]] (subject extraction) a.´ Whoi do you think [ ti´ [ti will be coming]] b. Whati do you think [ti´ that [Kim gave away ti]] (object extraction) b.´ Whati do you think [ti´ [Kim gave away ti]] Bayer (1984) claims that there is no subject-object asymmetry for certain dialects in German, particularly in Bavarian (382). 382. a. Weri meinst du [ ti daβ [ti ihn geküβt hat]] who think you that him kissed has b. Ich weiβ nicht [weri daβ [ti ihn geküβt hat]] I know not who that him kissed has The same has been claimed for pro-drop languages, such as Italian and Spanish (383) (Szczegielniak, 1999) and Polish (Stepanov and Georgopoulos, 1995). 383. Chii credi [ti che [ti verra a visitarci]] who think 384. that will visit (Italian) (Pesetsky 1982) a. Kogoi myślisz , ti że Maria pızyprowadzi ti who (you) think that Mary bring (Polish) (Szczegielniak, 1999) ‗Who do you think that Mary will bring‘ 38 It should be pointed out that (380a) can be judged as ‗ambiguous‘ by some native speakers 162 b. Kogoi myślisz, ti że ti pızyprowadzi Marię who (you) think that will bring Mary ‗You think that who will bring Mary‘ Russian data, that I have collected, also exhibit the absence of subject-object asymmetry as in (385). If in Polish, Italian, and certain dialects of German extraction is possible of both subject and object, in Russian neither subject nor object can be extracted out of embedded clauses with čto in C, supporting judgements offered by Stepanov and Georgopoulos (1995). 385. a. *Ktoi ty dumaeš‘ [ ti´ čto [ti kupil košku ]] who you think that (subject extraction) bought cat * ‗Who do you think that bought a cat?‘ a´. Ktoi ty dumaeš‘ [ ti´ [ti kupil who you think košku ]] bought cat b. *Čtoi ty dumaeš‘ [ ti´ čto [prines Peter ti]] what you think (object extraction) that brought Peter ‗What do you think that peter brought?‘ b´. Čtoi ty dumaeš‘ [prines Peter ti]] what you think brought Peter Therefore, how can one explain this phenomenon? Why are extractions out of clauses with a complementizer impossible? Why does this particular placement of čto influence the interpretation of the sentence? This chapter aims at answering these questions and proposing the analysis which might be closer to explaining the nature of the phenomenon than the ones that already exist. First, I will present some analyses proposed in the literature and show their insufficiency in accounting Russian data. 163 5.1. Indicative complementizer ‘čto’ 5.1.1. An NIC approach One of approaches that have been suggested in order to account for that-trace effect was offered by Pesetsky (1982), where he makes use of the Nominative Island Condition (NIC) proposed by Chomsky (1980:13): NIC: A nominative anaphor in S cannot be free in S´ containing S. Following the NIC proposal, a wh-trace behaves like an anaphor which should be governed by a wh-element (Szczegielniak, 1999). This condition is violated in (386a), since it has to be deleted in order to satisfy Doubly Filled Comp filter, making the sentence ungrammatical. Doubly Filled Comp filter: No CP can have both an overt specifier and an overt complementizer generated in C 386. a. *Whoi do you think [ ti´ that [ti bought a radio ]] b. Whoi do you think [ ti´ [ti bought a radio ]] Therefore, following the Doubly Filled Comp filter the trace ti´ should be deleted in (386a), which in turn violates the NIC, whereas example in (386b) does not violate any of two conditions, leaving the sentence grammatical. The examples in (387) do not pose any problems, since object traces are immune to NIC and the ti can be deleted to satisfy Doubly Filled Comp filter (Szczegielniak, 1999). 387. a. Whati do you think [ti´ that [Roger bought ti]]. b. Whati do you think [ti´ [Roger bought ti]]. 164 Russian data show a potential problem for the NIC approach. While English has a subject/object asymmetry in the case of wh-extraction across that-complementizer, Russian, on the other hand, does not show such asymmetry; moreover, it does not allow such extractions at all (388) (Stepanov & Georgopoulos, 1995). 388. a. Kogoi ty dumaeš‘, ti (*čto) privedet Elena ti? who you think that will-bring Elena ‗Who do you think that Elena will bring?‘ b. Ktoi ty dumaeš‘, ti (*čto) ti videl Elenu? who you think that saw Elena ‗Who do you think saw Elena?‘ Following the NIC approach, the example in (388b) shows a violation of NIC when the trace is deleted to satisfy Doubly Filled Comp filter, and thus the sentence is ungrammatical. This approach fails to explain ungrammaticality in (388a) where kogo is an object and a non-nominative anaphora, and thus should be immune to NIC, yet it is still ungrammatical. This suggests that Russian objects behave the same way as subjects, i.e. an anaphor cannot be free in CP, which is problematic for the NIC approach. Therefore, I put this analysis aside and turn my attention to another one. 5.1.2. An ECP account (Rizzi, 1990) Rizzi (1990) suggests another account of ‗that trace‘ effect employing the Government and Binding framework and utilizing the notions of antecedent and head government as per Relativized Minimality (1990). Relativized Minimality: . . .X. . .Z. . .Y. . . X α-governs Y only if there is no Z such that (i) Z is a typical potential α-governor for Y, (ii) Z c-commands Y and does not c-command X. 165 He proposes that traces are subject to the Empty Category Principle (Rizzi, 1990:32), which requires that empty categories be properly governed. (i) A non-pronominal empty category must be: (i) Properly Head Governed (ii) Antecedent Governed or Theta Governed Szczegielniak (1999) mentions that there two views on ‗proper‘ government. Kayne (1984) introduced a canonical notion of government, in which a language has a specified ‗canonical‘ direction of government (OV vs VO). In this view, X properly governs Y if X canonically governs Y. Rizzi presents a different version of ‗proper‘ government where X governs within X‘. (Rizzi, 1990:31). Rizzi defines Head Government as in (ii) and Antecedent Government as in (iii). (ii) Head Government: X head governs Y iff (i) X € {A, N, P, V, Agr, T} (ii) X m-commands Y39 (iii) no barrier intervenes (iv) Relativized Minimality is respected 39 Chomsky (1986) defines m-command as: X m-commands Y if and only if X does not dominate Y, Y does not dominate X, and the maximal projection of X dominates Y X m-commands everything that XP dominates, except for X‘, which dominates X XP X’ YP X Y’ Y 166 (iii) Antecedent Government: X antecedent-governs Y iff (i) X and Y are co-indexed (ii) X c-commands Y (iii) no barrier intervenes (iv) Relativized Minimality is respected Rizzi argues that according to the definition of government only SVO languages would have that-trace effect, inasmuch as in VSO or SOV languages Infl governs the subject. Moreover, Rizzi argues that in languages with Agreement in COMP, that-trace effect should not be present at all, given that C becomes a potential governor. This has been confirmed by the data from Bavarian German and West Flemish, where C is filled with appropriate morpho-syntactic features like tense or agreement (Szczegielniak, 1999). In order to explain the antisymmetry in (389), Rizzi combines the formulation of ECP presented in (ii) and Relativized Minimality, which states that a governor cannot govern inside the domain of another governor, i.e. X cannot govern Y if there is a closer potential governor Z for Y (…X…Z…Y). 389. a. Whoi do you think [ ti´ that [Bill saw ti]] b. *Whoi do you think [ ti´ that [ti left]] Thus in (389b) the antecedent ti´ governs ti, but the latter is not properly head governed. That does not govern ti, since it is not endowed with IP features. Thus, the subject trace is not properly head governed and as a result the ECP is violated (Rizzi, 1990:33). Object extraction in (389a) does not violate ECP, because the verb fulfils both requirements: it properly head-governs and theta-governs the object trace. The lack of a complementizer makes (389b) grammatical, as in (390). This allows Rizzi to distinguish two complementizers: one of which is realized as that and another one as AgrP. 390. Whoi do you think [ti´ [ti left]] Rizzi argues for that to be inert for government, while Agr belongs to the class of governors. An occurrence of Agr must be licensed by co-indexation with its specifier, 167 which must be filled by a wh-operator or a trace (Rizzi, 1990:52). Following from the said above, (391) will be represented as follows: 391. Whoi do you think [ ti´ Agr [ti Infl left]] The subject trace in (391) is properly head-governed by Agr, which is in the head of Comp and antecedent-governed by the Spec of Comp (Rizzi, 1990:53). Following Rizzi‘s approach, one would predict that object extraction out of an embedded clause over čto will not violate ECP, inasmuch as it would be properly headgoverned and antecedent/theta-governed. Yet, the example in (392b) is still ungrammatical. 392. a.*Ktoi ty dumaeš‘ [ ti´ čto [ti kupil who you think radio ]] that bought radio ‗Who do you think bought a radio?‘ a.´Ktoi ty dumaeš‘ [ ti´ [ti kupil who you think radio ]] bought radio b.*Čtoi ty dumaeš‘ [ ti´ čto [Roger kupil ti]] what you think that Roger bought ‗What do you think that Roger bought?‘ b.´Čtoi ty dumaeš‘ [ ti [Roger kupil ti]] what you think Roger bought Notice that the absence of the complementizer čto in (392a´ and 392b´) makes the sentences perfectly grammatical, but does it suggest a presence of some AgrP which would satisfy ECP? I will address this issue later in the Chapter. Based on the data presented in (392), we will put Rizzi‘s analysis aside and consider other proposals. 168 5.1.3. Richards‟ account (1999) Richards proposes another account on that-trace effect, assuming a copy theory of movement and suggesting that subject/object asymmetry arises in languages like English, where subjects are hard to extract because PF must receive clear instructions about which copy of movement to pronounce (Szczegielniak, 1999). He suggests that only positions which check strong features are spelled out at PF, and if an element checks more than one strong feature, which occurs on separate heads, PF does not get clear instructions which copy to pronounce. This explains subject/object asymmetry in English (393), where who, being a wh-subject, checks a strong EPP feature in [Spec;TP] ((393a), whereas the object does not ((393b). 393. a. *Whoi do you think [ ti´ that [ti bought a radio ]] b. Whati do you think [ ti´ that [Roger bought ti]] c. Whoi do you think [ ti [ti bought a radio ]] Based on the above discussion, (393c) should be ungrammatical, inasmuch as who checks a strong EPP feature, suggesting that the analysis proposed holds that a subject should never be extracted. In order to account for subject extraction in an example like (393c), Richards suggests that null complementizers are actually affixes which require a host, preventing pied-piping of the whole subordinate clause to satisfy wh-features on the matrix CP. We end up with a situation where either the PF spell-out condition or the null complementizer stranding condition40 has to be violated in order to move wh- 40 The complementizer stranding condition suggests that the complementizer that cannot be stranded by the movement of IP in either direction. The whole CP can move to the front by undergoing leftward movement (b), for example, via tapicalization; however the IP alone cannot move, leaving that behind (c). (Sun-Woong Kim, 2008) a. Mary told herself [that John is a fool] at least twice a day. b. [That John is a fool], Mary told herself at least twice a day. c. *[John is a fool], Mary told herself [that] at least twice a day. 169 elements out of embedded clauses. Richards claims that the null complementizer stranding condition outranks the PF spell-out condition and that we have subject movement over a null complementizer (Szczegielniak, 1999). This approach would predict that Russian objects check strong EPP features, since extraction out of an embedded clause with a complementizer is impossible (394). 394. *Čtoi ty dumaeš‘ [ ti čto [Roger kupil ti]] what you think that Roger bought ‗What do you think that Roger bought?‘ Indeed, Russian objects can check EPP features, as has been noted by (Bailyn, 2004). This notion is known as Generalized Inversion, where an object moves to check EPP features, forcing the verb to move over subject, which, in its turn, stays in-situ. 395. [Etu knigu] čitaet Ivan (Bailyn, 2004:4) this book reads Ivan ‘Ivan reads this book.‘ However, the example in (394) does not seem to be the case of object shift: Bailyn (2004:33) points out the requirement of verb-raising in inversion when an XP that satisfies the EPP is a non-Nominative. Richards‘ analysis would also predict that languages like Polish and Italian, where subject/object asymmetry is absent, do not have a strong EPP feature on IPs, since extraction of both subject and object do not pose any problems. In addition, his proposal does not take into account Criterial freezing, which suggests that when an element moves to a position to check features, it freezes and cannot move further. I present an account offered by Rizzi and Shlonsky (2007) later in the chapter, which proposes a solution to Criterial freezing, but first I outline main points of Szczegielniak‘s (1999) proposal. 170 5.1.4. Szczegielniak (1999) Szczegielniak (1999) makes another attempt to account for that-trace effect by expanding Rizzi‘s proposal which suggests the existence of two complementizers, one of which is realized as that and the other as AgrP. Szczegielniak (1999:13) argues that English null complementizers pattern with the German and West Flemish agreeing complementizers (396 and 397), which clearly indicate that features like agreement can be easily checked in the CP. 396. a. Damid ich komm (Munich Bavarian) sothat I come b. damidsd kommsd sothat-2SG come 397. a. Da-t marie goa-t (West Flemish, Shlonsky 1994, Haegeman 1992) that -3SG Marie fo-3SG b. da-n Marie en Valére goa-n that-3PL Marie and Valére Expanding Rizzi‘s account, Szczegielniak (1999) offers two complementizer positions in CP: C1, which bears agreement features and located lower than C2; and C2 which does not bear any [φ] features. 398. C2P C1P Spec C2 -wh -φ TP Spec C1 +wh +φ Spec T 171 He suggests that Russian complementizers are of the C2 type where no feature checking takes place. English, on the other hand, exhibits two types of complementizers, with that belonging to the C2 type, whereas a null complementizer is of the C1 type, bearing subject [φ] features. Furthermore, Szczegielniak suggests successive phase cyclic movement in languages with an agreeing complementizer where movement of the subject to a CP phase periphery position is possible. This leads to a conclusion that languages with a C 1 type will not exhibit that-trace effect. Modifying Chomsky‘s (1995) proposal, Szczegielniak suggests that an embedded C can serve as a non-feature checking landing site for a phase XP, which undergoes overt movement if that XP has moved to the periphery of a phase. 399. a.*Whoi do you think [CP ti´ that ti bought a radio] b. Whati do you think [CP ti´ that Roger bought ti] 400. a. Who do you think [CP bought a radio] Szczegielniak assumes that embedded C in (399a, b) cannot have any features because of the overt complementizer (C2 type). In cases when there is no overt complementizer (400), C1 attracts subject from [Spec;CP], placing it in a phase periphery position. It is crucial to understand the derivation for the examples in (399), presented in (401): 401. a. [SpecT Roger2 [Spec2 vP whati [Spec1 vP t2 [v bought]…t1]]] b. [SpecT whoi [Spec1 vP ti [vP bought a radio]]] Szczegielniak argues that in (401a), when V is introduced, the subject is merged in [Spec;VP], and later the wh-object moves out of VP to the outer [Spec;vP], at which stage both the subject and the object are at a phase periphery position. When T is introduced the subject raises to [Spec;TP] to check EPP feature; thus, when CP is introduced, the subject is no longer in a phase periphery position, whereas the object is. It is for this reason that an object is allowed to undergo successive cyclic movement, while the subject is ‗frozen‘ in [Spec;TP]. Subject wh-words can be fronted if they are first moved to [Spec;CP], an agreeing Complementizer is needed to attract the Subject for feature checking. Movement proceeds automatically via the edges of phases (‗phase 172 hoping‘ in Szczegielniak‘s terminology). The movement to a phase periphery position must be independently triggered by the feature composition of the C or v head, and the landing side must have a feature to check, since PF spelling-out happens at the last feature checking position (Szczegielniak, 1999:16). In English, a null complementizer attracts a subject to [Spec;CP], which thus occupies a phase periphery position and can move further, while the C2 complementizer that blocks the subject from moving to a phase periphery position and thus blocks it from successive movement (Szczegielniak, 1999). According to Szczegielniak, with a C1 type complementizer, a subject moves directly to [Spec;CP] to check EEP features while avoiding [Spec;TP] where the EPP features are normally checked. Szczegielniak does not give clear explanations as to why a subject skips [Spec;TP] and ends up in [Spec;CP]. Szczegielniak argues that only C1 type complementizers license a Spec position and at the same time can have an overt complementizer which allows subject extraction (402). 402. Ktoi myslisz, who (you) think [CP ti´ ze ti przyprowadzi Marie] that bring (Polish) Mary ‗Who do you think brought Mary?‘ Now, how can his approach be implemented for Russian? Szczegielniak (1999) suggests that a higher C2P in Russian is not a phase, and clauses with any overt complementizer block any kind of extraction. Thus, phase hopping goes through every phase position, and if any step is blocked then the whole movement is invalid. Szczegielniak does not give an explanation for what happens in the examples with a null complementizer; moreover, he points out that it is not an instance of agreeing null complementizers which heads a phase, as in English. Szczegielniak (1999) emphasizes that in Russian every overt complementizer blocks any kind of extraction. I would disagree with this statement, since there are instances in Russian, when a complementizer allows both subject and object extraction out of an embedded clause, as in the case of a subjunctive complementizer čtoby-‗that‘ (403). 173 403. a. Kto ty hočeš‘ *(čtoby) prišël na svadbu? who you want thatSUBJ came on wedding ‗Who do you want to come for the wedding?‘ b. Kogo ty hočeš‘ *(čtoby) who you want ja priglasila? thatSUBJ I invited ‗Who do you want me to invite?‘ This yields yet another ‗abnormality‘: the absence of subject/object asymmetry and allowance of both elements to be extracted. I investigate this phenomenon in section 5.2. Leaving Szczegielniak‘s approach aside for now, I address another proposal offered by Rizzi and Shlonsky (2007) for treating subject/object asymmetry 5.1.5. Rizzi and Shlonsky (2007) Rizzi and Shlonsky (2007) propose a different analytic path to subject/object asymmetry which avoids some problems with the approaches discussed above. In their paper, they adopt two ideas presented in Rizzi (2006): a) An element moved to a position dedicated to some scope-discourse interpretive property, a criterial position, is frozen in place (Criterial freezing) b) Classical EPP, the requirement that clauses have subjects, can be restated as a criterial requirement, the Subject Criterion. (Rizzi and Shlonsky, 2007) Subject Criterion is understood as a functional head Subject, distinct from and higher than T and other heads in the functional structure of the clause (Cinque, 1999), which attracts a nominal to its Spec and determines the subject-predicate articulation (Rizzi and Shlonsky, 2007). Thus, in accordance with (b) thematic subjects move to the Criterial subject position, where, according to (a), they are frozen by a Criterial Freezing principle. 174 (404a) is ruled out by Criterial Freezing, while in (404b), no problem arises for object extractions, since there is no Object Criterion, parallel to Subject Criterion (Rizzi and Shlonsky, 2007). 404. a. *Qui crois-tu [que [tqui va gagner]]?41 who think you that (French) will win (Rizzi and Shlonsky, 2007) ‗Who do you think will win?‘ b. Qui crois-tu [que [Paul va aider tqui ]]? who think you that Paul will help ‗Who(m) do you think that Paul will help?‘ This kind of approach to subject/object asymmetry provides an account for immovability of subjects, although there are quite a few languages allowing subject extractions out of embedded clauses, e.g. Polish and Italian. 405. Chii credi [ti che [ti verra a visitarci]] who think 406. that Ktoi myslisz, will (Italian) visit (Pesetsky 1982) [CP ti´ ze ti przyprowadzi Marie] who (you) think that bring Mary (Polish) (Szczegielniak 1999) ‗Who do you think brought Mary?‘ Rizzi and Shlonsky (2006; 2007) provide two main strategies for forming such questions: 1) Fixed subject strategies: the subject does not move and remains in its frozen position in [Spec;Subj], and an A´ construction involving subject is obtained with either no movement at all or with movement of a larger constituent (clause pied-piping). 41 Note that when a complementizer qui is used instead of que, the sentence is grammatical. Qui crois- tu qui va gagner? who think you that will win ‗Who do you think will win?‘ This phenomenon has been analysed by Taraldsen (2001) as qui=que+Expl. An Expletive is able to check Subject Criterion, and thus it allows the thematic subject to be extracted. I will discuss this issue later in the chapter. 175 2) Skipping strategies: the subject moves by skipping the freezing position and is extracted directly from its thematic position or from some other predicate-internal position. The most straightforward case study of the first strategy is the use of a resumptive pronoun for A‘ constructions involving embedded subjects (Rizzi and Shlonsky, 2007). Rizzi and Shlonsky study subject-object asymmetry in Hebrew, where an object resumptive pronoun can easily appear in-situ (407a) or be fronted to a topic or a topic like position in CP (407b-c), while subject resumptive pronoun should remain in-situ (408) (Rizzi and Shlonsky, 2007). 407. a. Kaniti raca et ha-šulxan še Xana amra še Dalya ma‘amina še Kobi oto. (I)bought ACC the-table that Hannah said that Dalya believes that Kobi wanted him ‗I bought the table that Hannah said that Dalya believes that Kobi wanted.‘ b. Kaniti et ha-šulxan še Xana amra še Dalya ma‘amina še oto Kobi raca t. (I)bought ACC the-table that Hannah said that Dalya believes that him Kobi wanted c. Kaniti et ha-šulxan še Xana amra še oto Dalya ma‘amina še Kobi raca t. (I)bought ACC the-table that Hannah said that him Dalya believes that Kobi wanted 408. a. Kaniti et ha-šulxan še Xana amra še Dalya ta‘ana še hu ya‘ale harbe kesef. (I)bought ACC the-table that Hannah said that Dalya claimed that he wil cost a lot money ‗I bought the table that Hannah said that Dalya claimed that will cost a lot of money.‘ b.*Kaniti ya‘ale et ha-šulxan še Xana amra še hu Dalya ta‘ana še harbe kesef. (I) bought ACC the-table that Hannah said that he Dalya claimed that will cost a lot money 176 The data in (407) was presented by Borer (1984) as example of the successive cyclic nature of wh-elements. Borer attributes the ungrammaticality in (408b) to subject relative operators: a language specific lexical property (Rizzi and Shlonsky, 2007). Rizzi and Shlonsky explain the ungrammaticality in examples like (408b) by the fact that the subject resumptive pronoun satisfies the Subject Criterion in [Spec;Subj] and ends up being frozen in the position, while the object resumptive pronoun is not, given that there is no Object Criterion. Rizzi and Shlonsky (2007) come across data where a sentence with a null relative operator is perfectly grammatical as in (409), which demonstrates conclusively that Hebrew must possess some device for extracting a subject without moving it first to [Spec;Subj], where it gets frozen. 409. Kaniti et ha-šulxan še Xana amra še Dalya ta‘ana še ya‘ale harbe kesef. (I)bought ACC the-table that Hannah said that Dalya claimed that will cost a lot money ‗I bought the table that Hannah said that Dalya claimed that will cost a lot of money.‘ Rizzi and Shlonsky (2007) suggest that the reason for the subject resumptive pronoun being forced to move to [Spec;Subj] is the weak nature of resumptive pronouns. Pronouns in (407) and (408) are considered to be weak, inasmuch as they are associated with an inanimate relative head. A weak pronoun cannot violate a positional constraint; it is assumed that a weak pronoun is licensed in a [Spec;Head], plausibly [Spec;Subj] for weak subject pronouns. As a result, a weak subject pronoun must move to [Spec;Subj] to satisfy the requirement, where it gets frozen. Relative operators, on the other hand, are not weak pronouns, and they can ‗skip‘ the freezing position as in (409). The technique of ‗skipping,‘ proposed by Rizzi and Shlonsky (2007) as a second strategy, is discussed after addressing clausal pied-piping Another case study of fixed subject strategy is clausal pied-piping discussed by Rizzi and Shlonsky (2007) on example of Imbabura Quechua. It has been noted that Criterial freezing prohibits satisfaction of the criteria in ‗passing,‘ i.e. the same element cannot satisfy two or more criteria. Although, it has 177 been seen in some languages that clausal pied-piping can solve the problem of simultaneous satisfaction of several Criterion as well as subject extraction. Imbabura Quechua, as well as Hebrew, shows subject/object asymmetry. An object can be extracted by wh-movement directly to the matrix Comp (410a), or by whmovement to an embedded Comp, and then by pied-piping of the whole embedded CP to the left (410b) (Rizzi and Shlonsky, 2007). 410. a. [Ima -ta] -taj Maria -ka Juzi miku-shka -ta kri-n? what ACC Q Maria TOP José eat-NOMINALIZER ACC believe-AGR ‗What does Maria believe that José ate?‘ b. [ima -ta wawa miku –chun] -taj Maria kri what ACCchild eat FIN -n? Q Maria believe-AGR ‗What does Maria believe (that) the child eat?‘ Lit. ‗[What the child eat] does Maria believe?‘ A subject wh-extraction out of an embedded clause is possible only by means of embedded clause pied-piping as in (411b); while in (411a) a Subject moves to [Spec;Subj] to satisfy Subject Criterion, becoming frozen, and thus cannot move further to the left without violating Criterial Freezing (Rizzi and Shlonsky, 2007:9). 411. a.*[pi] -taj Maria –ka chayamu-shka -ta kri -n? who Q Maria TOP arrive-NOMINALIZER ACC believe AGR ‗Who does Maria believe (that) has arrived?‘ b. [Pi chayamu-shka -ta] -taj Maria___ kri who arrive-NOMINALIZER Acc Q Maria -n ? believe AGR ‗Who does Maria believe (that) has arrived?‘ (411b) satisfies both Subject and Q Criteria without violating Criterial Freezing, i.e. whsubject pi moves to [Spec;Subj] inside the embedded clause to satisfy Subject Criterion, after this Criterion is satisfied, the whole embedded clause [pi chayamu-shkata] moves to [Spec;CP] to satisfy Q Criterion. Rizzi and Shlonsky (2007) have one issue with this strategy: when a wh-Subject moves to [Spec;Subj], it freezes there and the pied-piped phrase must have a whelement in its [Spec;CP]. Assuming that IP movement (leaving C stranded) does not 178 solve the problem, it would then follow that IP is not a Phase and thus inaccessible to long distance movement. In order to solve this problem, Rizzi and Shlonsky introduce a mechanism which allows local subject movement to C, based on example from Imbabura Quechua, where precedence of wh-subject over the Q head taj indicates movement of the wh-subject further. 412. [pi] -taj shamu -rka? who Q left AGR ‗Who left?‘ Rizzi and Shlonsky (2007) suggest that the Q feature in the embedded clause is not Criterial, but instead a pure formal feature which drives successive cyclic movement of a wh-operator in order to reach the criterial Q position. The legitimacy of the movement from the perspective of the Last Resort guideline is clear from cross-linguistic evidence for ‗internal movement‘ (a la van Riemsdijk, 1984), the strategy of moving a whoperator to the edge of pied-piped constituent transforming it into a complex phrase. The second strategy for subject extractions, proposed by Rizzi and Shlonsky, permits the thematic subject to skip the [Spec;Subj] position and avoid Criterial Freezing. One of the ways this manifests is by the filling of the criterial position by another element: an expletive, for example. Thus, in English copular constructions an expletive there satisfies the Subject Criterion, and a thematic wh-argument is available for further movement (413a). If there is no expletive, the subject moves to [Spec;Subj], where it satisfies the Subject Criterion, making no further movement possible (413a) (Rizzi and Shlonsky, 2007). 413. a. What do you think that there twhat is in the box? b.*What do you think that twhat is in the box? This strategy is used by Null Subject languages, like Italian, where a preverbal subject position is filled by pro leaving a thematic subject in a lower, predicate-internal position. The thematic subject is accessible to a movement directly from this position (414) (Rizzi and Shlonsky, 2007). 179 414. a. Chi credi che vincerà? who think that will win ‗Who do you think that will win?‘ b. Chi credi [che [pro Subj vincerà tchi]] who think that will win In this approach, an expletive is used for satisfying the Subject Criterion, thus allowing the thematic subject to escape the Criterial Freezing. In order to explain how a wh-subject moves to the left periphery, escaping Criterial Freezing in questions like (415), Rizzi and Shlonsky develop a derivation of local subject questions, for which they propose (for non-null subject languages) that projection Fin42 can be endowed with a relevant nominal feature and unvalued Phi features which can satisfy Subject Criterion. 415. Who came? They propose the following derivational steps: Step 1. Subject merges with the rest of the clause: Subj […[whsubj…]] Step 2. Fin, endowed with relevant nominal quality and unvalued Phi features, is merged directly with the structure prior to any movement to [Spec;Subj]: Fin+Phi [Subj […[whsubj…]]] Step 3. Subject Criterion is satisfied by Fin+Phi and thematic subject wh-element can freely move to its final scope position, [Spec;FocP] presumably, in the Rizzi and Shlonsky analysis, endowed with Q in the left periphery. Whosubj Foc [tsubj Fin+Phi [Subj …[tsubj…]]] 42 Recall that in Rizzi and Shlonsky‘s approach the CP is split into ForceP, FocP, WhP, TopP and FinP. 180 The Phi features of Fin are unvalued, and valuation is achieved when the subject moves through [Spec;Fin] on its way to the final position. In the end, Fin+Phi functions as a kind of bypassing device for the thematic subject by satisfying the Subject Criterion and allowing the subject to move higher (Rizzi and Shlonsky, 2007). Structurally, the derivation can be presented as follows: 416. Who came? Force Foc FinP Step 2 who Fin‘ SubjP Fin [-Phi] Phi features Step 3 to value the Spec t who Step 1 VP Subj V‘ twho came The same approach can be applied to subject extraction out of embedded clauses (Rizzi and Shlonsky, 2007). 417. Who do you think came? Who do you think [twho Fin+Phi [Subj [twho came]]] In (417), Fin+Phi is merged directly after the subject before wh-movement, thus satisfying Subject Criterion and allowing who to move to the left periphery, which values Phi features going through [Spec;Fin]. This approach is very similar to Agr in C in Rizzi‘s (1990) account. In the ECP account, Agr is in null C and properly governs the subject trace, thus allowing subject extraction out of the embedded clause, satisfying the ECP. The difference between the two approaches is the absence vs. presence of a wh-subject trace and the role of the 181 devices. In the ECP approach wh-trace is in [Spec;TP], while in the current approach it is in [Spec;Subj]. According to ECP, the role of the devices is to provide a proper governor and in the current approach to provide an expletive like element for satisfying the Subject Criterion (Rizzi and Shlonsky, 2007). Rizzi and Shlonsky (2007) discuss the issue of that-trace effect and come to a conclusion that in Standard English the complementizer that is incompatible with Fin+Phi strategy, i.e. there is no bypassing device for a thematic subject: it must move to [Spec;Subj] to satisfy the Subject Criterion, and gets frozen. Movement of the whsubject out of embedded clause is rendering the sentence in (418) ungrammatical. 418. *Who do you think that twho visited him? Describing the derivation step by step, we end up with the following steps: Step 1. Subject merges with the rest of the clause: Subj […[whsubj…]] Step 2. Fin, bearing that features (finiteness), is merged directly with the structure in [Spec;Fin]: Fin that [Subj […[whsubj…]]] Step 3. Subject Criterion is not satisfied, since Fin is not endowed with Phi features, and thematic subject has to move to [Spec:Subj], to satisfy Subject Criterion. That moves to ForceP to check force features. Rizzi and Shlonsky suggest that that expresses both finiteness (cannot be used with nonfinite IP) and declarative force. So, if the CP system minimally contains Force Fin IP, then the normal derivation of a that clause is where that is first merged in Fin, to express finiteness, subsequently moving to Force to check the force feature, in order to end up with the derivation in (419): 182 ForceP 419. Force‘ FinP that Fin‘ IP tthat In this derivation, that cannot be at the same time the head of a declarative clause and function as an expletive-like element satisfying the Subject Criterion. In order to explain a derivation lacking an overt C in English as in (30), Rizzi and Shlonsky (2007) propose that such derivation does not involve a complete Force structure, but a ‗truncated one.‘ 420. Who do you think came? They propose to truncate it at Fin, i.e. there is no Force in this derivation and a declarative interpretation is assigned by default. In this derivation, no conflict arises in the roles of Fin, and a Phi+Fin strategy can be deployed. Thus a possible derivation will be the following: Who do you think came? Truncated Force FinP Fin‘ Phi features Spec t who to value the 421. SubjP Fin [-Phi] VP Subj V‘ t who came 183 Rizzi and Shlonsky (2007) offer another more radical possibility: a deeper truncation which affects both CP and the SubjP layers. Therefore, an IP layer will be closing off with the AgrP, which is responsible for Case assignment. This would straightforwardly predict the absence of the Criterial Freezing effect, since SubjP is absent and there is no need for skipping device. 422. Who do you think came? Force FinP Truncated Fin‘ Spec SubjP Fin Subj‘ Spec AgrP Subj VP t who VP came Rizzi and Shlonsky (2007) provide a very detailed analysis of subject extractions out of embedded clauses. They do not discuss Polish data where presence of the complementizer ze does not force that-trace effect and a subject moves to the left periphery. 423. Kto myslisz ze przyprowadzi Marie? who (you) think that brought (Polish) Mary ‗Who do you think brought Mary?‘ Can we assume here that ze might have +Phi features which satisfy Subject Criterion? 184 To summarize, I have given a brief description of a number of proposals offering analysis of that-trace effect. I partially employ solutions proposed by Szczegielniak (1999) and Rizzi and Shlonsky (2007) in order to account for that-trace effect in Russian. However, before referring to the analysis of wh-subject extraction out of embedded clauses with that-trace effect, I study subjunctive clauses, since in this kind of constructions the subject/object asymmetry still persists although with different constraints: a subjunctive complementizer must be present. 5.2. Subjunctive complementizer ‘čtoby’ In section 4.4.1, we came across an example of a Russian subjunctive complementizer čtoby - ‗that‘, which allows extraction of both subject and object out of an embedded clause (424), contra to indicative complementizer čto-‗that‘, discussed in previous section, which blocks any extractions out of embedded clauses in Russian (425). 424. a. Kto ty hočeš‘ čtoby prišël na svadbu? who you want thatSUBJ came on wedding ‗Who do you want to come for the wedding?‘ b. Kogo ty hočeš‘ čtoby ja priglasila? whoACC you want thatSUBJ I invited ‗Who do you want me to invite?‘ 425. a.*Kogo ty dumaeš‘, čto privedet Elena? who you think that will-bring Elena ‗Who do you think that Elena will bring?‘ b.*Kto ty dumaeš‘, čto videl Elenu? whoACC you think that saw Elena ‗Who do you think saw Elena?‘ Interestingly, indicative čto can be easily omitted in (425), rendering the sentence grammatical; whereas subjunctive čtoby cannot be dropped. Its presence is obligatory 185 for the grammaticality. The same is observed in English subjunctive clauses, where presence of a subjunctive complementizer is obligatory: 426. I demand *(that) she leaves Subjunctive clauses have been argued to constitute more transparent domains than indicative ones, which make them syntactically transparent with respect to some syntactic relations (Quer, 1996). Subjunctive clauses in Russian are selected by so called volitional verbs or Sverbs (following Progovac‘s (1993) definition); these verbs are mainly verbs of wishing and requesting, such as xotet‟ - ‗to want‘, želat‟ - ‗to desire‘, poprosit‟ - ‗to ask for‘ and ugovorit‟ - ‗to persuade‘. The verb in the subjunctive clause must be morphologically in the past tense, although, it should be pointed out, that the event indicated in the embedded clause is not situated in the past either with respect to the event in the matrix clause or to the speech. 427. Mikael xočet čtoby Maria vyučila norveţskij. Mikael wants thatSUBJ Maria learnPAST Norwegian ‗Mikael wants Mary to learn Norwegian.‘ In (427), the subjunctive clause employs a verb vyučit‘-‗to learn‘ in the morphological past tense, but the event described in the embedded clause might happen in the future. It has been argued in the literature that the tense properties of subjunctive forms are defective and rely on the tense specification of the matrix (Quer, 1996). Contrary to subjunctive clauses, no restrictions are imposed on the tense marking of the embedded verb in indicative ones (428). 428. Ja dumaju čto Peter videl/ uvidit/vidit Elenu? I think that Peter saw/will see/sees Elena Picallo (1984, 1985) argues that even displaying morphological past tense, subjunctive predicates do not have an independent temporal interpretation, i.e. defective, opposite to indicative ones, which are contentful. This property prevents them from appearing independently in an assertive root clause (Quer, 1996). 186 429. *Ivan zvonil by. Ivan called –SUBJ, IMPF Picallo characterizes the subjunctive Infl as [-Tense, +Agr], whereas indicative one as [+Tense, +Agr]. The directivity of subjunctive Infl forces it to rely on the content of the higher Infl for its tense interpretation, which materializes in an anaphoric-like link between the temporal features of the subjunctive Infl and those of the root Infl (Quer, 1996:667). The tense dependency of the subjunctive is argued to be a reason of several transparency effects affecting the embedded clauses, one of them is the obviation effect, signifying that a pronoun cannot be co-referent with a non-pronominal that it ccommands. Subjunctive clauses embedded under volitional predicates do not allow coreference between the matrix and the embedded subjects. 430. Petjai xočet čtoby on*i/j ob etom ne znal? Peter want thatSUBJ he about this not knew ‗Peter does not want him to know about it.‘ In (430), the subject of the matrix clause Petja cannot be co-indexed with the subject of the subordinate clause on-‗he‘. This restriction is not observed in indicative clauses with a complementizer čto. 431. Petjai skazal čto oni/j vsё Petja said that he nam rasskaţet. everything to-us will tell ‗Petja said that he would tell us everything.‘ Chomsky (1981) attempts to explain the phenomenon of obviation, suggesting that the obviative interpretation of the pronominal follows from some version of the Avoid Pronoun Principle, ‗which imposes the choice of an empty pronominal over a lexically specified one where possible.‘ (Avrutin & Babyonyshev, 1997). In Russian, an overt pronoun can be avoided by using PRO instead. 187 432. Petja hočet PRO najti otvet na etu golovolomku? Peter wants PRO to find answer on this puzzle ‗Peter wants to find an answer on the puzzle.‘ According to Chomsky, the existence of acceptable (432) somehow blocks the option of using an overt pronoun in this construction, leading to a conclusion of adopting the Avoid Pronoun Principle. Avrutin & Babyonyshev find this analysis problematic since there is no reason to avoid the use of a lexical pronoun. Moreover, they come up with a Russian counterexample to the principle proposed by Chomsky (433) (Avrutin & Babyonyshev, 1997:233). 433. a. Volodja ugovoril Nadjui čtoby onai poexala v Evropu. Volodya persuaded Nadyai thatSUBJ shei went to Europe ‗Volodya persuaded Nadya to go to Europe.‘ b. Volodja ugovoril Nadjui PROi poexat‘ v Evropu. Volodya persuaded Nadyai PROi to go to Europe ‗Volodya persuaded Nadya to go to Europe.‘ (433) demonstrates the optionality in use of an overt pronoun as well as PRO; moreover, the possibility of having (433b) does not block the option of using an overt pronoun, as it has been suggested by Chomsky in favour of the Avoid Pronoun Principle. Finally, the fact that (433a) clearly reveals the possibility of a co-indexation of an object in the matrix clause with an overt pronoun in a subjunctive one makes the Avoid Pronoun Principle inappropriate as an explanation of the obviation phenomenon in subjunctive clauses. Another account for the obviation effect was developed in a number of languages, such as Spanish by Picallo (1984, 1985), Serbo-Croatian by Progovac (1993), etc., and became known as the domain extension approach. The main idea of this approach is that the Tense node of subjunctive clauses is anaphoric in a sense that it depends on its interpretation on the tense of the matrix clause, and the binding domain for the pronoun in the subject position of the subjunctive clause is extended and incorporated into the matrix clause. As a result, the violation of Principle B occurs, since a pronoun is bound 188 by the subject of matrix clauses in its extended governing category (Avrutin & Babyonyshev, 1997). Avrutin & Babyonyshev applied this analysis to Russian and pointed out that it explains examples like in (430), repeated here as (434). According to the domain extension approach, the Tense node of a subordinate clause is analyzed as anaphoric, and the governing category of the pronoun on–‗he‘ is extended to the matrix clause, which leads to pronoun on–‗he‘ being locally bound by the subject in the matrix clause, Petja, thus resulting in a violation of Principle B. In indicative clauses, on the other hand, the tense node of the embedded clause is analyzed as independent, and therefore no domain extension takes place and the overt pronoun is not bound in its governing domain (435). 434. Petjai hočet čtoby on*i/j ob etom ne znal. Peter want thatSUBJ he about this not knew ‗Peter does not want him to know about it.‘ 435. Petjai znajet čto oni/j ob etom poţeleet. Peter knows that he about this will regret ‗Peter knows that he will regret about it.‘ Avrutin & Babyonyshev provide some data which, according to them, are problematic for this approach. The first problematic case is presented in example (436) where an overt pronoun of the subordinate clause is co-indexed with an object of the matrix one. The possibility of this sort of co-indexation is not expected under the domain extension approach. Here, the governing category of ona – ‗she‘ should be extended to include the matrix clause, and thus be locally bound by the matrix object, which results in a violation of Principle B. 436. Maria ugovorila Nadjui čtoby onai poexala v Evropu. Maria persuaded Nadyai thatSUBJ shei went to Europe ‗Maria persuaded Nadya to go to Europe.‘ Quer (1996:662) explains the possibility of (436) by arguing that the restriction on the reference of the embedded subject does not apply here, given that it is effective only with respect to the matrix subject and not to an internal argument in the main clause. 189 Picallo (1984) also discusses this case and suggests the Case Resistance Principle (Stowell, 1981), which forces an extra position of the embedded clause as a result of which the internal argument of the main clause Nadju does not c-command the embedded subject ona–‗she‘ allowing co-reference. The second example of problematic data, accounted by Avrutin & Babyonyshev (1997), involves a possessive element contained within a subjunctive subject. In indicative sentences, a possessive pronoun cannot be co-indexed with the subject (437a), whereas a reflexive possessive pronoun must be co-indexed (437b) (Avrutin & Babyonyshev, 1997:234). 437. a. *Volodjai provodil egoi ţenu v Evropu. Volodya saw off his wife to Europe ‗Volodya saw his wife off to Europe.‘ b. Volodjai provodil svojui ţenu v Evropu Volodya saw off self‟s wife to Europe ‗Volodya saw his wife off to Europe.‘ In a subjunctive clause, on the other hand, we get the opposite effect where (438a) is grammatical, while (438b) is not, which is not predicted under the domain extension approach. 438. a. Volodjai xočet čtoby egoi ţena poexala v Evropu. Volodya wants thatSUBJ his wife went to Europe ‗Volodya wants his wife to go to Europe.‘ b.*Volodjai xočet čtoby svojai ţena poexala v Evropu. Volodya wants thatSUBJ self‟s wife went to Europe Under the domain extension approach, the reflexive in (438b) would be bound by the matrix subject in its binding domain, in accordance with principle B, in the same manner as seen in (437b). The pronoun ego-‗his‘ in (438a) would violate Principle B, just as in (437a), which is does not occur here (Avrutin & Babyonyshev, 1997). Avrutin & Babyonyshev propose yet another counterexample to the domain extension approach in (439), where subjects of some predicates can appear with quirky case and the verbs do not agree with dative subjects. 190 439. Volode bylo veselo. VolodyaDAT was fun ‗Volodya was having fun.‘ Now consider the example in (440) where, according to the domain extension approach, the binding domain of the dative pronoun is extended to the matrix clause and thus it should be bound by the matrix subject, leading to a violation of Principle B, which would predict ungrammaticality. Yet the sentence is well-formed (Avrutin & Babyonyshev, 1997:236). 440. Volodjai xočet čtoby emui bylo veselo. Volodya wants thatSUBJ him-dat was fun ‗Volodya wants to be having fun.‘ Avrutin & Babyonyshev (1997:237) offer another piece of evidence which cannot be explained by the analyses summarized so far. It has been proposed that some form of co-indexation between two Tenses must take place in order to trigger obviation. The example in (441b) shows that obviation can occur even if the higher Tense (in the matrix clause) with which subjunctive Tense could be co-indexed is absent. This would predict that domain extension for a pronoun in the subjunctive clause should not take place. 441. a.Volodino ţelanie čtoby rabočij klass vzjal vlast‘ zasluţivaet uvaţenija Volodya‟s desire thatSUBJ working class took power deserves respect ‗Volodya‘s desire for the working class to take the power deserves respect.‘ b. *Volodinoi ţelanie čtoby oni vzjal vlast‘ zasluţivaet uvaţenija Volodya‟s desire thatSUBJ he took power deserves respect ‗Volodya‘s desire to take the power deserves respect.‘ In (441), the noun želanie-‗desire‘ is related to the verb želat‟-‗to desire‘ which is volitional verb and licenses subjunctive clauses. Notice that such a construction is impossible with other, ‗non-volitionally‘ derived nominals (442). This fact perhaps suggests that the subjunctive may be driven also by lexical or semantic factors, i.e. volitional verbs and nouns derived from them. 191 442. *Volodino slovo čtoby rabočij klass vzjal vlast‘ zasluţivaet uvaţenija Volodya‟s word thatSUBJ working class took power deserves respect ‗Volodya‘s word for the working class to take the power deserves respect.‘ Quer (1996:680), in his notes, addresses the cases of independent subjunctive forms in non-assertive root contexts, and points out that they often involve some element in C indicating directive force, which seems to suggest that a covert modal is binding the tense variable of a subjunctive verb (for Catalan): 443. Que marxi! (Catalan) that leave Subj.PRS.3SG ‗Let him/her leave?‘ Progovac (1993) adopts the domain extension approach for a number of languages and argues that domain extensions do not occur in Russian subjunctive clauses. For Progovac, domain extension happens due to Infl and Comp deletion at LF. The process of deletion is constrained by recoverability, meaning that if Infl or Comp contain unrecoverable material, i.e. complementizers or negation, they cannot be deleted. She claims that Russian subjunctive particle by is in Infl, contra to Serbo-Croatian, where the particle is in Comp. She argues that the presence of a subjunctive particle in Infl makes it unrecoverable and non-deletable at LF. In support of this idea, she points out that NPIs and reflexives, which are licensed within a subjunctive clause whose domain has been extended do not occur in Russian subjunctive clauses (examples from Comrie (1980)) 444. *On ne xočet, čtoby ja obrascalsja ni k komu. he not want thatSUBJ I turn not to whom ‗He does not want that I turn to anyone.‘ So far, it looks like neither the Avoid Pronoun Principle nor the Domain Extension Approach are appropriate as an explanation of the obviation phenomenon in Russian subjunctive clauses. Taking the discussion above into consideration, Avrutin and Babyonyshev (1997) propose their perspective on the obviation phenomenon and tense defectiveness. 192 They suggest that there is a covert movement of V to I to C taking place at LF. They follow Watanabe (1993), who suggests that this movement operation takes place universally, arguing that AgrO must raise to T and AgrS to C to complete the process of Case-Assignment (Avrutin and Babyonyshev, 1997:240). The result of this process is illustrated in (446), which gives LF representation of example in (445). 445. Volodja skazal, čto on poceloval Nadju. Volodya said that he kissed Nadya ‗Volodya said that he kissed Nadya.‘ At the LF level, the embedded V raises to T, then V-T complex raises to AgrS and then the whole V-T-AgrS complex adjoins to C. 193 446. CP1 AgrSP1 1 C AgrS 1 i AgrS1 C1 T1 DP Volodjai AgrS1 V skazal TP ti T1 VP tT (said) CP2 tV AgrSP2 C2 AgrS i2 AgrS i2 C2 čto (that) AgrS2 DP on (he) ti V poceloval TP2 (kissed) tT VP tV 194 DP Nadju Avrutin and Babyonyshev (1997) claim that a parallel process of V raising to T, then to Agr and adjoining of the whole complex to C occurs in sentences with subjunctive embedded clauses. There is a difference between subjunctive and indicative complements: a temporal interpretation of the embedded clause. Given the fact, that though the subjunctive tense is morphologically past, the event is interpreted as taking place in the future, i.e. later than the event in the matrix clause, they argue for the existence of a relationship of temporal ordering between the two events. They suggest that this relationship is the result of the two events being bound by one event operator, which takes them as its variables. They adopt the approach proposed by Harley (1995), who develops an articulated VP structure and argue that the highest verbal projection is an EventP (dominated by the TP), which divides a sentence into two domains: clausal syntax and lexical syntax, where the rules of lexical composition operate the way that material contained within a single EventP is perceived to be one word (Avrutin and Babyonyshev, 1997:242). The highest projection of V (EventP) associated with all event properties: time, place, presence/absence of the initiator etc. In order to have an event in its scope an operator must have the head of the EventP (V) in its scope. Avrutin and Babyonyshev (1997) propose that the operator which takes the two events as its variables is a subjunctive complementizer čtoby. In order to bind the two events, the operator must have them in its scope at LF. Since čtoby is lower in the tree than the VP of the matrix clause, the event in the matrix clause is not in the scope the complementizer. To solve this, Avrutin and Babyonyshev suggest that čtoby moves at LF to a position where both events are in its scope. The movement of the operator is considered to be a last resort operation, thus the operator čtoby moves to the first available position, where it c-commands the matrix verb, gets the first event in its scope and cannot move any further. The result of this process is illustrated in (448), which gives an LF representation of the example in (447). 447. Volodja xočet, čtoby on poceloval Nadju. Volodya wants that-SUBJ he kissed ‗Volodya wants him to kiss Nadya.‘ 195 Nadya 448. CP1 AgrSP1 C1 1 1 AgrSi C C2 AgrSi2 DP Volodjai AgrS1 T1 T1 V AgrS1 TP ti CP2 tV V xočet (wants) AgrSP2 tC C2 čtoby DP on (he) (that) T2 VP=event tT AgrSi2 T2 AgrS2 TP2 ti V poceloval tT (kissed) VP=event tV 196 DP Nadju The embedded V raises to T, the V-T complex raises to AgrS, and the V-T-AgrS complex adjoins to the embedded complementizer. Then, the complementizer čtoby raises to adjoin to the matrix V, and the verb raising in the matrix clauses proceeds the same way as in the embedded clause. As a result, all verbal elements of the complex sentence occur within the matrix complementizer. In indicative clauses, on the other hand, the čto complementizer is not an operator and does not bind its variables, thus it does not raise to matrix V and does not move to the matrix CP at LF. Finally, to explain the impossibility of co-indexation of a subjunctive subject pronoun and a matrix subject, Avrutin and Babyonyshev (1997:245) propose that (447) is ruled out by Principle B, which applies to AgrS of the subjunctive clause. The pronominal subject of a subjunctive clause is co-referential with the subject of the matrix clause, i.e. the two are co-indexed. The AgrS of the subjunctive clause agrees with the subject, as does AgrS of the matrix clause. The relationship of agreement is realised as co-indexation between the subject and the AgrS. So, the matrix subject and the matrix AgrS are coindexed; the same goes for subjunctive AgrS and subject. By transitivity, the matrix AgrS and the subjunctive AgrS are co-indexed as well, resulting in two subjects and two AgrS‘s bearing the same index. Therefore, the subjunctive AgrS is bound by the matrix AgrS; matrix AgrS c-commands subjunctive AgrS, resulting in Violation of Principle B.43 To sum up what we know so far about subjunctive clauses in Russian, we come to the following: i) subjunctive clauses are selected by volitional verbs; ii) the verb in subordinate clauses has defective tense, i.e. the clause cannot be marked as present or future, appearing only in the past; iii) a nominative pronoun of a subordinate clause cannot be co-indexed with a subject of a matrix clause, known as an obviation effect, which is argued to be a result of tense dependency of subjunctive clause. We have shown that both the Avoid Pronoun Principle and the Domain Extension Approach are inappropriate as an explanation of the obviation phenomenon in Russian subjunctive clauses, and mo0reover for the approach outlined in Avrutin and Babyonyshev (1997), which explains the obviation effect and temporal interpretation by means of covert movement at LF. 43 Due to the space limit I will not concentrate here on explanation of absence of obviation phenomenon in cases mentioned in this section, for more details see Avrutin and Babyonyshev (1997). 197 Having sufficient information regarding peculiarities of indicative and subjunctive complementizers and possible analyses offered to account for these phenomena, we can propose a syntactic analysis for the complementizers in Russian. 5.3. Analysis of Russian Complementizers ‘čto’ and ‘čtoby’ Before going any further let us recall some data of indicative and subjunctive clauses. Remember that the extraction of a wh-subject and wh-object are equivalently unacceptable in questions with an indicative complementizer as in (449a-b); whereas it is possible in questions with a subjunctive complementizer (449c-d). 449. a.*Kto ty dumaeš‘ čto kupil who you think radio? tha bought radio ‗Who do you think bought a radio?‘ b. *Čto ty dumaeš‘ čto Roger kupil? what you think that Roger bought ‗What do you think that Roger bought?‘ c. Kto ty xočeš čtoby prišël na svadbu? who you want thatSUBJ came on wedding ‗Who do you want to come for the wedding?‘ d. Kogo ty xočeš čtoby ja priglasila? who you want thatSUBJ I invited ‗Who do you want me to invite?‘ Szczegielniak (1999) proposes two complementizer positions in CP: C1, bearing agreement features and C2, not bearing any [φ] features, suggesting that Russian complementizers are of the C2 type where no feature checking takes place, and as a result the extraction out of an embedded clause with a complementizer is impossible. It seems that the data in (449) 198 contradict Szczegielniak‘s statement regarding Russian. This brings up to the following question: Can we assume that Russian demonstrates two types of C in Szczegielniak terminology? Oshima (2003) suggests two types of complementizers for Russian: čto1, which we meet in indicative clauses, and čto2 which mainly occurs in Subjunctive clauses. She suggests that the subjunctive complementizer čto2 overtly attracts the mood marker by (450), unlike the indicative čto1. 450. Ja xoču čtoby zavtra vse bylo po preţnemu. I want that(SUBJ) tomorrow everything was as before ‗I want tomorrow everything to be as before.‘ Progovac (1993), following Brecht (1974), argues (for Serbo-Croatian) that by is a modal cliticizing onto COMP. Recall, that for Russian she suggests that the subjunctive particle by is in Infl and not in Comp, which precludes a domain extension for any Infl-defined domain. Avrutin and Babyonshev (1997) suggest that by is a part of a subjunctive complementizer čtoby. The same has also been argued by Oshima (2003). Before adopting any of above mentioned proposals we need to figure out what by actually is and how it adjoins to čto. By is usually treated as a mood particle, mainly because it is a hallmark of conditional mood in Slavic languages. It is assumed to introduce subjunctive mood, when by follows immediately after a complementizer (451); in contrast to a conditional, where it does not have a specific position (452). 451. Ja xoču čtoby ty mne pozvonil. I want thatSUBJ you me called ‗I want you to call me‘ 452. Ja xotel by poznakomit‘sja s I wanted part to get yo know nej. with her ‗I would like to get to know her.‘ 199 Oshima (2003) points out that the marker by denotes counterfactual situations and occurs clause-medially. By always occurs cliticized to a complement čto2 when the subjunctive clause is a complement to volitional verbs, as in (451); and it follows the subject in consequent clauses, as in (453), occuring adjacent to esli - ‗if‘ in antecedent clauses, as in (453). 453. Esli by on prishel ko mne včera, if my by obo vsёm dogovorilis. part he came to me yesterday, we part about all settled ‗If he had come to me yesterday, we could have settled everything.‘ I will follow Migdalski (2006) who argues that the by element exemplified in (452) and the one in (453) is the same element; a subjunctive vs. conditional reading depends on the verb by follows. If it follows a volitional verb, the construction is interpreted as subjunctive. Furthermore, Brecht (1997) noticed that when two subjunctive predicates are coordinated, čtoby is present in the first one, while the second one is introduced by by. 454. Ty velel čhtoby ja uexal v Minsk odin, a Vasja by ostalsja s you ordered thatSUBJ i got toboj? to Minsk alone and Vasja part remain with you ‗Did you order that I leave for Minsk and Vasja remain with you? (Brecht, 1997:35-36) It has been also observed by Dornisch (1998) for Polish that the subjunctive particle (‗auxiliary‘ in his terminology) as a clitic overtly raises to C in wh-questions (455); this seems to be true for Russian as well (456). 455. Co by Anna komu poleciła? (Dornish, 1998:221) what part Anna whom recommended ‗What would Anna recommend to whom? 456. Čto by mne kupit‘? what part me to buy ‗What should I buy?‘ 200 Dornisch points out that ‗in the unmarked pattern of multiple wh-questions formation, native speakers prefer to place the modal clitic in C, rather than in Infl‘ (Oshima, 2003:11). I argue that this is true for Russian as well, cf. the examples in (456 and 457). 457. *Čto mne by kupit‘? what me part to buy ‗What should I buy?‘ In a wh-question, particle by cannot be placed lower than subject in a wh-question, which is in [Spec;TP] (457), but has to follow a wh-element immediately as in (456), which suggests that particle must occupy a higher position than Infl, i.e. be situated in CP. Oshima (2003), following Terzi (1992), Krapova (2001) and Rivero (1994 etc), posit a projection M(ood) which accommodates subjunctive particles and functions as a realization of a feature [irrealis]. Oshima refers to the MP as MirrP, in order to distinguish it from a MoodP. Under the current framework, the syntactic structure of the CP with the complementizer is as follows. An indicative complementizer merges in Fin‘ and then moves to Force‘ as in (458); the Mood head is not activated, since it is not required by the matrix verb. 458. VP ForceP Force‘ čtoi FinP Fin‘ TP ti 201 The subjunctive complementizer has a different structure: the indicative complementizer čto merges in Fin; the particle by merges in MoodP; čto moves via Mood, where it merges with the particle by and moves to Force. 459. VP ForceP volitional verb Force‘ MoodP čtoby Mood FinP Fin‘ čto+by TP y čto (that) The volitional verb in the root clause selects a [+mood] compliment. The Mood of the embedded clause is spelled out in the higher position after čto moves through Mood. To summarize, the data suggest that Russian has only one complementizer, indicative complementizer čto, which merges in Fin and moves further to Force. Subjunctive complementizer čtoby is not a true morphological complementizer; it is derived by means of syntactic movement. The indicative complementizer čto moves through Mood, merges with a particle by and moves further to Force. This follows Szczegielniak‘s (1999) account of Russian complementizers, where he proposes only one complementizer of C 2 type. However, what makes it possible to extract both subject and object wh-element over the subjunctive complementizer while this extraction is blocked by the indicative complementizer? In order to suggest syntactic analysis of Russian wh-questions with embedded clauses, I make use of an analysis proposed by Rizzi and Shlonsky (2006; 2007). Let us recapitulate the main points of their analysis: 202 - The Subject Criterion can be satisfied by Fin+Phi and a subject wh-element can freely move to its final scope position. The Phi features of Fin are unvalued and valuation is achieved when the subject moves through [Spec;Fin] on its way to the final position. - The complementizer that is incompatible with a Fin+Phi strategy; hence that-trace effect occurs in sentences with subject extraction. - The absence of a complementizer in a clause with subject extraction can be explained either by: a) a truncation at Fin, i.e. there is no Force in this derivation and declarative interpretation is assigned by default, or by b) a deeper truncation which affects both CP and the SubjP layer. Recall that Russian indicative clauses do not exhibit Subject/Object asymmetry, i.e. extraction of either the subject or object out of an embedded clause with an indicative complementizer čto is impossible (460a-b), while the use of a null complementizer does not postulate any problems for extraction (460a´-b´). 460. a. *Kto ty dumaeš‘ čto kupil who you think košku? that bought cat ‗Who do you think bought a cat?‘ a´. Kto ty dumaeš‘ kupil who you think košku? bought cat b. *Čto ty dumaeš‘ čto Peter prines? what you think that Peter brought ‗What do you think Peter brought?‘ b´. Čto ty dumaeš‘ Peter prines? what you think Peter brought First, we need explain possibility of a subject movement to the left periphery in local subject questions. I follow Rizzi and Shlonsky (2007) by proposing a projection Fin, which is endowed with a relevant nominal feature and unvalued Phi features, which can satisfy Subject Criterion. 203 461. Kto kupil košku? who bought cat ‗Who bought a cat?‘ In (461), the subject merges with the rest of the clause, then Fin+Phi is merged with the structure prior to any movement to [Spec;Subj]; the Subject Criterion is satisfied by Fin+Phi and thematic subject wh-element can freely move to its final scope position, endowed with Q features in the left periphery. The Phi features of Fin are unvalued and valuation is achieved when the subject moves through [Spec;Fin] on its way to the final position. We get the derivation in (462), which gives representation of example in (461): 462. ForceP FinP kto ( who) Fin‘ Spec twho SubjP Fin [-Phi] VP Subj V‘ tkto kupil košku (bought) (cat) Now, the extraction of wh-subject out of the embedded clause in (463) is ungrammatical due to nature of čto, which is incompatible with Fin+Phi strategy (the same has been claimed by Rizzi and Shlonsky (2007) for English complementizer that). 204 463. *Kto ty dumaeš‘ čto kupil košku? who you think that bought cat ‗Who do you think bought a cat?‘ Subject Criterion is not satisfied, since Fin is not endowed with Phi features, and thematic subject has to move to [Spec:Subj], to satisfy Subject Criterion, were it freezes. That moves to ForceP to check force features. 464. ForceP Foc FinP čto (that) Fin‘ SubjP Fin tčto VP Subj kto (who) V‘ twho kupil košku (bought) (cat) The impossibility of subject extraction over the indicative complementizer čto has been clearly shown above. Now let us have a look at extraction of wh-object. 465. *Čto ty dumaeš‘ čto Peter prines? what you think that Peter brought ‗What do you think Peter brought?‘ 205 (465) demonstrates the impossibility of wh-object extraction over čto. Following Rizzi and Shlonsky‘s analysis, there is no reason for this movement to be illegal. There is no Object Criterion, parallel to Subject Criterion that would freeze the wh-object. Rizzi and Shlonsky‘s approach fails to explain example in (465). It seems that the complementizer that blocks the possibility of a wh-element to move to phase periphery position [Spec;ForceP]; thus, there is no escape hatch from the embedded phase and successive movement is impossible. Recall that in Chomsky‘s (1999) phase hopping theory, the successive cyclic movement involves movement from one phase periphery position to another, where [Spec;CP] is a phase periphery position and the whmovement has to move through it. This leads to the following statement: Whenever Head of Force is occupied by indicative „that‟, [Spec;ForceP] becomes inaccessible to a wh-element and further movement is impossible. This generalization works for all wh-elements. 466. a.*Kogda ty dumaeš‘ čto on pozvonit? when you think that he call ‗When do you think he will call?‘ b.*Začem ty dumaeš‘ čto on mnoj interesovalsja? what for you think that he with me interested ‗Why do you think he was asking about me?‘ Next, (467) presents data worth our attention, where no overt complementizer is present and the extraction of both a wh-subject and wh-object is possible. 206 467. a. Kto ty dumaeš‘ kupil košku? who you think bought cat ‗Who do you think bought a cat?‘ b. Čto ty dumaeš‘ prines what you think Peter? brought Peter „What do you think that peter brought?‘ In order to account for these data I apply analyses proposed by Rizzi and Shlonsky (2007) which were outlined in 5.1.5. One of the solutions they offer suggests deep truncation, which affects both CP and the SubjP layers. In their proposal, an IP layer will be closing off with AgrP, which is responsible for Case assignment. Let us consider Russian data in (468 and 469). 468. Kak ty dumaeš‘, (*čto) kto what you think pridet na večerinku? (*that) who will come on party ‗Who do you think will come to the party?‘ 469. a. Kakogo čërta ty skazal, čto Peter ne pridet? why the devil you said that Peter not come-FUT ‗Why the hell did you say that Peter would not come?‘ (possible answer: because I did not want to upset you) b.*Kakogo čërta ty skazal, Peter ne pridet? why the devil you said Peter not come (468) is an example of a partial wh-movement, where a wh-element stays in a lower CP and cannot move higher, inasmuch as the position in the matrix clause is occupied by an expletive-like wh-element kak. Notice, that here the complementizer that is not present. In (469a), the expression kakogo čërta – ‗why the hell‘ can modify only the matrix clause; moreover it requires use of čto in the embedded clause (469b), since kakogo čërta cannot be base generated in an embedded clause and thus cannot leave a trace in [Spec, CP], the presence of a complementizer is required. The examples above suggest that the CP domain cannot undergo deep truncation. 207 A second solution offered by Rizzi and Shlonsky (2007) proposes truncation of Force. In this derivation no conflict arises in the roles of Fin, and a Phi+Fin strategy can be deployed. This solution seems to be the most attractive. It explains the example in (470), where Fin endowed with [+Phi] features is merged, and a wh-element moves to the left periphery though [Spec;FinP], enabling Fin to check EPP feature on [Spec;SubjP]. 470. a. Kto ty dumaeš‘ kupil košku? who you think bought cat ‗Who do you think bought a cat?‘ There might be one problem with this solution when we look at examples with partial whmovement, where no overt complementizer takes place. 471. Kak ty dumaeš‘, (*čto) kto pridet what you think that na večerinku? who will-come on party ‗Who do you think will come to the party?‘ If we assume the truncation of Force, there will be no position of the wh-subject kto –‗who‘ to move, unless we assume that kto does not have to move here to [Spec;ForceP], I will address this issue in Chapter 6. To summarize, in this chapter, I have studied the phenomenon of that-trace effect in Russian indicative and subordinate clauses. I have discussed existing approaches and pointed out the insufficiency of these approaches in accounting Russian data. The data show that Russian indicative clauses do not exhibit Subject/Object asymmetry and that extraction of either object or subject out of an embedded clause with a complementizer is impossible. Following Rizzi and Shlonsky, I employ the projection Fin which is endowed with relevant nominal features and unvalued Phi features which satisfy Subject criterion. 208 The Russian data have suggested that whenever the complementizer čto is present, the movement of any wh-element is blocked, contrary to subjunctive complementizer čtoby, where such a movement is obligatory. Next chapter deals with partial wh-movement in Russian, examples of which we met in 5.3. 209 210 CHAPTER 6. WH-SCOPE MARKING CONSTRUCTIONS In this chapter, I focus on the syntactic properties of yet another type of interrogatives in Russian, known in the literature as wh-scope marker questions (à la van Riemsdijk, 1982) or questions with partial wh-movement (à la Mc Daniel, 1989). 472. Kak ty dumaeš‘, kto pridet How you think na uţin? who will come to dinner ‗Who do you think will come for dinner?‘ The peculiarity of these type of questions is that they trigger an answer which does not involve supplying the value to the wh-phrase kak – ‗how‘ which has matrix scope, but instead an answer, which involves supplying the value for the wh-phrase kto – ‗who‘ which undergoes movement in its own clause (Stepanov, 2000). Thus, (472) triggers a response similar to the one triggered by a long-distance question in (473). 473. Kto ty dumaeš‘ pridet na uţin? who you think will come to dinner ‗Who do you think will come for dinner?‘ Answer: I think that Peter and Marit will come for dinner. In (472), first wh-element kak, which appears in the matrix clause, marks the scope, in van Riemsdijk‘s (1982) terms, of the second wh-phrase, which moves overtly to a clause initial position in the lower clause. 211 The phenomenon of scope marking has been the focus of investigation in a number of languages, for example, German (474), Hungarian (475), Polish (476), Malay (477), Hindi (478), etc. 474. Was glaubst du wen Irina liebt? (Fanselow, 2007:442) what believe you who-acc Irina loves ‗Who do you belive that Irina loves?‘ 475. Mit gondolsz hogy kivel who-acc think beszélt Mari? that who-with talk (Horvath, 1997) Mary ‗Who do you think that Mary talked?‘ 476. Jak myślisz, kogo Janek lubi? (Stepanov, 2000:1) how think-you whom John loves ‗Who do you think John loves?‘ 477. Ali memberitahu kamu tadi Ali told apa (yang) Fatimah baca?(Cole&Hermon, 2000:105) you just now what that Fatimah read ‗what did Ali tell you just now (that) Fatimah was reading? 478. Siitaa-ne kyaa socaa, ki ravii-ne kis-ko dekhaa? (Mahajan, 2003) Sita-erg what thought that Riva-erg what-dat saw ‗Who did Sita think that Riva saw?‘ This phenomenon has also been studied in Russian by a few linguists, among others Stepanov (2000), Gilderen (2001), although it has not been studied to the extent in the above-mentioned languages (perhaps due to a limited distribution of these types of question), this phenomenon deserves our attention. One could propose treating example (472) as a sequence of questions involving some kind of parenthetical. Stepanov (2000) notes that such sequential/parenthetical questions are available universally. English is one of the languages which allows the following sequence (Stepanov, 2000:4): 479. What do you think? Who does John like? 212 In (479), each question forms a separate intonational unit. Questions involving parentheticals are also available in German. Compare a whscope marking question in (480) with the one involving a parenthetical in (481). 480. Was glaubst du, wen sie liebt? what believe you who she loves ‗Who do you believe that she loves?‘ 481. Was glaubst du, wen liebt sie? what believe you who loves she ‗Who do you think, who does she love?‘ It is very easy to see the difference between the two questions in German. The V2 effect appears in the question involving the parenthetical, which suggests that the second clause ‗wen liebt sie?‘ has matrix properties. The second clause in (480), on the other hand, does not show the V2 effect, indicating that this clause is subordinate to the matrix predicate glaubst (Stepanov, 2000). The contrasts visible in German are not observed in Russian. Stepanov enumerates some empirical diagnosis which can help us to distinguish wh-scope marking questions from parenthetical ones. It has been previously noted by Dayal (1996) and Hovarth (1997) that it is always possible to obtain a bound variable reading in whscope marking questions. Applying the same diagnosis to Russian, Stepanov (2000:4) gets the following: 482. Kak sčitaet [kaţdyj iz studentov]i kuda egoi mogut otpravit‘? how thinks every from student where him can send-they? ‗Where does every student think that they can send him?‘ In the second clause of (482), the pronoun ego – ‗him‘ receives a bound variable reading when located within the scope of the quantifier každyj iz studentov – ‗every student‘ in the higher clause. The availability of the bound variable reading indicates the standard ccommand relationship between the elements of the first and second clauses. Note that such 213 a bound variable is not available in sequential/parenthetical questions in English (Stepanov, 2000:4): 483. *What does [every student]i think? Where will they send himi? Based on the data presented above, Stepanov concludes that a wh-scope marking question must be represented by a single phrase marker, and therefore it cannot be involved in sequential/parenthetical structures. Furthermore, the presence of c-command suggests that the second clause in Slavic wh-scope marking questions is in subordination to the matrix predicate, i.e. the clausal scope of the wh-scope marker must be situated higher than that od the ‗true‘ wh-element. 6.1. Some properties of wh-scope marker in Russian Some of the properties of Slavic wh-scope marking questions are common among constructions found in other languages, for example German, Polish, Hindi, etc. (Stepanov, 2000). A. Any wh-phrase can be associated with the wh-scope marker: 484. Kak ty dumaeš‘ gde/ kak/kogda/začem on pročital lekziju? how you think where/how/when/ why he read (Russian) leacture ‗Where/how/when/why did he give lecture?‘ 485. Was glaubst du wann/warum sie gekommen ist? what think you when/why she come ‗When/why do you think she came?‘ 214 is (German) B. Wh-scope marking is only possible with a matrix verb which can subcategorize for a non-interrogative sentential complement, i.e. verbs of ‗thinking.‘ 486. a.Kak ty dumaeš‘, kogda on pridet? how you think (Russian) when he will come ‗When do you think he will come?‘ b.*Kak ty sprosiš, kogda on pridet? how you ask 487. when he will come a.Was denkt Mark wen wir einladen sollen? what thinks Mark who we invite (German) should ‗Who does Mark think we should invite?‘ b.*Was fragt Mark wen wir einladen sollen? what asks Mark who we invite should ‗Who does Mark ask we should invite?‘ C. Any number of embedded wh-phrases can be associated with the wh-scope-marker 488. Kak ty dumaeš‘, kto kogda zvonil? how you think (Russian) who when called ‗Who do you think called and when?‘ 489. Jak myślisz, kto co czytają? (Polish, Stepanov 2000:7) how think-you who what read ‗Who do you think read what?‘ D. A wh-scope marker and a true wh-element cannot occur in the same clause. This requirement is known as ‗anti-locality‘, introduced by Müller (1997). 490. *Kak čto vy dumaete ona kupila? how what you think (Russian) she bought ‗What do you think she bought? 215 491. *Was ist sie warum gekommen? what is she why (German, Mycock, 2004:375) come ‗Why has she come? E. The wh-scope marker occupies clause-initial matrix position in wh-fronting languages. 492. Kak ty dumaeš‘, kto pridet na uţin? how you think (Russian) who come to dinner ‗Who do you think will come for dinner?‘ 493. Wat tinke jo wa‘t ik sjoen haw? what think you who.that I seen (Frisian, Hiemstra, 1986:97) have? ‗Who do you (that) I have seen?‘ 494. Was glaubst du wann sie gekommen ist? what think you when she come (German) is ‗When do you think she came?‘ 495. Jak myślisz, gdzie Maria tańczyła? (Polish, Stepanov 2000:7) how think-you where Maria danced ‗Where do you think Maria danced?‘ F. A matrix clause containing a wh-scope marker cannot be negated or contain a negative operator. 496. *Kak vy ne dumaete, kuda on pošel? how you not think (Russian) where he went ‗Where do not you think that he went?‘ 497. *Was glaubst du nicht, mit wem Maria gesprochen hat? what think you not with whom Maria spoken has (German) (Mycock, 2004:379) ‗Who don‘t you think Maria has spoken to?‘ 498. *Jak nie myślisz, kto kocha Marię? how not think (Polish, Lubańska, 2004: 76) who loves Mary ‗Who do not you think that loves Mary?‘ 216 499. *Koi bhii nahii kyaa soctaa thaa, ki no-one kon aayegaa? (Hindi, Mycock, 2004:379) what thinks was that who will come ?‗Who did no-one think will come?‘ G. Possibility of the second clause of a wh-scope marking interrogative to be a yes/no question, rather than a wh-question. 500. Kak vy sčitaete, budet li how you think will zavtra doţd‘? (Russian, Stepanov, 2000:8) part tomorrow rain ‗What do you think about whether it will rain tomorrow?‘ 501. Tum kyaa socte ho ki you what think Mary-ne Hans-se baat kiyaa yaa nahiiN? that Mary-erg Hans-ins talked or (Hindi) not (Mycock, 2004:377) ‗Do you think Mary talked to Hans or not? 502. Jak myślisz, czy how think Piotrek przyszedl? (Polish, Stepanov, 2000:8) whether Priotrek came ‗What do you think about whether Peter came?‘ H. A wh-scope marker may appear in any clause which intervenes between a true whphrase in an embedded clause and the highest occurrence of a wh-scope marker. 503. Kak ty dumaeš‘, kak Ivan sčitaet, kogo how you think vybirut prezidentom?44 (Russian) how Ivan believes who-acc will elect president ‗Who do you think that John think will be elected as the president?‘ 504. Was glaubst du, was Peter meint, mit wem Maria geschprochen hat? what believe you what Peter think with whom Mary talked (German) has ‗Who do you believe Peter thinks Mary talked with?‘ I. The wh-scope marker cannot be separated from the true wh-phrase by more than one clause. 44 Some speakers consider these kind of constructions to be degraded. 217 505. *Kak vy dumaete, čto Ivan sčitaet, how you think čto pročitali studenty? what Ivan believes what read (Russian) students ‗What do you think that John believes that the students read?‘ 506. *Jak Janek sądzi, że Piotrek myśli, co studenci prseczytali? (Polish) how Janek judge that Peter thinks what students read ‗What did John believe Peter thinks that the students read?‘ Properties referred to in A-F are common for wh-scope marking constructions found in different languages, while properties mentioned in G-H are found in just a few languages, and the ones mentioned in I are unique to Slavic wh-scope marking constructions. Another characteristic which is easily derived from the data above is that the whscope marker in Russian constructions is a wh-element meaning how, contrary to other whscope marking languages, where the wh-scope marker is a wh-element meaning what. This property of Russian language plays a crucial role for the analysis we will propose. 6.2. Previous analysis of wh-scope marking constructions It is clear that the true wh-phrase in a subordinate clause and a wh-scope marker in a matrix clause are linked in a wh-scope marking construction; the question that remains is what the nature of the link between the clauses is. Several approaches have been proposed in the literature to account for this phenomenon; I address some of them. 218 6.2.1. The Direct Dependency Approach The Direct Dependency Approach (henceforth DDA) is considered to be a commonly acceptable analysis of wh-scope marking questions. It was first proposed by van Riemsdijk (1982) and received further development by McDaniel (1989). Under the DDA approach, it is assumed, beginning with van Riemsdijk (1982), that a wh-scope marking question receives an interpretation similar to the one of the corresponding long-distance question. This quality of wh-scope marking constructions has been outlined earlier in examples (472) and (473). Following van Riemsdijk‘s assumption of similarity with long-distance questions, McDaniel (1989) proposes an account of whscope marking in terms of chains of wh-elements established in S-structure. According to this proposal, the sequence of the scope marker, the true wh-phrase and its trace form a whchain at S-structure, which is subject to locality restrictions (Rizzi, 1992). Furthermore, McDaniel suggests that the wh-scope marker is semantically empty and functions as an expletive-like element. Thus, according to DDA, kak in Russian does not literally mean how despite its lexical make-up, but is an element which has the syntactic feature [+wh] and has no semantic meaning. 507. Kak ty dumaeš‘, čto on kupil? how you think what he bought ‗Who do you think he bought?‘ According to DDA, an expletive-like element is base generated in the matrix [Spec;CP], which is the only Comp that has [+wh] feature in a wh-scope marking question, and functions as a some kind of place holder for the true wh-phrase which is located in the lower clause. The true wh-element moves to [-wh] [Spec;CP] in the lower clause and is licensed in overt syntax by forming a chain with the scope marker, thereby satisfying WhCriterion45. Later, at LF, the true wh-phrase moves to the matrix [Spec;CP] in order to 45 Wh-Criterion is defined as: a wh operator and a head endowed with the wh feature must be in a spec-head configuration at S-structure (Rizzi, 1996) 219 replace the wh-scope marker and to obtain the scope over the whole sentence (Dayal, 1994:143). 508. Was glaubst du mit wem Maria geschprochen hat what think you with who Maria spoken has ‗Who do you think Maria has spoken to?‘ CP C‘ Spec wasi (what) CP IP C glaubstk (think) CP NP du (you) VP CP V tk CP CP C‘ CP Spec CP mit wemi (with who) At LF IP CP NP Maria (Maria) VP PP ti V geschprochen hat (spoken has) Under the DDA approach, the structure in (507) undergoes the following steps: Step 1. Kak is base-generated in [Spec;CP], which is endowed with [+wh] features. Čto is base generated in VP as a complement to kupil. Čto moves from its base position to [Spec;CP] and forms a chain with scope marker in matrix clause, thus satisfying WhCriterion. [CP[+wh] Kak ty dumaeš‘, [CP[-wh] čtoi on kupil ti]? how you think what he bought 220 Step 2. At LF, čto moves to matrix [Spec;CP], replacing semantically empty kak, and as a result, we receive the following structure at LF: [CP[+wh] Čto ty dumaeš‘, [CP[-wh] ti on kupil ti]? what you think he bought This gives us the interpretation similar to the one of the corresponding long-distance question. Although DDA seems to explain relationship between the wh-scope marker and true wh-element, there are a few arguments against adopting this approach. One of the arguments, mentioned by Stepanov (2000) comes from property G described in 6.1., i.e. the second clause of a wh-scope marking interrogative can be a yes/no question, rather than a wh-question. Below I repeat the Russian example in (509). 509. Kak vy sčitaete, budet li how you think will zavtra doţd‘? (Russian, Stepanov, 2000:8) part tomorrow rain ‗What do you think about whether it will rain tomorrow?‘ Possible answers: ‗yes, it will rain‘ or ‗no, it won‘t rain‘ This implies that there is no wh-element in the second clause which can replace the expletive scope marker at LF. Furthermore, possible answers for the question in (509) suggest that the proposition in the second clause becomes the focus of the question. The second problem for McDaniel‘s (1989) proposal, according to Stepanov (2000), arises from the non-replacement of the wh-scope marker in yes/no questions, leading to a violation of the principle of Full Interpretation, which does not allow expletive elements to enter LF (Chomsky, 1986). Adopting Chomsky‘s (1995) minimalist framework, one would assume featuredriven movement here, which leads us to distinguishing yet another problem for the DDA (Stepanov, 2000), there is no reason for the true wh-phrase to raise at LF in order to replace the wh-expletive. Finally, this approach fails to explain the reason for a true wh-element to move to [-wh] position, assigned for it in the lower [Spec;CP]. 221 Based on the discussion above, it is clear that DDA cannot be used in order to account for Russian wh-scope marking. Another approach that is worth our attention is the Indirect Dependency approach proposed originally by Dayal (1994). 6.2.2. The Indirect Dependency approach The Indirect Dependency approach (henceforth IDA) was proposed in order to account for properties observed in Hindi (wh-in-situ language), i.e. the wh-scope marker occurs in argument position of the matrix verb, where it is base-generated. 510. jaun kyaa soctaa hai meri kis-se john what thinks baat karegii? (Dayal, 1994:138) Mary who-with will talk ‗Who does John think Mary will talk to?‘ The main idea of the IDA is that the scope marker is a true wh-element which is generated in argument position and is co-indexed with the complement CP, which is adjoined at the IP level. Dayal (1994, 2000) proposes the following structure for IDA: IP 511. CPi CP IP VP VP NP NP jaun (John) NPi IP V kyaa soctaa hai (what) (thinks) meri (Mary) NP kis-sej (who with) 222 V baat karegii (will-talk) Finally, at LF, the wh kyaa and kis-se move to [Spec;CP]. 512. CP CP CPi CP IP Spec kyaai (what) VP NP jaun (John) NPi ti Spec kis-sej (who with) VP NP meri (Mary) V soctaa hai (thinks) IP NP tj V baat karegii (will-talk) In Dayal‘s approach there is no direct dependency between the wh-expressions. Instead, they form two local dependencies, indirectly connected by co-indexation of the trace of wh1 with the CP that dominates wh2 and the effect of long-distance dependency arising from their co-indexation (Dayal, 2000:161). Dayal (2000:162) defines the semantics for the Hindi example that would allow answers specifying values for the embedded wh-question. She suggests that the wh-word in the matrix clause is an ordinary wh-expression used to question over propositions and the adjoined complement is a question over individuals. The crucial step in the interpretation of scope marking is in defining the semantics of co-indexation between the matrix preverbal position and the adjoined complement. A wh-expression is interpreted here as an existential quantifier. She suggests that the matrix question should only let in those propositions that also belong in the denotation of the complement. Given that all natural language quantification is restricted overtly or covertly, this can be achieved by treating the complement as a restrictor of the matrix wh (Dayal, 2000:162).The main idea of the semantic part of this approach is that the embedded clause in (510) meri kis-se baat karegii - ‗Mary who-with will talk‘ must provide a restriction on the matrix essential quantifier in the form of the wh-scope marker kyaa (Stepanov, 2000:26).46 46 See Dayal (2000) for a detailed semantic analysis. 223 Stepanov (2000) points out that the Indirect Dependency approach deals with the problems faced by the Direct Dependency Approach. First of all, the IDA allows the lower clause to be a yes/no question, since it is an interrogative structure itself. Secondly, IDA solves a potential problem associated with LF movement of the true wh-element to the [Spec;CP] of matrix clause, inasmuch as this movement does not take place in IDA. Finally, the IDA deals with this problem for languages with overt movement - something which the DDA fails to explain. Recall that the DDA does not explain why a wh-element in the second clause moves to [-wh] position in an embedded [Spec;CP]. Under the IDA the explanation is straightforward: the embedded clause is a question by itself, which means that a wh-element moves to the left periphery of the clause where it checks both wh and interrogative Q features. The Indirect Dependency approach seems to work perfectly for German and Hindi, and, moreover, it can explain the properties described in section 5.1. for Slavic languages, Russian in particular, which the DDA fails to account for. There is one property of the Slavic wh-scope marker that presents a problem for Dayal‘s analysis, namely lexical makeup of the wh-scope marker. Recall that Dayal‘s approach was origionally developed in order to account properties observed in Hindi, and then later it was extended to German. Both German and Hindi utilize a wh-word meaning what as a wh-scope marker (kyaa for Hindi and was for German). Based on this similarity, Dayal proposes that the scope marker is base generated in the argument position and co-indexed with the complement CP (Dayal, 1994). I repeat here the structure for easy reference: 513. jaun kyaa soctaa hai meri kis-se john what thinks baat karegii? Mary who-with will talk ‗Who does John think Mary will talk to?‘ 224 (Dayal, 1994:138) IP CPi IP CP IP VP VP NP NP jaun meri (John) NPi V kyaa (Mary) soctaa hai (what) NP kis-sej (who with) (thinks) V baat karegii (will-talk) This proposal works fine for languages utilizing a wh-scope with a meaning what, while it poses problems for Slavic languages, using an adjunct wh-element meaning how as a scope marker. 514. Kak ty dumaeš‘, kogda oni nam pozvonjat? how you think when they to-u will-call-they ‗When do you think they will call us? Note that kak cannot be substituted by čto (meaning what) as a scope marker; the sentences will be ungrammatical (515). Interestingly, if we use čego/čë discussed in section 4.2., instead of čto, the sentence appears to be less degraded. 515. a.*Čto ty dumaeš‘, kogda oni nam pozvonjat? what you think when they us-to call-they b.?Čego/čë ty dumaeš‘, kogda oni nam pozvonjat? what you think when they us-to call-they The lesser degradation of (515b) can be explained by the nature of čego/čë, being interpreted as why, rather than what and belonging to the adjunct wh-phrase group in its meaning, the same as kak. 225 Dayal‘s syntactic analysis fails to account for Russian data, given that the whadjunct kak cannot be generated in the argument position, Dayal‘s proposal requires some modification to accommodate a Slavic wh-scope marker. Stepanov (2000) considers the meaning of the wh-scope marker in Slavic to be a problem for the semantic part of Dayal‘s analysis. Recall that her proposal is about taking the wh-scope marker to be an existential quantifier over sets of propositions. Wh-scope marker in Hindi happens to be the same wh-word as the wh-word for what in questions that inquire about propositions as in (516). 516. Jaun kyaa socta hai (Dayal, 2000:162) John what think-PR ‗What does John think?‘ The Russian counterpart of Hindi example in (516) also employs a wh-word meaning what as in (517). 517. Čto (ob etom) dumaet Ivan? what about this think (Stepanov, 2000:32) Ivan ‗What does John think (about that)?‘ Example (517) shows that čto in Russian is well capable of quantifying over propositions, just like kyaa in Hindi (516). However, as it has been shown in example (515), the substitution of kak with čto leads to ungrammaticality. Therefore, even if we assume that čto in Russian is an existential quantifier over sets of propositions, it still cannot function as a scope marker (Stepanov, 2000). In order to find out if kak can function as an existential quantifier over sets of propositions, Stepanov replaces čto in (517) with kak. 518. *Kak (ob etom) dumaet Ivan? how about this (Stepanov, 2000:33) think Ivan Intended: ‗What does John think (about that)?‘ 226 Comparison of (517) and (518) shows that it is not possible to have kak in the same context as čto. The only reading (518) obtains here is where kak functions as a regular adverbial questioning the manner of John‘s thinking. The data presented in above provide evidence against Dayal‘s predicted parallelism between the semantic function of the wh-phrase as an existential quantifier over sets of propositions and that of a wh-scope marker (Stepanov, 2000:33). Russian examples (517) and (518) illustrate the differences in the distribution of the wh-scope marker and existential quantifier over propositions, i.e. the wh-word kak, meaning how, that participates in wh-scope marking, is not capable of quantifying over propositions, whereas wh-word čto, meaning what, capable of quantifying over propositions, resist being a scope marker. This leads Stepanov (2000) to the conclusion that the distribution of the two whelements is not parallel in Dayal‘s sense. Stepanov does not consider the above discussion to be an argument against the IDA; however, Dayal‘s approach can still be correct. One can assume that kak in wh-scope marking questions may function as an existential quantifier over propositions regardless of the fact that it cannot be used in Russian counterpart to What does John think? To summarize, the Indirect Dependency approach suggests that both wh-scope marker and wh-element in the embedded clause are true wh-elements; the scope marker is generated in argument position and is co-indexed with the complement CP. There is no direct dependency between wh-expressions, but instead, they form two local dependencies, i.e. both clauses are questions by themselves and movement of wh-elements to [Spec;CP] is feature-driven. IDA allows the lower clause to be a yes/no question and does not suggest movement at LF of the wh-element in an embedded clause to the matrix [Spec;CP] for replacement of the wh-scope marker. In order to allow answers specifying values of the embedded wh, Dayal suggests that the second clause in a wh-scope marking question functions as a semantic restriction on the existentially quantifying wh-scope marker. Dayal‘s approach meets some challenges when applying IDA to Russian data, due to the lexical make-up of the wh-scope quantifier which is different from other languages. In the next section, I suggest an alternative analysis for wh-scope marking constructions. 227 6.2.3. The Mixed Dependency Approach Mahajan (1996) develops the Mixed Dependency Approach (henceforth MDA), according to which the higher wh-phrase is an expletive associated with the embedded interrogative clause and is indirectly linked to a true wh-element. Later this approach is refined in Fanselow and Mahajan (2000), suggesting that there is a direct link between the matrix whelements and the embedded clause (Malhotra and Chandra, 2007). Under this analysis, the wh-scope marker is a wh-correlate of a ‗sentential expletive‘ in declarative sentences, associated with finite clauses in the same regard as it in English is associated with CPs as in It is clear that John left (Stepanov, 2000:14). Mahajan (1990), studying Hindi wh-scope marking constructions, comes to the conclusion that kyaa, meaning what, is a wh-correlate of yeh, meaning it. He draws this conclusion based on the fact that both kyaa and yeh can occur in the same object position of the matrix verb (Stepanov, 2000:14): 519. a. siitee-ne kyaa socaa ki ravii-ne kis-ko dekhaa sita-erg what thought that Rivi-erg who saw ‗What did Sita think, who Ravi saw?‘ b. siitee-ne yeh jaantaa hai ki sita-erg it know ravii-ne kis-ko dekhaa that Ravi-erg who saw „Sita knows (it) who Ravi saw?‟ Mahajan shows in his analysis that at D-structure the expletive yeh forms a constituent with the clause that follows it, and this constituent is located in preverbal position as in (520) (Stepanov, 2000:14). 520. raam-ne [yeh [ki mohan-ne gaaRii Thiik kii]] socaa Ram it that Mohan car fixed 228 think Later in the derivation, the embedded clause undergoes extraposition but yeh remains in the preverbal object position (521) (Stepanov, 2000:14). 521. raam-ne yeh socaa [ki mohan-ne gaaRii Thiik kii] Ram it think that Mohan car fixed Based on the data presented above, Mahajan concludes that the wh-scope marker kyaa is a wh-correlate of expletive yeh, i.e. kyaa is a kind of sentential expletive that has [+wh] feature. In a wh-scope construction, the embedded question is a complement of the whelement kyaa, and together they form a constituent, a complex DP, which forms the object of the matrix verb, then kyaa moves to its preverbal object position, presumably for case checking, and then, at LF level, the CP replaces the wh-scope marker by pied-piping raising at LF, as illustrated in (522). 522. siitee-ne kyaa socaa ki raam-ne kis-ko dekhaa Sita-erg what thought that Raam-erg who ‗What did Sita think, who Ram saw?‘ 229 saw (Stepanov, 2000:15) CP C Spec IP VP DP sitaa-ne (Sita) V DPi DP kyaai (what) V socaa (thought) CPi C ti Spec IP C ki (that) At LF VP DP raam-ne (Ram) NP kis-ko (who) V dekhaa (saw) Replacement of the wh-scope marker by the CP at LF is reminiscent of the Direct Dependency Approach, which also treats the wh-scope marker as an expletive. Stepanov (2000) mentions that a similar analysis of wh-scope marking constructions was proposed by Herburger (1994), where the wh-scope marker is the head of a DP taking a CP complement, which is obligatory extraposed at LF level, and as a result it appears as the second clause in a wh-scope marking construction. DP is a regular whelement which undergoes wh-movement at LF. Stepanov (2000) proposes the following schematic structure of Hindi example (522) after the LF wh-movement and extraposition took place: 230 523. CP IP Spec IP CP DPi D siitee-ne ti socaa raam-ne kis-ko dekhaa (Sita-erg ti thought) (Raam-erg who saw) tCP kyaa (what) Stepanov (2000:15), inspired by the approach proposed by Mahajan (1990), Fanselow and Mahajan (1996), and Herburger (1994), proposes a similar analysis for Slavic languages, where the wh-scope marker forms a constituent with the second clause at D-Structure resulting in the following representation: 524. vy dumaete [XP kak[CP kogo you think how Ivan ljubit]]? who-acc John love Stepanov suggests that kak in Russian occupies the head position and takes the second clause as a complement. He proposes the following steps of the derivation: Step 1. Wh-element moves to [Spec;CP] in the embedded clause: [kogoi Ivan ljubit ti] who-acc John loves Step 2. CP is combined with the wh-scope marker into a constituent [XPkak + [CPkogoi how Ivan ljubit ti]] who-acc John loves 231 Step 3. Assuming that kak is a head, Stepanov suggests that kak moves overtly to the matrix Comp to check strong Q features. (Stepanov‘s suggested head movement is a problem for the Head Movement Constraint). [CPKakj [IPvy dumaete [XP tj [CP kogoi how you think Ivan ljubit ti]] ]] who-acc John loves Step 4. At LF, the entire XP that originates in the object position of the matrix verb moves to matrix [Spec;CP]. [CP [XP kak [CP kogo how ljubut Ivan]]i vy dumaete ti ] who-acc loves Ivan you think In order to explain movement at LF, Stepanov (2000) adopts the copy theory of movement (Chomsky 1995), under which a moved element leaves a copy of itself in the base generated position. The structure of Step 3, after the overt movement took place can be presented as follows (Stepanov, 2000:24): [CPKakj [IPvy dumaete [XP kakj [CP kogoi how you think how Ivan ljubit whoi]] ]] who-acc John loves who-scc Following Chomsky (1995) and Nunes (1995), Stepanov (2000) assumes that for the purposes of semantic computation the wh-scope marker behaves as if it has not moved to the matrix CP. The semantic component makes use of the lower copy of the moved whscope marker kak and does not ‗see‘ the higher copy (Stepanov, 2000:24). This means that the overt movement of kak must be driven only by formal requirement, such as ‗strength‘, in Chomsky (1995) minimalist framework. Under the copy theory of movement, the lower wh-scope marker still forms the constituent with the second clause at LF, which moves to [Spec;CP] at the level of LF. Furthermore, Stepanov utilizes Dayal‘s Indirect Dependency Approach, which suggests that the second clause in a wh-scope marking question functions 232 as a semantic restriction on the wh-scope marker, defined as an existential quantifier. In Stepanov‘s proposal, the wh-scope marker forms a constituent with the second clause, thus restricting it at the level of LF (Stepanov, 2000). To summarize, the Mixed Dependency Approach combines elements of the Direct Dependency Approach and the Indirect Dependency approach. The wh-scope marker is treated as a kind of expletive with a [+wh] feature, and the embedded clause contains a true wh-phrase. The embedded question is a complement of the wh-expletive, and together they form a constituent, a complex DP, which forms the object of the matrix verb. Later, at the level of LF, the CP moves to the matrix [Spec;CP] position where it substitutes the whexpletive (Mahajan, 1996), or the DP (wh-expletive) moves together with CP to the matrix [Spec;CP] (Herburger, 1994). Stepanov (2000) utilizes this approach and offers his analysis of Russian wh-scope marking questions. He suggests that the wh-scope marker kak forms a constituent XP with the embedded clause, its complement, and the entire XP moves at LF to [Spec;CP]. In the next section, I offer an analysis of wh-scope marking construction in Russian, which adopts one of the above-presented approaches. 6.3. Analysis of Russian wh-scope marking constructions Before analyzing the wh-scope marking construction, it is necessary to understand whether Russian wh-scope marker kak is an expletive or a true wh-element. Recall that most of the languages we came across utilize a wh-scope marker meaning what, except Russian and Polish which use kak, meaning how. Moreover, it has been shown in 6.2.2. that Russian kak cannot quantify over propositions, in contrast to the wh-scope marker meaning what in other languages. 233 6.3.1. The status of the Slavic wh-scope marker As it has been seen from the discussion above, Russian uses kak for extending scope and it seems to be the minimal specification of a wh-phrase, just as do is the minimal specification of the verbal head (Mycock, 2004). 525. Kak ty dumaeš‘, kogda oni pročitajut how you think stat‘ju? when they will-read-through article ‗When do you think they will finish the article?‘ If Russian kak is an expletive, it should display properties showing lack of semantic content. Mycock (2004) provides evidence for the wh-scope marker what (crosslinguistically) being an expletive by drawing a parallel between English the expletive it and the wh-scope marker what. Let us go though some properties of the expletive it: A. The expletive pronoun it has no Ɵ-role and is simply a ‗place-filler‘ for the subject position (Danon, 2010). Recall that according to the Direct Dependency Approach, a whscope marker is referred to as a ‗place holder‘ for a true wh-element. B. The expletive pronoun it cannot be replaced by a wh-word (526) (Danon, 2010). Wh-scope marker cannot be replaced by another wh-word; however, only a wh-word meaning kak, for Russian, is licensed in this kind of construction (čego can appear in this type of constructions as in (515), but the sentence is degraded). 526. a. It surprised Max that Matilda won the elections. b. *What surprised Max that Matilda won the elections? 234 527. Kak/*čto/ *kto ty dumaeš‘, kogda oni pridut? How/what/who you think 528. (Russian) when they will come Was/*wann glaubst du, wann sie gekommen ist? what/when think you when she come (German) is C. Both expletive and wh-scope marker can appear in intervening clauses, forming unbounded dependencies; this is also true for Russian kak. 529. Kak ty dumaeš‘, kak Ivan sčitaet, kogo how you think vybirut prezidentom? (Russian) how Ivan believes who-acc will elect president ‗Who do you think that John think will be elected as the president?‘ 530. Raam-ne kyaa socaa, ki ravii-ne kyaa kaha, ki son sa aadmii aayaa thaa? Ram-erg what think-past that ravi-erg what say-past that which man ‗Which man did Rami think that Ravi said came?‘ came be-past (Hindi, Mahajan, 2000:322) D. Expletives cannot be stressed; the same is true for the Russian wh-scope marker kak. Mycock (2004) points out that the same property is found in Hungarian. In Hindi, on the other hand, wh-scope marker kyaa can be stressed only when the true wh-element in the embedded clause is stressed as well. E. Expletives cannot be passivised. According to Mycock (2004), this characteristic is true for Hindi. It is inapplicable to Russian, since the wh-scope marker here is an adjunct whelement kak, which cannot be passivised. We have shown that a wh-scope marker has some common propoerties with an expletive pronoun. Now how does this differ from a true wh-element? It is assumed that a true wh-element is an interrogative quantifier. Horvath (1997) notes that ‗nothing‘ can be an answer to a wh-fronting construction; an answer not possible for wh-expletive questions. The same is true for Hungarian (531), Hindi (Mycock, 2004) and Russian (532): 235 531. a. Mit mondott Janos, hogy ki-vel tancolt? (Horvath, 2000:301) what-acc say-past.3sg John –nom that who-with dance-past.3sg ‗With whom did John say that he had danced?‘ b.*Emmit nem momdott, hogy ki-vel tancolt. nothing.acc not say.past.3sg that who-with dance.past.3sg ‗He did not say anything with whom he had danced.‘ 532. a. Kak ty dumaeš‘, s how you think kem vstretilsja Mark? with whom met (Russian) Mark ‗Who do you think Mark met with? b.*Nikak ne dumaju s nothing not think kem Mark vstretilsja. with whom Mark met According to Mycock (2004), the data presented above are inconsistent with Dayal‘s proposal that a wh-scope marker is a quantifier over propositions. The same has been shown by Stepanov (2000) and outlined in 6.2.2. Therefore, it seems that a wh-scope marker is not a propositional variable, but an expletive in the constructions under discussion. Furthermore, we come across interesting examples in (533) for Russian and (534) for Mandarin Chinese, where the wh-scope marker appears to be optional: 533. a.Kak ty dumaeš‘, kogo ja videla? how you think (Gelderen, 2001:90) whom I saw ‗Who do you think I saw?‘ b.Ty dumaeš‘, kogo ja videla? you think whom I saw ‗Who do you think I saw?‘ 534. Ying yiwei Min buy-asp what? (Mycock, 2004:373) Ying think Min buy-asp what ‘What does Ying think Min bought?‘ 236 (533b) shows that, although kak is omitted, the question still gets the interpretation similar to the one of the corresponding long-distance question as in (533a). Interestingly, it seems that wh-expletive kak is also optional with yes/no questions in the second clause: 535. a. Kak ty dumaeš‘, budet li how you think zavtra doţd‘? will part tomorrow rain ‗What do you think about whether it will rain tomorrow?‘ b. ty dumaeš‘, budet li you think zavtra doţd‘? will part tomorrow rain ‗What do you think about whether it will rain tomorrow?‘ Dropping of the wh-scope marker does not lead to ungrammaticality, unlike wh-questions with true wh-elements (536). This fact also suggests that a wh-expletive is semantically empty. 536. a. Počemy ty dumaeš‘ čto eto opasno? why you think that this dangerous ‗Why do you think that this is dangerous?‘ b.*Ty dumaeš‘ čto eto opasno? you think that this dangerous Intention: ‗Why do you think that this is dangerous?‘ It is clear from the data that a true wh-element cannot be dropped without losing its meaning. The answer in (536) can be only ‗yes or no‘. It clearly follows from the discussion above that the wh-scope marker kak in Russian is not a true wh-element, contrary to the Indirect Dependency approach, but rather some kind of an expletive, which has [+wh] and Q features. The wh-expletive raises to the left periphery of the matrix clause for feature checking. From now on I refer to the whscope marker as a wh-expletive. 237 6.3.2. Syntactic structure of wh-scope marking constructions So far we have established that the Russian wh-scope marker is in fact some kind of whexpletive, which is endowed with the same features as true wh-element; it has [wh] and Q features which have to be checked by raising the wh-expletive to the left periphery of the matrix clause. It has been suggested by the Direct Dependency Approach that the wh-expletive is base generated in the matrix [Spec;CP]; however, the Indirect Dependency approach and the Mixed Dependency Approach argue an argument position of the matrix verb as a base position for the wh-scope marker. So, first we need to understand where the wh-expletive is generated. Let us recall one of the properties of a wh-scope marker discussed in 6.1.: a matrix clause containing a wh-scope marker cannot be negated or contain a negative operator. 537. *Kak vy ne dumaete, kuda on pošel? how you not think (Russian) where he went ‗Where do not you think that he went?‘ (537) is an example of a negative island effect (Rizzi, 1990; 1992). Rizzi proposes that negation is a selective barrier to extraction and only referential (bearing Ɵ-roles), but not non-referential expressions, can escape this barrier. Negation can act as a potential antecedent governor for the trace of the extracted wh-phrase, which leads to a violation of Relativized Minimality. If, according to DDA, a wh-expletive was located in [Spec;CP], example (537) would have been grammatical. Based on this fact I dismiss the idea that a wh-expletive can be base generated in [Spec;CP]. I follow the proposal suggested in the IDA and MDA where a wh-scope marker is base generated inside vP. Recall that according to these approaches, a wh-expletive is located in an argument position of the matrix verb, which works fine for wh-scope marker meaning what, but does not quite working for Russian, which utilizes wh-adjunct kak. 238 I propose that the wh-expletive in Russian occupies a position inside vP and takes an embedded CP as a complement. I assume that wh-scope is in Spec position due to the incompatibility of a moved wh-phrase with a wh-scope marker in the matrix clause, contrary to Stepanov‘s approach (2000), which assumes Head position for the wh-element. 538. *Kak čto vy dumaete ona kupila? how what you think she bought ‗What do you think she bought? Following Stepanov (2000), I propose the following steps for the derivation of the whscope marker construction. 539. Kak vy dumaete čto how you think ona kupila? what she bought ‗What do you think she bought? Step 1. Wh-element moves to [Spec;CP] in the embedded clause: [čtoi ona kupila ti] what she bought Step 2. Wh-expletive takes embedded CP as a complement and they form a constituent [Kak [čtoi ona kupila ti]] how what she bought Step 3. Kak moves overtly to the [Spec;CP], contrary to Stepanov (2000), where it moves to Head of CP [CPKakj [IPvy dumaete [XP tj [CP čtoi ona kupila ti]]]] how you think what she bought 239 Now, recall that under the Cartographic Approach (Rizzi, 1997), the left periphery of the clause is seen as a structural zone defined by a system of functional heads and their projection. Thus, Rizzi distinguishes Force, Topic, Focus, (Wh – for embedded clauses), and Fin within the C system. If in Italian, Wh projection appears only in an embedded clause, I argue that in Russian it can also appear in a matrix clause; subsequently, Wh becomes a part of C system in Russian wh-questions. Interestingly, this is not a case in Russian wh-scope marking constructions. Consider the example below: 540. *Kak ty a ne ja dumaju, čto ona kupila? How you and not I think what she bought The data in (540) appear to demonstrate that a wh-expletive and a focus cannot co-occur in the same structure, suggesting that they compete for the same position, i.e. Focus. Mycock (2004) also suggests that the wh-expletive is focused, drawing this conclusion based on the data coming from wh-in-situ languages, in which the position of the wh-expletive appears consistent with it being a focused element in the matrix clause. She proposes to analyze the wh-expletive as bearing the discourse function FOCUS47. Mycock (2004:383) takes up the notion that discourse functions are integrated into the meaning of a sentence according to the Extended Coherence Condition, which states that Focus and Topic must be linked to the semantic predicate argument structure of the sentence in which they occur, either by functionally or by anaphorically binding an argument. (Dalrymple, 2001:390) Therefore, following from the discussion above, the wh-expletive must be linked to an element bearing a grammatical function, and according to the analysis outlined, the whexpletive is linked to the embedded clause, bearing the grammatical function COMP 47 Mycock (2004) provides an analysis of the wh-scope marking construction in the non-derivational framework of Lexical Functional Grammar. However, I will not discuss it here, but rather only outline the main idea of Mycock‘s analysis. 240 (Mycock, 2004:384). Finally, she suggests that an embedded clause functions as a controller in a wh-expletive construction. This idea is similar to the one outlined in the Indirect Dependency and Mixed Dependency Approaches, where an embedded complement is treated as a sort of restrictor of a matrix wh. Returning to the discussion of the position of a wh-expletive in the C domain, I suggest that kak moves to a [Spec;Foc] position (Recall that Rizzi (1997;1999) suggests [Spec;Foc] as a landing site for wh-elements), where it checks relevant features, whereas a true wh-element in the second clause, moves to a designated WhP in the left periphery of the embedded clause. Following from the above discussions I assume the syntactic structure of a whscope marking construction as in (541): 541. Kak ty dumaeš‘ kogo ona priglasila? how you think whom she invited ‗Who do you think she invited?‘ ForceP FocP FinP IP kaki (how) vP DP ty (you) WhP V dumaeš‘ (think) ForceP wh WhP ti FinP wh IP kogoj vP (whom) DP ona (she) V priglasila (invited) 241 Wh-ObjP tj In this view, the wh-expletive kak is base generated in [Spec;WhP] within vP and it takes an embedded complementizer as a complement, forming with it a constituent (a la Stepanov (2000), Mahajan (1996), etc.) . Later in the derivation, the wh-expletive moves to [Spec;FocP], where it checks the relevant features. In order to account for the semantic part of the wh-scope marking construction, I follow Stepanov (2000), who proposes that by forming a constituent with the wh-expletive, the second clause restricts the wh-scope marker at the level of LF, thus yielding the interpretation of a long distance movement question. I support Stepanov‘s (2000) utilization of the copy theory of movement proposed by Chomsky (1995) where he suggests that for the purposes of semantic computation the wh-scope marker behaves as if it has not moved to the matrix CP. Under the copy theory of movement the wh-scope marker still forms a constituent with the embedded clause at LF and moves together by pied-piping to the left periphery. Furthermore, I propose that this constituent moves to [Spec;ForceP] where it checks interrogative Q feature,48 and gets the scope over the whole construction. 48 Recall that in Stepanov‘s analysis the XP constituent moves to [Spec;CP] at LF. 242 542. ForceP FocP Spec FinP IP kaki (how) vP DP ty (you) WhP V dumaeš (think) ForceP WhP wh ti FinP IP wh kogoj vP (whom) At LF DP ona (she) V priglasila Wh-ObjP tj (invited) saw Now, recall the discussion in 5.3., which addresses examples without an overt complementizer as in (543). 543. a. Kto ty dumaeš‘ kupil košku? who you think bought cat ‗Who do you think bought a cat?‘ One of the solutions, offered by Rizzi and Shlonsky (2007), is truncation of Force. Partial wh-movement seemed to be a problem for this solution. According to their analysis absence of an overt complementizer in (544) suggests truncation of Force, which means that the whelement kogo – ‗whom‘ does not have a position to move for checking the Q feature. 243 544. Kak ty dumaeš‘ kogo ona priglasila? how you think whom she invited ‗Who do you think she invited?‘ Based on the analysis proposed in this section, and assuming Rizzi and Shlonsky‘s Force truncation in the clauses without an overt complementizer, the structure of (544) is presented in (545), where at LF the whole constituent moves to [Spec,ForceP] of the matrix clause where the Q feature is checked. 545. ForceP FocP Spec FinP IP kaki (how) vP DP ty (you) V dumaeš (think) truncated WhP ForceP WhP wh ti FinP IP wh kogoj At LF vP (whom) DP ona (she) V priglasila Wh-ObjP tj (invited) saw 244 Thus, Rizzi and Shlonsky (2007) suggestion of Force truncation in clauses without an overt complementizer works for Russian as well. To summarize, this chapter has outlined properties of Slavic wh-scope marking questions and summarized previous analyses offered in order to account for the phenomenon of a whscope marking construction. According to the Direct Dependency Approach, a wh-scope marker is semantically empty and treated as some kind of expletive, which functions as a place holder for a true wh-element in the embedded clause, which moves to replace the wh-expletive at the level of LF. The wh-expletive is base generated in [Spec,CP] and endowed with a [+wh] feature. This approach suggests a direct link between the wh-expletive and true wh-phrase in the embedded clause. A few problems arising from this analysis allowed us to dismiss this approach and turn to the other two. Under the Indirect Dependency Approach, there is no direct link between the whexpletive and the wh-word; rather there is a link between the wh-scope marker and the embedded question. A wh-scope marker is a true wh-element which is base-generated in an argument position within VP and restricted by the embedded clause. This approach assumes that a wh-scope construction consists of two local wh-dependencies, which are coindexed, and the effect of long-distance dependency arises from their co-indexation. The Mixed Dependency Approach combines elements of the DDA and IDA, where the wh-scope is an expletive, forming a constituent with the embedded clause. In order to account for Russian wh-scope marking constructions, I offer an analysis based on the Mixed Dependency Approach. After studying the properties of Russian whscope marker, I come to the conclusion that Russian kak is an expletive endowed with a [wh] feature. Kak is base generated in vP: it takes the embedded clause with a true whphrase as a complement and they together form a constituent. Kak moves to [Spec;FocP] where it checks relevant features. By forming a constituent with the wh-expletive, the second clause restricts the wh-scope marker at the level of LF, thus yielding the interpretation of a long distance movement question. 245 246 CHAPTER 7. CONCLUDIING REMARKS This dissertation has offered a comparative analysis of Russian wh-questions with other languages involving multiple wh-movement, and has argued against the common perception of Russian as a language where a wh-phrase does not undergo wh-movement or show any Superiority effects. The data and analyses presented in this work argue for the opposite. Before discussing syntactic properties of multiple wh-movement in Russian, I have devoted the Introduction to outlining the ideas that underlie the present dissertation. I have offered a brief discussion of the Cartographic approach, Relativized Minimality and Criterial freezing and showed their development in the literature. In Chapter 2, I have presented traditional views on multiple wh-fronting languages, Russian in particular. Russian has been argued by Stepanov (1997) to pattern with wh-insitu languages in the absence of an overt wh-movement. Furthermore, it has been claimed by Stepanov (1997) and Bošković (2002) that the reason for fronting wh-elements in Russian is purely focus checking. Before going into a detailed discussion of properties of wh-movement in Russian, I have discussed the notion of Superiority effect, Empty category Principle, single pair and pair list readings, Weak Crossover and D-linking. Based on the information available in the literature, I have outlined properties of Bulgarian and SerbCroatian wh-questions, which, later in the chapter, have been compared to the properties of Russian wh-questions. The initial information gathered from the literature has indicated that Russian does not allow single pair readings, which means that overt syntactic wh-movement takes place. It has been also pointed out by Rudin (1996), Zavitnevich (2001) and Stepanov (1997) that Russian shows a preference for a particular wh-word order in some multiple questions, but 247 no further research has been done until now. The information collected contradicts previous assumptions made on Russian which has supported our intention to study wh-movement in Russian. The facts discussed in Chapter 2 have encouraged us to consider the possibility for Russian to be a language exhibiting Superiority effects and to study Russian by comparing it to Bulgarian, the language displaying a rigid word order. In Chapter 3, I have followed Krapova and Cinque‘s (2005) detailed analysis of the Bulgarian left periphery, where they distinguish a specific ordering of wh-phrases which reflects the order prior to whmovement. The data collected have revealed that the order of Russian wh-phrases is very similar to the one in Bulgarian, which shows that [+human] wh-elements precede [-human], subjects precede objects and adjuncts, and the order of wh-elements in embedded clauses is much stricter and more differentiated. I have also presented Meyer‘s (2004) study of Superiority effects in Russian, which supports a few conclusions I have drawn out of my data. Having the data at hand has allowed me to proceed with a syntactic analysis of whquestions. I have discussed four main analyses of multiple wh-questions mentioned in the literature, i.e. the adjunction analysis (Rudin, 1988), the wh-cluster analysis (Grewendorf, 2001), the multiple [Spec;CP] analysis (Pesetsky, 2000, 2002; Richards, 1997), and split CP analysis (Rizzi, 1997; Grohman, 1998; Laenzlinger and Soare, 2005, etc.). Discussing Bulgarian data, I have presented Krapova and Cinque‘s (2005) syntactic analysis of Bulgarian wh-questions, where they adopt Rizzi‘s (1997) Cartographic approach, and suggest that the order of wh-elements in Bulgarian reflects the order prior to wh-movement within the IP structure (Mittelfeild). The Superiority effect in their analysis is captured under a modified version of Relativized Minimality, where links of a chain do not count as interveners, but the whole chain. Following Krapova and Cinque‘s (2005) lead, I have looked at the word order in Russian and compared it to the order of wh-elements distinguished in this chapter. These data have allowed me to argue that the order in the IP structure reflects the order after the movement, which supports an analysis of Russian wh-questions according to the Cartographic approach. 248 Finally, the Chapter 3 concludes with a syntactic analysis of Russian wh-questions. Following Rizzi‘s (1997) approach, I have proposed to extend the C system and distinguish additional projections, i.e. ForceP, FocP, TopP, FinP and a separate projection WhP for whelements. Following Krapova and Cinque (2005), I have assumed that the order of whelements is realized in the IP (Mittelfeld) prior to movement and is preserved under the notion of Relativized Minimality. In order to account for a not-really-free order of whelements in wh-matrix questions, I have proposed that the order is ‗ruled‘ by internal makeup, syntactic function of wh-elements and prominence of information, based on which order is realized in Mittelfeld, preserved under the notion of Relativized Minimality and emerges in CP. In Chapter 4, I drew my attention to a peculiar group of wh-phrases in Russian which I referred to as ‗why-wh-elements‘. First, I have provided an analysis for two nonhomonymous lexical items počemu and začem which distinguished purpose from reason varieties of why and have shown that they demonstrate some syntactic differences, leading to the conclusion that the two whys occupy different positions in the syntactic derivation. Počemu is argued to be externally –merged in the functional projection CauseP, which is located in the CP domain; while začem is base-generated in VP, like other wh-elements, and moves to the left periphery to check wh-features. Next, I focused on aggressively non-D-linked wh-phrases of kakogo čërta – „why the hell‘ type and nominal wh-elements čto and čego meaning ‗why.‘ I have shown that these types of wh-phrases are different from počemu and začem, i.e. they cannot occur in situ or with verdical predicates. I have discussed Obenauer and Poletto‘s (2000) analysis of Italian rhetorical wh-questions (questions with aggressively non-D-linked wh-elements) where they claim that the position of a wh-element in such questions is higher than in true wh-questions. Russian data did not support their claim, since in Russian a Topic cannot be higher than a wh-element. Furthermore, the data presented in 4.2.5 have revealed that aggressively non-D-linked wh-elements follow ‗true‘ wh-elements in coordinated multiple wh-questions. While discussing wh-questions with aggressively non-D-linked wh-elements, I have come across examples with coordinated wh-phrases, which have required attention and have been addressed in 4.3. I have considered the analysis proposed by Grimshaw (1978), 249 where she considers coordination prior to the wh-movement, and the analysis offered by Kazenin (2002) and Camacho (2003), where coordination is treated as an instance of two CPs coordination followed by sluicing, and I have dismissed both of them. Gribanova‘s (2009) proposal of coordination in CP involving quantifier absorption analysis seemed to look most plausible, but it did not account for the Superiority effect and pair-list reading emerging in coordination of more than two wh-elements. Finally, following Merchant (2008), I have come to the conclusion that Russian coordinator ‗i‟ found between whphrases in Russian is a spurious coordinator, which is different from a canonical coordinator. I have argued that wh-elements are coordinated above IP after they have moved to their positions in the left periphery; later they move to a [Spec;FocP] where the spurious ‗i‘ is realised as a discourse marker bearing some focus feature. Wh-elements form a cluster and move to ForceP, to check a [Q] feature. Chapter 5 has addressed the phenomenon of that-trace effect in Russian indicative and subjunctive clauses. In this chapter, we have discussed major existing approaches, such as NIC approach (Pesetsky, 1982), ECP account (Rizzi, 1990), Richards‘ account (1999), Szczegielniak‘s proposal (1999) and Rizzi and Shlonsky‘s proposal (2006). The data have shown that neither indicative nor subjunctive clauses exhibit Subject/Object asymmetry and that while extraction of either object or subject out of an indicative embedded clause is impossible, extraction out of a subjunctive clause does not cause any problems. Discussing subjunctive complementizers in 5.2, I have studied the phenomenon of obviation effect and outlined three analyses devoted to find an explanation to this phenomenon, namely the Avoid Pronoun Principe (Chomsky, 1981), the Domain Extension Approach (Picallo, 1984; Rizzi, 1989; Progovac, 1993), and Avrutin and Babyonyshev‘s analysis (1997). In 5.3., I have argued that Russian has only one complementizer, the indicative complementizer čto, which merges as Fin and moves further to Force. The subjunctive complementizer čtoby, on its part, is not a true morphological complementizer; it is derived by means of syntactic movement, i.e. the indicative complementizer čto moves through Mood, merges with a particle by and moves further to Force. The impossibility of the subject/object extraction out of the indicative embedded clause has brought us to the conclusion that whenever Head of Force is occupied by that, 250 [Spec;ForceP] becomes inaccessible to a wh-element, thus there is no escape hatch from the embedded phase and further movement is impossible. This does not happen in a subjunctive clause, where, presumably, the fact that čto moves through MoodP and gets a [+Mood] feature allows the complementizer to co-occur with a wh-element. This Chapter has also addressed the phenomena of Criterial Freezing and Subject Criterion. I follow Rizzi and Shlonsky (2007) in suggesting that the projection Fin is endowed with the relevant nominal features and unvalued Phi features which can satisfy the Subject Criterion, allowing the wh-subject to ‗skip‘ the [Spec;Subj] position and avoid the Criterial Freezing. Chapter 6 has been devoted to wh-scope marking constructions in Russian. This chapter has offered a comparative analysis of properties of wh-scope marker in Russian and other wh-scope marking languages, such as German, Polish and Hindi. The data have revealed that a wh-scope marker in Russian and Polish constructions is a wh-element meaning how, contrary to other wh-scope marking languages, where a wh-scope marker is a wh-element meaning what. The discussion of why Slavic languages employ a wh-scope marker meaning why and not what as in other languages is left for further research. I have summarized previous analyses of wh-scope marking constructions discussed in the literature, namely the Direct Dependency Approach (McDaniel, 1989), the Indirect Dependency Approach (Dayal, 1994) and the Mixed Dependency Approach (Mahajan, 1996). I have followed Mycock (2004) in defining the status of Russian wh-scope marker kak and have come to the conclusion that kak is an expletive endowed with a [wh] feature and base generated in vP; it takes the embedded clause with a true wh-phrase as a complement and they together form a constituent. Kak moves to [Spec;FocP] where it checks relevant features. This dissertation has presented another perspective on the syntactic structure of whquestions in Russian, arguing against the traditional view on wh-movement. The data collected have shown that wh-questions in Russian do involve wh-movement, they do exhibit Superiority effects and they have a much more complicated syntactic structure than it has been previously assumed. This dissertation did not address any semantic issues of wh-elements. It would been of special interest to study the semantic properties of aggressively non-D-linked wh251 elements discussed in Chapter 4 and wh-scope marking constructions discussed in Chapter 6. Also, pursuing the Cartographic approach, it would be of great value to study the word order in Russian in more detail. I leave these issues for further research. 252 BIBLIOGRAPHY Alboiu, G. 2002. The Features of Movement in Romanian. University of Bucharest Press, Bucharest. Avrutin, S. and M. Babyonyshev. 1997. Obviation in Subjunctive Clauses and AGR: Evidence from Russian, Natural Language and Linguistic Theory, 15:229-262. Bailyn, John Frederick. 2004. Generalized Inversion. Natural Language & Linguistic Theory, 22:1–49. Kluwer Academic Publishers. Barrs, Andrew. 2000. Minimalism and asymmetric wh-interpretation. In Step by step: Essays on minimalist syntax in honor of Haward Lasnik, ed. by Roger Martin, David Michaels and Julian Urigereka, 31-52. Cambridge, MA: MIT Press. Bayer, Joself. 1984. COMP in Bavarian Syntax. Linguistic Review, 3:209-274. Belletti, Adriana. 2004. Aspects of the low IP area. In L. Rizzi (ed.) The Structure of IP and CP. The Cartography of Syntactic Structures. vol. 2, OUP. Beletti, Adriana and Luigi Rizzi. 1988. Psych-verbs and q-theory. Natural Language and Linguistic Theory, 6:291-352. Billings, Loren and Catherine Rudin. 1996. Optimality and Superiority: A new approach to overt multiple-wh ordering. Formal Approaches to Slavic Linguistics: the College Park Meeting, J. Toman (ed.), 35-60. Michigan Slavic Publications, Ann Arbor. Borer, Hagit. 1984. Restrictive relatives in Modern Hebrew. Natural Language and Linguistic Theory, 2:219-260. Bošković, Ţeljko. 1997. Sometimes in [Spec, CP], sometimes in-situ. Ms. University of Connecticut, Storrs, CT. Bošković, Ţeljko. 1998. Wh-Phrases and Wh-movement in Slavic. Position paper. Comparative Slavic Morphosyntax. 253 Bošković, Ţeljko. 1998a. Multiple wh-fronting and economy of derivation. In Proceedings of the WestCoast Conference on Formal Linguistics, 16:49-63. Stanford University, Stanford, CA. Boškovic, Ţeljko. 2000. Sometimes in [Spec, CP], sometimes in situ. In Step by Step: Essays On Minimalism in Honor of Howard Lasnik, R. Martin, D. Michaels and J. Uriagereka (eds), 53-87. Cambridge, Mass.: MIT Press. Bošković, Ţeljko. 2001. On the Interpretation of Multiple Questions. Linguistic Variation Yearbook. John Benjamin‘s, Amsterdam. Bošković, Ţeljko. 2002. On multiple wh-fronting. Linguistic inquiry, 33/3:351-383. Brecht, Richard D. 1974. Deixis in Embedded Structures. Foundations of Language, 11:489-518. Brecht, Richard D. 1977. Čtoby or čto and by. Folia Slavica, 1:33-41. Camacho, José. 2003. The structure of coordination: Conjunction and agreement phenomenon in Spanish and other languages. Dordrecht: Kluwer. Chaves, Rui P. and Denis Paperno. 2007. On the Russian Hybrid Coordination Construction. CSLI Publications. Chomsky, Noam. 1973. Conditions on transformations. In A Festschrift for Morris Halle, ed. Stephen Anderson and Paul Kiparsky, 232-286. New York: Holt, Rinehart and Winston. Chomsky, Noam. 1980. On binding. Linguistic Inquiry, 11:1-46. Chomsky, Noam. 1981. Lectures on Government and Binding. Foris. Chomsky, Noam. 1986. Knowledge of language. New York: Praeger. Chomsky, Noam. 1995. The Minimalist Programme. MIT Press Chomsky, Noam. 1999. Derivation by Phase. MIT Occasional Papers in Linguistics. Chomsky, Noam. 2000. Minimalist inquiries. In Martin, R., Michaels, D. and Uriagereka, J. (eds.). Step by step: Minimalist syntax in honor of Howard Lasnik, 3:89-155. Cambridge, MIT Press. Chomsky, Noam. 2001. Derivation by Phase, in Ken Hale: A life in Language, M. Kenstowicz (ed.), 1-52. Cambridge, Mass: MIT Press. Chomsky, Noam and Howard Lasnik. 1977. Filters and Control. Linguistic Inquiry, 8:425– 504. 254 Cinque, Guglielmo. 1990. Types of A‟ Dependencies. MIT Press, Cambridge, Mass. Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic perspective. New York, Oxford University Press. Cinque, Guglielmo. 2002. Functional Structure in the IP and DP. In The Cartography of Syntactic Structures,vol.1. Oxford University Press, New York. Cinque, Guglielmo. 2002a. Complement and Adverbial PPs: Implications for Clause Structure, talk given at the 25th Glow Colloquium, Amsterdam, April 9-11. Cinque, Guglielmo. 2006. Restructuring and Functional Heads. The Cartography of Syntactic Structures, vol.4, Oxford University Press, New York. Citko, Barbara. 2008. Wh-questions with conjoined wh-pronouns. Handout. Berkeley Linguistics Society, February 2008, University of California, Berkeley. Cole, Peter and Gabriella Hermon. 2000. Partial wh-movement: evidence from Malay. In Wh-Scope Marking, Lutz Uli, Gereon Müller &Arnim von Stechow (eds). 101-130. Collins, Chris. 1990. Why and how come, MIT Working Papers in Linguistics, 15:31-45, MITWPL. Comrie, Bernard. 1980. Clause Structure and Movement Constraints in Russian. In C. Chvany and R. Brecht, eds., Morphosyntax in Slavic, 98-113. Columbus, Ohio: Slavica Publishers. Cowart, Wayne. 1997. Experimental Syntax: Applying Objective Methods to Sentence Judgements. Thousand Oaks, California: Sage. Dalrymple, Mary. 2001. Lexical functional grammar, Syntax and Semantics, volume 34. Academic Press. Danon, Gabi. 2010. Expletives, raising and the ECP. Bar-Ilan University. Dayal, Veneete. 1994. Scope marking as Indirect wh-dependency. In Natural Language Semantics, 2: 137-170. Dayal, Veneete. 1996. Scope marking: In defence of Indirect Dependency. Lutz, U. & G. Müller (eds.) Papers on wh-scope marking. SFB 340, Universität Tübingen, 107-130. Dayal, Veneete. 2000. Scope marking: cross-linguistic variation in indirect dependency. In Wh-Scope Marking, Lutz Uli, Gereon Müller &Arnim von Stechow (eds), 157-193. Dayal, Veneeta. 2006. Multiple wh-questions. In M. Everaert and H. van Riemsdijk (eds.), Syntax Companion 3, Blackwell, Oxford. 255 Dikken, Marcel den and Anastasia Giannakidou. 2002. From hell to polarity: ―Aggressively non-D-linked‖ wh-phrases as polarity items. Linguistic Inquiry, 33:31–61. Dong, Zheng-Min. 1995. On phonologically null prepositions: a reply to Fowler and Yadroff. Journal of Slavic Linguistics, 3(2):378–386. Dornisch, Ewa. 1997. Auxiliaries and Functional Projections in Polish. In Formal Approaches to Slavic Linguistics (FASL) IV: The Cornell Meeting 1995, eds. Wayles Browne, Ewa Dornisch, Natasha Kondrashova and Draga Zec, 183-209. Ann Arbor: Michigan Slavic Publications. Dornish, Ewa. 1998. Multiple Wh-Questions in Polish: the Interactions between WhPhrsaes and Clitics. Ph.D. dissertation, Cornell university, Ithaca. Franselow, G, & Mahajan, A. 1996. Partial movement and successive cyclicity. In U. Lutz & G. Muller (eds.), Papers on Wh Scope Marking. Stuttgart/Tubingen: Universitat Stuggart/ Universitat Tubingen/IBM Deutschland, 131-161. Fanselow, G. 2007. Partial Wh-Movement. In M. Everaert and H. van Riemsdijk (eds.), The Blackwell Companion to Syntax. Blackwell Publishing, Malden, MA, USA. Featherston S. 2005. That-trace in German. Lingua, 115:1277-1302. Fowler, George and Michael Yadroff. 1993. The argument status of accusative measure nominals in Russian. Journal of Slavic Linguistics, 1(2):251–279. Franks, Steven. 2006. Another look at li placement in Bulgarian. The Linguistic Review, 23:161–211. Franks, Steven and Katarzyna Dziwirek. 1993. Negated adjunct phrases are really partitive. Journal of Slavic Linguistics, 1(2): 280–305. Franks, Steven and Nina Rojina. 2008. Idiosyncrasies of Russian kakogo čërta ‗why the hell‘, presented at the 3rd Slavic Linguistics Society Conference. Gazdar, Gerald, Klein, Ewann, Pullum, Geoffrey K. and Sag, Ivan A. 1985. Generalized Phrase Structure Grammar. Oxford: Blackwell, and Cambridge, Ma: Harvard University Press. Gazdik, Anna. 2010. I-structure, S-structure and Multiple Questions in French and Hungarian, presented at 15th International Lexical Functional Grammar Conference. 256 Gelderen, Veronique van. 2001. Partial wh-movement in Russian. In Ton van der Wouden & Hans Broekhuis (eds.) Linguistics in the Netherlands 2001. Amsterdam: John Benjamin, 89-100. Gracanin-Yuksek, Martina. 2010. Towards a New Typology of Coordinated Wh-Questions. Ms. Grebenyova, Lidia. 2004. On Superiority, T-to-C Movement and Interpretation of Questions. University of Maryland at College Park LING 895 General Papers. Grewendorf, Günther. 1988. Aspekte der deutschen Syntax. Tübingen: Narr. Grewendorf, Günther. 2001. Multiple wh-fronting. Linguistic Inquiry, 32:87-134. Gribanova, Vera. 2009. Structural adjacency and the typology of interrogative interpretations. Linguistic Inquiry, Volume 40, Number 1:133-154. Grimshaw, Jane. 1978. On the syntax of conjoined Wh Words in English. University of Massachusetts Occasional Papers on Linguistics, 3:1-10. Grohmann, Kleanthes K. 1998. Syntactic inquiries into discourse restrictions on multiple interrogatives. Ms. (generals paper), UMD. Appeared revised in Werner Abraham (ed.), Groninger Arbeiten zur germanistischen Linguistik 42. Rijksuniversiteit Groningen: Germanistisch Instituut, 1-60. Grohmann, Kleanthes K. 2001. Symmetries in Locality. Manuscript, Johann Wolfgang Goethe-Universität Frankfurt am Main. Haegeman, Liliane. 1992. Theory and Description in generative Syntax: A case study of West Flemish. CUP. Haegeman, Liliane. 1994. Introduction to Government and Binding Theory. Oxford: Blackwell. Hagstrom, Paul. 1998. Decomposing Questions. Doctoral Dissertation, Massachusetts Institute of Technology. Haida, Andreas & Sophie Repp. To appear. Monoclausal question word coordinations across languages. Proceedings of NELS 39. Haider, H. 1983 Connectedness effects in German. Groninger Arbeiten zur Germanischen Linguistik, 23:82-119. Harley, Heidi. 1995. Subjects, Events and Licensing. Ph.D. dissertation MIT. 257 Herburger, E. 1994. A semantic difference between full and partial wh-movement in German. Paper presented on 68th Annual meeting of the Linguistic Society of America, Boston, MA. Hiemstra, Inge. 1986. Some aspects of wh-questions in Frisian. In NOWELE (NorthWestern European Language Evolution), 8:97-110. Higginbotham, James and Robert May. 1981. Questions, quantifiers and crossing. The Linguistic Review, 1:40-81. Hornstein, Norbert. 1995. Logical form: from GB to Minimalism. Cambridge, MA: Blackwell. Horvath, Julia. 1986. Focus in the theory of grammar and the syntax of Hungarian. Dordrecht: Foris. Horvath, Julia. 1997. The status of ‗wh-expletives‘ and the partial wh-movement construction of Hungarian. Natural Language and Linguistic Theory, 15:509-572. Horvath, Julia. 2000. On the Syntax of Wh-scope Marker Construction: Some Comparative Evidence. In Uli Lutz, Gereon Müller & Arnim von Stechow (2000), 271-316. Huang, Jim. 1982. Logical Relations in Chinese and the Theory of Grammar, PhD Dissertation, MIT. Izvorski, Roumyana. 1993. On wh-movement and focus-movement in Bulgarian. Presented at CONSOLE 2, University of Tübingen. Kayne, Richard. 1984. Connectedness and Binary Branching. Foris, Dordrecht. Kayne, Richard. 1994. The Antisymmetry of Syntax. Linguistic Inquiry Monograph TwentyFive. MIT Press. Kazenin, Konstantin. 2002. On coordination of wh-phrases in Russian. Ms. University of Tübingen and Moscow State University. Kim, Jeong-Seok. 1997. Syntactic focus movement and ellipsis: A minimalist approach. Doctoral dissertation, University of Connecticut, Storrs. Kim, Sun-Woong. 2008. On the Directionality of Movement and its Relationship to Stranding. Presented at ICKL 2008. Ko, Heejeong. 2005. Syntax of Why-in-Situ: Merge into [Spec,CP] in the Overt Syntax. Natural Languages and Linguistic Theory, 23(4):867-916. 258 Krapova, Iliana. 2001. Subjunctives in Bulgarian and Modern Greek. In M. L. Rivero and A. Ralli, (eds.), 105-126. Krapova, Iliana and Guglielmo Cinque. 2005. On the Order of Wh-phrases in Bulgarian Multiple Wh-fronting. University of Venice Working papers in Lingusitics. vol. 15:171-197. Kuno, Susumu and Jane J. Robinson. 1972. Multiple WH-Questions. Linguistic Inquiry, 3: 463-487. Laenzlinger, Christopher and Gabriela Soare. 2005. Multiple wh-fronting in Romanian: a Cartographic approach. Ms. University of Geneva. Lasnik, Howard, and Mamoru Saito. 1984. On the nature of proper government. Linguistic Inquiry, 15:235-289. Lasnik, Howard, and Mamoru Saito. 1992. Move α: Conditions on its application and output. Cambridge, Mass.: MIT Press. Lipták, Aniko. 2003. Conjoined questions in Hungarian. In Multiple wh-fronting, Cedrik Boeckx and Kleanthes Grohmann, (eds.), 141-160, Amsterdam: J.Benjamins. Lubańska, Maja. 2004. Wh-scope marking in Polish. Poznań Studies in Contemporary Linguistics, 39:73-88. Mahajan, Anoop. 1990. The A/A-bar distinction and movement theory. Doctoral dissertation. MIT, Cambridge, MA. Mahajan, Anoop.1996. Wh-Expletives and the Syntax of Partial Wh-Movement. In Papers on Wh-Scope Marking, Lutz Uli and Gereon Muller (eds.), Stuttgart/Tubingen: Universitat Stuggart/ Universitat Tubingen/IBM Deutschland, 163-177. Mahajan, A. 2000. Towards a Unified Treatment of Wh-Expletives in Hindi and German. In Wh-Scope Marking, Lutz Uli, Gereon Müller &Arnim von Stechow (eds), Amsterdam: John Benjamins, 317-332. Malhotra Shiti and Pritha Chandra. 2007. A Short Note and Wh-scope-marking. In University of Maryland Working Papers in Linguistics, 16:46-62, ed. A. Omaki, I. Ortega-Santos, J. Sprouse and M. Wagers. College Park, MD: UMWPiL. McDaniel, D. 1989. Partial and multiple wh-movement. Natural Language and Linguistic Theory, 7:565-604. Merchant, Jason. 2008. Spurious coordination in Vlach multiple wh-fronting. Handout. 259 Mid-America Conference, 26-28 October 2008, University of Kansas. Meyer, Roland. 2004. Superiority effects in Russian, Polish and Czech: Judgments and Grammar. Ms. University of Leipzig/ University of Regensburg. Meyer, Roland. 2004a. Syntax der Ergänzungsfrage. Empirische Untersuchungen am Russischen, Polnischen und Tschechischen (= Slavistische Beiträge 436). München: Otto Sagner. Migdalski, Krzysztof. 2006. The Syntax of Compound Tenses in Slavic. LOT 130. Miyagawa, Shigeru. 1993. LF Case-checking and minimal link condition. In Case and Agreement II, MIT Working Papers in Linguistics, 19:213-254. Department of Linguistics and Philosophy, MIT, Cambridge, Mass. Müller, Gereon. 1997. Partial wh-movement and optimality theory. In The Linguistic Review, 14:249-306. Mycock, Louise. 2004. The wh-expletive construction. Proceedings of the LFG04 Conference University of Canterbury. Nilsen, Østein. 2000. The Syntax of Circumstantial Adverbials. Oslo, Novus Press. Nunes, Jairo. 1995. The copy theory of movement and linearization of chains in the Minimalist Program. Ph.D. diss., U. of Maruland, College Park, MD. Obenauer, Hans-Georg. 1994. Aspects de la syntaxe A-barre, Thèse de doctorat d‘état, Université de Paris VIII. Obenauer, Hans-Georg. 2004. Nonstandard wh-questions and alternative checkers in Pagotto. In H. Lohnstein and S. Trissler, (eds.), Syntax and Semantics of the Left Periphery, Interface Explorations, 9:343-384. Berlin: Mouton de Gruyter. Obenauer, Hans-Georg and Cecilia Poletto. 2000. ‗Rhetorical‘ wh phrases in the left periphery of the sentence. University of Venice, Working papers in Linguistics Vol 10, n.1:121-151. Ochi, Masao. 2004. How come and other adjunct wh-phrases: A cross-linguistic perspective. Ms. Osaka University. Oshima, Shin. 2003. Subjunctives and Subject Obviation. Journal of Inquiry and Research, No. 78. Oţegov, S.I. and N.J. Švedova. 1992. Tolkovyj Slovar‟ Russkogo Jazyka. Izdatel‘stvo ―Az.‖ 260 Pesetsky, David. 1982. Complementizer-trace phenomena and the nominative island condition. The Linguistic Review 1, 3. Pesetsky, David. 1987. Wh-in-situ: Movement and Unselective Binding. E. Reuland and A. ter Meulen, (eds.), The Representation of (In)definiteness, 98-129. Cambridge, MA: MIT Press. Pesetsky, David. 2000. Phrasal movement and its kin. Linguistic Inquiry. Monograph Thirty-Seven. Picallo, Carmer. 1984. The Infl node and the Null Subject parameter. Linguistic inquiry, 15:75-102. Picallo, Carmer. 1985. Opaque Domains. PhD dissertation, City University of New York, New York. Progovac, L. 1993. Subjunctive: The (Mis)behaviour of Anaphora and Negative Polarity. The Linguistic Review, 10:37-59. Quer, Josep. 1996. Changing moods: Complement licensing and mood alternations in questions. Utrecht: Utrecht Institute Linguistics OTS. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartik. 1985. A Comprehensive Grammar of the English Language. London: Longman. Richards, Norvin. 1997. What moves where when in which language. Doctoral dissertation, MIT, Cambridge, Mass. Richards, Norvin. 1999. Pied-piping and Islands: the That-trace effect. GLOW Newsletter: 44-45. Richards, Norvin. 2001. Movement in Language: Interactions and architectures. Oxford University Press, Oxford. van Riemsdijk, Henk. 1982. Correspondence and the Empty Category Principle. Tilburg papers in Language and Literature 12. U. Of Tilburg. van Riemsdijk, Henk. 1984. On pied-piped infinitives in German relative clauses. In Toman, J. (ed.). Studies in German grammar, 165-192. Dordrecht, Foris. Rivero, M. L. 1994. Clause Structure and V-movement in the Languages of Balkans, Natural Language and Linguistic Theory, 12:63-120. Rizzi, Luigi. 1990. Relativized minimality. MIT Press, Cambridge, Mass. Rizzi, Luigi. 1992. Argument/adjunc(ta)symmetries, In Proceedings of NELS, 22. 261 Rizzi, Luigi. 1996. Residual verb second and the Wh-Criterion. In Belletti, A. and Rizzi, L. (eds.). Parameters and functional heads, 2:63-90. New York, Oxford University Press. Rizzi, Luigi. 1997. The fine structure of the left periphery. In L. Haegeman (ed.) Elements of Grammar, 281-337. Dordrecht: Kluwer. Rizzi, Luigi. 2001. On the position ―Int(errogative)‖ in the left periphery of the clause. In Current Studies in Italian Syntax: Essays Offered to L. Renzi, Cinque, G. & G. Salvi (eds.). New York: Elsevier, 287-296. Rizzi, Luigi. 2001a. ―Relativized Minimality Effects‖, In The Handbook of Contemporary Syntactic Theory, M.Baltin and C.Collins (eds.), 89-110. Oxford, Blackwell. Rizzi, Luigi. 2002. Locality and Left Periphery. In A. Belletti (ed.) Structures and Beyond. The Cartography of Syntactic Structures, vol. 3, OUP. Rizzi, Luigi. 2004. On the Form of Chains: Criterial Positions and ECP Effects. Ms. University of Siena/Geneva. Rizzi, Luigi. 2006. On the Form of Chains: Criterial Positions and ECP Effects. In WhMovement: Moving on, L. Cheng & N. Corver, Cambridge (eds.), MA: MIT Press, 97133. Rizzi, Luigi. 2007. On some Properties of Criterial Freezing, in CISCL Working papers on Language and Cognition, 1:145-159. Rizzi, Luigi and Ur Shlonsky. 2006. Satisfying the subject criterion by a non-subject: English locative Inversion and heavy NP shift, ms., University of Siena and University of Geneva. Rizzi, Luigi and Ur Shlonsky. 2007. Strategies of subject extraction. In Interfaces + Recursion = Language? Chomsky's Minimalism and the View from Syntax-Semantics, H. M. Gärtner, U. Sauerland (eds.), 115-160. Berlin: Mouton de Gruyter. Rochemont, Michael. 1986. Focus in generative grammar. Philadelphia: John Benjamins. Rojina, Nina. 2007. Four types of ‗whys‘ in Russian. Generative Grammar in Geneva 5: 133–142. Ross, John Robert. 1967. Constraints on Variables in Syntax. Doctoral dissertation, MIT. [Page references are to the edition published as Infinite Syntax!, Norwood, N.J.: Ablex Publishing, 1986.] 262 Rudin, Catherine. 1988. On multiple questions and multiple wh-fronting. Natural Language and Linguistic Theory 6: 445-501. Rudin, Catherine. 1996. Multiple questions south, west, and east: A government-binding approach to the typology of wh-questions in Slavic languages. International Journal of Slavic Linguistics and Poetics. Sannikov, Vladimnir. 1989. Russkie sočinitel‟nye konstruktsii: semantika, pragmatika, sintaksis. Moskva: Nauka. Shlonsky, Ur. 1994. Agreement in Comp. The Linguistic Review 11, 351-375. Shlonsky, Ur and Gabriela Soare. 2009. Where‘s ‗Why‘?. Ms. Université de Genève. Skrabalova, Hana. 2006. Two types of wh-questions with conjoined wh-items in Czech. Presented at the Colloquium Coordination and Ellipsis, University of Paris. Stepanov, Arthur and C. Georgopoulos. 1995. Structure building and the conceptual interface: An analysis of Russian long-distance wh-questions. In M. Lindseth and S. Franks (eds.) FASL 5 Proceedings, 275-294. Stepanov, Arthur. 1997. On wh-fronting in Russian. In Proceedings of the North Eastern Linguistic Society 28. University of Massachusetts, Amherst. Stepanov, Arthur. 2000. Wh-scope marking in Slavic. Studia Linguistica 54(1):1-40. Stepanov, Arthur and Wei-Tien Dylan Tsai. 2007. The Hows and Whys of How and Why: Unselective binding and merger of wh-adjuncts. Ms. Stjepanoviç, Sandra. 1995. Short-distance movement of wh-phrases in Serbo-Croatian matrix clauses. Ms., University of Connecticut, Storrs. Stowell, Timothy. 1981. Origins of Phrase Structure. Doctoral Dissertation, Massachusetts Institute of Technology, Cambridge. Szczegielniak, Adam. 1999. That-trace effects‘ cross-linguistically and successive cyclic movement. Papers on Morphology and Syntax, Cycle One MIT Working Papers in Linguistics, Vol 33:369-393. Karlos Arregi, Benjamin Bruening, Cornelia Krause, and Vivian Lin (eds). MITWPiL. Taraldsen, Knut T. 2001. Subject extraction, the distribution of expletives and stylistic inversion. In Hulk, A. and Pollock, J.-Y. (eds.). Subject inversion in Romance and the theory of universal grammar, 6:163-182. New York, Oxford University Press. 263 Terzi, A. 1992. Pro in Finite Clauses: A Study of the Inflectional Heads of the Balkan Languages, PhD dissertation, City University of New York, New York. Watanabe, Akira. 1993. AGR-based Case theory and its interaction with the A'-system. Doctoral dissertation, MIT, Cambridge, Mass. Zavitnevich, Olga. 2001. On wh-questions and on nature of wh-words in Russian in comparison with English, Working Papers in English and Applied Linguistics, Research Centre for English and Applied Linguistics, University of Cambridge, UK. Zoerner, Cyril. 1995. Coordination: the syntax of &P. Doctoral dissertation, Cornell University, Ithaca, NY. 264