University of Southern California Dissertations and Theses

Agreement on the left edge: the syntax of left dislocation in Spanish

(USC Thesis Other)

Agreement on the left edge: the syntax of left dislocation in Spanish

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content AGREEMENT ON THE LEFT EDGE: THE SYNTAX OF LEFT DISLOCATION IN SPANISH by Omar Beas ____________________________________________________________________ A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (LINGUISTICS) August 2007 Copyright 2007 Omar Beas ii Dedication To my parents Hugo and Maria Helvezia To my little sister, Cristina iii Acknowledgements Two special people in my dissertation committee helped me to initiate and carry out this dissertation. I am very grateful to Maria Luisa Zubizarreta and Patricia Schneider Zioga, who helped me to lead this project to its completion. María Luisa Zubizarreta, my academic advisor, was a source of inspiration before I came to USC and during my graduate studies. I borrowed a great deal of time and knowledge from her linguistic expertise and sharp mind. I must say that many ideas I explored in this dissertation started from her previous work on clause structure and Romance syntax. Gracias Maria Luisa por tu paciencia y por motivarme siempre a seguir adelante. During the last two years of my studies at USC, I made progress because, along with Maria Luisa, I counted on the assistance of Patricia Schneider Zioga. She spent many hours discussing previous versions of this manuscript in form and content. It has been a great pleasure working with Patricia. Certainly, I must also say thank you Patricia for your excellent suggestions and for teaching me so much in such a short period of time. I would also like to thank Mario Saltarelli, Hajime Hoji, and Jean Roger Vergnaud, who were also members of my committee. Mario Saltarelli gave me important feedback in the initial stages of this project and helped me elaborate my ideas. I also had the privilege of being a student of Hajime and Jean Roger. In their classes, I benefited from Hajime’s research program and from Jean Roger’s passion for linguistics. Their classes showed me how fascinating generative grammar is and how linguistics is a science and a piece of art at the same time. iv I must also thank the Spanish and Portuguese Department at USC, which provide me with the necessary financial support to complete my studies through several teaching assistantships in its language program. Special thanks are due to Gayle Vierma, Director of the Basic Language Program who always gave me excellent advice for my personal and professional life. Many thanks are due to my professors in Peru: Rodolfo Cerrón Palomino, Liliana Sanchez and Jose Camacho; who stimulated my passion for linguistics and syntax. To my friends in Peru and in the United States, thank you for keeping in touch: Milagros Lucero, José Riqueros, Vidal Carbajal, Angela Amaya, Silvia Suarez, Jorge Terukina, Emily Hinch, Roberto Mayoral, Ana Sanchez Muñoz, Veronica Medda, Alvaro Cerrón Palomino, Mauricio Sanabria, William Carter, Ruth Gutiérrez, Michael Rushforth, and Monica Cure. And finally, Cynthia, I want to say thank you for the happiness and peace your love has brought me. After you became part of my life, I learned to look forward to the future with confidence. v Table of Contents Dedication ii Acknowledgements iii Abbreviations vi Abstract viii Chapter 1: An Introduction to the Study of Clause Structure 1 1.1 Characterizing Word Order in Spanish 2 1.2 Theoretical Framework 8 1.3 The Standard Analysis of Spanish Clause Structure 17 1.4 Outline of the dissertation 33 Chapter 2: Lexical Preverbal Subjects: Dislocation and Canonical Patterns 36 2.1 Preverbal Lexical Subjects, EPP and pro-drop languages 37 2.2 Preverbal Subjects in Canonical and Dislocated Positions 41 2.3 Evidence from Scope Interactions 59 2.4 Summary and Conclusions 80 Chapter 3: Non declaratives and Second Position on the Left Edge 81 3.1 The Licensing Requirements of the Left Edge 82 3.2 Raising to the Left Edge: Verb Movement in Spanish 91 3.3 Interrogative Clauses in Modern Spanish 99 3.4 Conclusions 123 Chapter 4: Towards an Analysis of Spanish Declaratives 126 4.1 Typological Considerations on Clausal Typing and Second Position 126 4.2 The Licensing of Declarative c*º 140 4.3 Second Position in Modern Spanish: Declarative Clauses 154 4.4 Assertion Markers and the Affirmative sí 175 4.5 Conclusions 187 Chapter 5: Agreement in the Left Periphery and Dislocated Subjects 189 5.1 Agreement and the Comp-Infl Articulation 191 5.2 L-Subjects and Movement to the Lower Left Edge 213 5.3 The Makeup of the Lower Left Edge 228 5.4 Conclusions 242 References 243 vi Abbreviations 1 first person 2 second person 3 third person ACC accusative ADV adverb(ial) AFF affirmation AGR agreement COMP complementizer DAT dative DECL declarative F feminine FOC focus FUT future IMP imperative IND indicative INDF indefinite INF infinitive IPFV imperfective IRR irrealis M masculine NEG negation, negative vii NOM nominative OBJ object PL plural PRS present PROH prohibitive PST past PTCP participle Q question particle/marker REFL reflexive SBJ subject SBJV subjunctive SG singular TOP topic viii Abstract This dissertation is an investigation of two interrelated clausal left edge properties in the Spanish syntax: the positional effects of left edge elements (akin to the Tobler Mussafia and Wackernagel effects) and the availability of constituent dislocation. I focus on the syntax of preverbal lexical subjects and their relation to the several positions in the clause. Based on the projectionist approaches developed by Rizzi (1997), Poletto (2000), among others, this dissertation sketches a more reduced version of the complementizer system dubbed the Narrow Left Edge. Accordingly, the left periphery is composed by two core projections: the Point of View Projection (CP) and the Assertion Projection (c*P). In particular, I claim that the phenomenon of second position placement is manifested in Modern Spanish (cfr. Suñer 1994 and Fontana 1993) and must be related to the licensing of the c*P projection in declaratives and nondeclaratives. In addition, the mechanics of word order dislocation shows that constituents can be displaced to the left edge by movement or by base generation. These two effects can be explained if an Agree relation is postulated between the Assertion Phrase c*º in the lower left edge, and Tº in the inflectional layer. This dissertation is structured into five chapters: Chapter 1 situates the study of Spanish word order within the theoretical framework of the Minimalist Program (Chomsky 1995, 2001). Chapter 2 establishes that there are two patterns of preverbal lexical subjects: canonical subjects and dislocated subjects. Chapter 3 investigates the restriction against the simultaneous occurrence of preverbal subjects and wh- ix words in interrogative clauses. I argue that Spanish displays this restriction as a manifestation of the second position constraint. Chapter 4 develops an analysis of Spanish declaratives. Taking into account the distribution of assertion markers and pronominal elements in the language, I argue that they also constitute evidence for the activation of the second position in the clause. In Chapter 5, I make a novel observation in the study of the dislocated pattern of subjects by identifying the second position of the clause with dislocation derived by movement. 1 Chapter 1 An Introduction to the Study of Clause Structure and Dislocation in Spanish Chomsky (1986) states two main questions for the study of language as an internalized system: (1) a. What is the knowledge of language? b. How is this knowledge of language acquired? The potential answers to these questions have played a fundamental role in the research program in generative grammar in the last twenty years and have given rise to different levels of abstraction in the linguistic enterprise. Whereas the question in (1)a inquiries about how to define the knowledge of language, and in this sense, aims to comply with the criterion of descriptive adequacy; the question in (1)b seeks to understand why this knowledge of language is possible in accordance to the criterion of explanatory adequacy. Retrospectively, syntactic research since the mid 1970s has attempted to show the viability of this program of investigation. For instance, the Government and Binding approach, developed in the 1980s, was particularly successful in providing a wide range of empirical coverage of linguistic data consistent with the descriptive adequacy criterion. The knowledge of language was characterized as modular and as a set of filters or grammatical conditions on the output. For instance, we find the X- bar Theory, which dictates conditions on how phrase structure is built; the Thematic Theory, which establishes how predicates and related to their arguments; and the Binding Theory and the Case Theory, which explain how nominal expressions are 2 formally licensed in form (Case) and interpretation (binding). In order to characterize the fact that human languages display the property of displacement (namely they are realized in one point of a phonetic string but are interpreted in a different one), Government and Binding approaches assume a general transformation rule: Move-α. This rule constitutes the central generative engine that allows any category to move anywhere anytime. This dissertation is inscribed within the program outlined in (1) and attempts to provide plausible answers to these questions through the study of the clausal properties of Spanish and of the phenomenon of word order dislocation. In the following sections, I will inscribe the phenomena of word order dislocation within the tenets defined by the questions (1) and within the Minimalist Program of Chomsky (1995) and the recent developments of Chomsky (2000, 2001), which provides a better conceptual apparatus to answer them. 1.1 Characterizing Word Order in Spanish In the generative approach to language, an answer to the question in (1)a is to hypothesize that what speakers know is an I-language, a set of intuitions or senses about how the linguistic expressions are structured. A first step to address this question is to consider the knowledge of speakers about which expressions are possible in their system of knowledge. Consider the following paradigms: (2) English a. He likes syntax b. He doesn’t like syntax 3 (3) Quechua a. Pay shamu-n kunan-ka he come-3SG today-TOP “He comes today.” b. Pay mana shamu-n-chu kunan-ka he NEG come-3SG-NEG today-TOP “He does not come today.” Sentences in (2) and (3) are instances of how the negation is structured in English and Quechua. To say that English and Quechua speakers know that (2)b and (3)b correspond to the negative counterpart of (2)a and (3)a respectively is to claim that they know the meaning and the form of negation. In other words, this knowledge entails how to interpret the negative clitic n’t and where to place n’t in a sentence. Whereas English speakers will place n’t after the person inflection contained in the auxiliary do, Quechua speakers will place the negative marker mana “not” before the person inflection and they will reduplicate it with the suffix chu. A first approximation is then to characterize the set of sentences or structures that belong to a certain I-language, namely the set of grammatical expressions or expressions that a grammar can generate. Spanish illustrates another case in the domain of word order in which for a given sentence, there are many possible permutations. The notion of unmarked or neutral order of a sentence can be characterized in terms of neutrality with respect to the more “marked” semantic or pragmatic effects triggered by focalization of new 4 information, emphasis, and perhaps topichood. Thus, a given ordering of constituents is said to correspond to the unmarked order of the language if that sequence does not include any of the aforementioned effects. One way in which this neutrality in word order is typically detected is by its compatibility with an informational question such as what happened? Therefore, the constituents of a sentence are neutral if they constitute a felicitous sequence to such informational question. In what follows, I will presuppose the application of this test when referring to the descriptive notions of unmarked order, neutral order or base order. For instance, the sentence in (4) with a ditransitive verb overtly realizing all its arguments (subject, direct object and indirect object) corresponds to the base order of constituents. In this sentence, none of the constituents has a special informational status: (4) Me=parece que aquí autografió Violeta un libro a Esther. DAT=seems COMP here autograph:PST Violeta a book DAT Esther “It seems to me that Violeta autographed a syntax book for Esther here.” The paradigm in (5) exemplifies the phenomenon known as dislocation, namely, the detachment of a constituent from its base position. As illustrated in (5), dislocation in Spanish can be multiple, including subjects, direct objects and indirect objects which can appear interchangeably in the preverbal field of the verb autografiar: (5) a. Me=parece que Violeta, a Esther, un libro se=lo=autografió. DAT=seems COMP Violeta DAT Esther a book DAT=ACC=autograph:PST b. Me=parece que Violeta, un libro, a Esther se=lo=autografió. DAT=seems COMP Violeta a book DAT Esther DAT=acc=autograph:PST 5 c. Me=parece que a Esther, Violeta, un libro se=lo=autografió. DAT=seems COMP DAT Esther Violeta a book DAT=acc=autograph:PST d. Me=parece que un libro, Violeta, a Esther se=lo=autografió. DAT=seems COMP a book Violeta DAT Esther DAT=acc=autograph:PST e. Me=parece que a Esther, un libro, Violeta se=lo=autografió. DAT=seems COMP DAT Esther a book Violeta DAT=acc=autograph:PST f. Me=parece que un libro, a Esther, Violeta se=lo=autografió DAT=seems COMP a book DAT Esther Violeta DAT=acc=autograph:PST "That book, Esther, it seems to me that Violeta autographed it for her." As attested by the different possibilities to combine subjects and nonsubjects, Spanish permits a high degree of variation in the preverbal position, data that has also been reported in other Romance languages such as Italian (Rizzi 1997). However, given that the description of the knowledge of language also includes the set of linguistic expressions that speakers intuit as impossible, a particular grammar must also be able to characterize the speakers’ intuitions of those linguistic expressions or structures that are impossible or do not correspond to their I-language. Therefore, English and Quechua speakers not only know how to negate a sentence, a significant aspect of their intuitions is the knowledge of which sentences are impossible in their I-language: (6) a. *He no likes not syntax. b. *Pay shamu-n mana kunan-ka. he come-3SG NEG today-TOP “He does not like syntax.” 6 English speakers are also able to characterize sentence (6)a as impossible. A similar reaction to sentence (6)b occurs with Quechua speakers who reject the English style to negate a sentence. Similarly, in spite of the freedom of the preverbal field in Spanish, it is possible to find several restrictions. For instance, there is an asymmetry between absolute verb initial sequences in unaccusatives and unergatives unmarked word order. As exemplified below, unaccusatives in Spanish display a preference for the order with the verbal predicate preceding the lexical subject (VS) whereas unergatives display a preference for the order with the lexical subject preceding the verbal predicate (SV). The paradigms in (7) and (8) illustrate these observations (the symbol ‘%’ marks sentences with a very low level of acceptability): (7) a. Llegó el músico. arrive:3;PST the musician “The musician arrived.” b. %El músico llegó. the musician arrive:3;PST “The musician arrived.” (8) a. %Trabajó esta secretaria en mi oficina. work:3PST this secretary in my office “This secretary worked in my office.” 7 b. Esta secretaria trabajó en mi oficina. this secretary work:3PST in my office “This secretary worked in my office.” In addition, with transitives or ditransitives, the preverbal field must be occupied by at least one constituent, a null topic or a clitic. Zubizarreta (1998, 2007) points out that the VSO order is declining, but VSO is still attested in structures where some element other than the subject plays the role of subject of predication in initial position. For instance, the sentences in (9) contain a constituent in preverbal position; (10) is a resultative construction, presumably with a null topic in preverbal position and those sentences in (11) illustrate clitic initial sentences: (9) a. Juan compra un periódico todos los días. Juan buy:3;PRS a newspaper all the days “Juan buys a newspaper everyday.” b. Juan regaló un libro a María. Juan give:3;PST a book ACC Maria “Juan gave a book to Maria.” (10) Acaba de ganar España el mundial de fútbol. just of win:INF Spain the championship of soccer “Spain just won the soccer world championship. (11) a. Se=compra Juan un periódico todos los días. REFL=buy:3;PRS Juan a newspaper all the days “Juan bought himself a newspaper.” 8 b. Le=regaló Juan un libro a María. DAT=give:3;PST Juan a book ACC Maria “Juan gave a book to Maria.” However, if the preverbal position is empty, sentences with an absolute verb initial order are perceived as stilted or downright ungrammatical (Gutiérrez Bravo 2002, 2007 and Suñer 1994) as illustrated in (12): 1 (12) a. %Compra Juan un periódico todos los días. buy:3;PRS Juan a newspaper all the days “Juan buys a newspaper everyday.” b. %Regaló Juan un libro a María. give:3;PST Juan a book ACC Maria “Juan gave a book to Maria.” If the construction of a particular grammar is successful to characterize the set of expressions allowed by the grammar (i.e., grammatical sentences) as well as those expressions that are impossible (i.e., ungrammatical sentences), this first level of linguistic inquiry meets the condition of descriptive adequacy. 1.2 Theoretical Framework Beyond the construction of a theory about the knowledge reflected in an I- language, a higher level of abstraction results from the answer to question (1)b, dubbed as the Plato’s problem. Plato’s problem draws on the observation that 1 Although Gutiérrez Bravo (2002) describes the data as “Mexican Spanish”, my own judgments and those of the Latin American Spanish informants I consulted are consistent with his description of the facts. However, I must point out that other speakers accept absolute verb initial sequences. 9 children develop a highly rich system of knowledge, independently of external factors and with a minimal amount of exposition to the linguistic input. This empirical confirmation is surprising given that this process occurs in a relatively short period of time, in absence of explicit or formal instruction, and with a remarkable speed. Moreover, the nature of the external stimuli is often imperfect and incomplete but children develop a very sophisticated understanding of language, a system of knowledge that allows them to recognize what is a possible or an impossible structure in their I-language. From this perspective, any theory of language must conform to the criterion of explanatory adequacy in order to explain how speakers arrive at such rich stage of linguistic knowledge represented by an I- language from a very diverse source of stimuli. One perspective to face the challenge posed by Plato’s problem has been to hypothesize the existence of a Universal Grammar. Universal Grammar is theory of the initial state of faculty of language before any exposure to external stimuli, and prior to the development of an I-language. Presumably, this theory characterizes a language specific biological program, independent of other cognitive systems such as memory, intelligence and logical or mathematical reasoning. 1.2.1 The Interface Levels Although well formedness conditions can be established to rule in grammatical sentences and rule out ungrammatical sentences to achieve descriptive adequacy, the Minimalist Program also assumes that syntax is an economical computational system. Therefore, from the possible set of grammatical derivations, 10 the grammar only derives those derivations that are optimal. If a derivation only consists of legitimate objects at the PF and LF interfaces, it is said to comply the Principle of Full Interpretation: (13) Principle of Full Interpretation The LF and PF representations must consist solely of well formed legitimate elements (interpretable) at the relevant levels. A syntactic object is legitimate if its derivation only contains the minimal set of instructions required for its semantic and phonological interpretation in the conceptual/intentional and articulatory/perceptual performance levels. In other words, given that the output of the generative procedure consists of pairings of sounds and meanings, one of the core postulates in minimalism is that the only levels of representation are those that correspond to the properties relevant to the phonological interpretation (PF) and logico-semantic interpretation (LF): (14) Lexicon Numeration Overt syntax spell-out 3 PF LF articulatory/perceptual conceptual/intentional systems systems This is a radical departure from the approach developed in the 1980s. The computational procedure is not divided in separate modules and there are no levels of representation other than the interface levels of PF and LF. Typically, in the 11 minimalist approach, the computational system consists of a single computational procedure (the overt syntax), a lexicon and two points of interface of the linguistic system with the cognitive systems: PF and LF. A derivation consists of a set of lexical items L, taken from the lexicon and consisting of phonological, semantic and formal properties arranged in a numeration N. Grammatical operations combine to modify the elements of a numeration N in the overt syntax. It is hypothesized that these grammatical operations are triggered by the necessity to license properties illegitimate to the interface levels. Once a structure has reached the point where all its components can be optimally interpreted by PF and LF (namely, no illegitimate objects exist), it is said to have reached the point of spell out. Accordingly, the derivation must be assigned a phonological interpretation (at PF) and a semantic interpretation (at LF) in compliance with the Principle of Full Interpretation. 1.2.2 Phrase Structure Construal The core operation to build structure is Merge, hypothesized as a primitive in the construction of the theory of the faculty of language. This operation is a concatenation process that selects two objects α, β from the numeration to form a bigger object Z: (15) Z= {α, {α, β} α 3 α β In (15), Merge has applied to α and β to create Z. In this case, Z is said to be the label of the concatenation process, namely Z corresponds to the projection of α, one 12 of the selected components. For instance, the derivation in (16) is obtained by the successive application of Merge. Given the numeration N={ DP 1 , DP 2, DP 3 , Vº, vº}, the computational system creates structure by recursive applications of Merge. First, the computational system selects Vº and DP 3 to form Z 1 ={V’, {Vº, DP 3 }}, then the procedure selects Z 1 and DP 2 to create Z 2 ={VP, {DP 2 , Z 1 }; Z 2 and vº to create Z 3 ={v’, {vº, Z 2 }} and finally it selects Z 3 and DP 1 to create Z 4 ={vP, {DP 1 , Z 3 }: (16) vP v v v v DP 1 v’ v v v v vº VP 1 1 1 1 DP 2 V’ v v v v Vº DP 3 Basic constituents such as subjects (DP 1 ) and objects (DP 2 and DP 3 ) are generated inside the thematic layer delimited by the vP/VP complex. At this level, DP 1 , DP 2 and DP 3 are interpreted as arguments of the verbal predicate and their θ-roles (agent, experiencer, theme, goal, patient) are automatically assigned (“checked”) by the concatenation process. Thus, in minimalism, θ-roles do not constitute a “level” or a specific module that can be independent or separated of the computation process, but they are part of the derivation as condition on the interpretation of the relation predicate/argument. 1.2.3 Checking Theory An intriguing property of natural languages is displacement. Languages typically display arrangements of phonetic strings that do not correspond to the structure in which they are interpreted. Within minimalism, the displacement property is characterized as the composite operation Move. Basically, Move involves 13 a “retrieval process” by the computational system. Assume that in the numeration N={α, β, δ}, α is marked with an index meaning “more than one occurrence of α” as suggested in Chomsky (1995). Once Z has been created from α and β, the grammar accessing to this index must represent the multiple occurrences of α by an additional procedure. This results from the application of the operation Copy to α and Remerge (or “External Merge”) of α. At the PF level, only one of the copies of α is pronounced, therefore the operation Delete must also apply. The application of Move is costly because Move involves the activation of other licensing strategies and a heavier computational load. For instance, it involves more than one application of the operation Merge for a single syntactic object besides Copy and Delete. In minimalism, the existence of all these operations describing the property of displacement must be justified for some reason. Within the Minimalist Program, Move is forced by the Principle of Last Resort: (17) Principle of Last Resort Operations do not apply unless required to satisfy a constraint. The minimal number of operations is applied to satisfy the constraint (Weinberg 1999). An approach pursued in the early 1990s considered that a movement operation is allowed if and only if there is some morphological feature that needs a special structural configuration. For instance, Chomsky (1995) assumed that the existence of an index meaning “more than one occurrence of α” is actually some kind of morphological requirement or property expressed as a feature. Thus, licensing of a 14 morphological feature F is called “checking” and can occur in two configurations: specifier/head or head/head. If δ is the functional head where checking must apply in one of these two configurations, we have: (18) a. Specifier/head configuration b. Head/head configuration δ ... 3 3 YP F δ δ F 3 3 δ F ... Y F δ F Feature checking consists of the matching of the feature F in δ with the feature F of a lexical element projected as a phrase (YP) or projected in the form of a head (Yº) in the local domain of δ F . After checking occurs the feature F disappears (namely, it is licensed). In order to express this idea, I will adopt Rizzi’s Criterial Freezing (Rizzi and Shlonsky 2005): (19) Criterial Freezing An element moved to a position dedicated to some scope-discourse interpretative property, a criterial position, is frozen in place. The condition in (19) states that once a phrase ZP has undergone checking, a further checking operation cannot affect ZP. This occurs presummably for Economy reasons in the interpretation of chains. If the checking position of ZP is considered a strong position where pronunciation must apply in the PF component, chains with two strong links are not legible by the computational system, unless an additional strategy may allow surmounting one of the strong links (for details see Boeckx 2003 for resumption, and Rizzi and Shlonsky 2005 for subject extraction). 15 1.2.4 (Un)interpretability, Strength and the EPP Why does the computational system of the grammar need to check features? A plausible answer has been to claim that features require checking because they cannot be interpreted at the PF and LF levels. In compliance with the Principle of Full Interpretation, some features are LF/PF interface related features and are interpretable (for instance the categorial features of lexical items and the φ-features of nominals), but others are uninterpretable (for instance, the φ-features of verbs and adjectives and the Case features of inflectional heads). The licensing of uninterpretable features indicates that a condition in the derivation must be satisfied. Grammatical operations such as Move apply following the Principle of Last Resort. To the extent that the licensing conditions for feature uninterpretability depend on language specific considerations, Chomsky (1993) proposed that some uninterpretable features are strong whereas others are weak. Strong features are uninterpretable at LF and PF, triggering overt movement before the spell-out. In contrast, weak features are invisible to PF and are checked at LF after spell-out. In this system, the examples from Spanish and Chinese in (20) are analyzed as a function of the strong/weak nature of the uninterpretable [qu] feature: (20) a. ¿quién cree Mabel que compró los libros? who believe:3 Mabel that buy:3;PST the books “Who does Mabel believe bought the books?” 16 b. Mabel xiangxin shei mai-le shu. Mabel believes who buy-ASP books “Who does Mabel believe bought the books?” The differences between Spanish and Chinese in the formation of interrogatives are expressed in terms of a morphological property. In Spanish, the uninterpretable [wh] feature of quién “who” is strong whereas the uninterpretable [wh] feature in shei “who” in Chinese is weak. As a result, Spanish requires overt movement before spell out and Chinese can wait until LF to check the weak feature [wh] through covert movement. Similarly, the contrast between the example from Early Modern English in (21) and the examples from Modern English in (22) can be characterized in terms of the strong/weak feature distinction: (21) Thou thinkest not of this now. (Launce, Two Gentlemen of Verona) (22) a. Mabel mistrusts her. b. Mabel does not mistrust her. In the Early Modern English, the finite verb was located in a higher position than the one occupied by the negation, but in Modern English, this does not happen anymore. In Modern English, the finite verb must be below the position of the negation. Assume that the strength of the agreement/inflectional features in Tº is responsible for the observed pattern. If the negation adjoins to the vP/VP complex, one can say that φ-features in Tº were strong in Early Modern English forcing V-to-T raising. If this were not the case, those features would remain unchecked and the derivation 17 would not be legible at the interface levels. In contrast, Modern English Tº contains uninterpretable weak φ-features. Weak features can survive after spell out and are checked by covert movement of only the relevant matrix of features required to satisfy a grammatical condition. 1.3 The Standard Analysis of Spanish Clause Structure Syntactic research on the Spanish clause structure in the 1980s and early 1990s has attempted to account for the variability of word order in the language through the postulation of several formal mechanisms and through the adoption of many assumptions, many of them sometimes not completely justified. Although there are many variants, I will call the predominant analysis in the 1990s “Standard Analysis”, an approach developed within the basic assumptions of the Minimalist Program (Chomsky 1993, 1995). In the Section 1.3.1, I introduce the core development of the analysis of preverbal subjects and its relation with postverbal subjects within the dislocation analysis, a subvariant of the Standard Analysis. In Section 1.3.2, I discuss the basic approach to dislocation of direct objects and indirect objects (henceforth nonsubjects). Finally, in Section 1.3.3, I summarize the main problems and drawbacks created by the Standard Analysis. 1.3.1 The Analysis of Preverbal and Postverbal Subjects A controversial topic within the research of Spanish syntax has been to determine which construal strategy (movement or base generation) and which 18 properties correspond to grammatical subjects in preverbal and postverbal position. To begin with note that in this language, subjects can be preverbal or postverbal: (23) a. María me=devolvió el libro que le=presté. Maria DAT=return:3;PST the book COMP DAT=lend:1;PST “Maria returned to me the book that I lent her.” b. Me=devolvió María el libro que le=presté. DAT=return:3;PST María the book COMP DAT=lend:1;PST “Maria returned to me the book that I lent her.” c. Me=devolvió el libro que le=presté María DAT=return:3;PST the book COMP DAT=lend:1;PST María “Maria returned to me the book that I lent her.” A crucial assumption in the analysis of these examples within the Standard Analysis is the adoption of the VP-internal subject hypothesis. Accordingly, the subject in (23)b and (23)c is generated inside the VP and remains there whereas the inflected verb and other material are displaced over this position. Therefore, the sentence (23)b is derived by verb movement as in (24)a and the sentence in (23)c is derived by verb movement and displacement of the direct object over the position of the subject as in (24)b: 19 (24) a. VSO order b. VOS order CP CP v v Cº TP Cº TP v 1 Vº-Tº VP Vº-Tº VP me=devolvió v me=devolvió 1 María V’ el libro J VP v v t i …el libro María V’ v t i …t J Under the assumption that the agreement φ-features in Tº are strong, finite verbs should move to Tº for morphological theory internal reasons. Suñer (1994) provides empirical evidence for V-to-T movement and claims that the finite verb does not reach left edge nodes. Hence, the order VSO follows straightforwardly by leaving the subject in [Spec VP] in situ and raising the verb to the Tº head. The derivation of VOS can also be easily accommodated. For instance, Ordóñez (1997) has shown that the object in VOS sequences is the result of the adjunction of the direct object el libro que le presté “the book that I lent her” to the maximal projection of the verbal phrase. From this position, the object c-commands the subject in the surface structure. 2 Several complications arise to determine how postverbal subjects receive Nominative Case. Note that the mechanism of feature strength is seriously challenged by the empirical data. If we assume that the Case features of subjects in Spanish are strong and that these features need to be checked in a Spec/Head configuration with Tº, VSO and VOS sequences should not be attested. This is 2 See Zubizarreta (1998a: 125-133) for the discussion of empirical data that leads to a perspective different than Ordóñez’s idea of scrambling. 20 because subjects would have to obligatorily move to the specifier of Tº as in English. Given that verb movement affects only the Tº node and does not affect higher nodes as suggested by Suñer (1994) among others, only the generation of a SVO sequence would be expected. Similarly, if one postulates that the Case-related features of the subject are weak in Spanish, it is not easy to arrive at a satisfactory answer to the many surface orderings in this language. If Case features are weak, there could be two options: First, the subject moves covertly as a maximal projection to check Nominative in a [Spec, Head] configuration or second, the relevant features of the subject are attracted to the Tº head. The first option is clearly empirically inadequate as demonstrated by Zubizarreta (1998a). If subjects move covertly from [Spec, VP] to [Spec, TP] to check Nominative Case, then it is difficult to explain the asymmetries in the interpretation of bare plurals and the effects of Condition C. Consider first the paradigm in (25): (25) a. A menudo juegan niños en este parque. often play:3PL children in this park “Children often play in the park.” b. *Niños juegan a menudo en este parque. children play:3PL often in this park “Children often play in the park.” Bare plurals in Spanish are interpreted only existentially unlike their counterparts in English where it is possible to have existential and generic readings. In addition, they 21 can only appear in postverbal position, but not in preverbal position as shown in (25). If subjects were to move covertly, Zubizarreta argues, they should be ambiguous and display a nonexistential reading, contrary to the facts. A similar argument can be constructed for the interpretation of object pronouns in (26), involving the Condition C: (26) a. Esta mañana la madre de Juan lo=castigó. this morning the mother of Juan ACC=punish:3PST “This morning, Juan’s mother punish him.” b. *Esta mañana lo=castigó la madre de Juan. this morning ACC=punish:3PST the mother of Juan “This morning, Juan’s mother punish him.” In example (26)a, the pronoun lo “him” may corefer with the DP Juan embedded within a preverbal subject. In contrast, coreference is disallowed if the DP Juan is inside a postverbal subject as in (26)b. Assuming that binding principles apply at LF, if postverbal subjects had to move to [Spec, TP] after spell-out, the contrast in (26) should not occur because (26)b would have the same structural representation as (26)a where coreference is possible. The second option, adopted by Zubizarreta (1998a), consists of feature attraction, an operation invisible to PF. The head Tº only attracts the D-feature of the DP subject, not the whole DP. This approach solves the problems raised by the 22 paradigms in (25) and (26), and opens the discussion of the nature of the [Spec, TP] and preverbal subjects. 3 If [Spec, VP] is a Case position, [Spec, TP] is expected to be a nonCase position or an Ā-position, which may be occupied by phrases other than the DP subject. 4 Within the Standard Analysis, an interpretation of the problem was developed in the 1990s to account for the nature of preverbal subjects and the [Spec, TP] position. Accordingly, several researchers assume that there is no link between a subject generated postverbally and a subject generated preverbally because postverbal subjects and preverbal subjects correspond to different numerations and derivations (Olarrea 1996, Ordóñez and Treviño 1999, Ordóñez 1997, 2000, Uribe Etxebarria 1995). This conception, which I will dub the dislocation analysis, is based on the idea that VSO corresponds to the unmarked word order. VSO is generated by leaving the lexical subject in [Spec, VP] and moving the lexical verb from Vº to Tº as in (27)a: 3 This is an option very akin to the application of the operation Agree in recent versions of the Minimalist Program developed since Chomsky (2000, 2001) with the difference that Agree obviates the overt/covert distinction present in earlier models. 4 Another option is to claim that [Spec, TP] is sometimes a Case position (A-position) and sometimes a nonCase position (Ā-position). Although this approach is present in some variations of the Standard Analysis (Suñer 1994, Campos 1997), it leaves unexplained why a grammar needs to resort to two different mechanisms of Case licensing with respect to preverbal subjects. 23 (27) a. Unmarked order: VSO b. Marked order SVO CP v v NP i […] Cº InflP v v InflP Vº»Inflº VP v : v Vº»Inflº VP ! NP V’ v ! v pro i V’ z-- ... t i …O v ... t i …O In contrast to English, which checks Nominative Case in [Spec, TP] by the application of A-movement and displays a rigid SVO in declaratives, the alternative orderings displayed by Spanish constituents suggest that the subject in (27)a does not need to move for Case theoretical reasons. Instead, structural Case is assigned in situ by Tº under feature attraction. Given that movement is hypothesized to apply as a Last Resort and Nominative has been checked in situ, the source of SVO in (27)b must arise via a different operation. According to the dislocation analysis, preverbal lexical subjects are generated as adjuncts within the left periphery. Therefore, the derivation of preverbal subjects in (27)b includes an empty pro 5 , absent in the numeration that generates postverbal subjects. The problem is restated in recent minimalist versions that attribute the displacement property to the existence of the “Generalized EPP feature.” 6 The Generalized EPP feature represents an instruction to the grammar to treat functional heads marked with the notation “EPP” as heads requiring a specifier. In particular, 5 See Roselló (2000) and Cardinaletti (1997) for a discussion on the exact location of pro in Romance. I will not take a position with respect to this point because the analysis introduced in the main text does not depend crucially on the adoption of one alternative over the other. 6 As indicated in Section 1.2.4, the EPP is not properly speaking a feature. It might be understood as a “subfeature of a feature” (Pesetsky and Torrego 2001). 24 Alexiadou and Anagnostopoulou's influential 1998 work on the syntax of subjects in pro-drop languages suggests that pro-drop languages allowing the alternation VSO/SVO have a different parametrization conditions than those observed in non pro-drop languages. In pro-drop languages such as Spanish and Greek, they argue, the EPP is checked by Xº-movement of the verb; therefore, no additional movement of the subject to [Spec, TP] is needed. In contrast, in non pro-drop languages such as English and French, the EPP-feature is checked either by XP-movement of the NP subject or by insertion of an expletive. Consequently, the alternation of VSO and SVO orders in pro-drop languages in Alexiadou and Anagnostopoulou’s view suggests that preverbal subjects must be generated by left dislocation, expressed in terms of adjunction to the left edge of the constituent corresponding to the subject. 1.3.2 The Analysis of Nonsubjects and Dislocation Consider the following paradigms involving direct and indirect objects. Examples in (28) correspond to direct objects and indirect objects in their base position (θ-position or Case position) and those in (29) correspond to the same direct and indirect objects in a dislocation structure: (28) a. Omar compró estas esculturas. Omar buy:3;PST these sculptures “Omar bought these sculptures.” b. Omar regaló una escultura a sus amigos. Omar give:3;PST a sculpture DAT his friends “Omar gave a sculpture to his friends.” 25 (29) a. Estas esculturas las=compró Omar. these sculptures ACC=buy:3;PST Omar “These sculptures, Omar bought them” b. A sus amigos les=regaló Omar una escultura. DAT his friends DAT=give:3;PST Omar a sculpture “To his friends, Omar gave a sculpture.” The examples in (29) illustrate the construction known as “Clitic Left Dislocation” (CLLD), pervasive in Spanish and Romance languages. In CLLD, a constituent appears displaced within the left edge of the clause and is connected with the rest of the clause by a resumptive clitic, with which the dislocated constituent shares features such as person, number, gender, Case and specificity. The analyses of CLLD have typically focused on direct objects and can be classified into two main families. On the one hand, in some analyses the dislocated constituent is conceived as generated by movement (Cinque 1977, Rivero 1980, Rizzi 1997, Cecchetto 2000, Aoun and Benmamoun 1998, Escobar 1997, and Villalba 2000). On the other hand, other analyses assume that the dislocated constituent is base generated in its surface position (Cinque 1990, Iatridou 1991, Suñer 2006, Ordóñez and Treviño 1999 among others). Movement and base generation analyses have problems to account for the source of the alternative order of (29), illustrated with a root context in (30), but also possible in embedded contexts. In these examples, the dislocated arguments corresponding to the direct and indirect object can appear in presence of a preverbal 26 subject and an intonational break is clearly detected for the sentence to be felicitous (represented by a comma): (30) a. Estas esculturas, Omar las=compró. these sculptures Omar ACC=buy:3;PST “These sculptures, Omar bought them” b. A sus amigos, Omar les=regaló una escultura. DAT his friends Omar DAT=give:3;PST a sculpture “To his friends, Omar gave a sculpture.” If the dislocated constituents in (30) are understood as base generated constituents, the data posses a problem to base generation and movement approaches because it is not clear why this strategy is available in (30), but not in (29). Similarly, if the constituents in (30) arise by movement, the Standard Analysis must relax the strong claim about preverbal subjects as dislocated constituents because dislocated preverbal subjects are expected to block movement to the left edge. However, if this weak version of the Standard Analysis were postulated, the problems in the characterization of the [Spec, TP] and Nominative Case that were pointed out in Section 1.3.1 would emerge again. This problem is critical especially because although the constructions with direct and indirect objects in (29) are taken as CLLD, they have been shown to display different properties. For instance, Campos and Zampini (1990) have claimed that in contrast to direct objects, indirect objects in the dislocation construction are generated exclusively by movement. In addition, they claim that the associated clitic 27 in indirect objects is an agreement marker and not a resumptive clitic as in the case of direct objects. Another difference has been pointed out by Zubizarreta (1998a) and Suñer (2000a, 2000b), precisely in the interpretation of the dislocated constituents with respect to the preverbal/postverbal position of subjects. Zubizarreta notes that whereas a dislocated direct object in the CLLD construction can reconstruct below preverbal subjects, reconstruction does not take place with postverbal subjects. Suñer complements this observation by noting that this contrast is absent with dislocated indirect objects in CLLD, in which reconstruction is possible with preverbal and postverbal subjects. Although the analysis in this dissertation mainly focuses in the dislocation of subjects, future research is needed to extend the analysis to nonsubjects. Here, I will only point out that the tendency in the field and within the Standard Analysis is to assume that the analysis of dislocated subjects is fundamentally similar to the analysis of nondislocated subjects, a claim that I will show to be inaccurate. 1.3.3 Discussion The approach depicted in Section 1.3.1 and Section 1.3.2 is not exempted from problems, both on the empirical side and on the theoretical side. The Standard Analysis fails to characterize Spanish in terms of descriptive adequacy and explanatory adequacy. In empirical terms, the characterization of unmarked word order versus marked word order in the Standard Analysis does not respond to the general intuition that speakers feel in which SVO represents the neutral arrangement of constituents in 28 a sentence. This is obvious from the complications to characterize the status of preverbal subjects. If unmarked or neutral order is said to correspond to the arrangement of the arguments of a verbal predicate when they occupy a Case or a θ- position, and preverbal subjects are dislocated topics in the left edge; the intuition of SVO as neutral can never be characterized. In addition, a second empirical problem concerns about the description of the position of the inflected verb with respect to the left edge. As mentioned before, researchers such as Suñer (1994) have claimed that the inflected verb in Spanish can only reach inflectional nodes. 7 However, there is evidence for the conclusion that this is not true. This is probably due to the fact that the investigation initiated under the assumptions of the Standard Analysis, required the refinements of descriptive paradigms and syntactic tools about the left periphery. For instance, if the position of the complementizer que ‘that’ in Spanish is the head of CP, once one assumes a projectionist approach such Rizzi’s (Rizzi 1997), in which the CP field consists of several sublayers of functional projections, the claim about the positioning of the inflected verb in Spanish requires either (dis)confirmation or refinement. Although it is not the purpose of this dissertation to fill this descriptive gap, I want to present empirical evidence for the idea that left edge nodes are available as a landing site of verb movement based on two studies on the topic: Lois (1989) and Bok Bennema (2001). The argument is based on Cinque’s study of adverbial placement. In short, Cinque distinguishes between several types of adverbs among 7 The evidence provided by the paradigms reported by Suñer is actually incomplete. She shows that the inflected verb can reach the inflectional layer under theoretical and empirical considerations, but she does not show convincingly that the inflected verb must remain there. 29 them manner adverbs (i.e. lógicamente ‘logically’), which mark the left edge of the vP-VP complex; modal adverbs (probablemente ‘probably’) which are specifiers of the highest inflectional projection and several other adverbs that I will refer to as “middle adverbs” (inmediatamente ‘immediately,’ siempre ‘always,' ya ‘already,’ frecuentemente ‘frequently.’), which appear within the inflectional layer or its projections. Adopting Cinque’s idea that adverbs can be used to detect layers of the functional system, verb movement to the left edge can be argued to be available in Spanish if one examines the relative position of the inflected verb with respect to manner, middle and modal adverbs. If verb movement is impossible in Spanish, manner, modal and middle adverbs must precede the inflected verb. On the contrary, if verb movement is possible, one should observe that the inflected verb can precede the position of the relevant adverbials. As illustrated in (31), main verbs in Spanish must precede manner adverbs, indicating that they leave the vP/VP layer: (31) Manner Adverbs a. Cristina lógicamente contestó esa pregunta. Cristina logically answer:PST that question "*Cristina answered that question in a logical manner." "Obviously, Cristina answered that question." 30 b. Cristina contestó lógicamente esa pregunta. Cristina answer:PST logically that question "Cristina answered that question in a logical manner." "*Obviously, Cristina answered that question." The adverb lógicamente must follow the main verb in (31) to give rise to the interpretation where Cristina did the action of answering certain question in a logical manner. This interpretation is only available in (31)b. Crucially, if the adverb precedes the verb -(31)a, it is either interpreted as a parenthetical (dependent on whether there is additional material as for instance the preverbal subject Cristina) or as having a sentential scope, in which case lógicamente conveys a different meaning, namely that Cristina obviously answered the question. In addition, Spanish main verbs can either precede or follow middle and modal adverbs. Sentences with middle adverbs in (32) represent the sequences with the verb preceding or following the adverbs and sentences in (33) represent these two possibilities with modal adverbs: (32) Middle Adverbs a. Cristina leyó inmediatamente un libro de lingüística. Cristina read:PST immediately a book of linguistics "Cristina immediately read a linguistics book." b. Cristina inmediatamente leyó un libro de lingüística. Cristina immediately read:PST a book of linguistics "Cristina immediately read a linguistics book." 31 (33) Modal Adverbs a. ... probablemente le=hablaste a tu hermano. probably DAT=talk:PST DAT your brother “You probably talked to your brother." b. ... le=hablaste probablemente a tu hermano. DAT=talk:PST probably DAT your brother “You probably talked to your brother." From these paradigms, one can conclude that the inflected verb in Spanish must move to some projection within the inflectional layer. However, it also suggests that the inflected verb could appear displaced beyond the inflectional layer and reach the CP nodes, a fact ignored in the Standard Analysis. A more critical problem derives from the incomplete characterization of these facts within the Standard Analysis. In the literature on Germanic languages, the phenomenon known as verb second (V2) has been characterized as movement of the finite verb to the left edge. If the empirical paradigms indicate that the finite verb in Spanish can also appear displaced within the left periphery, one must also contemplate the possibility that Spanish displays some form or variant of the verb second phenomenon. Although this option has been considered to be absent for historical reasons (Fontana 1993), rejected (Suñer 1994) or simply assumed (Mallén 1993), no careful characterization of the phenomenon exists in the field. The arguments reduced to the comparison of Spanish with German, the prototypical example in the V2 literature, 32 leading to biased conclusions, especially if the verb second phenomenon is considered a uniform phenomenon. Even though German and Spanish display obvious differences, these do not necessarily disqualify the claim that a V2 mechanism might be involved in Spanish. As a matter of fact, several researchers such as Vikner (1995), den Besten (1983), Platzack (1983), and Cardinaletti and Roberts (1990) among others have observed that even within Germanic and Scandinavian languages, there is variation in the verb second parameters illustrated by German. Yet, these languages are argued to display a second position effect. 8 Finally, under theoretical considerations the approach advocated by the Standard Analysis, especially in versions based on the “Generalized EPP property”, seems counterintuitive and not accord with the minimalist spirit. One still needs to justify whether unification of several mechanisms under the label “EPP” leads to a better characterization of the faculty of language. Although the generative enterprise has been relatively successful to tackle issues of descriptive and explanatory adequacy with the arrival of minimalism the focus of linguistic research shifted to how to evaluate the different conceivable versions of the Principles and Parameters models. Probably, the best alternative to the problem is one elaborated on the nonexistence of an “EPP” mark. This alternative should be able to recognize the underlying mechanisms responsible for the displacement property without losing descriptive or explanatory force. 8 Romance languages replicate a similar situation: V2 does not constitute a homogeneous phenomenon. As pointed out by Poletto (2000), Old Spanish (see also Fontana 1993) and Old French V2 were not confined to root contexts and Old Italian also displayed V2. Several Northern Italian Dialects also display variations in the parameters illustrated for German although they still display V2. 33 1.4 Outline of the dissertation Our investigation of the Spanish left edge departs from an elaboration of the projectionist approaches by Rizzi (1997), Poletto (2000) and others. Accordingly, the complementizer system is a field that hosts several functional projections, a mirror of the articulation of the thematic system (Larson 1988), and inflectional system (Emonds 1978, Pollock 1989, Chomsky 1993 and others). Informational and discoursive roles such as new/old information, point of view, declarative, interrogative and imperative force as well as modality are encoded in different heads: (34) ForceP ... IntP … TopP* ... FocP ...TopP*... FinP ... IP Higher left periphery Lower left periphery Based mainly on conceptual considerations 9 , I will sketch a more reduced version of Rizzi’s system in this dissertation called “the Narrow Left Edge,” In the Narrow Left Edge system, there are two fields: the higher left edge and the lower left edge. Each field is headed by only one projection. The Point of View or Perspective Projection (CP) heads the higher left edge and the Assertion Phrase c*P heads the lower left edge: (35) ... CP ... c*P ... IP HIGHER LEFT EDGE LOWER LEFT EDGE Within the Narrow Left Edge, there is no independent projection to represent the force or the clausal type of a sentence. Instead, clausal typing is the result of the existence of the arrangement of a primitive set of features into a template. The Point 9 For a more detailed treatment of the system envisioned in the Narrow Left Edge, the reader is referred to Beas (2007). 34 of View Phrase (CP), located in the higher left edge is in charge of encoding the orientation of the speech act towards the speaker or towards the hearer. Informally, c*º can be equated to the combination of Rizzi’s Finite Phrase and Focus Phrase within the lower left edge. Finally, the new system seeks to capitalize on Baker’s observation that agreement and dislocation are closely interrelated (Baker 1996). Accordingly, higher and lower Topic Phrases and their recursivity results from adjunction of a maximal projection to the specifier of CP (higher topics) or movement of a maximal projection to the specifier of c*P (lower topics). Higher and lower topics are not a property of a certain functional projection, but they are a property of the syntax of the lower and higher left edge. This dissertation is organized into six chapters as follows. In Chapter 2, I examine the syntax of preverbal subjects in declarative clauses in Modern Spanish. I argue that two patterns of preverbal subjects must be recognized: a nondislocation pattern and a dislocation pattern. Importantly, the nondislocation pattern shows A- related properties, suggesting the obligatoriness of movement to the [Spec, TP] for Case theoretical reasons. I call this pattern “Canonical”, and I claim that it corresponds to the neutral order of constituents. Given the logic of Last Resort, I also conclude that in the VSO word order, the subject is also in a canonical position and hence VP-external, a conclusion reached by Zubizarreta (1994). In Chapter 3 and Chapter 4, I investigate the properties of nondeclaratives and declarative clauses respectively. Based on the left edge projections of Rizzi’s 1997 influential paper, I sketch a Narrow Left Edge, a reduced version of the left 35 periphery. Based on Roberts (2005), I develop a typological argument to argue for the availability of a dedicated position in the Spanish left periphery. This syntactic position is the head of the Assertion Phrase. The head c* is responsible for the creation of a positional effect similar to the one found in classical V2 languages. In Chapter 5, I analyze the dislocation pattern and the properties of the c*º position. Contra the Standard Analysis, I argue that the relation between the resumptive and dislocated subjects is not only one of base generation. Subjects can be attracted by movement to the specifier of c*º, a dislocation position. These observations follow from the idea that c*º is an agreeing head within the left edge, establishing an Agree relation with the structure. Importantly, the satisfaction of Agree in c*º creates an intervention effect to further extraction or movement operations targeting the left edge. This is the source of the phenomenon known as multiple dislocation. Given that movement is blocked after the c*º head has fulfilled its licensing requirements, only adjunction with resumption is allowed. Multiple dislocation is then the result of the combination of agreement in c*º and multiple adjunction once the specifier of c*º has been occupied by a maximal projection. 36 Chapter 2 Lexical Preverbal Subjects: Dislocation and Canonical Patterns Within the Standard Analysis, a widespread approach to the characterization of preverbal subjects is provided by the dislocation analysis, which can be stated as in (1): (1) A preverbal lexical subject is base generated outside its thematic position within the left periphery. The empirical claim that the dislocation analysis puts forward is that for every sequence SV(O), preverbal lexical subjects must be generated by left dislocation. In other words, preverbal lexical subjects are always represented as adjuncts to some functional projection, presumably located within the CP field as depicted in 0: (2) Preverbal Subjects in the Standard Analysis CP 2 Preverbal IP Subject 2 Iº VP 2 e V’ 2 Vº As adjuncts, preverbal lexical subjects are related to a hypothetical element e in thematic position. In some versions of the dislocation analysis, the element e is a covert pronominal (i.e., pro) moving to [Spec, TP] for Case theoretical reasons (Alexiadou and Anagnostopoulou 1998) and in other versions, e is overtly realized as the subject agreement morphology, which absorbs Case and incorporates into Iº (Ordóñez 1997). However the nature of the element e is realized in the dislocation 37 analysis, preverbal lexical subjects are typically thought as indirectly related to a thematic position and to a Case theoretic position. The main goal of this chapter is to explore the accuracy of the idea that lexical preverbal subjects can be unambiguously described as left dislocated constituents in Spanish. In fact, I demonstrate in Section 2.2 and Section 2.3 that at least two patterns must be recognized in Spanish preverbal subjects: those in canonical positions and those that function as dislocated constituents. 2.1 Preverbal lexical subjects, EPP and pro-drop languages One way to test whether the properties of preverbal subjects are in fact those attributed to left dislocated constituents is to consider Alexiadou and Anagnostopoulou's influential 1998 work on the syntax of subjects in pro-drop languages. In contrast to English, which obligatorily requires subjects to fill the [Spec, TP] position, pro-drop languages such as Greek do not show the same obligatoriness. This important difference is due to a parameterization of the way the EPP-feature is checked. In pro-drop languages, the EPP is checked by Xº-movement of the verb; therefore, no additional movement of the subject to [Spec, TP] is needed. In contrast, in non-pro-drop languages, the EPP-feature is checked either by XP- movement of the NP subject or by insertion of an expletive in passives, raising, and unaccusatives. 10 Consequently, the alternation of VSO and SVO orders in pro-drop 10 Importantly, Alexiadou and Anagnostopoulou assume that movement is not triggered for morphological reasons as has been assumed in earlier versions of the Minimalist Program. Accordingly, the mechanism of Nominative Case licensing is the same in pro-drop and non- pro-drop languages. In both cases subjects are licensed in their Case-features and φ-features in [Spec, VP] (in-situ). If subjects must move in a non pro-drop language such as English, it is because Tº requires a specifier (i.e., bears the EPP property) and there is no alternative or 38 languages suggests that some mechanism other than the one active for non pro- drop languages such as English must be responsible for the generation of preverbal subjects. According to Alexiadou and Anagnostopoulou, this mechanism is reflected through the strategy of left dislocation. Left dislocation represents preverbal lexical subjects as constituents generated by adjunction to a non-Case position (i.e., in Ā- position) 11 and linked to a resumptive pro in θ-position. The first argument for the dislocated nature of preverbal subjects provided by Alexiadou and Anagnostopoulou is based on the distribution of certain middle field adverbials such as yesterday, tomorrow, and last year in relation to preverbal subjects. On the basis of the observation that these adverbs modify an IP and have the inflectional domain within their scope when they appear in sentence initial and final positions, it is reasonable to conclude, as Alexiadou and Anagnostopoulou (1998: 502, ex. 15a) do, that when preverbal subjects precede these adverbs, they must be dislocated and they must be in a position other than the [Spec, IP]: (3) O Petros xtes meta apo poles prospathies sinandise ti Maria. the Peter yesterday after from many efforts meet:PST the María "After so many efforts, yesterday Peter met María." Other important type of evidence for the dislocated nature of preverbal subjects in Greek supplied by Alexiadou and Anagnostopoulou has to do with the interpretation more economical way to fulfill this condition. Pro-drop languages, however, check the EPP- property by moving the verb to Tº, an operation that does not expand or create more structure. 11 For the ease of exposition, this is the sense in which I will distinguish between A- and Ā- positions: a position where a certain expression is assigned Case is an A-position, and a position where a certain expression is not Case marked is an Ā-position. 39 of certain elements relative to their structural position. They observe that in pro- drop languages the scope properties that a quantifier phrase displays in its initial/thematic position are normally preserved after the application of overt movement. Consequently, whereas preverbal subjects in English are ambiguous as in (4)a, their Greek counterparts are not ((4)b and (4)c), suggesting their left dislocated nature as reported by Alexiadou and Anagnostopoulou. The scope ambiguities observed in English are due to the fact that there is a copy or a trace of the DP subject in [Spec, VP]. Thus, whereas English resorts to copying the subject in its VP- internal position onto the [Spec, TP] to check the EPP (i.e., "A-movement"), Greek does not need to resort to this strategy because the verb movement has already checked the EPP and Agree has checked Nominative for the subject. Therefore, unambiguous scope in Greek SVO can be understood as the lack of a copy or a trace within the VP. As represented below, a pronominal like element e with a covert realization (i.e., pro) or with an overt realization (i.e., the agreement subject morphology) may occupy the [Spec, VP] position: (4) a. [ IP Some student i [ VP t i filed every article]]. AMBIGUOUS b. [ IP Kapios fititis [ VP e i stihiothetise kathe arthro]]. some student filed every article "Some student filed every article." NOT AMBIGUOUS 40 c. [ IP Ena pedhi [ VP e i diavase kathe arthro]]. a child read every article "A child read every article." NOT AMBIGUOUS Sentence (4)a in English either means that there is a student such that he filed every article or that for every article available, there is a different student who filed each of them. In contrast, the Greek sentence in (4)b only accepts the first interpretation. Similarly, one would expect that the scope interaction between an indefinite preverbal subject in Greek and a head such as a modal ((5)) or negation ((6)) should also give rise to unambiguous interpretations. Empirical data corroborate this point: (5) a. Prepi na parousiasi ena pedhi tin ergasia mexri to telos tis vdomadas. must SBJV present a child the essay until the end the week AMBIGUOUS b. Ena pedhi prepi na parousiasi tin ergasia mexri to telos tis vdomadas. a child must SBJV present the essay until the end the week NOT AMBIGUOUS (6) a. Dhen eroteftikan poli andres ti Maria. NEG fell.in.love many men the Mary AMBIGUOUS b. Poli andres dhen eroteftikan ti Maria. many men NEG fell.in.love the Mary NOT AMBIGUOUS 41 Alexiadou and Anagnostopoulou (1998: 506, ex. 23) also observe that indefinite preverbal subjects obligatorily receive a strong (partitive/specific) interpretation but do not receive an existential interpretation: (7) a. Ena pedhi diavase to "Paramithi horis Onoma." a child read the fairy.tale without name "A certain child/one of the children read Fairytale without a title." ONLY SPECIFIC b. diavase ena pedhi to "Paramithi horis Onoma." read a child the fairy.tale without name “A child read ‘Fairytale without a title’.” OR “A certain child/one of the children read ‘Fairytale without a title’.” SPECIFIC OR NONSPECIFIC In contrast to preverbal indefinite subjects in English, which are ambiguous between a specific interpretation and a nonspecific or existential one, Greek preverbal indefinite subjects in (7)a can only be taken as presupposing the existence of an entity, and are thus partitive/specific. However, if the subject is post-verbal as in (7)b, the existential reading is accessible. The availability of unambiguous specific readings is then correlated with subjects in preverbal position, an indication of dislocation in Greek. 2.2 Preverbal Subjects in Canonical and Dislocated Positions In Alexiadou and Anagnostopoulou’s logic, Spanish, as a pro-drop language that displays verb movement and VSO/SVO alternation should show properties of 42 preverbal subjects similar to those in Greek. In this section, I examine the status of preverbal subjects in Spanish and show that they do not mirror their Greek counterparts despite claims to the contrary (Ordóñez 1997, Ordóñez and Treviño 1999, Olarrea 1996, Uribe Etxebarria 1995, among others). Instead, the empirical paradigms suggest that some environments are compatible with a dislocated subject analysis and others are not. 2.2.1 Specific and Nonspecific Preverbal Subjects Recall that the basis of Alexiadou and Anagnostopoulou’s argument for a dislocation analysis rests on the observation that preverbal indefinite subjects in Greek obligatorily receive only a strong (partitive/specific) interpretation and are never interpreted existentially (see (7)). The obligatoriness of strong/specific readings in preverbal subjects is taken as lack of raising or/and reconstruction within VP (See Fox 1993 and Diesing 1992). I demonstrate that there is no obligatory specificity requirement for preverbal lexical subjects in Spanish and that in fact it is possible to find environments where preverbal subjects are nonspecific. In order to test for the specificity of a constituent in Spanish, I will consistently apply the observation that specific indefinites are incompatible with restrictive clause modification in the subjunctive mood (Leonetti 2003, 2004; Moreno Quibén 2000). Therefore, if specificity is obligatory in preverbal subjects, these subjects would be expected not to tolerate modification by a relative clause in subjunctive. 43 To begin with, it is important to note that the ambiguity between a specific or nonspecific interpretation of a given indefinite seems sensitive to whether the context is opaque (Jackendoff 1971). 12 I illustrate this point with examples from Spanish: (8) a. Un lingüista busca un trabajo en el MIT. a linguist looks.for a job in the MIT “a linguist is looking for a job at MIT.” b. Un lingüista maneja este auto todos los días. a linguist drives this car all the days “A linguist drives this car everyday.” If preverbal subjects in the examples in (8) are modified by a relative clause in the subjunctive, only the subjects of intentional/propositional verbs clearly accept specific/nonspecific readings. For instance, the subject of buscar in (8)a clearly accepts modification by a relative clause inflected either in indicative mood (specific reading) or in subjunctive mood (nonspecific reading) as illustrated in (9): 12 In general, an opaque context is one in which existential generalization or substitution of coreferential terms is blocked by the nonspecific interpretation of the indefinite. Intentional or propositional verbs (such as buscar ‘to look for’; or aceptar ‘to accept’; aconsejar ‘to advise’; creer ‘to believe’; desear ‘to wish’; intentar ‘to try’; merecer ‘to deserve’; pedir ‘to ask for’; perseguir ‘to chase’; prometer ‘to promise’; proponer ‘to propose’; querer ‘to want’; sugerir ‘to suggest’) typically legitimate opaque contexts. 44 (9) a. Un lingüista que se=gradú-a con honores busca un a linguist that REFL=graduate-3PRS;IND with honors looks.for a trabajo en el MIT. job in the MIT “A (specific) linguist that graduates with honors looks for a job at the MIT.” SPECIFIC READING b. Un lingüista que se=gradú-e con honores busca un a linguist that REFL=graduate-3PRS;SBJV with honors looks.for a trabajo en el MIT. job in the MIT “Some linguist that graduates with honors looks for a job at the MIT.” NON SPECIFIC READING In contrast, the subject of a nonintentional/nonpropositional verb such as manejar in (8)b shows a tendency to reject a modification in the subjunctive ((10)): (10) a. Un lingüista que se=gradú-a con honores maneja a linguist that REFL=graduate-3PRS;IND with honors drives este auto todos los días. this car all the days “A linguist that graduates with honors drives this car everyday.” 45 b. ??Un lingüista que se=gradú-e con honores maneja a linguist that REFL=graduate-3PRS;SBJV with honors drives este auto todos los días. this car all the days “A linguist that graduates with honors drives this car everyday.” In embedded clauses, a similar effect is obtained: subjects of nonintentional verbs reject modification by a relative clause inflected in the subjunctive: (11) a. *Creo que un lingüista que se=gradú-e con believe:1 that a linguist that REFL=graduate-3PRS;SBJV with honores maneja este auto todos los días. honors drives this car all the days “I believe that a linguist that graduates with honors drives this car everyday.” b. *Me=parece que un lingüista que se=gradú-e con me=seems that a linguist that REFL=graduate-3PRS;SBJV with honores maneja este auto todos los días. honors drive:3 this car all the days “It seems to me that a linguist that graduates with honors drives this car everyday.” Moreover, if there is an operator in the complementizer position such as si "whether" in (12)a, the nonspecific reading of the preverbal indefinite subject is licensed: 46 (12) Me=pregunto si un lingüista que se=gradú-e me=wonder whether a linguist that REFL=graduate-3PRS;SBJV con honores maneja este auto diariamente. with honors drives this car daily “I wonder whether a linguist that graduates with honors drives this car everyday.” NONSPECIFIC READING POSSIBLE A similar outcome is obtained with if-clauses. As shown by (13), the preverbal subject can be interpreted as nonspecific even in absence of subjunctive mood: 13 (13) Si un lingüista se=gradú-a con honores, recibirá if a linguist REFL=graduate-3PRS;IND with honors, receive:FUT apoyo económico. support economic “If a linguist graduates with honors, he will receive financial support.” NONSPECIFIC READING POSSIBLE Note that in addition to their specific interpretation, sentences with indefinite NPs that express generic properties of the entities of which they are predicated can also receive nonspecific interpretations: 13 In Spanish, even when the complementizer si “if” or “whether” looks similar to the affirmative marker element sí “yes” in (12) and (13), they have a different morphosyntactic status and prosody. The affirmative marker is more restricted in its distribution and can be separated from the verb by a small set of elements. This is evidently not so in the case of the si complementizer. In addition, assertion elements can bear emphatic stress while complementizers cannot. 47 (14) a. un lingüista detesta las ceremonias. a linguist hates the ceremonies “A linguist hates the ceremonies.” b. un lingüista que voluntariamente lleg-a tarde a linguist that voluntarily arrive-3PRS;IND late detesta las ceremonias. hates the ceremonies “A linguist that voluntarily arrives late hates the ceremonies.” c. un lingüista que voluntariamente llegu-e tarde a linguist that voluntarily arrive-3PRS;SBJ late detesta las ceremonias. hates the ceremonies “A linguist that voluntarily arrives late hates the ceremonies.” A typical observation is that constituents that receive a specific interpretation are incompatible with existential constructions. As illustrated in (15), the derived subject una lingüista “a linguist” cannot be marked with the accusative marker ‘a’ which is known to produce a specific interpretation. Nonetheless, this subject is compatible with a preverbal position: 48 (15) a. Había (*a) una lingüista en esta habitación, pero ahora had:IPFV ACC a linguist in this room but now ya no la=veo. anymore NEG her=see "There was a linguist in this room but I do not see her anymore." b. Una lingüista había en esta habitación, pero ahora a linguist had:IPFV in this room but now ya no la=veo. anymore NEG her=see “There was a linguist in this room but I do see her anymore.” Schneider Zioga (2002, 2007) has observed that nonveridicality/irrealis contexts are semantically incompatible with topichood and specificity. 14 This observation is consistent with the availability of nonspecific readings for preverbal subjects when future tense is added. Contrast (10) with (16): (16) a. Un lingüista que se=gradú-a con honores manejará a linguist that REFL=graduate-3PRS;IND with honors drive:FUT este auto todos los días. this car all the days “A linguist that graduates with honors will drive this car everyday.” 14 See Section 2.3.3 for data suggesting that affirmative/negative particles do not constitute irrealis/non-veridicality contexts in syntactic terms. Their role seems to be oriented toward marking what was asserted or denied in the focus structure of a sentence. 49 b. Un lingüista que se=gradú-e con honores manejará a linguist that REFL=graduate-3PRS;SBJ with honors drive:FUT este auto todos los días. this car all the days “A linguist that graduates with honors will drive this car everyday.” Similarly, indicative mood is incompatible within the scope of irrealis markers such as ojalá "perhaps." For instance, in sentence (17)a, the subject un lingüista, modified by a relative clause in indicative, becomes ungrammatical or less acceptable in presence of ojalá. In contrast, if the relative clause is inflected in subjunctive as in (17)b, the sentence is perfect, indicating the availability of a nonspecific reading for the preverbal subject: (17) a. *Ojalá que un lingüista que se=gradú-a con perhaps that a linguist that REFL=graduate-3PRS;IND with honores manej-e este auto todos los días. honors drive-3PRS:SBJV this car all the days “Perhaps a linguist that graduates with honors will drive this car everyday.” 50 b. Ojalá que un lingüista que se=gradú-e con perhaps that a linguist that REFL=graduate-3PRS;SBJV with honores manej-e este auto todos los días. honors drive-3PRS:SBJV this car all the days “Perhaps a linguist that graduates with honors will drive this car everyday.” In conclusion, Spanish displays several environments where it is possible to find preverbal subjects with a nonspecific interpretation (opaque contexts, generics, existential constructions and irrealis/nonveridicality contexts). 15 Given this evidence, the claim that preverbal subjects in Spanish are in fact always dislocated must be relaxed: some preverbal subjects are associated with specificity and topicality and some are not. 2.2.2 Evidence from Adverb Placement According to Alexiadou and Anagnostopoulou, the placement of middle field adverbs following preverbal subjects indicates left dislocation. This is illustrated in (18): (18) 2 Dislocated position ZP= “middle field adverb” for the subject 2 ADVERB IP 2 Canonical position I’ for the subject (…) 15 Note that indefinite preverbal subjects also display specific and nonspecific readings when affected by modals. For details, see Section 2.3.2. 51 Again, observe that this is not always the situation in Spanish, as preverbal subjects can precede or follow middle field adverbs: 16 (19) a. Omar ayer invitó a Cristina. Omar yesterday invite:PST ACC Cristina “Yesterday, Omar invited Cristina.” b. Ayer, Omar invitó a Cristina. yesterday Omar invite:PST ACC Cristina “Yesterday, Omar invited Cristina.” It is also the case that intonational or prosodic indicators do not help to clarify the identity of these two preverbal locations. For instance, the presence of an intonational break between the preverbal subject and ayer "yesterday" in (19)a is not obligatory and cannot be taken as immediate evidence for dislocation or nondislocation of preverbal subjects. As an illustration, Suñer (2003: 343, ex. 3a and 3a’) notes that middle field adverbs preceding preverbal lexical subjects do not get a parenthetical reading. According to her, this may be taken as “an indication that [...] [preverbal lexical subjects] are not dislocated”: (20) a. El maestro probablemente encontrará rápidamente el error the teacher probably find:FUT quickly the mistake "The teacher will probably find the mistake quickly." 16 According to Schneider Zioga (p.c.), the pattern in (19) also holds in Greek. 52 b. *El maestro rápidamente encontrará probablemente el error. the teacher quickly find:FUT probably the mistake "The teacher will probably find the mistake quickly." In fact, the intonational break does not even distinguish cases with clearly dislocated constituents as CLLDed objects. In these cases, the intonational break between the dislocated constituent and the adverb (represented by ‘#’) is also optional: (21) A Cristina (#) ayer la=invitó Omar. ACC Cristina (#) yesterday her=invite:PST Omar “Cristina, Omar invited her yesterday.” One way to investigate whether the sequences subject-adverb and adverb- subject in (19) correspond to two different structures is by considering the context question in which each instance appears. Consider the context questions in (22) and the paradigm involving clitic left dislocated objects in (23): (22) a. ¿Qué pasó? what happen:PST “What happened?” b. ¿Qué hizo Omar? what do:PST Omar “What did Omar do?” c. ¿Quién invitó a Cristina? who invite:PST ACC Cristina “Who invited Cristina?” 53 d. ¿A quién invitó Omar? ACC who invite:PST Omar “Whom did Omar invite?” (23) a. A Cristina ayer la=invitó Omar. ACC Cristina yesterday her=invite:PST Omar “Cristina, Omar invited her yesterday.” b. Ayer a Cristina la=invitó Omar. yesterday ACC Cristina her=invite:PST Omar “Cristina, Omar invited her yesterday.” A crucial difference between these two paradigms is that the dislocated objects in (23) and the preverbal subject Omar in (19)a never provide a proper answer to a wide focus question such as ¿Qué pasó? regardless of the order between a Cristina and ayer. For instance, (23)a and (23)b are compatible with the context question ¿Quién invitó a Cristina? in (22)c, which has a narrow focus interpretation on the subject Omar. Similarly, the subject in (19)a is only felicitous as an answer to a context question that scopes over the VP such as ¿Qué hizo Omar? (e.g., (22)b) or as an answer to a context question prompting narrow scope over the object such as ¿A quién invitó Omar? (e.g., (22)d). In contrast, only sentence (19)b, with the adverb preceding the preverbal subject, is neutral with respect to intonational and prosodic factors that affect the information structure of a proposition, because (19)b can perfectly answer a wide focus question such as that in (22)a. In other words, example (19)b corresponds to 54 the unmarked or neutral word order of constituents. This means that if we do not consider the adverbial in these paradigms, example (19)b becomes ambiguous in its context question: it can be a natural answer to (22)a (wide focus on the whole proposition) or (22)b (narrow focus on the verb and its object). These observations imply that it is plausible that Spanish actually encodes different properties and structural positions for what superficially are understood as “preverbal subjects.” In particular, the syntax of Spanish might better be described as containing at least two options for the positioning of preverbal subjects: a dislocated pattern and a canonical position. For instance, note that the possibility of preceding or following an adverbial does not always occur as in (19). Consider the situation with the adverbial ojalá “maybe”, which arguably occupies the functional space between TP and CP, triggering irrealis mood (subjunctive) on the verbal inflection. If ojalá is present, the preverbal subject cannot precede this adverbial: (24) a. Ojalá Cristina/una sintactista invit-e a Omar perhaps Cristina/a syntactician invite-3PRS;SBJV ACC Omar a su graduación. to her graduation "Perhaps Cristina/a syntactician will invite Omar to her graduation." 55 b. *Cristina/*una sintactista ojalá invit-e a Omar Cristina a syntactician perhaps invite-3PRS;SBJV ACC Omar a su graduación. to her graduation "Perhaps Cristina/a syntactician will invite Omar to her graduation." As shown by the grammaticality of (24)a, ojalá triggers obligatory subjunctive mood on the verb, a context which was shown in Section 2.1 to be opaque. In other words, (24)a suggests that the position occupied by indefinite subjects such as un sintactista “a syntactician” cannot be described as a dislocated position because no specificity requirement is forced. Furthermore, given that ojalá divides the CP and IP systems, the location of lexical subjects in examples such as (24)a must correspond to some specifier within the inflectional layer. Another argument for the existence of a nondislocated position can be constructed based on the distribution of temporal adverbs with respect to nonreferential elements. A widespread observation has been that nonreferential elements resist dislocation (Cinque 1990, Baker 1996, Schneider Zioga 1993, 2007). 17 From an initial inspection, this observation seems to hold in Spanish for clear cases of dislocation of direct objects, but not for subjects and indirect objects: 18 17 For example, Baker (1996) has convincingly shown that in languages such as Mohawk where the requirement on obligatory dislocation of constituents is strictly followed, there is no equivalent to words such as nobody and that in order to express a similar meaning a periphrastic form is used instead. Crucially, this form is referential and not quantificational. 18 Section 2.2.1 shows that irrealis contexts are special in Spanish because they do not trigger an obligatory condition on specificity. The paradigm in (28) leads to a similar conclusion: nonstructural cases such as the dative tend to obviate the requirement on 56 (25) a. *A nadie la=invitó Omar. ACC nobody her=invite:PST Omar “Omar invited nobody.” b. A nadie le=habló Mimi. DAT nobody DAT=speak:PST Mimi “Mimi spoke to nobody.” c. Nadie invitó a Omar. nobody invite:PST ACC Omar “Nobody invited Omar.” Let us assume with Alexiadou and Anagnostopoulou that middle field adverbs divide the clause structure into two different positions. Thus, the field following the temporal adverb ayer “yesterday” in (19) corresponds to the TP projection, whereas the field above the same temporal adverb indicates higher projections including left edge nodes. The paradigm in (19) is repeated as (26) for convenience: (26) a. Omar ayer invitó a Cristina. Omar yesterday invite:PST ACC Cristina “Yesterday, Omar invited Cristina.” b. Ayer, Omar invitó a Cristina. yesterday Omar invite:PST ACC Cristina “Yesterday, Omar invited Cristina.” specificity. See Schneider Zioga (2002, 2007) for a discussion of the relation between the antiagreement effect and the lack of structural Case in Chamorro and Palauan. 57 If preverbal subjects must be dislocated in Spanish, we would expect that the replacement of the preverbal subject Omar by a nonspecific indefinite such as nadie “nobody” in (26)a would be ungrammatical in parallel to clearly dislocated objects. Furthermore, a nonspecific indefinite should be able to replace the preverbal subject Omar in (26)b, showing strict adjacency between the finite verb and the preverbal subject. This would only be possible if a canonical position is available. These two expectations are corroborated by the empirical data. Observe that nadie cannot appear as a dislocated subject ((27)a), but nadie requires can still occur in canonical subject position as in (27)b: (27) a. *Nadie ayer invitó a Cristina. nobody yesterday invite:PST ACC Cristina “Yesterday, nobody invited Cristina.” b. Ayer, nadie invitó a Cristina. Yesterday nobody invite:PST ACC Cristina “Yesterday, nobody invited Cristina.” Moreover, indefinites that precede middle field adverbs must be obligatorily specific as shown by their incompatibility with modification by a relative clause in subjunctive: 58 (28) a. *Finalmente, un sintactista que trabaja-ra conscientemente finally a syntactician that work-PST;SBJV consciously ayer ganó una beca. yesterday win:PST a scholarship “Finally, a syntactician that works consciously won a scholarship yesterday.” b. Finalmente, ayer un sintactista que trabaja-ra finally yesterday a syntactician that work-PST;SBJV conscientemente ganó una beca. consciously win:PST a scholarship “Finally, a syntactician that works consciously won a scholarship yesterday.” OBLIGATORY SPECIFIC READING The adverbial ayer "yesterday" must precede nonspecific indefinite subjects as in example (28)b; otherwise, the sentence becomes ungrammatical as in (28)a. It is important to consider that in (28)a, the intended scope of the adverb ayer "yesterday" is the whole clause and not only the verb contained within the relative clause modifying the subject. Therefore, the reported judgment in this example does not include a pause after or before the adverbial ayer "yesterday", which would give rise to a parenthetical reading. Summarizing, in Section 2.2 I have shown evidence for two patterns of preverbal subjects in Spanish syntax. The empirical data suggest that preverbal subjects do not need be dislocated. This is evident from the absence of obligatory specific interpretations of preverbal subjects and adverb placement. Presumably, this 59 would indicate the existence of two distinct syntactic positions for preverbal subjects: one with dislocation of preverbal subjects and another with preverbal subjects in canonical position. 2.3 Evidence from Scope Interactions Recall that Greek showed a uniform wide scope pattern for QP subjects in preverbal position with respect to another QP or an Xº head with quantificational force, suggesting dislocation for these constituents. In contrast, in English, where similar structures give rise to wide or narrow scope readings, this is not what happens. Contra Alexiadou and Anagnostopoulou’s contention that Spanish preverbal subjects behave in a pattern parallel to Greek, we find a pattern that does not correspond completely to either Greek or English: (29) Indefinite + QP Indefinite + Modals Indefinite + Negation Indef>QP QP>Indef Indef>Mod Mod>Indef Indef>Neg Neg>Indef Greek ٭ ٭ ٭ English Spanish ٭ In other words, Spanish displays a mixed pattern. On the one hand, it displays an English pattern in the interaction between indefinites in subject position plus QPs, and indefinites in subject position plus modals. On the other hand, Spanish also displays a Greek pattern in the combination of indefinites in subject position plus negation. In sections 2.3.1 through 2.3.3, I will illustrate each of these different situations. 60 2.3.1 QP/QP Interactions First, the interaction of an indefinite and an every-NP phrase in object position gives rise to ambiguity, unexpected if preverbal subjects must be dislocated, as in Greek. Note that the ambiguous readings for indefinite subjects in the VSO order remain ambiguous in the SVO order. This can be illustrated with the two Spanish indefinites un ‘a’ and algún ‘some’ interacting with cada ‘each’. Consider first the case of algún in interaction with an every NP object quantifier. The example (30)a illustrates that indefinite subjects in postverbal position are ambiguous between a wide scope and a narrow scope reading with respect to the quantifier cada. As shown by the interpretations of (30)b, the object quantifier can scope over the preverbal subject position and retain the ambiguity of scope: (30) a. En este país, estudió algún lingüista cada lengua nativa in this country study:PST some linguist each language native exhaustivamente. exhaustively “In this country, some linguist studied each native language exhaustively.” AMBIGUOUS 61 b. En este país algún lingüista estudió cada lengua in this country some linguist study:PST each language nativa exhaustivamente. native exhaustively “In this country, some linguist studied each native language exhaustively.” AMBIGUOUS A similar pattern is obtained with the indefinite un interacting with an every NP phrase in object position. Both wide and narrow scopes of the object quantifier over the preverbal subject and postverbal position are available: (31) a. Finalmente, ganó un lingüista cada beca en su finally win:PST a linguist each scholarship in his universidad. university “Finally, a linguist won every scholarship from his university.” AMBIGUOUS b. Finalmente, un lingüista ganó cada beca en su finally a linguist win:PST each scholarship in his universidad. university “Finally, a linguist won every scholarship from his university.” AMBIGUOUS 62 Suñer reports examples with similar judgments. Example (32) illustrates the interaction between algún NP in subject position and the quantificational element cada NP in direct object position (Suñer 2003: 344, ex. 5): (32) a. En la biblioteca departamental, sacó prestado algún in the library departmental, take_out:PST borrow:PTCP some estudiante cada libro. student each book “In the departmental library, some student borrowed each book.” AMBIGUOUS b. En la biblioteca departamental, algún estudiante sacó in the library departmental, some student take_out:PST prestado cada libro. borrow:PTCP each book “In the departmental library, some student borrowed each book.” AMBIGUOUS Suñer (2003: 345, ex. 6) provides the contrasts in (33) to exemplify the interaction of the indefinite subject un interacting with cada NP in object position: (33) a. Montaba guardia un vigilante en cada esquina. stand:IPFV guard a policeman in each corner “A policeman was looking after each corner.” AMBIGUOUS 63 b. Un vigilante montaba guardia en cada esquina. a policeman stand:IPFV guard in each corner “A policeman was looking after each corner.” AMBIGUOUS As Suñer points out, in one interpretation, the preverbal indefinite subject is taken as specific, namely, algún and un are interpreted as having wide scope over the QP- cada NP as an instance of quantification over events. Therefore, for (33)b, it holds that there is a specific student x such that for each book y, x borrowed y at different times. It is important to note that the preferred reading for (32)b and for (33)b is the one which takes the indefinite as nonspecific, with narrow scope in relation to the QP-cada NP, which distributes over it. Under this interpretation, the sentence means for each book y, there is a different student x such that x borrowed y. This is an indication that some preverbal subjects in Spanish are not in dislocated positions or that they are not too high in the structure to prevent quantificational objects to scope over their positions. Leonetti (2000) remarks that appropriate contexts of elicitation of sentences with a verb in the subjunctive are those that entail a process or that express a purpose. 19 Therefore, the nonspecific interpretation of indefinite subjects in the b examples of (30), (31), (32) and (33) can be forced if they are modified by a relative clause inflected in the subjunctive and the proper context of elicitation is added. In 19 Despite this observation, I have still found variation in (34) (and (35)b below) in some speakers. 64 the following examples, the additional context entailing a process or continuum is highlighted in italics and preverbal subjects are taken in its nonspecific reading: (34) a. En algún rincón del mundo, algún lingüista que publiqu-e in some corner of_the world, some linguist COMP publish-3;SBJV su tesis estudiará cada lengua nativa. his thesis study:FUT each language native “In any corner of the world, some linguist that publishes his thesis studied every native language.” b. Después de muchos intentos, algún estudiante que aprueb-e after of many attempts some student that pass-3PRS;SBJV mi clase sacará prestado cada libro. my class take_out:FUT borrow:PCTPeach book “After so many attempts, some student passing my class will borrow each book.” c. Finalmente, un lingüista que termina-ra la tesis ganó finally a linguist that finish-3PST;SBJV the thesis win:3PST cada beca en esta universidad. each scholarship in this university “Finally, a linguist that finished his thesis won every scholarship in this university.” 65 d. Día tras día, un vigilante que apreci-e su day after day a policeman that appreciate-3PRS;SBJV his trabajo montará guardia en cada esquina. job stand:3PST guard in each corner “Day after day a policeman that appreciates his job will be looking after each corner” Furthermore, the nonspecific interpretation of indefinites is available when they are in the scope of a quantificational element. 20 Consider the paradigm in (35) in which the nonspecific readings of the direct objects are unavailable: (35) a. Cristina contrató a un lingüista. Cristina hire:PST ACC a linguist “Cristina hired a linguist.” b. Finalmente, un lingüista que trabaja-se en este departamento finally a linguist that work-3PST;SBJV in this department contrató a Cristina. hire:PST ACC Cristina “Finally, a linguist that worked in this department hired Cristina.” Assuming that the scope properties of quantificational elements are constrained by Economy Principles (Fox 2000), the indefinite object un lingüista in (35)a and the indefinite preverbal subject in (35)b do not interact with the nonquantificational element Cristina. If preverbal subjects must be referential in Spanish, it would be 20 In Spanish, the addition of the prepositional accusative ‘a’ to animate direct objects facilitates a specific interpretation (Zubizarreta 1994 and Leonetti 2003, 2004). 66 expected for them not to interact with other constituents, either because they are nonquantificational or because they are base generated very high in the structure. Accordingly, the nonspecific interpretation should be unavailable. This expectation is however not borne out because nonspecific interpretations are possible: (36) a. Cada lingüista que recib-a una beca agradecerá each linguist that receive-3PRS;SBJV a scholarship thank:FUT a un profesor de esta universidad. ACC a professor from this university “Each linguist that received a scholarship will thank a professor from this university” NONSPECIFIC AVAILABLE b. Un lingüista que recib-a una beca agradecerá a linguist that receive-3PRS;SBJV a scholarship thank:FUT a cada profesor de esta universidad. ACC each professor of this university “A linguist that receives a scholarship will thank a professor from this university.” NONSPECIFIC AVAILABLE Examples in (36) display nonspecific readings for the indefinite subjects or objects as shown by the possibility of modification by a relative clause in the subjunctive. Assuming with Diesing (1992) that nonspecific readings are represented by the syntactic component within VP at LF, these examples suggest that the nonspecific indefinite and the QP are within each other’s scope: the former is reconstructed back within VP such that the latter can scope over it at LF. 67 2.3.2 Scope Interaction with Modals Let us go back to the situation discussed by Alexiadou and Anagnostopoulou where an Xº interacts with an indefinite. The first case they discuss is the interaction of a modal with indefinite subjects. According to Barbiers (2002: 11), the epistemic interpretations correspond to a “class of interpretations involving a speaker-oriented, or, in the case of embedded clauses, matrix-subject oriented qualification or modification of the truth of a proposition.” In contrast, the deontic/root interpretation involves “the will, ability, permission or obligation to perform some action or bring about some state of affairs.” This is illustrated by the examples in (37), which show that both readings may coexist independently of any scopal interaction: (37) John must be at home at six o’clock a. Epistemic reading: “Given what I, the speaker, know), I predict that John will be at home at six o’clock.” b. Deontic/root reading: “John is obliged to be at home at six o’clock.” In the literature, there have been at least two proposals to account for the ambiguity between epistemic and deontic interpretations of modals. Ross (1969) and Perlmutter (1970) and much of the generative tradition argued for two different structures for each reading in (37). The deontic reading is analyzed as a control structure, whereas the epistemic reading corresponds to a raising structure. On the other hand, based on the incompatibility of epistemic modals with yes/no questions and wh-questions, McDowell (1987) proposed that these different interpretations are correlated to 68 different positions that modals may occupy at LF. Epistemics are basically quantifiers that can raise to a clause initial position at LF, whereas deontics are not quantifiers and must stay in their base position. If an interrogative operator occupies the head or the specifier of Cº, the epistemic reading is blocked by some version of the Doubly-Filled Comp Filter. 21, 22 Moreover, Wumbrand (1999) has recently observed that both epistemic and deontic modals in English allow two interpretations with respect to a quantificational subject: one where the QP subject takes wide scope with respect to the modal (QP > modal) and another where the QP subject takes narrow scope with respect to the modal (modal > QP). Following Alexiadou and Anagnostopoulou's logic, one might expect that in pro-drop languages, where dislocation of the subject should be obligatory, only the wide scope reading of the subject over the epistemic or deontic reading of the modal would be possible (QP > epistemic/deontic modal). Given that left dislocated constituents are generated very high in the structure, they cannot be lowered below the position occupied by modals and they cannot even be reached by 21 Note that devices such as the Doubly-Filled Comp Filter do not have a place in syntax under the minimalist assumptions that I am adopting here. It seems probable that they are replaced by an EPP-checking mechanism. Two important developments in this direction are those of Alexiadou and Anagnostopoulou (1998) and of Aoun and Benmamoun (1998). In Alexiadou and Anagnostopoulou’s system, EPP-checking is a syntactic mechanism in which uninterpretable features are checked by Xº-movement or XP-movement. In contrast, Aoun and Benmamoun (1998) state that the EPP can be taken as part of a set of phonological filters that apply to the syntactic derivation. 22 Given that the adoption of one or another proposal does not affect the general argument made in this paper, I will not take a position between these two points of view, but refer the reader interested in the topic to Wurmbrand (1999) and, especially, Barbiers (2002). 69 raising of the modals. 23 However, despite being a pro-drop language, Spanish disconfirms Alexiadou and Anagnostopoulou’s theory. Under the deontic or epistemic reading of a modal, Spanish preverbal subjects allow wide scope (i.e., indefinite > modal) as well as narrow scope readings (i.e., modal > indefinite) over the modal. As noted by Jackendoff (1971) for English, modals induce ambiguity with respect to specific indefinites or nonspecific indefinite objects. Preverbal indefinite subjects in Spanish behave in the same way; they can be interpreted as specific indefinites or nonspecific indefinites. Consider example (8)b, repeated as (38) in which a nonintentional verb triggers a specific interpretation of the preverbal subject un lingüista: (38) Un lingüista maneja este auto todos los días. a linguist drives this car all the days “A linguist drives this car everyday.” ONLY SPECIFIC In (38) the nonintensional verb manejar, “to drive”, allows only the specific reading for its preverbal subject. In contrast, (39) shows that when a modal is included with the same nonintensional verb, its subject readily shows ambiguity between a specific and a nonspecific interpretation, as attested by the possibility of modification by relative clauses in indicative and subjunctive: 23 Given the data reported by Alexiadou and Anagnostopoulou ((5)), this expectation cannot possibly be confirmed for Greek, which according to them “only allows the deontic interpretation” (meaning wide scope of the indefinite subject with respect to the deontic modal). 70 (39) a. Un lingüista puede manejar este auto todos los días. a linguist can drive:INF this car all the days “A linguist can drive this car everyday.” SPECIFIC OR NONSPECIFIC b. Un lingüista que se=gradú-a con honores puede a linguist that REFL=graduate-3PRS;IND with honors can manejar este auto todos los días. drive:INF this car all the days “A linguist that graduates with honors can drive this car everyday." c. Un lingüista que se=gradú-e con honores puede a linguist that REFL=graduate-3PRS;SBJV with honors can manejar este auto todos los días. drive:INF this car all the days “A linguist that graduates with honors can drive this car everyday.” As an illustration of the possibility of wide and narrow scope readings of an indefinite subject with respect to epistemic and deontic modals, take the examples in (40) and (43). Consider first the sentence in (40)b, to be read as a continuation of the context in (40)a: 71 (40) Indefinites and modals: Epistemic context a. Context: In this bookstore, there are no copies of Introduction to Government and Binding Theory by Haegeman, Introducing Transformational Grammar by Ouhalla or Transformational Grammar: A First Course by Radford. By contrast, there are many copies of Some Concepts and Consequences of the Government and Binding Theory by Chomsky. Therefore, I believe that it is true that: b. Un lingüista debe haber comprado libros de texto en esta a linguist must have:INF bought books of text in this universidad university “A linguist must have bought textbooks in this university” AMBIGUOUS Example (40) refers to a situation in which the truth value of the sentences in (40)b is dependent on the evidence available to the speaker and her point of view as established in (40)a. Accordingly, subjects in preverbal or postverbal position are ambiguous in a reading in which “in view of the evidence available in (40)a, it is necessarily the case that some linguist or other bought textbooks” (must > a linguist), and a reading where “there is a specific linguist such that in view of the evidence available in (40)a, it is necessarily the case that he bought textbooks” (a linguist > must). 72 Suñer (2003: 345, ex. 8 and 9) reports similar data for the modal poder which is ambiguous between an epistemic (meaning can) and a deontic reading (meaning may): (41) a. Puede presentar un chico su trabajo este viernes. can present:INF a guy his work this Friday “A guy can/may present his work this Friday.” AMBIGUOUS: un > pudo; pudo > un b. Un chico puede presentar su trabajo este viernes. a guy can present:INF his work this Friday “A guy can/may present his work this Friday.” AMBIGUOUS: un > pudo; pudo > un However, the data is more complex. Suñer also notes that with perfect tenses, and independently from indefiniteness and subject position, there is a strong preference for taking the indefinite as only having wide scope: (42) a. Pudo haber asesinado un estudiante a su loro. could have:INF assassinate:PTPC a student ACC his parrot “A student could have assassinated his parrot.” NOT AMBIGUOUS: ??un > pudo; pudo > un b. Un estudiante pudo haber asesinado a su loro. a student could have:INF assassinate:PTCP ACC his parrot “A student could have assassinated his parrot.” NOT AMBIGUOUS: ??un > pudo; pudo > un 73 The subtle details of these cases deserve further research. What seems clear as Suñer remarks is that Spanish does not show that preverbal subjects are obligatorily interpreted as specifics or partitives with respect to the modal and as such cannot be taken as obligatorily dislocated. To give additional support to this claim, observe the example in (43)b elicited in the deontic context of (43)a and with the interpretations in (44): (43) Indefinites and modals: Deontic context a. Context: According to the rules in this department of linguistics, it is required for students to choose a candidate after the first year. Therefore, it is true that… b. Después del primer año, todos los estudiantes de lingüística deben after of:the first year all the student of linguistics must nombrar a una candidata para que recib-a un premio. nominate:INF ACC a candidate for that receive-3PRS;SBJV a prize “After the first year all the linguistics students must nominate a candidate in order for her to receive a scholarship.” (44) Interpretation of (43) THREE-WAY AMBIGUOUS a. It is necessarily the case that all the linguists nominate a different president. (Deontic reading with “relative” surface scope of all the linguistics students > a candidate). 74 b. It is necessarily the case that there is one specific president such that all the linguists nominate her. (Deontic reading with “relative” inverse scope of a candidate > all the linguistics students). c. For all the linguists it is necessarily the case that each of them nominated a candidate. (Deontic reading with absolute scope of all the linguistics students over the proposition –and a candidate). Example (43)b is three-way ambiguous depending on the relative height of the subject's position in relation to the modal and the indefinite object. In the interpretations in (44)a and (44)b, the surface scope (i.e., all the linguistics students > a candidate) and inverse scope (i.e., a candidate > all the linguistics students) are asserted to occur relative to a situation in which the students all nominated one specific candidate or they each nominated a different candidate. In contrast, (44)c entails a stronger reading because it is asserted that in order for a certain candidate to receive a scholarship, the candidate must be a member of a certain type of students, namely linguistics students. Given the ambiguous interpretations of preverbal subjects interacting with modals, and interacting with deontics, it is not possible to maintain that preverbal subjects can only be analyzed as dislocated. In particular, the pattern exhibited by 75 Spanish seems to suggest that a raising analysis of the epistemic/deontic distinction might be possible. 2.3.3 Scope Interaction with Assertion Markers Given the two previously discussed contexts (QP/QP interactions and modals/QP interactions), it might be expected that Spanish would show an English pattern. However, this is not so: whereas in (45)a negation can have either wide scope over the subject –no>muchos lingüistas, or narrow scope ?muchos lingüistas>no, the same does not happen in the case of (45)b. In other words, Spanish also has a Greek pattern: (45) a. No leen muchos lingüistas libros de texto. NEG read:3PL many linguists books of text “Many linguists do not read textbooks.” AMBIGUOUS: no > muchos; muchos > no b. Muchos lingüistas no leen libros de texto. Many linguists NEG read:3PL books of text “Many linguists do not read textbooks.” NOT AMBIGUOUS: *no > muchos; muchos > no Suñer (2003: 346, ex. 11) also describes similar data in which a preverbal quantificational subject has only wide scope over the negation: (46) a. No aprobaron muchos estudiantes el examen final. NEG pass:3PL many students the exam final “Many students did not pass the final exam.” NOT AMBIGUOUS 76 b. Muchos estudiantes no aprobaron el examen final. Many students NEG pass:3PL the exam final "Many students did not pass the final exam." NOT AMBIGUOUS Negation is also a licenser for ellipsis in Spanish. When the remnant is a subject, it is restricted to referential elements. Thus the example in (47)c with the nonspecific indefinite quantifier alguien is disallowed. On the other hand, if the remnant subject is an indefinite as in (47)b, it has to be obligatorily specific . Material deleted as a result of ellipsis is represented by a double strikethrough: (47) a. Cristina leyó 'Syntactic Structures' y el lingüista no leyó Cristina read:3;PST Syntactic Structures and the linguist NEG read:3;PST 'Syntactic Structures'. Syntactic Structures "Cristina read Syntactic Structures and the linguist did not." b. Cristina leyó 'Syntactic Structures' y un lingüista no leyó Cristina read:3;PST Syntactic Structures and a linguist NEG read:3;PST 'Syntactic Structures'. Syntactic Structures "Cristina read Syntactic Structures and a linguist did not." c. Cristina leyó 'Syntactic Structures' y alguien no leyó Cristina read:3;PST Syntactic Structures and somebody NEG read:3;PST 'Syntactic Structures'. Syntactic Structures "Cristina read Syntactic Structures and the linguist did not." 77 In fact, assertion markers in general provide the context for the dislocation pattern of subjects in Spanish, not only negation. For instance, ellipsis data are licensed by the assertion items with a polarity value such as sí ‘yes’; no ‘no’; también ‘also’ and tampoco ‘neither’ seem to affect the interaction of two QPs that are otherwise understood as ambiguous: (48) a. Esta noche, un policía no/sí vigiló cada banco this evening a policeman NEG/AFF look_after each bank “This evening, a policeman did not/DID look after each bank.” NOT AMBIGUOUS: un > cada; *cada > un b. En este país, algún lingüista no/sí está estudiando cada lengua in this country, some linguist NEG/AFF is studying each language nativa. native “In this country, some linguist is not /is really studying each native language.” NOT AMBIGUOUS: algún > cada; *cada > algún A similar effect can be found with modals. When negation is placed in a structure similar to our three-way ambiguous example in (43), the epistemic readings resulting from the interaction of the preverbal subject and the QP object disappear. The preverbal subject is interpreted deontically with respect to the modal: 78 (49) Todos los lingüistas no necesitan admirar a una mujer. all the linguists NEG need:3PL admire:INF ACC a woman “All the linguists need not admire a woman.” (50) Interpretation of (49) a. It is not the case that necessarily all the linguists admire each a different woman. NOT AVAILABLE (epistemic reading with all the linguists > a woman) b. It is not the case that necessarily there is one specific woman such that all the linguists admire her. NOT AVAILABLE (epistemic reading where a woman > all the linguists) c. For all the linguists it is not the case that they are obliged to admire a woman. AVAILABLE (deontic reading where all the linguists > a woman) Therefore, the position of the subject with respect to affirmative/negative particles is relevant because when the subject precedes these particles, it shows properties attributed to dislocated constituents. Recall that in Section 2.2.1, following Schneider Zioga (2007, 2002), it was noted that specificity and topichood are incompatible with irrealis or veridicality contexts. The following pattern suggests that affirmative/negative particles cannot be understood to provide such contexts because they require any subject that precedes them to be interpreted as obligatorily specific: 79 (51) a. un lingüista no publicó su tesis este año. a linguist NEG publish:3PST his thesis this year “A linguist did not publish his thesis this year.” OBLIGATORILY SPECIFIC b. un lingüista que sab-e dibujar árboles sintácticos no a linguist that know-3PRS;IND draw:INF trees syntactic NEG publicó su tesis este año. publish:3PST his thesis this year “A linguist that knows how to draw syntactic trees did not publish his thesis this year.” c. *un lingüista que sep-a dibujar árboles sintácticos no a linguist that know-3PRS;SBJV draw:INF trees syntactic NEG publicará su tesis este año publish:3FUT his thesis this year “A linguist that knows how to draw syntactic trees will not publish his thesis this year” As (51)b and (51)c illustrate, a subject that precedes an assertion particle disallows modification by a relative clause in subjunctive, an indication that a nonspecific reading is impossible. In conclusion, Spanish displays a Greek pattern when its preverbal subjects are followed by assertion particles such as affirmation or negation. This pattern is not totally unexpected if we recognize that the position of these particles in the structure 80 can vary crosslinguistically. Presumably, in the case of Spanish, the projection containing these elements is located very high in the structure, above TP and the projection containing modals. I will go over these issues in Chapter 4. 2.4 Summary and Conclusions In this chapter, I have illustrated that although Spanish is a pro-drop language that allows verb-movement and the alternation VSO/SVO, it does not respond uniformly to the tests Alexiadou and Anagnostopoulou suggested as indicators of the left dislocated nature of preverbal subjects. In particular, Spanish allows for preverbal subjects in canonical and dislocation positions. On the one hand, the empirical evidence from scope interactions between two quantificational elements and the relative scope between a quantificational element plus a modal lead us to think that preverbal subjects in Spanish show properties attributed to nondislocated constituents. On the other hand, the interpretation of preverbal subjects with respect to assertion particles (affirmation/emphasis and negation) implies properties associated with dislocation. Furthermore, the evidence provided by the placement of adverbs also suggests that these two types of preverbal subjects are correlated with different syntactic positions, and presumably are licensed by different mechanisms. This hypothesis will be explored in Chapter 4, but in order to prepare the ground for this discussion, it is necessary to understand the composition of the left periphery in Spanish. This is the topic of the following chapter, Chapter 3. 81 Chapter 3 Nondeclaratives and Second Position in the Left Edge of Modern Spanish In Chapter 2, I showed that there are two patterns of preverbal lexical subjects in SVO sequences: a canonical pattern and a dislocated pattern. A step forward to understand why these two patterns exist along with dislocation of nonsubjects is to investigate the structure of the left edge as the inner mechanism responsible for word order and dislocation. The aim of this chapter is to gain a better grasp of how marked clausal types are syntactically licensed by the Spanish left edge. In particular, I will concentrate on the position of the finite verb and auxiliaries in interrogatives. The discussion of the mechanics of the Spanish left edge in clausal marked types will initiate an argument for the existence of a second position effect in Spanish, an argument to be completed in Chapter 4 with an analysis of declarative clauses. Accordingly, the left edge is conceived as the articulation of two syntactic positions: Cº (the Perspective or Point of View Phrase) and c*º (the Assertion Phrase). The special licensing conditions of these two functional heads will play a fundamental role in understanding the phenomenon of dislocation and word order in Spanish. The structure of this chapter is as follows. In Section 3.1, I consider some theoretical aspects related to the representation of the complementizer system within the Narrow Left Edge system. Section 3.2 elaborates on the position of the finite verb, auxiliaries and modals in Spanish. It will be shown that left edge nodes in Spanish are possible landing sites of verb movement, contra the current consensus in 82 the field. Section 3.3 reconsiders the analysis of wh-questions and subject inversion as the result of wh-fronting and verb raising to the left periphery. I analyze the impossibility of preverbal lexical subjects to occur with a fronted wh- (the Preverbal Subject Constraint) as the effect of licensing the lowest projection of the left edge, the Assertion Projection (c*P). Section 3.4 is the conclusions. 3.1 The Licensing Requirements of the Left Edge In this section, I sketch the conception of the left edge that I call “the Narrow Left Edge.” Section 3.1.1 spells out some basic assumptions in the analysis and introduces the notion of ph-lexicalization (Zubizarreta 2001). Section 3.1.2, I sketch the Narrow Left Edge system. Importantly, the notion of clausal typing is decomposed in a system of uninterpretable and interpretable features distributed in the left edge heads. I present the basic structure for simple declaratives and focused declaratives, which will play a role in the discussion of interrogative types in Section 3.3. 3.1.1 Some Basic Assumptions: Features, Economy and Markedness Chomsky (1995, 2001) suggested that the inventory of features manipulated by the computational system can be classified into three groups: semantic features, phonological features and formal features. The interpretable features on a given category can be read off from their associated interface level. Semantic features are interpretable at the LF level, but uninterpretable at the PF level. Conversely, phonological features are interpretable at the PF level, but uninterpretable at the LF 83 level. Formal features cannot be interpreted either at the LF level or at the PF level and must resort to a special licensing relation. Within the minimalist program developed in Chomsky (1995, 2000), uninterpretable features are the exclusive triggers of syntactic licensing operations. For example, the operation of Agree is dependent on the activation of a probe by an uninterpretable feature u F, which searches for a matching goal to license u F. Under certain conditions, the existence of an uninterpretable EPP feature pied piping the goal category can follow the Agree operation. The goal is thus copied within the local domain of the probe by the composite operation Move (Agree + Copy + Merge). I will not adopt these basic assumptions here. Instead, I will follow Nash and Rouveret’s idea that checking of uninterpretable features is not the only driving force of movement and that interpretable features can also trigger syntactic operations: (1) Interpretable features may trigger syntactic computations. Roberts (2005: 123) discusses the status of the declarative feature in Rizzi’s Force Phrase. Considering declarative clauses as unmarked clauses with respect to nondeclarative clauses (interrogatives, exclamatives, desideratives), Roberts argues that the [+declarative] feature does not exist at all. This conclusion is derived from the idea that the feature content of Rizzi’s Force Phrase can be underspecified. An underspecified Force head selecting for a [+finite] Finite Phrase conveys the interpretation of the clause as a declarative. In other words, declarative clauses are interpreted in the absence of any feature specification in root Forceº. 84 I will also follow Nash and Rouveret (2002)’s theory of proxies. A proxy projection is a contentless functional category, not included in the numeration, but created in the course of the derivation. Proxies are created as a consequence of fission of unchecked uninterpretable features in a given head. Thus, the creation of a proxy is justified as a step in the derivational process to satisfy a formal requirement. Given the Principle of Full Interpretation, proxys as well as already checked uninterpretable features are invisible to the computational system: (2) Principle of Full Interpretation The LF and PF representations must consist solely of well formed legitimate elements (interpretable) at the relevant levels. Thte creation of a proxy obeys to Economy Principles. In particular, Nash and Rouveret suggest that there is a constraint on the number of syntactic operations targeting a head, requirement expressed in their Single Licensing Condition: (3) Single Licensing Condition (SLC) A functional category can enter into a licensing relation with the feature content of only one terminal node in its checking domain. According to Nash and Rouveret, the term ‘licensing relation’ in the definition (3) refers to the checking of an uninterpretable feature by an attracting matching feature. Given the assumption that interpretable features can also trigger operations, I would like to suggest with Nash and Rouveret that Zubizarreta’s notion of ph(onological)- lexicalization is also subject to the Single Licensing Condition. Ph-lexicalization 85 refers to the substantivization of an interpretable feature by head movement or by movement of a maximal projection to a specifier position: (4) Ph(onological)-lexicalization A feature in the head of a functional projection is ph-lexicalized if the head or the Spec of that projection is associated with appropriate ph- features at the point of spell out (Zubizarreta 2001: 189). Ph-lexicalization is an operation that licenses the interpretable content of a functional head by aligning this functional head with proper material containing phonological/lexical features. Within the set of checking operations, ph- lexicalization is a checking operation that explicitly requires the visibility of phonological features, interpretable at PF. 3.1.2 Features in the Left Edge and Clausal Typing The projectionist approaches to the left edge assume that the complementizer system consists of several projections. For example, Rizzi (1997) assume the system in (5)a and Poletto (2000) the system in (5)b: (5) a. Rizzi’s left periphery ForceP ... IntP … TopP* ... FocP ...TopP*... FinP ... IP b. Poletto’s left periphery H Topic...Scenesetting...ForceP...LeftDisl...FocP...WH...FinP...IP Within the projectionist approaches, the different semantic effects created in the interface between the informational component and the grammatical component are encoded in specific projections. For instance, Poletto has a special projection for 86 hanging topics (H Topic) and left dislocated constituents (Left Disl), whereas Rizzi creates two topic positions (TopP*). In addition, in both systems the Force Phrase (ForceP) determines the clausal type of a proposition, namely if it is declarative or nondeclarative (interrogative, imperative or exclamative for example). The Int(errogative) Phrase (IntP) and the Focus Phrase (FocP) in Rizzi’s system and the FocP and the WHP in Poletto’s system are in charge of the representation of nonpresupposed information or new information.. Finally, the Finite Phrase (FinP) for Rizzi and Poletto represents the finiteness, modality and tense distinctions of the inflectional layer. Instead of adopting a projectionist approach, I would like to assume a reductionist approach to the left edge. I will call this system “the Narrow Left Edge” because it only consists of two heads: the Point of View head or Big CP projection (Cº) and the Assertion head or Small c*P projection (c*º): (6) The Narrow Left Edge ... CP ... c*P ... IP In contrast to Rizzi’s and Poletto’s approach to the left periphery (Rizzi 1997), within the Narrow Left Edge system, the notion of “Topic” as an independent functional projection does not play any role. Instead, “topic” is only a way to constitute a predication relation as part of the interplay of CP and c*P. Based on Baker’s observation that agreement and dislocation are closely interrelated (Baker 1996), higher and lower Topic Phrases and their recursivity (‘*’ in Rizzi’s system) result from adjunction of a maximal projection to the specifier of CP (higher topics) 87 or movement of a maximal projection to the specifier of c*P (lower topics). Higher and lower topics are not a property of a certain functional projection, but they are a property of the syntax of the lower and higher left edge. The c*º head is the locus of assertion, an operator position. Given that there is a close relationship between an asserted proposition and finiteness, and nonfinite propositions never express assertions, I will assume that the c*º head bears a [finite] feature in simple declaratives. In focused declaratives, I follow Zubizarreta (1998a)’s articulation of the notion of focus in terms of assertion structures to propose the existence of a [Q] feature with quantificational force. I will assume that the [finite] feature and the [Q] feature are interpretable features, which give substantive content to the Assertion Phrase at the logico semantic interface. 24 The Cº head represents the locus of the notion of perspective of the speaker or point of view, which intends to characterize the intuitive notion of orientation of the speech in clausal types expressing “the identification of the speaker with a person who participates in the event/state described in a clause” (Kuno and Kaburaki 1977: 628). For instance, whereas declaratives seem unmarked for a perspective or point of view and also unmarked for an orientation, interrogatives exclamatives and imperatives take the perspective of the speaker and are nonspeaker oriented. I propose that these observations can be formally characterized by leaving the Cº head underspecified as [0] or by adding a [command] feature that ranges over the featural 24 The adoption of [finite] in c*º as an interpretable feature probably departs from Rizzi’s characterization of the left edge in terms of a Finite Phrase. Rizzi notes that the content of the Finite Phrase encodes a wide gamut of semantic and grammatical information including modality, tense and agreement, which are presumably uninterpretable. 88 content of the c*º head (namely, the [finite] or the [Q] features). As before, the [command] feature has a substantive value and its insertion in the structure has repercussions at the interpretative levels. Summarizing, the Narrow Left Edge system is composed of two heads with substantive content as illustrated in (7): (7) Interpretable features in the left edge a. Cº = {[command]} b. c*º = {[Q], [finite]} I also suggested that the Cº head can be underspecified before the point of spell out. This means that the Cº head can be contentless. Given that the computational system manipulates only substantive categories, if the specifier of CP does not lexicalize a Cº [0] head, the phrase marker in (8)a is interpreted as (8)b at LF and PF: 25 (8) Default value for Cº a. CP b. c*P 2 2 SPEC C’ c*º TP 2 [finite]/[Q] Cº c*P [0] 2 c*º TP [finite]/[Q] Lexicalization then gives an interpretation to the Point of View head Cº in phonetic and semantic terms. Otherwise, a Cº head lacking substantive content in semantic terms and with no overt specifier is expected to be invisible and deleted by the computational system, presumably, for optimality reasons. 25 In the Narrow Left Edge system, a specifier in a Cº [0] head lexicalizes Cº as a hanging topic. Hanging topics appear in matrix clauses. 89 Let us discuss the notion of “clausal typing” encoded in Rizzi’s and Poletto’s system in the Force Phrase. Lisa Cheng proposed the Clausal Hypothesis stated in (9) (Cheng 1991). She observes that clausal types such as interrogatives among others are identified syntactically or morphologically across languages: (9) Clausal Typing Hypothesis Every clause needs to be typed (Cheng 1991) Although I will assume (9), I will differ in the mode of realization of the typing process. In Cheng’s approach, the typing of a clause is given by explicit licensing of the head or the specifier of the complementizer phrase by syntactic or morphological means. In contrast, I claim that the formal manifestation of the notion of clausal type results from the interaction of interpretable and uninterpretable features in the left periphery. Consider the following two sets of features defined on one or more than one functional head: U, the set of uninterpretable features (U-features) and I, the set of interpretable features (I-features): (10) a. U={ u [WH] , u [AGR], u [CASE]} b. I={ i [COMMAND], i [Q], i [FINITE]} It is reasonable to think that the Cº and c*º nodes could host uninterpretable features. The presence of features such as [wh], [AGR], [T], or [Case] in a given grammar might be determined by parametric variation in the lexicon. I would like to propose that the presence of these uninterpretable features triggering checking operations in combination with the set of interpretable features in Cº and c*º defines the notion of 90 “clausal typing.” Thus, the clausal type of a linguistic expression is defined from the intersection of U and I, expressed in the Feature Typing Alignment Hypothesis: (11) Feature Typing Alignment Hypothesis The clausal typing of a linguistic expression results from the alignment of interpretable features (i-features) and uninterpretable features (u-features) on one or more heads within the (narrow) left edge. The hypothesis in (11) suggests that clausal typing and the force of a clause are not primitive notions. Instead, they can be decomposed in more primitive units defined over the interpretable and uninterpretable features available to the assertion and point of view heads. The alignment of interpretable and uninterpretable is given by selectional relations. Declaratives are the result of the association of the default value [0] in Cº with the I-featural content of c*º. If [finite] in c*º is selected, a simple declarative is derived. In contrast, if [Q] in c*º is selected, a focused construction is derived: (12) a. Declarative b. Focused declarative CP CP 2 2 Cº c*P Cº c*P [0] 2 [0] 2 c*º c*º [finite] [Q] The selection of [Q] in contrast to [finite] expresses the quantificational force present in clauses containing focused constructions, but absent in simple declarative clauses. Only focused constructions introduce a variable and a value for it: 91 (13) a. Cristina comió la manzana. Cristina eat:3PST the apple “Cristina ate the apple.” b. LA MANZANA comió Cristina (no la pera) the apple eat:3PST Cristina not the pear “Cristina ate THE APPLE (not the pear).” If Cº carries a specified value, it is possible to derive in a similar way other clausal types. I will delay the discussion of how other clausal types could be generated until Section 3.4, where I suggest a possible extension to other clausal types such as imperatives, desideratives and exclamatives. For a more detailed elaboration of the Narrow Left Edge, see Beas (2007). 3.2 Raising to the Left Edge: Verb Movement in Spanish Suñer (1994) claims that verb movement to the inflectional system must occur in Spanish, but this operation does not reach the left periphery. She provides several pieces of evidence that the verb can be in Tº. However, as the inflected verbal elements are not shown to obligatorily remain in Tº, the argument is incomplete. 26 Further elaborations of the left edge also motivate a revision of the empirical paradigms. For instance, once one assumes a proposal where the CP system consists of several sublayers of functional projections, Suñer’s claim requires either (dis)confirmation or refinement. In particular, the landing site of the inflected verb in 26 See the end of Section 3.3.3., examples (61)-(62) for the discussion of Suñer’s alleged pieces of evidence for the unavailability of verb movement to the left edge. 92 Spanish might not be the highest functional head in the CP system. Yet, this landing site could still be within the left periphery, higher than the inflectional layer. In sections 3.2.1 through 3.2.3, I will review evidence for this point of view drawn mainly from the distribution of low, medial and high adverbs with respect to finite verbs, auxiliaries and modals. 3.2.1 Verb Raising outside the Thematic Domain For expository purposes, assume the structure in (14) where F1º and F2º are functional heads of the articulated inflectional layer and CP represents the left edge of the clause: (14) CP Left edge 2 Cº F1 2 ModalAdv F1 Inflectional layer 2 F1 F2 2 MiddleAdv F2 2 F2 VP 2 MannerAdv V’ Thematic layer 2 Vº In his theory of adverbials, Cinque (1999) argues that adverbs occupy the specifiers of the functional projections they modify. In (14), manner adverbs (i.e. lógicamente ‘logically’) mark the left edge of the VP; whereas modal adverbs (probablemente ‘probably’) are specifiers of the highest inflectional projection –F1. Following Bok Bennema (2001), I will call the adverbs that occupy an intermediate position 93 between F1º and F2º: “middle adverbs” (inmediatamente ‘immediately,’ siempre ‘always,' ya ‘already,’ frecuentemente ‘frequently.’) Main verbs in Spanish can precede manner adverbs, as it is the case of main verbs in French. This ordering indicates that they leave the vP/VP layer: (15) Manner Adverbs a. Cristina lógicamente contestó esa pregunta. Cristina logically answer:PST that question "*Cristina answered that question in a logical manner." "Obviously, Cristina answered that question." b. Cristina contestó lógicamente esa pregunta. Cristina answer:PST logically that question "Cristina answered that question in a logical manner." "*Obviously, Cristina answered that question." The adverb lógicamente must follow the main verb in (15) to give rise to the interpretation where Cristina did the action of answering certain question in a logical manner. This interpretation is only available in (15)b. Crucially, if the adverb precedes the verb -(15)a, it is either interpreted as a parenthetical (dependent on whether there is additional material as for instance the preverbal subject Cristina) or as having a sentential scope, in which case lógicamente conveys a different meaning, namely that Cristina obviously answered the question. For additional examples and discussion, see Zubizarreta (2007). 94 3.2.2 Verb Raising to the Left Periphery Nevertheless, in contrast to French where main verbs always precede middle and modal adverbs (Pollock 1989), their verbal counterparts in Spanish can either precede or follow these adverbial elements. Given (14), one possible interpretation of the data is to argue that verb raising may be optional within the inflectional layer. It is also possible to argue that verb raising from F2 to F1 contributes to detopicalize a potential initial position of the middle adverb. I will not offer a solution to this particularity in the Spanish clause structure, but I will refer the reader to Bok Bennema (2001) for an alternative based on the ordering of Merge between the verbal predicate and the adverbials. Sentences with middle adverbs in (16)a, (17)a and (18)a represent the sequence with the verb preceding the adverbial and sentences in (16)b, (17)b and (18)b represent the ordering with the verb following the adverb: 27 (16) Middle Adverbs a. Cristina leyó inmediatamente un libro de lingüística. Cristina read:PST immediately a book of linguistics "Cristina immediately read a linguistics book." b. Cristina inmediatamente leyó un libro de lingüística. Cristina immediately read:PST a book of linguistics "Cristina immediately read a linguistics book." 27 Sentences (16)b, (17)b and (18)b with the middle adverb preceding the finite verb bear strong accent and are probably focused. This might indicate that the ordering middle adverb- --finite verb is a derived order. I will not pursue this issue here because it does not affect the main point of the text. 95 (17) Middle Adverbs a. Cristina compone siempre sus canciones. Cristina compose always her songs "Cristina always composes her songs." b. Cristina siempre compone sus canciones. Cristina always compose her songs "Cristina always composes her songs." (18) Middle Adverbs a. Cristina cuestiona frecuentemente la verdad. Cristina questions frequently the truth “Cristina frequently questions the truth." b. Cristina frecuentemente cuestiona la verdad. Cristina frequently questions the truth “Cristina frequently questions the truth." An analogous pattern in the distribution of main verbs occurs with modal adverbs, the ones dividing the inflectional layer from the left periphery. In (33), the modal adverbial probablemente 'probably' can also precede or follow the main verb: (19) Modal Adverbs a. ... probablemente le=hablaste a tu hermano. probably DAT=talk:PST DAT your brother “You probably talked to your brother." 96 b. ... le=hablaste probablemente a tu hermano. DAT=talk:PST probably DAT your brother “You probably talked to your brother." From the paradigms involving the distribution of manner, middle and modal adverbs; one can conclude that the inflected verb in Spanish must move outside the VP to the highest inflectional projection. In addition, the distribution of the inflected verb preceding modal adverbs such as probably suggests that the inflected verb can appear displaced beyond F1º and within the left periphery. 3.2.3 Auxiliaries in the Left edge If one assumes the adverbial structure in (14), there also seems to be evidence for the activation of the left edge by the placement of some auxiliaries. Observe that manner adverbs must follow the auxiliary and the participial verb. In a non parenthetical reading, the adverbial muy bien “very well” cannot intervene between the auxiliary haber and the participial verb: (20) a. Funes había recordado muy bien cada detalle. Funes have:3;IPFV remember:PTCP very well each detail “Funes had remebered every detail very well.” b. *Funes había muy bien recordado cada detalle. Funes have:3;IPFV very well remember:PTCP each detail “Funes had remebered every detail very well.” Given the facts discussed in Section 3.2.2, I will claim that the participle in examples such as (20)a has undergone raising to the inflectional layer. Assuming that the 97 auxiliary in these examples has been base generated within the IP, the distribution of modal adverbs to be examined suggest that auxiliary raising to the left edge is possible contra the Standard Analysis. Recall that modal adverbs divide the frontier between the inflectional and complementizer systems. The paradigm in (21) shows that the modal adverb probablemente “probably” can intervene between the auxiliary había “had” and the main verb in its participial form when a lexical subject is in preverbal position, suggesting movement to the left edge. The distribution of modal adverbs, auxiliaries in these examples is parallel to their English counterparts as suggested by the translations: (21) a. Ótanes había probablemente fallecido ayer. Ótanes have:3;IPFV probably pass:PTCP.away yesterday “Ótanes had probably passed away yesterday.” b. El CMV habría probablemente causado esta lesión cerebral. the CMV have:3COND probably caused this injury brain “The CMV would have probably caused this brain injury.” However, in contrast to English, the auxiliary can appear in second position if a nonsubject is fronted to clause initial position. In (22), the auxiliary appears in second position when an adverb occupies the clause initial position: (22) Mañana habrá probablemente fallecido Ótanes. tomorrow have:3;FUT probably pass:PTCP.away Ótanes “Tomorrow, Ótanes will have probably passed away.” 98 The adverb mañana “tomorrow” does not get any special intonation in these examples and behaves like a topic. Importantly, if an object is displaced to the clause initial position, a clitic with the same reference as the displaced constituent must accompany the auxiliary in second position (see (23)a). Otherwise, the sentence is ungrammatical as in (23)b: 28 (23) a. Esta lesión cerebral la=habría probablemente causado el CMV. this injury brain it=have:3COND probably caused the CMV “This brain injury, the CMV would have probably caused it.” b. *Esta lesión cerebral habría probablemente causado el CMV. this injury brain have:3COND probably caused the CMV “This brain injury, the CMV would have probably caused it.” The presence of the object clitic and the auxiliary preceding the modal adverb probablemente may indicate the activation of a position within the left edge in (23)a. However, a modal adverb can also precede the auxiliary in the verbal complex. The example (24)a contains a lexical subject in preverbal position and (24)b includes an object in preverbal position: (24) a. El CMV probablemente habría causado esta lesión cerebral. the CMV probably have:3COND caused this injury brain “The CMV would have probably caused this brain injury.” 28 In contrast to the grammatical akwardness of (23)b, lexical verbs seem to allow this inversion pattern very easily. I will leave this problem for further clarification. 99 b. Esta lesión cerebral probablemente la=habría causado el CMV. this injury brain probably it=have:3COND caused the CMV “This brain injury, the CMV would have probably caused it.” There is a crucial difference betwen the paradigms with auxiliaries preceding modal adverbs and the paradigms with auxiliaries following modal adverbs. In the former case, the adverb predicates over the entire proposition, but in the latter situation, the constituent can be oriented to the constituent in initial position (subject or non subject). Given the empirical evidence discussed, I will adopt the simplest solution. I will conclude that auxiliaries sometimes raise to the left edge and sometimes they do not do so. 3.3 Interrogative Clauses in Modern Spanish In this section, I examine Modern Spanish interrogative clauses in order to show that Modern Spanish resorts to the Assertion head, c*º to license marked clauses. In overt contradiction with analyses based on the Standard Analysis of clause structure (see Chapter 1, Section 1.3), I demonstrate that the restriction known as Preverbal Subject Constraint (PSC) is better explained in terms of the absence or presence of verb movement to c*º and fronting to the [Spec, c*P] of a wh-constituent or a lexical DP. It is crucial to recognize that if preverbal lexical subjects occupy the lower left edge (i.e., [Spec, c*P]) a fronted wh-phrase needs to be specific or D- linked. 100 In Section 3.3.1, I introduce the basic empirical paradigms related to the well known generalization against the cooccurrence of interrogatives and preverbal lexical subjects (the Preverbal Subject Constraint). Section 3.3.2 questions Torrego’s and Suñer’s conclusions based on the examination of several sets of novel data. It will be argued that the specific or D-linked nature of the fronted wh- attenuates the effects of the Preverbal Subject Constraint. Finally, Section 3.3.3 analyzes the empirical data within the Narrow Left Edge approach. I will argue that the suspension of the effets of the Preverbal Subject Constraint is due to the failure of verb movement to the left periphery, a positional effect akin to what is observed in V2 languages. 3.3.1 The Preverbal Subject Constraint with Interrogatives A major observation in the literature on Spanish interrogatives is that preverbal subjects cannot cooccur with certain types of wh- (Torrego 1984, Suñer 1994, Suñer and Lizardi 1995, Goodall 1991, Ordóñez 1997, 2000, Olarrea 1996, Uribe Etxebarria 1995, Baauw 1998, Inclán Nichol 1997, Toribio 2000, and Zubizarreta 2001). In particular, many of the prevailing ideas on the topic in the field take Esther Torrego 1984’s observation as a starting point. Torrego noted that preverbal lexical subjects in wh-interrogatives surface as a function of the argumental or nonargumental status of the extracted wh-constituent. This restriction can be stated as in (25): 101 (25) The Preverbal Subject Constraint (PSC) Argumental nonsubject wh-phrases cannot coappear with preverbal subjects. Consider (26) and (27), from Inclán Nichol (1997). These paradigms show that if the wh-phrase object qué "what" appears fronted in the left edge, subject/verb inversion is obligatory in matrix and embedded clauses: 29 (26) a. ¿Qué compró Carmen? what buy:PST Carmen “What did Carmen buy?” b. *¿Qué Carmen compró? what Carmen buy:PST “What did Carmen buy?” (27) a. Carmen quiere saber qué compró Juan. Carmen want:3SG know:INF what buy:PST Juan “Carmen wants to know what Juan bought." b. *Carmen quiere saber qué Juan compró. Carmen want:3SG know:INF what Juan buy:PST “Carmen wants to know what Juan bought." Although several researchers have found variations on the type of nonargumental wh-phrase that disobeys the Preverbal Subject Constraint (Ron 1998, and especially 29 There are dialects such as Caribbean Spanish (Puerto Rican and Dominican) where the constraint described in (25) does not apply at all, probably due to historical changes in their grammar. See Toribio (2002) and Suñer (1994) who have reported that examples parallel to (26)a and (27)a are perfectly grammatical in these dialects. 102 see Gutiérrez Bravo 2002), adjunct wh-phrases such as dónde "where", cuándo "when” and por qué "why" typically allow subjects to appear postverbally (i.e., (28)) and preverbally (i.e., (29)): 30 (28) Nonargument wh-phrases>verb>subject ordering a. Me=preguntaba que por qué llamó Carmen a su novio. DAT=wonder:IMPFV COMP for what call:PST Carmen DAT her boyfriend “I was wondering why Carmen called her boyfriend." b. Me=preguntaba que dónde compró Carmen su Rolls Royce. DAT=wonder:IMPFV COMP where buy:PST Carmen her Rolls Royce “I was wondering where Carmen bought her Rolls Royce." c. Me=preguntaba que cuándo vendió Carmen su Rolls Royce. DAT=wonder:IMPFV COMP when sell:PST Carmen her Rolls Royce “I was wondering when Carmen sold her Rolls Royce." (29) Nonargument wh-phrases>subject>verb ordering a. Me=preguntaba que por qué Carmen llamó a su novio. DAT=wonder:IMPFV COMP for what Carmen call:PST DAT her boyfriend “I was wondering why Carmen call her boyfriend." b. Me=preguntaba que dónde Carmen compró su Rolls Royce. DAT=wonder:IMPFV COMP where Carmen buy:PST her Rolls Royce “I was wondering where Carmen bought her Rolls Royce." 30 It is not clear if the source of this variability becomes an effect of grammaticality or preference. See also Goodall (2004) for the role of non linguistic factors associated with the presence/absence of the Preverbal Subject Constraint. 103 c. Me=preguntaba que cuándo Carmen vendió su Rolls Royce. DAT=wonder:IMPFV COMP when Carmen sell:PST her Rolls Royce “I was wondering when Carmen sold her Rolls Royce." As expected, long distance extraction of nonsubject arguments like the direct object qué "what" in (30)b also triggers inversion: (30) a. Martha piensa que Julieta compró veneno. Martha think:3SG COMP Julieta buy:3PST poison “Martha thinks that Julieta bought poison." b. ¿Qué piensa Martha que compró Julieta? what think:3SG Martha COMP buy:PST Julieta “What does Martha thinks that Julieta bought?” Suñer (1994) made a key contribution to the discussion of the core paradigms and the generalizations related to the restriction against preverbal subjects in interrogatives. She noted that the argument/adjunct distinction was insufficient to explain why preverbal subjects are ruled out in interrogatives. Suñer notes that some non θ-marked arguments indeed force a postverbal subject: (31) a. ¿Cuánto pesará esta beba? how.much weigh:FUT this baby “How much will this baby weight?” 104 b. *¿Cuánto esta beba pesará? how.much this baby weigh:FUT “How much will this baby weight?” Suñer’s analysis of these empirical facts postulates the [+argumental] feature for constituents that receive a θ-role (qué in (26)a) and those constituents that do not receive a θ-role (cuánto in (31)a for instance). The ungrammaticality of examples such as (26)b and (27)b is due to a minimality effect on the [+argumental] feature. Consider the derivations of examples (32)a and (32)b represented in (33): (32) a. ¿Qué compró Carmen? what buy:PST Carmen “What did Carmen buy?” b. *¿Qué Carmen compró? what Carmen buy:PST “What did Carmen buy?” (33) a. CP b. CP 2 2 Qué j C’ Qué j C’ [+arg]2 [+arg] 2 Cº TP Cº TP 2 2 pro T’ Carmen i T’ 2 * [+arg] 2 Tº VP Tº VP compró 2 compró 2 [+arg] Carmen V’ [+arg] t i V’ [+arg]2 2 Vº t j Vº t j In the derivation (33)a, the wh-constituent qué bears an [+argumental] feature and shares it with Cº by agreement. Since verbs select arguments, the finite verb 105 contained in Tº must also be [+argumental] and must also be the licensor for Cº by head/head agreement. Although the structure with a null pro in (33)a is licensed without any complications, when the [Spec, TP] is occupied by an overt subject marked as [+argumental], the licensing relation between the inflected verb and the fronted wh cannot be established. In Suñer’s explanation, this is because an intervention effect occurs on the [+argumental] feature. Ungrammaticality then arises as an effect of minimality. Given that adjuncts are marked as [-argumental], Suñer claims that they do not create an intervention effect if a [+argumental] subject appears in preverbal position. An obvious problem with Suñer’s stipulation is the lack of motivation of the [+argumental] feature, only designed to maintain Torrego’s argument/adjunct asymmetry. Consider examples (26)a and (31)a repeated below as (34)a and (34)b respectively: (34) a. ¿Qué i compró Carmen t i ? what buy:PST Carmen “What did Carmen buy?” b. ¿Cuánto i pesará esta beba t i ? how.much weigh:FUT this baby “How much will this baby weight?” One conceptual problem due to the use of the term “argumental” is that it is designed to describe paradigms where the tail of a chain is θ-marked (the trace of qué in (34)a) and to describe paradigms where the tail of a chain is not θ-marked (the trace of 106 cuánto in (34)b for instance). This identification leaves unanswered the question of why wh-questions with a θ-marked tail require postverbal subjects the same as wh-questions with a non θ-marked tail. Crucially, a further elaboration of the technical term of argumenthood by Cinque (1991), leaves examples such as (31) unexplained, Suñer’s key contribution. Consider the example (31)b repeated as (35): (35) *¿Cuánto i esta beba pesará t i ? how.much this baby weigh:FUT how much will this baby weight?” Rizzi’s original idea understands the argumenthood of a predicate as a function of the ability of being θ-marked. Cinque’s 1991 elaboration of Rizzi’s terminology considers θ-marked constituents to be [+referential]. Therefore, an argument is characterized by being [+referential]. If we adopt Cinque’s remarks, the wh-word cuánto in (35) cannot be “an argument of” the verbal predicate pesar “to weight” because its trace is [-referential]. In Suñer’s system, this would mean that cuánto would bear a [-argumental] feature. Given that cuánto would rather be [-argumental], [+argumental] preverbal subjects would be expected to cooccur, contrary to the facts as attested by the ungrammaticality of (35). Despite these insufficiencies, Suñer’s remarks on the core paradigms opened the possibility to question the Preverbal Subject Constraint and its associated explanations, which have been typically formulated as a reflex of the argument/adjunct asymmetry. 107 We will illustrate in Section 3.3.2 how Suñer’s system faces additional empirical problems some of them with an impact on the basic assumptions of the Standard Analysis. 31 I will claim that preverbal lexical subjects do not cooccur with wh-words as a function of the argumental or nonargumental status of the fronted wh- phrase. This contention will lead us to reformulate the Preverbal Subject Constraint in (25) in terms of whether or not the wh-phrase is nonreferential/nonspecific. 3.3.2 D-linking/Specificity and the Absence of Inversion There are additional empirical reasons to question the argument/adjunct approach to the restriction of preverbal subjects in interrogatives. I propose that the right generalization needs to be reformulated in very different terms, related to the presence or absence of two factors: the specificity of the wh-word and the D-linking properties of the wh-word. 32 A first set of data leading to a revision of the Preverbal Subject Constraint involves cases where a preverbal lexical subject may occur with a specific wh- phrase. The examples reported by several researchers such as Zubizarreta (2001), 31 For instance, Suñer assumes two different Case licensing positions [Spec, VP] and [Spec, TP]. This duality is possible in her system because postverbal subjects can remain at [Spec, VP] when an interrogative is fronted (see also Gutiérrez Bravo 2007). She also claims that the inflected verb cannot reach left peripheral nodes in interrogatives. Moreover, an issue that Suñer does not discuss is the influence of specificity or D-linking of the wh-phrases occurring in the left edge. 32 Specificity (Enς 1991), D-linking (Pesetsky 1987) and referentiality (Cinque 1991, Rizzi 1990) are different notions covering subtle aspects of the empirical paradigms. For expository purposes, I will not make any distinction among these terms in this chapter. For a useful discussion of these terms and the empirical data they cover, I refer the reader to Kiss (1993). 108 Inclán Nichol (1997), among others consist of θ-marked or referential arguments in Cinque 1991’s terminology do not force a postverbal subject: (36) a. ¿A cuál estudiante Lupe llamó el fin de semana? ACC which student Lupe call:1;PST the end of week “Which student did Lupe call this weekend?” b. ¿A quién de los presentes Lupe saludó en la fiesta? ACC who of the present Lupe meet:1;PST in the party “Who of the present people did Lupe meet at the party?” Cases such as (36) falsify the alleged argument/adjunct asymmetry. Preverbal subjects can cooccur with a D-linked wh-phrase as in (36)a or they can appear with a partitive wh-phrase as in (36)b even when both wh-words are argumental. Another environment where preverbal lexical subjects coappear with a fronted wh-word is manifested in the presence of assertion or polarity markers. Inclán Nichol (1997) reports data from Peninsular Spanish in which preverbal subjects are licensed with interrogatives if an assertion marker is introduced. For instance, the paradigms in (37) and (38), preverbal lexical subjects are allowed in presence of an affirmative marker or a negative marker: (37) a. ¿A quién Lupe sí llamó? ACC who Lupe AFF call:1;PST “Whom did Lupe positively call?” 109 b. ¿A quién Wiles sí le= resolvió el problema? DAT who Wiles AFF DAT=solve:3;PST the problem “Whom did Wiles positively solve the problem?” (38) a. ¿A quién mi amiga favorita no invitó en su cumpleaños? ACC whom my girl.friend favorite NEG invite:3;PST to her birthday “Whom didn’t my favorite girl friend invite to her birthday?” b. ¿A quién mi profesora favorita no le=regaló un libro? DAT whom my teacher favorite NEG DAT=give:3;PST a book “Whom didn’t my favorite teacher give a book?” Interestingly, the observations reported for Peninsular Spanish can be extended to Latin American Spanish dialects. For additional examples I refer the reader to Zubizarreta’s recent work (Zubizarreta 2007). There are two fundamental properties involved in the analysis of these paradigms. Both wh-words and lexical subjects occurring in preverbal position in these constructions must be specific or D-linked for the sentence to be felicitous: (39) Properties of constructions in paradigms (36)-(37) a. The ‘quality’ of the wh-word: the wh-word must be specific or D- linked. b. The ‘quality’ of the preverbal subject: The preverbal subject intervening in wh-questions must be specific or D-linked. 110 In the rest of this section, I will discuss in detail the property in (39), related to the quality of the wh-words. The property in (39)b, corresponding to the quality of the lexical subject appearing in preverbal position will be the topic of Section 3.3.3. Let us illustrate the observation that wh-words appearing with preverbal lexical subjects must be specific (see Inclán Nichol 1997) or D-linked. 33 One test to show this point can be drawn from the possible answers to wh-questions when a preverbal or a postverbal lexical subject occurs. For instance, (40) can only be answered by indicating the existence of an individual, presupposed in the context (i.e. (40)a), but they cannot be answered with a non specific indefinite (i.e. (40)b). In this sense, questions involving an affirmative marker parallel clefts: (40) ¿A quién Carmen sí llamó? DAT whom Carmen AFF call:PST “Who did Carmen call?" a. A Cristina. DAT Cristina b. *A nadie. DAT nobody In contrast, both possibilities are available with constructions where subject/verb inversion occurs. One can answer the question in (41) with a referential element or a nonspecific indefinite, an indication that wh-phrases alternating with postverbal lexical subjects are ambiguous between a specific, (a context with a set of 33 A piece of evidence for the D-linking/specificity condition comes precisely from dialectal variation. The equivalent of the examples with direct objects in (36), (37)a and (38)a sometimes are reported to admit doubling by a clitic. See Suñer (1992). 111 individuals is presupposed) and nonspecific interpretation (in absence of a presupposed context): (41) ¿A quién llamó Carmen? DAT whom call:PST Carmen “Who did Carmen call?" a. A Cristina. DAT Cristina b. A nadie. DAT nobody As discussed recently by Zubizarreta (2007), wh-questions that contain a positive or negative polarity item are not informational questions but contrastive questions. Therefore, the answer of a contrastive question cannot negate the presupposition associated with the wh-word. In clear cases where the preposed wh- is not specific or nonD-linked, we expect preverbal lexical subjects to be ruled out. The paradigms in (42) and (43) illustrate two representative cases: the wh-word que “what” and the aggressively non D-linked wh-word a quién diablos “who the hell”: (42) a. ¿Qué compró mi novia en esta tienda? what buy:3PST my girlfriend in this store “What did my girlfriend buy in this store?” 112 b. Nada. nothing “Nothing.” (43) a. ¿A quién diablos golpeó mi novia en la fiesta? ACC who devil beat:3PST my girlfriend in the party “Who the hell did my girlfriend beat at the party?” b. A nadie. ACC nobody “Nobody.” As shown above, the informational question with the wh-word qué in (42)a can be answered with the nonspecific indefinite nada in (42)b. Similarly, the aggressive non D-linked wh- in (43)a can be answered with the nonspecific indefinite nadie in (43)b because it is a truly informational question, where a contrast set is not available. Interestingly, the nonspecific what “que” and the aggressive non D-linked wh- a quién diablos “who the hell” also disallow the occurrence of preverbal lexical subjects when an assertion or polarity marker is present. The paradigm in (44) illustrates the first case, the impossibility of a preverbal lexical subject with the nonspecific what “que”: (44) a. *¿Qué Wiles sí le= resolvió a Lupe? what Wiles AFF DAT=solve:PST DAT Lupe "What did Wiles solve to Lupe?" 113 b. *¿Qué mi profesora preferida sí le=enseñó ayer? what my teacher favorite AFF DAT=teach:3;PST yesterday “What did my favorite teacher taught yesterday?” Aggressively non D-linked wh-words illustrate a similar pattern. Given that they convey informational focus and never express presupposed information; they are expected not to occur with preverbal lexical subjects. Therefore, subject inversion is forced (see (45)a) independently of whether or not an assertion marker is present ((45)b versus (45)c): (45) a. ¿A quién diablos invitó Cristina a la conferencia? ACC who devil invite:3PST Cristina to the lecture “Who the hell did Cristina invite to the lecture?” b. *¿A quién diablos Cristina invitó a la conferencia? ACC who devil Cristina invite:3PST to the lecture “Who the hell did Cristina invite to the lecture?” c. *¿A quién diablos Cristina sí invitó a la conferencia? ACC who devil Cristina AFF invite:3PST to the lecture “Who the hell did Cristina invite to the lecture?” Provisionally, let us reformulate the Preverbal Subject Constraint in terms of the absence of specificity or D-linking of the wh-word: (46) The Revised Preverbal Subject Constraint (provisional) If a nonspecific or a nonD-linked wh-word is preposed, lexical subjects in Spanish cannot appear in preverbal position. 114 In other words, the revised version of the Preverbal Subject Constraint correlates the postverbal position of the subject with the “quality” of the wh-word being fronted. “Subject inversion” takes place if the wh-word in interrogatives is nonspecific or nonD-linked. Otherwise, preverbal lexical subjects are expected to cooccur with a fronted wh-word. In the next section, we will see that the revision of the Preverbal Subject Constraint in stated in (46) needs to be refined to include the second property listed in (39): the quality of the lexical subject in preverbal position. Given the results obtained in Chapter 2, this refinement should also include a discussion of what ‘preverbal’ means in these contexts. 3.3.3 A V2 Analysis of Interrogatives in Modern Spanish In this section, I will discuss the second property occurring in constructions where a preverbal lexical subject can coappear with a fronted wh-: (47) a. ¿A cuál estudiante Lupe llamó el fin de semana? ACC which student Lupe call:1;PST the end of week “Which student did Lupe call this weekend?” b. ¿A quién de los presentes Lupe saludó en la fiesta? ACC who of the present Lupe meet:1;PST in the party “Who of the present people did Lupe meet at the party?” (48) a. ¿A quién Lupe sí llamó? ACC who Lupe AFF call:1;PST “Whom did Lupe positively call?” 115 b. ¿A quién Wiles sí le= resolvió el problema? DAT who Wiles AFF DAT=solve:3;PST the problem “Whom did Wiles positively solve the problem? In addition to the quality of the wh- (it must be specific or D-linked), I will show that the quality of the preverbal lexical subject also matters. Preverbal lexical subjects in (47) and (48) must display the property in (49): they must be specific or D-linked: (49) The preverbal lexical subject in paradigms (47)-(48) The ‘quality’ of the preverbal subject: The preverbal subject intervening in wh-questions must be specific or D-linked. The following paradigm, reported by Zubizarreta (2007), indicates that preverbal lexical subjects cooccurring with fronted wh-questions must be specific or D-linked: (50) a. *Qué diarios alguien no lee nunca? what newspapers someone NEG read:3PRS never b. Qué diarios alguien que tú conoces sí lee what newspapers someone COMP you know:3PRS;IND AFF read:3PRS siempre? always c. *Qué diarios alguien que tú conozcas sí lee what newspapers someone COMP you know:3PRS;SBJV AFF read:3PRS siempre? always 116 The obligatoriness of the specificity/D-linking requirement for the preverbal lexical subjects in these examples indicates that they are dislocated as argued in Chapter 2. The example (50)a is ungrammatical because the intervening preverbal subject alguien ‘someone’ is a nonspecific indefinite. If a relative clause inflected in indicative is added as in (50)b, a specific reading is forced and the sentence becomes acceptable. In contrast, ungrammaticality arises if the relative clause in question is inflected in the subjunctive, forcing the nonspecific reading of the indefinite alguien as in (50)c. Note that the wh-words in examples such as (51) are nonspecific but they still allow preverbal lexical subjects to occur: (51) a. ¿Qué cosa alguien no lee nunca? what thing someone NEG read:3PRS never b. ¿Qué cosa alguien que tú conoces sí lee siempre? what thing someone COMP you know:3PRS;IND AFF read:3PRS always Crucially, (51) falsify the provisional version of the Revised Preverbal Subject Constraint stated in (46) and repeated as (52): (52) The Revised Preverbal Subject Constraint (provisional) If a nonspecific or a nonD-linked wh-word is preposed, lexical subjects in Spanish cannot appear in preverbal position. The example in (51) contains a nonspecific wh-word in the preverbal field but still preverbal lexical subjects are attested. Therefore, (52) needs to be revised to include 117 the data in (51). In order to do this, the notion of “preverbal position” needs to be qualified for lexical subjects. As suggested by the paradigm below, lexical subjects marked as indefinite and specific are allowed to coappear in preverbal position with fronted nonspecific wh-words (see (53)b), but lexical subjects marked as indefinite and nonspecific are ungrammatical (see (53)c), a fact consistent with (50)c: (53) a. ¿Qué cosa nadie traduce al español? what thing nobody translate:3PRS to:the Spanish b. ¿Qué cosa nadie que sabe inglés traduce what thing nobody COMP know:3IND English translate:3PRS al español? to:the Spanish c. *¿Qué cosa nadie que sepa inglés traduce what thing nobody COMP know:3SBJV English translate:3PRS al español? to:the Spanish Within the approach to subject positions developed in this dissertation (for details see Chapter 2), there are at least two patterns of preverbal positions for lexical subjects. Some lexical subjects are arranged in a Canonical position where no specificity requirement operates. In contrast, other lexical subjects appear as part of a dislocation pattern where specificity is obligatory. I have hypothesized that preverbal subjects in the canonical pattern are located in the [Spec, TP]. Given that dislocation 118 is manifested as a property of the left edge, I would like to suggest that the lexical subjects discussed in previous paradigms are located in the specifier of the c*º head, the Assertion Phrase: (54) a. Canonical Pattern b. Dislocation Pattern TP c*P 2 2 DP T’ DP c*’ 2 2 Tº c*º We can thus reformulate the Revised Preverbal Subject Constraint by noting that if a nonspecific or a nonD-linked wh-word is preposed, subjects in the canonical pattern cannot appear in preverbal position. I suggest that the right formulation of the Preverbal Subject Constraint must be phrased as the New Preverbal Subject Constraint: (55) The New Preverbal Subject Constraint If a nonspecific or a nonD-linked wh-word is preposed, preverbal lexical subjects in Spanish cannot appear in the canonical position. Consider again the elaboration of the left edge introduced in Section 3.1. Assume that focused constructions and interrogatives share an interpretable [Q] feature in c*º. What makes an interrogative c*º different from a focus c*º is that only interrogative c*º is specified with an uninterpretable [wh] feature. The derivations in (56) show how the CP/c*P complex is articulated in focused declaratives and interrogatives: 119 (56) a. Focused Declarative b. Interrogative CP CP 2 2 Cº c*P Cº c*P [0] 2 [0] 2 c*º c*º i [Q] i [Q] u [wh] Since, the feature [Q] has a semantic/phonological nature; [Q] cannot be deleted at LF by the Principle of Full Interpretation. Therefore, the licensing relation for c*º must be phonologically related as argued in Section 3.1 by means of the operation of ph-lexicalization: (57) [Q]-ph-lexicalization A verbal head ph-lexicalizes the [Q] feature in c*º. Note that the resulting structure entails the projection of a specifier for the c*º head, which reflects the generalization that if a language displays verb movement in interrogatives, this language must also display overt wh-movement (Bury 2003). Recall that the results obtained in Section 3.2 suggested that the distribution of the inflected verb in Spanish might include left edge nodes. In addition, I have demonstrated in Chapter 2 that preverbal lexical subjects in Spanish can surface in a canonical pattern, involving the checking of Nominative Case in [Spec, TP]. Given the discussion of the empirical paradigms in interrogative clauses and the New Preverbal Subject Constraint, a reasonable hypothesis is to argue that the inversion structures in (32) –repeated as (58)- correspond to the derivations in (59): 120 (58) a. ¿Qué compró Carmen? what buy:PST Carmen “What did Carmen buy?” b. *¿Qué Carmen compró? what Carmen buy:PST “What did Carmen buy?” (59) a. c*P b. c*P 2 2 Qué j c*’ Qué j c*’ 2 2 c*º TP c*º TP compró 2 2 Carmen i T’ Carmen i T’ 2 2 Tº VP Tº VP 2 compró 2 t i V’ t i V’ 2 2 Vº t j Vº t j In inversion structures such as (58)a, the lexical subject is in canonical position or a Case position (i.e. [Spec, TP]). After verb raising to c*º in the lower left edge, the lexical subject surfaces as postverbal. The purpose of V-to-c*º is to ph- lexicalize the Q features on the c*º head. If this is true, one can also explain the ungrammaticality of (58)b as the failure to ph-lexicalize the assertion head in the lower left edge. I will state this condition in relation to the New Preverbal Subject Constraint in (55): (60) Preverbal lexical subjects in Spanish cannot appear in the canonical position in wh-interrogatives if [Q]-ph-lexicalization needs to apply. 121 Within the Standard Analysis, lack of raising to the left edge has been claimed to be a characteristic of Spanish syntax. Consider the classical examples reported by Suñer (1994: 345-347) which are intended to show the absence of verb movement to the left edge: (61) a. ¿Qué idioma todavía estudia Pepita en su tiempo libre? what language still study:3SG Pepita in her free time "Which language does Pepita still study in her free time?" b. ¿A cuántos todavía no les=has mandado tú tarjetas DAT how.many still NEG DAT=have:2SG sent you postcards para las fiestas? for the holidays “To how many haven’t you sent cards for the holidays?” Given that an adverb intervenes between the wh-word and the position of the inflected verb in (61), Suñer concludes that verb movement does not target the left periphery. However, the adverbs in these examples are focalized and the wh- constituents must be D-linked, a property already observed in wh-words preposed in the presence of the dislocation pattern of preverbal lexical subjects. For example, the use of todavía in (61)a presupposes that there might be other languages that Pepita studied in her free time but that at the present moment this is not the case anymore with the exception of at least one language. Similarly, (61)b presupposes that there is a group of people that will be sent postcards although some of them have not received any yet. 122 Furthermore, a non D-linked wh- does not trigger the same effect. For instance, it is well known that the wh-word what "qué" can hardly be taken as D- linked. Therefore, it is not surprising that the sentences with adverbials intervening between what and the verb in Suñer’s paradigm become severely degraded if not ungrammatical if the wh-word is nonspecific or non D-linked: 34 (62) a. *¿Qué probablemente canta un borracho a medianoche? what probably sing:3SG a drunk by midnight "What does a drunk probably sing at midnight?" b. *¿Qué aún no le=dió Mafalda a su mamá? what yet NEG DAT=give:PST Mafalda DAT her mom "What didn’t Mafalda give to her mom yet?" c. *¿Qué ayer compró Ximena en la carretera? what yesterday buy:PST Ximena in the highway "What did Ximena buy on the road yesterday?" 34 The example in (62)b is reported in Suñer (1994: ex. 25a) as grammatical. With the exception of Caribbean dialects where (62)b is perfect, this example is severely degraded among informants of Peninsular and Latin American dialects I consulted. They report that judgments improve when the reference of qué is specified within a previous discoursive context. For example (i) is much better than the out of the blue example in (62)b: (i) De todas estas cosas y juguetes... of all these things and toys "Among all these things and toys...” (ii) ¿qué aún no le=dió Mafalda a su mamá? what yet not DAT=give:PST Mafalda DAT her mom which didn't Mafalda give to her mom yet?" In other words, even examples such as (62)b conform to the observations made in the main text related to the quality of wh-s and preverbal lexical subjects in interrogatives. 123 If verb raising to the left periphery is closely related to a positional effect in V2 languages such as German, the failure or success of ph-lexicalization of the c*º head in Spanish interrogatives can be used to argue for a similar effect. I have claimed that this positional effect can be defined in c*º, a dislocation position in Spanish. 35 3.4 Conclusions In this chapter, I have described the left periphery in Spanish with special emphasis on interrogative clauses. I have argued that the Preverbal Subject Constraint, which disallows the occurrence of preverbal lexical subjects and wh- words in interrogatives, should be analyzed as a failure of verb raising to the left edge. This redefinition should include a specification of what “preverbal position” means for lexical subjects as in (63): (63) The New Preverbal Subject Constraint If a nonspecific or a nonD-linked wh-word is preposed, preverbal lexical subjects in Spanish cannot appear in the canonical position. This condition can be explained within an articulated version of the left periphery dubbed “the Narrow Left Edge.” In this system, there are two core functional projections: the Point of View Phrase (CP and the Assertion Phrase (c*P). The effects of the New Preverbal Subject Constraint are the result of the failure of the condition in (64), defined over c*º: 35 As will be argued in Chapter 4, German positional effects might be defined over the Cº head. 124 (64) [Q]-ph-lexicalization A verbal head ph-lexicalizes the [Q] feature in c*º. Verb movement to the left edge in Spanish interrogatives is one strategy used to comply with (64). Finally, I would like to suggest that the Narrow Left Edge system can be extended to other clausal types. Given that he notion of clausal typing/force is not understood as a primitive of the theory of functional projections, the different clausal types must be a manifestation of a particular arrangement of interpretable and uninterpretable left edge features. The combination of the features of the Cº and c*º heads gives rise to the following hypothesized clausal types: (65) Clausal typing Cº [0] i [command] u [wh] i [Q] focus construction desiderative interrogative c*º i [finite] declarative imperative exclamative For instance, the specification of [wh] in Cº derives interrogative and exclamative clauses depending on the featural selection in c*º. The establishment of a selectional relation expresses correctly the intimate relation between these interrogatives and exclamatives. Both clausal types use wh-words to identify their marked nature, but they differ in terms of whether the proposition introduced by the clause denotes quantificational force or not. Interrogatives denote quantificational force, which makes them similar to focused constructions and exclamatives do not denote quantificational force: 125 (66) a. Exclamative b. Interrogative CP CP 2 2 Cº c*P Cº c*P [wh] 2 [wh] 2 c*º c*º [finite] [Q] Finally, let us speculate that imperatives and desideratives are derived by the introduction of a [command] feature in Cº, which is combined either with a [finite] feature in c*º or a [Q] feature in c*º: (67) a. Imperative b. Desiderative/Prohibitive CP CP 2 2 Cº c*P Cº c*P [command] 2 [command] 2 c*º c*º [finite] [Q] The combinations of [command], [wh] and [Q] in (67) would give rise to imperative and desiderative clauses. In the next chapter, I will discuss the plausibility of the activation of th left edge in Spanish declarative constructions. In particular, I will argue that a second position effect can also be found in declaratives, not only in interrogative or marked clauses. 126 Chapter 4 Towards an Analysis of Spanish Declaratives: The Lower Left Edge In the this chapter, I will attempt to answer the question of whether the characterization of the phenomenon of second position in unmarked clausal types such as declaratives is restricted to the prototypical verb second languages or whether this phenomenon in fact corresponds to a more abstract mechanism. In particular, the underlying mechanism might be active in other languages but under different licensing conditions, a very pertinent question within the context of Modern Spanish. If the second position phenomenon is not considered a language specific property, but a clausal related property, this line of inquiry becomes not only justified, but it is actually necessary. 4.1 Typological Considerations on Clausal Typing and Second Position It is well known that some languages are sensitive to the presence of certain elements occurring in the second position of the clause. For instance, Jacob Wackernagel described in his 1892 study of Indoeuropean languages the typical effect by which certain unstressed or unstressable words are generally found in second position. Since then, this descriptive generalization is known as the Wackernagel Law: (1) Wackernagel Law Inherently unstressed/unstressable words are preferentially found in second position (P2). 127 The Wackernagel Law as stated in (1) is a positive condition. However, the distribution of unstressed particles such as object pronouns in Medieval Romance led to a relaxation of the Wackernagel Law. As observed by Tobler (1889) and Mussafia (1886), object pronouns do not seem to encompass any requirement on a second position. On the contrary, the generalization on Medieval Romance seems based on a restriction on the first or absolute linear position, the Tobler Mussafia Law: (2) Tobler Mussafia Law Inherently unstressed/unstressable words (i.e. object pronouns) cannot stand in first position (P1). Within the formal approach adopted here, the intuition behind (1) and (2) will consider the Wackernagel Law and the Tobler Mussafia Law as closely related aspects of the left edge. I will tentatively state (1) and (2) in terms of the descriptive generalization on the positional effects in (3): (3) Positional Effects If an inherently unstressed/unstressable word is present in the structure, a non inherently unstressed/unstressable word must stand in first position. The general question to be answered is how the positional effects expressed as “first/second position” are characterized in the left edge. It is well known that (1) and (2) are intended to capture several pieces of crosslinguistic variation. Once the positional effects in (3) are shown for Spanish, one should attempt to understand how these positional effects can be situated within the range of syntactic variation. 128 One can characterize the positional effects related to (3) within the Narrow Left Edge system as a function of the phonological visibility of a syntactic unit. As argued in Chapter 3, interpretable features contained in left edge heads require sometimes phonological support. This visibility requirement is expressed in terms of ph-lexicalization: 36 (4) Ph(onological)-lexicalization A feature in the head of a functional projection is ph-lexicalized if the head or the Spec of that projection is associated with appropriate ph- features at the point of spell out (Zubizarreta 2001: 189). Given the condition on ph-lexicalization, now consider the Narrow Left Edge system in (5). The Narrow Left Edge system is divided into two fields: the Higher Left Edge, headed by the Perspective/Point of View head or Cº; and the Lower Left Edge, headed by the Assertion head or c*º: (5) Fields in the Narrow Left Edge CP Higher Left Edge 2 SPEC C’ 2 Cº c*P Lower Left Edge 2 SPEC c*’ 2 c*º TP One factor of variation is related to the locus of lexicalization within the Narrow Left Edge. In declarative clauses, Higher Left Edge languages require special licensing of 36 Ph-lexicalization may be taken as a property of phases. Assuming that CP and v*P are phases (Chomsky 2000, 2001), ph-lexicalization can be argued to apply to v*º when the lexical verb in Vº incorporates to v*º. 129 Cº within the complementizer system. One of such languages will be shown to be German (see Section 4.2.1). In contrast, Lower Left Edge languages place a licensing requirement on c*º. Examples of such languages are Welsh, Breton and in particular, I will argue, Spanish. A second factor of variability has to do with the syntactic strategy utilized to comply with the ph-lexicalization requirement. As suggested by Zubizarreta (2001), Roberts (2005) among others, the natural choices are given by two of the core operations of the computational system: Move or Merge. This additional aspect of the licensing of functional projections allows us to distinguish between several language types. Consider the following typology connecting the locus of lexicalization and the lexicalization strategy as the parameters of variability: (6) Parametric variation and 1P/2P: Declaratives Locus of Lexicalization Higher Left Edge Lower Left Edge Move finite verb or auxiliary in German matrix clauses head movement of a particle (Spanish, Breton) or a verb (Spanish). Lexicalization strategy Merge complementizer in German embedded clauses a particle in Welsh or in Spanish As shown in (6), some Higher Left Edge languages lexicalize Cº in declaratives by Move or by Merge. A good exemplar is again German that lexicalizes Cº by verb or auxiliary movement in matrix clauses and lexicalizes Cº by Merge of the complementizer daβ in embedded clauses selected by a non bridge verb. In the case of Lower Left Edge languages, I will show that Welsh licenses c*º by Merge of a 130 head and Breton licenses c*º by Move. A more complex situation in the licensing of declaratives is shown by Spanish, another Lower Left Edge language. In Spanish, the c*º head can be lexicalized by Move and sometimes by Merge. In the next two sections, I will attempt to provide plausible answers to the question of how to characterize the crosslinguistic variation of positional effects and I will atempt to provide an answer to the question of why some languages display second position configurations and first position configurations. 4.1.1 Clausal Typing and Markedness Crosslinguistic variation in terms of positional effects is very well known. Some languages display the typical second position pattern (V2 languages such as German or V1 languages such as Breton and Welsh), others lack of any similar detectable effect (arguably East Asian languages) and others exhibit positional effects restricted to specific clausal types, but not to others (the so called Residual second position languages). One interesting observation to account for the range of variation is made by Ian Roberts. Assuming a markedness approach where interrogatives are more marked than declaratives, Roberts (2005:136) proposes two implicational universals to explain that languages display restricted or generalized first/second position effects across clausal types: (7) Roberts Generalization a. If a language has V2 in declarative clauses, then it has V2 in nondeclarative clauses. 131 b. If a language has declarative particles, then it has nondeclarative particles. Roberts claims that the generalizations in (7) explain the existence of full second position languages such as Welsh and German. In these languages, the licensing of their respective positional requirements is obtained either by a special particle (Welsh) or a constituent (German). Crucially, independently of the type of element, the licensing is needed both in declaratives and in interrogatives. Furthermore, (7) also predicts the existence of residual second position languages like English and French. In other words, the positional requirement might apply to marked clauses but not to unmarked clauses. As it is well known, English displays positional effects only interrogative clauses but not declarative clauses. Finally, Roberts also indicates that the implicational universals in (7) also explain that languages obeying the first/second positional restriction only in unmarked clauses but not in marked clauses are unattested. The direction of the implicational conditions in (7) is summarized in the table below: (8) Clausal Types and positional effects (Roberts 2005) Positional effect in declaratives? Positional effect in nondeclaratives? 1. Full 2P languages YES YES 2. Residual 2P languages NO YES 3. No positional effects NO NO 4. Unattested type YES NO Within the Narrow Left Edge system (see Chapter 3, Section 3.1), a similar result can be obtained if in addition to the bundle of uninterpretable features assigned to a functional head, the relevant interpretable features require a condition on 132 lexicalization. Let us follow Zubizarreta (2001) in assuming that this condition is characterized by the operation of ph-lexicalization (see also Chapter 3, Section 3.1.2), a PF instruction to the grammar to align the phonological features of a functional head in the context of another head with phonological content in the overt syntax or in the PF branch of the derivation: (9) Ph(onological)-lexicalization A feature in the head of a functional projection is ph-lexicalized if the head or the Spec of that projection is associated with appropriate ph- features at the point of spell out (Zubizarreta 2001: 189). The prototypical structure for a declarative before the point of spell out is depicted in (10): (10) Declarative CP 2 Cº c*P [0] 2 c*º [F] The diacritic [F] represents [finite] or [Q], the substantive content of the c*º head. Assuming that the positional effects mirror the exclusive activation of left edge heads, the ph-lexicalization requirement can apply to the c*º head or to the Cº head. The activation of the higher nodes of the clause structure in declaratives through ph- lexicalization predicts that the left edge must also be activated when nondeclaratives are constructed. For instance, consider the representations of a desiderative or an imperative sentence: 133 (11) a. Imperative b. Desiderative CP CP 2 2 Cº c*P Cº c*P [command] 2 [command] 2 c*º c*º [finite] [Q] If the clausal type is a desiderative or a imperative sentence, Cº must be specified as [command], and Cº must also combine with a feature of the c*º head. Whether ph- lexicalization activates Cº or c*º in declaratives, if an imperative or a desiderative clause is built, the distribution of positional requirements of full verb second languages is expected. In addition, residual second position languages are expected to be restricted to the specification of Cº, the higher node of the Narrow Left Edge. A ph-lexicalization condition on nondeclarative Cº does not automatically entail lexicalization of c*º in declaratives. Finally, a grammar allowing the absence of a positional requirement in nondeclaratives and the presence of a positional requirement in declaratives is also expected not to occur. This combination is predicted not to happen because one would claim that only declarative clauses are realized in such a grammar and that nondeclaratives cannot be constructed. 37 37 Some additional consequences may be derived from the markedness approach to the left periphery if it is correlated with the Subset Principle. The Subset Principle is a learning condition that stipulates that the most inclusive grammar compatible with the evidence available to the learner will be acquired first. In particular, it is predicted that unmarked clauses such as simple and focused declaratives will be acquired first. On the other hand, marked clauses will be more difficult to learn and will appear in a later stage of the acquisition process. 134 4.1.2 Phrase Structure, Interpretability and Verb Movement One intriguing question is why some languages display positional effects where the finite verb must be in second position (“V2 languages”, for instance German) and why others allow special particles to stand in initial position (“V1 languages”, for instance Welsh). In V2 languages, a speficier requirement must accompany V-movement, whereas in V1 languages, a specifier requirement is absent. In this section, I elaborate a possible explanation based on the findings by Bury (2003) and Roberts (2005). I will argue that these positional effects can be understood as the manifestation of conditions on the ph-lexicalization of the left edge heads containing interpretable features. Following Bury (2003), I will claim that the specifier requirement follows from wellformedness conditions on phrase structure building. Let us start the discussion with the necessity of a specifier requirement in V2 languages, which is absent in V1 langauges. Bury (2003) noted that there is an asymmetry between Move and Merge, the two core operations in the computational system. When Merge applies to two syntactic objects α and β, the label of the new structure can be the result of the projection of either α or β: (12) a. α b. β 2 2 α β α β In contrast, in minimalist models that rely on feature checking, the application of Move to an object α in a target structure δ, requires the projection of the target δ, but 135 not the projection of the moved element. Thus, (13)a is allowed but self- attachment of α as in (13)b is excluded: (13) a. δ b. * α 2 2 α i δ α i δ 2 2 δ β δ β 2 2 t i β t i β The contrast between (13)a and (13)b entails that moved material cannot project and head a structure. Crucially, Bury argues, the structure in (13)b is problematic because the head of the phrase is not uniquely defined: the top α has two heads, the moved α and the trace of α. Self-attachment of α and projection of α, however, are not excluded by any independent principle of phrase structure. Instead, the ungrammaticality of (13)b follows from the assumption that movement is driven for feature checking purposes and checking of features occurs in one of two configurations: Head/Head configuration and Spec/Head configuration. Importantly, if a Spec is projected for (13)b as in (14), the structure is fine: (14) α 2 XP α 2 α i δ 2 δ β 2 t i β Bury (2003) proposes that verb movement can be dispensed with in favor of the operation of s(yntactic/structural)-movement or p(ronunciation/PF)-movement. S- movement produces structures based on (14) with an obligatorily projected specifier 136 after self attachment of α ((14)). In contrast, p-movement structures do not require the obligatory projection of a specifier. If we conceptualize the extended projection of the verb as a chain, one can understand p-movement as the pronunciation of any position within this extended projection. As an illustration, consider (15)a where s-movement applied and (15)b where p-movement applied: (15) a. Created by s-movement b. Created by p-movement VP CP 2 2 XP V’ Cº TP 2 2 V i VP Tº VP 2 2 ZP V’ ZP V’ 2 2 t i Vº In (15)a, s-movement operated to create a specifier/head structure, V is pronounced in its landing site after self-attachment. In contrast, there is no structural displacement in (15)b and p-movement applies to the extended projection of V, namely the T and the C heads. Thus, p-movement involves the pronunciation of the verb in T or C. The structures depicted in (15) have important consequences for the analysis of verb second languages. In particular, (15)a corresponds to the structures created in Full V2 languages (i.e., Germanic) and (15)b corresponds to the structures attested in verb initial languages (i.e., Celtic). On the one hand, the application of self- attachment/s-movement in (15)a derives the specifier requirement, typical in Germanic Full V2 languages. 137 Within the Narrow Left Edge system, Bury’s idea of self attachment may be too strong. 38 This is because the content of the c* º head and the lexical verb can be dissociated and within the Narrow Left Edge, the content of c*º indeed has a repercusion at the LF component. The key idea in Bury’s system is to derive the specifier requirement from phrase structure conditions without resorting to feature checking or the postulation of an EPP feature. This vision may be maintained if the ph-lexicalization condition is assumed to apply to interpretable features. As argued in Chapter 3, interpretable features may also trigger syntactic operations (as in Nash and Rouveret 2002) and in particular, they may trigger ph-lexicalization, a condition on the alignment of an interpretable feature with proper phonological content. Let us assume with Epstein (1998) that the condition in (16) holds: (16) Substantivization of functional heads at LF After a syntactic operation applies to F by S, functional heads (F) become substantive (S) at LF. [ FP [ Fº Sº 1 [Fº] ... t 1 ]] become [ SP [ Sº Sº 1 [Fº] ...t 1 ]] at LF 38 In spite of the fact that Bury’s analysis is attractive, there are additional problems requiring a solution. For instance, it is difficult to derive in his system the contrast between Full V2 languages and Generalized V2 languages. Verb second in Full V2 languages such as German is a root phenomenon. In embedded clauses, verb second is only displayed in the absence of the complementizer of bridge verbs. In contrast, Generalized V2 languages such as Icelandic and Yiddish do not display this asymmetry and they require a verb second structure in root and embedded clauses even in the presence of an overt complementizer. Even if one stipulates, as Bury does, that structures created by self-attachment/s-movement cannot be selected, it is difficult to explain the situation in Icelandic and Yiddish, requiring further stipulations. 138 Now, (17)a represents a pre-spell out stage of the derivation where the complex T/V has moved to ph-lexicalize c*º. If the Specifier Requirement does not apply to (17)a to generate (17)b, the substantivization of c*º at LF will give rise to an ofending structure like (17)c: (17) a. Ph-lexicalization of c*º b. The Specifier Requirement c*P c*P 2 2 c*º TP SPEC c*’ [Tº+Vº] i 2 2 Move t i VP c*º TP 2 [Tº+Vº] i 5 t i t i c. (17)a after substantivization at LF VP 2 Vº i VP 2 t i VP 2 t i The structure in (17)c is ungrammatical because the higher VP cannot be defined with a single head. The head of the top VP can be Vº or any of the traces in the structure. Now consider the derivation of a structure when c*º is ph-lexicalized by Merge of a head Yº. After Yº is inserted in c*º, the derivation in (18)a is generated, but the substantivization condition can directly apply to (18)a to generate (18)c. The specifier requirement is not needed either and becomes optional: 139 (18) a. Ph-lexicalization of c*º b. The Specifier Requirement. c*P c*P 2 2 c*º TP SPEC c*’ Y i 2 2 Merge Tº c*º TP Y i 5 c. (18)a after substantivization at LF YP 2 Yº VP 2 Vº The reason of this difference is that (18)c does not violate any condition of phrase structure: Yº is defined as the head of YP and Vº is defined as the head of VP. It seems then that when Move does not involve the checking of uninterpretable features, the specifier requirement is needed for phrase structure and LF conditions: heads need be uniquely defined and functional heads become substantive at LF. Ph- lexicalization by Move creates such a context where interpretable features of a functional category are aligned with a proper phonological matrix. In contrast, if ph- lexicalization of an interpretable feature occurs by Merge, no specifier requirement is needed. Thus, the elaboration of Bury’s system within the postulates of the Narrow Left Edge is able to explain the crosslinguistic difference between languages with a ph-lexicalization requirement performed by Move and languages with a ph- lexicalization requirement performed by Merge. 140 4.2 The Licensing of Declarative c*º In the system adopted in this dissertation, the ph-lexicalization of the head of the Assertion Phrase (c*º) can take place by the application of Move or Merge (Zubizarreta 2001, Roberts 2005 among others). I will adopt the typology proposed by Roberts (op. cit.) to illustrate that languages can activate their clausal second position by using Merge, Move or by the creation of a [Spec, Head] relation: (19) Ph-lexicalization types in the left edge In Section 4.2.1, I illustrate how German resorts to verb raising to license second position structures involving the verb. In Section 4.2.2, I argue that a similar effect occurs in Welsh to license the second position. However, in contrast to German, Welsh employs specific particles to type a declarative clause, preempting verb movement. Finally, I introduce the case of Breton in Section 4.2.3, which exemplifies that in absence of declarative particles and verb raising, the second position can be licensed by Long Head Movement of a nonfinite form. 4.2.1 German I begin to illustrate the different ph-lexicalization strategies within the Narrow Left Edge syste with German, which represents the classical case in the V2 literature. In main/root clauses, German requires the verb to be linearly in second position. The counterpart of this requirement conspiring for the creation of V2 Root/embedded asymmetry Language 1. Creation of a [Spec, Head] in Cº YES German 2. Merge of a head in c*º NO Welsh 3. Adjunction of a head in c*º NO Breton 141 structures in this language is the obligatory presence of a specifier as illustrated by the examples in (20) from Travis (1991: 343) and Olsen (1985: 143): 39 (20) a. Die Kinder haben das Brot heute gegessen. the children have the bread today eaten “Today, the children have eaten the bread." b. *Heute die Kinder haben das Brot gegessen. today the children have the bread eaten “Today, the children have eaten the bread." 39 Sequences involving V3 are forbidden in German if multiple fronting generates the constituents preceding the verb or the auxiliary (van Riemsdijk and Zwarts 1974 for Dutch, van Riemsdijk 1997 for German and also Schneider Zioga 2007). If the fronted verb is preceded by a single moved constituent and further preverbal constituents are base generated, superficial V3 orders are allowed. This explains the contrast between (i) and (ii) reported by Bury (2003: 135): (i) a. *Frank gestern hat den Kuchen gegessen. Frank yesterday has the cake eaten “Yesterday, Frank has eaten the cake.” b. *Gegessen Frank hat gestern den Kuchen. eaten Frank has yesterday the cake “Yesterday, Frank has eaten the cake.” (ii) a. (Er sagte) der Hans der habe schon wieder Hunger. he say:PST the Hand the have already again hunger “(He said that) Hans is hungry again already.” b. (Christian meinte ) in der Stadt da werde er es nicht mehr lange Christian mean:PST in the city there would he it no more long aushalten. bear “(Christian said that) in the city, he could not stand it much longer there.” The paradigm in (i) involves multiple fronting of a subject and an adverb, giving rise to ungrammaticality by independent conditions on minimality. In contrast, the paradigm in (ii) shows that the constituents der Hans and in der Stadt are coreferential with the pronominals der and da respectively in their derived positions. 142 The constituent in initial position in German can also be a full VP as demonstrated by the VP-fronting construction. As shown by the examples in (21) reported by Webelhuth (1985), VP-fronting can pied pipe or not the complement of the verb: (21) a. [ VP Das Buch gelesen] j hat Hans nicht t j. the book read have:2 Hans not “Hans has not read the book.” b. [ VP Gelesen t i ] j hat Anna das Buch i nicht t j. read:PTCP have:2 Anna the book not “Anna has not read the book.” Note the auxiliary is in second position and intervening heads such as negation do not block the fronting of the VP. In contrast, embedded clauses show a different pattern indicating that V2 in German is lexically restricted. A V2 structure is only possible in the complement of bridge verbs and V2 is conditioned to whether or not the complementizer daβ ‘that’is overtly realized. The following examples from Cardinaletti and Roberts (1990) exemplify this point: (22) a. Bridge verb Ich glaube, (*daß) gestern hat Maria dieses Buch gelesen. I believe that yesterday have:3SG Maria this book read. "I believe Mary read this book yesterday." 143 b. Nonbridge verb *Ich bedauere, (daß) gestern hat Maria dieses Buch gelesen. I regret (that) yesterday has Maria this book read. “I regret that Maria has read this book yesterday." Example (22)a shows that V2 is only possible with bridge verbs and in the absence of the complementizer daβ “that.” In (22)b, V2 is impossible with nonbridge verbs, independently of the presence of the complementizer daβ ‘that.’ The root/embedded asymmetry just observed in combination with the specifier requirement can be explained by assuming that ph-lexicalization in German applies to the underspecified declarative Cº: (23) Underspecified declarative Cº must be ph-lexicalized in German. Consider the derivation of a root clause in (24) at the point where an underspecified Cº is merged in the structure. Given the absence of special particles in German and verbal agreement morphology, if the ph-lexicalization requirement holds in Cº, the lexical verb or the auxiliary must move to the left edge. Verb or auxiliary raising lexicalizes c*º on its way to Cº as a way to comply with the Principle of Full Interpretation: (24) CP 2 Cº c*P [0] 2 c*º TP [F] 2 T’ 2 Tº LEXICAL VERB 144 The characteristic requirement of having a constituent in the specifier position of Cº follows from the principles of phrase structure as argued by Bury (2000, 2003). Consider the derivation of V2 structures after verb raising to the Cº head in (25). First, as Bury argues, verb movement to Cº creates the compound head in (25)a and makes functional heads substantive at LF in (25)b (i.e., they are Vs and VPs): (25) a. Ph-lexicalization [ CP [ Cº Vº 1 [Cº] ... t 1 ]] b. Lexicalization of functional heads at LF After a syntactic operation applies to F by S, functional heads (F) become substantive (S) at LF [ CP [ Cº Vº 1 [Cº] ... t 1 ]] becomes [ VP [ Vº Vº 1 [Cº] ...t 1 ]] at LF c. Specifier requirement [ VP SPEC [ Vº Vº 1 [Cº] ...t 1 ]] Following Bury (2003), the structure in (25)b is ruled out by conditions on the well formedness of the phrase structure because both the trace of V 1 and V 1 would ambiguously be defined as heads of the VP. Therefore, if a specifier (SPEC) is projected as in (25)c, the structure is rescued because now V 1 can be uniquely defined as the head of the top VP and the trace of V 1 can be defined as the head of the lower VP. In embedded clauses, the correlation of verb second and the distinction bridge verb/nonbridge verb follows from the nature of Cº. The complement of bridge 145 verbs project a CP-c*P structure such as (26)a but the complement of nonbridge verbs only project a c*P structure such as (26)b: (26) a. Bridge verbs b. Nonbridge verbs CP c*P 2 2 SPEC C’ c*º TP 2 [finite]/[Q] Cº c*P [0] 2 c*º TP [finite]/[Q] As stated in (23), Cº is the locus of lexicalization in German, not c*º. However, the complementizer daβ ‘that’ is inserted in c*º as this head encodes properties related to the finiteness of the verb. With bridge verbs, the insertion of the complementizer in c*º preempts V2: verb or auxiliary raising to the left edge does not occur in presence of daβ. With nonbridge verbs, the overt realization of daβ also blocks verb or auxiliary raising. If daβ does not occur, the verb cannot raise to c*º precisely because the structure projected by the complement of nonbridge verbs lacks the higher Cº and ph-lexicalization applies to Cº in German. 4.2.2 Welsh An interesting case to consider is Welsh. According to Roberts (2005), there is evidence to argue that verb movement in this language does not target left peripheral nodes, as it is evident in German. This is apparent from the lack of a similar asymmetry to the one found in German main/embedded clauses (Roberts 2005: 20): 146 (27) Tybed a geith hi ddiwrnod rhydd wythnos nesa? wonder:1 Q get:FUT she day free week next “I wonder if she will get a free day next week.” Similar to the complementizer if in English, the special particle a introduces indirect questions in Welsh. Assuming that the pronominal subject hi ‘she’ is located in [Spec, TP], the order VSO in (27) is unexpected if the presence of overt material in COMP preempts verb movement to the left edge. Another argument, provided by Roberts (2005) has to do with the distribution of the auxiliary bod “be” and its special properties related to its position in COMP. If regular verbs display properties different than those of bod and bod is licensed in the left edge, regular verbs should not reach the left edge. The idiosyncratic nature of bod relies on being inflected for specific tense forms, absent in other verbal forms. In Welsh, present/future, conditional and preterite forms are synthetic whereas all other forms are periphrastic. However, bod has a synthetic form for the present and the imperfect. Interestingly, the variation of the morphological shape of bod correlates with the clausal type involved. The paradigm in the table (28) suggests that bod is syncretic with respect to finiteness and clausal typing. Observe the alternation Ø/y/d across affirmative, interrogative and negative clauses in the present tense and the alternation r/Ø/d in the imperfect tense (the symbol ‘Ø’ represents the absence of a morphological marking): 147 (28) a. Present tense b. Imperfect tense Roberts (2005) proposes that the imperfect and present tenses are only licensed in the left edge (specifically in Rizzi’s Finº functional projection). Given that bod is in complementary distribution with the declarative particles fe/mi and the focus particles a/y, and its distribution does not display a root/embedded asymmetry, it is reasonable to assume that bod raises to the left edge with the special purpose of licensing the clausal typing and finiteness. Otherwise, regular verbs must resort to the special particles to license COMP because they cannot raise to license the lower left edge. Therefore, one can conclude that the licensing of left edge is carried over into the use of specialized clausal typing particles. Assuming the Narrow Left Edge system of projections, the following table summarizes the distribution of the Welsh main particles with respect to the c*º and Cº projections: (29) Welsh particle system Affirmative Interrogative Negative dw i Ø ydw i (dy) dw i ti Ø wyt i dwyt ti mae o/hiØ ydy o/hi ndydy dan ni Ø ydan ni (dy)dan ni dach chi Ø (y)dach chi (dy)dach maen nhw Ø ydyn nhw dydyn Affirmative Interrogative Negative roeddwn i oeddwn i Ø doeddwn roeddet ti oeddet ti Ø doeddet ti roedd o/hi oedd o/hi Ø doedd o/hi roedden ni oedden ni Ø doedden ni roeddech chi oeddech chi doeddech roedden nhw oedden nhw doedden c*º-Particles a y yr fe mi Cº-Particles ni nad 148 The particles fe and mi type root declaratives and the particle yr types embedded declaratives. The root affirmative y appears in its phonological variant r with the tenses associated with bod as discussed above. Interrogatives and focused declaratives must be marked by the particle a. Finally, ni and nad introduce negative sentences in root and embedded contexts respectively and are presumably merged in c*º and then moved to Cº. The paradigm in (30) from Bury (2003: 139) and (31) from Roberts (2005) illustrates some facets of the particle system in Welsh. The paradigm in (30) illustrates how Welsh uses the particle mi in root declaratives, but requires the particle yr in embedded contexts: (30) a. Mi oedd John yn chwerthin DECL be:PST John in laugh “John was laughing.” b. Mae Mair yn gwybod [yr oedd John yn chwerthin] is Mair in known DECL be:PST John in laugh “Mair know that John was laughing.” Similarly, (31)a shows that Welsh introduces negative sentences in root contexts with the particle ni. If the clause is embedded as in (31)b, the special particle nad needs to be used: 149 (31) a. Ni ddarllenodd emrys y llyfr. NEG read:PST Emrys the book “Emrys did not read the book.” b. Dan ni ’n gabeithio nad ydach chi yn siomedig. are we ASP hope NEG are you PRED disappointed “We hope that you are not disappointed.” Welsh also displays special particles to type interrogatives: (32) a. Pa ddynion a werthodd y ci? which men Q sell:PST the dog “Which men sold the dog?” b. Y dynion a werthodd y ci. the men FOC sell:PST the dog “It is the men who have sold the dog.” Thus, Welsh types its clauses with special particles conforming to Roberts’ generalization in (7)b. Accordingly, the existence of declarative particles in a given grammar should entail the existence of non declarative particles. As shown in the preceding paragraphs, Welsh conforms to this typological generalization. In the next section, the study of Breton will illustrate a grammar in which declarative particles are absent, but nondeclarative clauses are still typed with a dedicated particle. 150 4.2.3 Breton Breton exhibits a similar system of clausal typing particles to the one found in Welsh. Consider the system of particles exemplified in the table in (33) and compare it with the one introduced in (29) for Welsh: (33) Breton particle system The paradigm in (34) illustrates the use of these particles. Typically, the Welsh particles introduce clauses with VSO orderings: (34) a. Gouzout a ra Lenaig [e lennas Yann al lizher] know PRT do:3SG Lenaig SUB read:PST Yann the letter “Lenaig knows that Yann read the letter.” (Stephens 1990:3) b. Goulenn a reas [hagen oac’h eveurus] ask:PST PRT do:3;PST INT be:2SG;PST happy “He asked whether you were happy.” (Roberts 2005) c. [Ne lena ketar vugale levrioh]. NEG read not children books “The children do not read books.” (Borsley and Stephens 1989) Subordinating e Interrogative hag-en Negative ne “direct” relative a “indirect” relative e 151 d. Ar plac'h yaouank [a weles dec'h]. the girl young REL see:2;PST yesterday “You saw the young girl yesterday.” “The young girl you saw yesterday.” (Timm 1988) In contrast to the situation in embedded clauses, verb initial orders in simple declaratives are disallowed in Breton and the insertion of a focus particle or a full constituent in first position does not alleviate the ungrammaticality: (35) a. *Lenn Anna al levr. read:PRS Anna the book “Anna reads the book.” (Roberts 2005) b. *A lenn Anna a levr . PRT reads Anna the book read:PRS “Anna reads the book.” c. *Al levr lenn Anna. the book read:PRS Anna “Anna reads the book.” Roberts notes that one important difference emerging from the comparison between Welsh and Breton stems on the presence of the declarative particles fe and mi to type declarative clauses in Welsh, strategy which is absent in Breton. The absence of declarative particles of the Welsh type may thus be the responsible factor of the pattern in (35). 152 Importantly, the focus particle a (glossed as “PRT”) in (35)b is subject to a second position requirement. A head different than the verbal head must front in the absence of a focus particle in order to avoid an offending verb initial order. These two strategies are summarized below: (36) Licensing of the Breton left edge a. In marked declarative constructions, the left edge is licensed by fronting of a constituent. b. In unmarked or neutral declarative constructions, the left edge is licensed by fronting of a head. Let us assume with Roberts (2005) that a is merged in the lower left edge, in our terms in the head c*º. Considering the first strategy, one can claim that the fronting of a constituent operates to the specifier of c*P. Therefore, when the particle a is used, a pattern similar to Germanic V2 but with the particle a in second position must be used: (37) Al levr a lenn Anna. the book PRT reads Anna “Anna reads the book.” In presence of the particle a, Breton allows a DP argument (see (38)a and (38)b) or a VP (see (38)c) in the specifier of c*P: (38) a. Annaig a gerc’ho bara evit koan. Annaig PRT fetch:FUT bread for dinner “Annaig will fetch bread for dinner.” 153 b. Bara a gerc’ho Annaig evit koan. bread PRT fetch:FUT Annaig for dinner “Annaig will fetch bread for the dinner.” c. Kerc’hat bara a raio Annaig evit koan fetch bread PRT fetch:FUT Annaig for dinner “Annaig will fetch bread for the dinner.” Given that Breton does not resort to any specialized particles to type declaratives, the marked status of the particle a becomes neutral if the construction known as Long Head Movement arises. Long Head Movement refers to a construction, in which an infinitival or a participle fronts to the head c*º occupied by a or an auxiliary. The movement of these heads may cross one or more intermediate heads on its way to c*º: (39) a. Lenn a ra Anna al levr. read:INF PRT do:3SG Anna the book “Anna reads the book.” b. Lennet en= deus Anna al levr. read:PCTP have:3SG Anna the book “Anna has read the book.” Although Long Head Movement in Breton resembles the German remnant topicalization construction, it can be shown that Long Head Movement displays very different properties. For instance, only Long Head Movement is blocked by negation 154 and is clause-bounded. For details, I will refer the reader to the discussion in Roberts (2005: 126). 4.3 Second Position in Modern Spanish: Declarative Clauses In Chapter 3, I demonstrated that the analysis of marked clauses in Modern Spanish shows that the second position in the language is active. Following the logic of the typological generalizations proposed by Roberts in (7)a repeated as (40), Spanish should be a residual V2 of the English type with the second position requirement applied only in marked clauses such as interrogatives, but not in unmarked clauses or declaratives: (40) If a language has V2 in declarative clauses, then it has V2 in nondeclarative clauses Nevertheless, given that Spanish left periphery is more complex than English and allows different subject positions (see Chapter 2), object positions, multiple dislocations and a higher flexibility for the positioning of the finite verb; this direction for the analysis of declaratives seems counterintuitive. One cannot equate English and Spanish without missing these important differences. At the same time, one must also recognize that the evidence for the activation of a second position in declaratives in Modern Spanish is not obvious. For instance, one cannot take Spanish to be a full V2 language of the German type. In fact, the consensus in the field is that Spanish cannot be a V2 language because it differs in crucial ways from German (for discussion see Suñer 1994). 40 A tacit reasoning 40 An exception is Mallén (1993) who simply assumes that verb second is active in the language, but without an explicit or formal justification. 155 within the field seems to have been that since Old Spanish displayed the effects of the Tobler Mussafia Law (Fontana 1993) and Old Spanish evolved into a grammar in which the effects of the Tobler Mussafia Law were not visible, the second position of the clause ceased to be active. However, under the conception that the second position phenomena actually represent very abstract properties of the left edge, the question of whether Modern Spanish left edge displays a positional effect remains open. This remark is critical especially if one considers that the second position phenomena is not a byproduct or specific attribute of object or weak pronouns. It is also important to consider that the phenomena under the label of “second position” are not uniform across languages (Vikner 1995, den Besten 1983, Platzack 1983, and Cardinaletti and Roberts 1990 for Germanic; Schneider Zioga 2007 for Kinande and Poletto 2000 for Old Romance). 4.3.1 Ph-lexicalization in the Spanish Left Edge In Section 4.2, I have illustrated three ways to license the functional projections responsible for the second position effects in declaratives: (41) a. The German strategy The head Cº requires ph-lexicalization by a finite verb or auxiliary. A [Spec, Head] relation is created. Root/embedded asymmetry. b. The Welsh strategy A particle ph-lexicalizes the head c*º by Merge. No root/embedded asymmetry. 156 c. The Breton strategy Absence of declarative clausal typing particles Ph-lexicalization of c*º by fronting of a head (Long Head Movement) From the analysis of interrogative clauses in Spanish, it seems clear that Spanish displays a positional effect not only in main clauses, but also in embedded clauses. This effect is independent of the distinction bridge/nonbridge verb. As an illustration, consider the minimal pair in (42): (42) a. Mimi se=preguntaba que qué compró Omar. Mimi REFL=wonder:IPFV COMP what buy:3PST Omar “Mimi was wondering what Omar bought.” b. *Mimi se=preguntaba que qué Omar compró. Mimi REFL=wonder:IPFV COMP what Omar buy:3PST “Mimi was wondering what Omar bought.” (43) a. Mimi quiere saber [qué compró Omar]. Mimi want:3SG know:INF what buy:PST Omar “Mimi wants to know what Omar bought." b. *Mimi quiere saber [qué Omar compró]. Mimi want:3SG know:INF what Omar buy:PST “Mimi wants to know what Omar bought." I assume that the absence of a root/embedded asymmetry in Spanish interrogatives indicates that the positional effect cannot be defined over the Cº head. If anything, it would have to be specified on c*º. Based on the empirical evidence that will be 157 shown in Section 4.3.2 and Section 4.3.3, I will claim that the requirement on ph- lexicalization of c*º is in fact true of Spanish declaratives: (44) a. c*º must be ph-lexicalized in Spanish b. c*’ 2 c*º TP 2 DP T’ 2 Tº VP ph-lexicalization 5 The source of variation in the licensing of positional effects in Spanish would arise from the type of strategy (Move or Merge) and the different licensers. Consider the structures in (45), which represent the interpretable content of c*º in simple declaratives and focused declaratives. The c*º head is hypothesized to be [finite] in simple declaratives and [Q] in focused declaratives: (45) a. Declarative b. Focused declarative CP CP 2 2 Cº c*P Cº c*P [0] 2 [0] 2 c*º c*º [finite] [Q] I will claim that the main strategy used in Spanish is Move. Importantly, interpretable features are ph-lexicalized in the Spanish left edge heads by different licensers. On the one hand, the [finite] feature can be activated by a verbal head containing the overt realization of φ-features. In particular, a clitic/verbal complex or 158 the subject agreement can ph-lexicalize the Assertion head. 41, 42 On the other hand, the [Q] feature requires ph-lexicalization by a verbal head. In addition to the Move strategy, the features [finite] and [Q] can also be associated with an assertion marker by movement. In particular, I will claim that the Assertion head can also be licensed by Move when the affirmative marker is “floating” in the left edge or can be licensed by Merge when a complementizer is realized in the sí que construction (“Anchored sí construction”). All the previous observations can be summarized in (46): (46) Activation of the Spanish Left Edge a. The c*º [finite] head is activated by Move applied to the overt presence of φ-features in (object clitics or subject agreement). b. The c*º [Q] head is activated by Move applied to the overt presence of a verbal head. c. In addition, the c*º [finite] and the c*º [Q] heads can be activated by Move (i.e., an affirmative marker in the floating construction) or by Merge (a complementizer in the anchored construction). The use of the Move strategy in (46)a and (46)b and (46)c will be shown to give rise to a [Spec, Head] relation. The Merge strategy in (46)c is more controversial and requires a better understanding of the syntax of the different classes of assertion markers. In particular, if we take the affirmative marker, it seems plausible to argue 41 Raising of agreement subject will be discussed in Chapter 5 within the analysis of subject dislocation. 42 See Schneider Zioga (2007)’s analysis of Kinande, a related proposal where V2 languages may arise when the element in second position is AGR. 159 that the strategy represented by (46)c is an innovation, perhaps the result of a historical change in progress. See Section 4.5 for speculations in this direction. 4.3.2 Word Order and Clitics in Absolute Verb Initial Sequences Despite the fact Spanish lacks the clausal typing particles typical of Welsh; I would like to suggest an interesting parallel between these two languages. A novel way to conceive clitics in Modern Spanish is to recognize that they have clausal typing properties. In Romance in general and in Spanish in particular, clitics can surface in a proclitic (they precede the verbal host) or an enclitic pattern (they follow the verbal host). Crucially, only the proclitic pattern correlates with finiteness and declarative type. If the enclitic pattern arises, the clause must be nonfinite and nondeclarative (infinitivals, imperatives, desideratives or exclamatives): (47) Correlation proclisis/clausal type Finiteness and declarativity are correlated with the proclitic pattern. The paradigm in (48) illustrates the proclitic pattern with an object clitic and with a reflexive clitic with a benefactive value. In both cases, the clause “typed” with the proclitic pattern is a finite declarative: (48) Proclisis and Clausal Typing a. Mimi lo=visitó en su Mustang nuevo. Mimi ACC=visit:3PST in her Mustang new “Mimi visited him in her new Mustang.” 160 b. Se=alquiló un departmento en Westwood. REFL=rent:3PST an apartment in Westwood “(S/he) rented an apartment in Westwood for (herself/himself).” In contrast, enclisis is correlated with nonfiniteness and with a nondeclarative value. This is shown in (49)a with an infinitival and in (49)b with a matrix imperative: (49) Enclisis and Clausal Typing a. Visitar=lo en su Mustang nuevo es imútil. visit:INF=ACC in her Mustang new is useless “Visiting him in her new Mustang is useless.” b. Alquíla=lo en Westwood. rent:IMP=ACC in Westwood “Rent it in Westwood.” I would like to propose that the finite verb in examples involving the proclitic pattern has been displaced to the left edge, in particular to c*º in simple declaratives and presumably to Cº in matrix imperatives. If the c*º position is ph-lexicalized by a complementizer, preempting verb raising; an assertion marker is needed to create a positional effect (see also Section 4.4). Another strategy when a complementizer fills the c*º head is to change the clausal type by altering the indicative verb morphology to subjunctive mood, giving rise to desideratives and embedded imperatives. 43 Crucially, absolute verb initial orderings in Spanish require the overt presence of a clitic/verbal complex or the overt presence of subject agreement in the Assertion head. If the canonical word order of constituents is defined in terms of the 43 See Beas (2007) for details. 161 absence of marked discourse pragmatic roles such as topic, focus and emphasis, Spanish canonical word order of constituents displays an asymmetry between verb types. Gutiérrez Bravo (2002: 3-4, 2007) reports that when transitives and unergatives are considered, lexical subjects must precede the verbal predicate: (50) a. Cristina llamó a Melissa el jueves. Cristina call:3;pst ACC Melissa the Thursday “Cristina called Melissa last Thursday.” b. José Luis bailó toda la noche. Jose Luis dance:3;PST all the night “Jose Luis danced all the night.” In contrast, the verbal predicate must precede the lexical subject with unaccusatives: (51) Llegó Monica a la reunión temprano. arrive:3;PST Monica to the meeting early “Monica arrived early to the meeting.” Leaving aside unaccusatives, transitive and unergative sequences with the lexical verb in first position are stilted and tend to be interpreted as yes/no questions, not as declaratives. Therefore, the example (50)a with a transitive verb is only acceptable with an absolute verb initial sequence if the clause is interpreted as an interrogative (unacceptability is marked as %): 44 44 Absolute verb initial orders cannot be appropriate answers to a wide focus question either. Given that sequences where the verb precedes all other constituents have been related to some form of detopicalization role, I would like to suggest that these sentences correspond to a different structure involving the placement of the finite verb on the head of the Point of View Phrase (Cº). This issue deserves further investigation. 162 (52) a. %Llamó Cristina a Melissa el jueves. call:3;pst Cristina ACC Melissa the Thursday “Cristina called Melissa last Thursday.” b. ¿Llamó Cristina a Melissa el jueves? call:3;pst Cristina ACC Melissa the Thursday “Did Cristina call Melissa last Thursday?” Crucially, absolute verb initial sequences with transitives and unergatives become fully grammatical if a clitic is added. In such instances, the outcome is not interpreted as neutral because pragmatic considerations come into play. Take example (53)a as the base sentence representing the canonical order of constituents. As expected, if the verb is initial as in (53)b, the sentence is degraded or awkward. If a clitic is inserted in first position, the sentence is completely acceptable: (53) a. Los prisioneros robaron las llaves de la cárcel. the prisoners rob:3PL;PST the keys of the jail “The prisoners robbed the keys of the jail.” b. %Robaron los prisioneros las llaves de la cárcel. rob:3PL;PST the prisoners the keys of the jail “The prisoners robbed the keys of the jail.” c. Se=robaron los prisioneros las llaves de la cárcel. REFL=rob:3PL;PST the prisoners the keys of the jail “The prisoners robbed the keys of the jail.” 163 d. Las=robaron los prisioneros. ACC=rob:3PL;PST the prisoners “The prisoners robbed them.” As illustrated by (53)c-(53)d, clitics in initial position do not have to be related to an internal argument of the verbal predicate. They can be unrelated to the argument structure of the predicate and play a detopicalization role (see (53)c) or they can refer back to a previously mentioned constituent (a “topic”), conceptually associated to an internal argument (see (53)d). I would like to propose that the oddness of absolute verb initial declarative sentences without a clitic and the higher acceptability of declarative sentences with a clitic rescuing an otherwise absolute verb initial sequence is due to the ph- lexicalization requirements of the [finite] feature present in the head of the Assertion Phrase: (54) [finite]-ph-lexicalization The [finite] feature in c*º ph-lexicalizes if the appropriate licensing material contains visible φ-features, unchecked at the point of the spell out. Consider the derivation of the sentence in (53)b versus the sentence in (53)c. Starting with (53)a, after verb has raised to Tº and the lexical subject los prisioneros has moved to its canonical position (i.e. [Spec, TP]) to check Nominative Case, the head c*º with the [finite] feature is inserted and the derivation reaches the stage in (55): 164 (55) c*’ 2 c*º TP [finite] 2 DP T’ los prisioneros 2 Tº VP robaron 5 las llaves de la cárcel The head c*º contains the interpretable feature [finite] and requires ph- lexicalization. Given the Criterial Freezing (see Chapter 1, Section 1.2.3), subject agreement on the verb has become inactive or frozen after raising of the DP subject los prisioneros to the [Spec, TP]. Therefore, the finite verb cannot move to the c*º head. If the finite verb could move, the ungrammatical structure (56)b would be created and the [Spec, Head] structure in (56)a would never arise: (56) a. δ b. * α 2 2 α i δ α i δ 2 2 δ β δ β 2 2 t i β t i β Thus, the only way a specifier can be created for the c*º head is if the verbal head is provided with some additional φ-features, materialized in the form of a clitic. Given the special status of clitics, their presence as part of a clitic/verbal complex in c*º must be licensed by their association with the proper argument they refer to. This link can be established through the presence of an overt or covert DP in the local environment of c*º. A [Spec, Head] relation is therefore created: 165 (57) a. c*P b. c*P 2 2 DP c*’ pro c*P 2 2 c*º c*º [finite] [finite] clitic/verbal complex clitic/verbal complex Although the clitic in the clitic verbal complex can be related to an object (see (53)d), this analysis also finds support in the case of subjects. In examples similar to (53)c, the clitic initial structures with transitives and unergatives share the property of being rescued by a clitic that takes the form of the lexically reflexive clitic and displays agreement with the subject. In (58)a and (58)b the pronominal subject agrees in person and number with the inflection contained on the finite verb and with the form of the initial clitic: (58) a. Yo me=caminé más de una milla. I 1;REFL=walk:1;PST more of a mile “I walked more than a mile.” b. Nosotros nos=caminamos más de una milla. We 1PL;REFL=walk:1PL;PST more of a mile “We walked more than a mile.” As mentioned before, these clitics reinforce the action denoted by the proposition. For example, in contrast to the cliticless counterparts of (58), which more or less express the proposition indicated in the English gloss, the structures including the clitic indicate the involvement of the agent who performs the action of caminar “to walk.” 166 In absence of resumption and clitics, arguments other than the lexical subject can appear in first position. However, in that situation, the initial constituent receives a contrastive focus intonation and the finite verb must appear immediately after the focused constituent or ungrammaticality occurs. Thus, the specification of c*º in focused declaratives is [Q], which also requires ph-lexicalization: (59) [Q]-ph-lexicalization The [Q] feature in c*º ph-lexicalizes if the appropriate lexical material associated with c*º does not contain active (“unchecked”) φ- features. Consider the paradigm in (60) based on Hernanz and Brucart (1987) and Zubizarreta (1994). If the direct object la carta “the letter” in (60)a is displaced to sentence initial position, the finite verb must precede the grammatical subject as in (60)b. 45 If not, ungrammaticality arises as exemplified in (60)c: 45 As expected, fronting of indirect objects replicates a similar pattern. Example (i)b and (i)c with reference to the neutral sentence in (i)a show that the linear order between the verb and the overt subject must be altered for the sentence to be grammatical: (i) a. Pedro escribió a Juan. Pedro write:3;PST DAT Juan "Pedro wrote to Juan." b. A JUAN escribió Pedro (no a Verónica). DAT Juan write:3;PST Pedro NEG DAT Verónica "TO JUAN, Pedro wrote (and not to Verónica)." c. *A JUAN Pedro escribió (no a Verónica). DAT Juan Pedro write:3;PST NEG DAT Verónica "TO JUAN, Pedro wrote (and not to Verónica)." The same situation obtains with PP complements. 167 (60) a. Juan escribió la carta. Juan write:3;PST the letter "Juan wrote the letter." b. LA CARTA escribió Juan (pero no el ensayo). the letter write:3;PST Juan but NEG the essay "THE LETTER, Juan wrote (but not the essay)." c. *LA CARTA Juan escribió (pero no el ensayo). the letter Juan write:3;PST but NEG the essay "THE LETTER, Juan wrote (but not the essay)." Consider the derivation of (60)b versus the derivation of (60)c. After the lexical subject has raised to [Spec, TP] and has checked the agreement features on Tº and its own Case related features, the [Q] feature on the c*º head is merged and the structure reaches the following form: (61) c*’ 2 c*º TP [Q] 2 DP T’ Juan 2 Tº VP escribió 5 la carta The derivation in (61) proceeds as follows. The condition on ph-lexicalization in (59) attracts the finite verb because the subject Juan in its canonical position has already deactivated subject agreement. Assuming that the φ-features of the direct object have been deactivated by some previous checking operation, the direct object cannot front to ph-lexicalize the head c*º because the [Q] would remain without lexical support. 168 Something similar would occur if the lexical subject occupies the specifier of c*º. Therefore, the only grammatical derivation takes the verbal head and raises it to the left edge, projecting an obligatory specifier. The intuition behind this analysis attempts to capitalize on the idea that only deactivated [AGR] can relate to the [Q] feature. The association of deactivated [AGR] and the quantificational focus feature [Q] is the source of the effect of emphasis or reinforcement detected in focused declaratives (for related ideas see Simpson and Wu 2002). 4.3.3 Modal Particles in the Left Periphery Spanish has a limited set of particles with different degrees of flexibility to appear in preverbal or postverbal position. Typically, they are related to adverbial elements, conjunctions and demonstratives. Consider the paradigm in (62) with several adverbs or adverbial particles that range over the aspectual/sentential portion of a declarative clause with a modal meaning: (62) a. Mónica posiblemente llegue tarde. Mónica possibly arrive:SBJ late "Possibly, Mónica will arrive late." b. Mi hermana obviamente se=divorció my sister obviously REFL=divorce:PST "Obviously, my sister divorced." 169 c. El cantante evidentemente está enfermo. the singer evidently is sick "The singer is evidently sick." A very important property of the adverbial particles in (62) is that they display selectional properties, typical of verbal heads for example. The following contrast shows that these adverbial particles must select for a finite clause: (63) a. Obviamente trabajar es bueno para tu salud. obviously work:INF is good for your health “It is obviously true that working is good for your health.” b. *Obviamente trabajar. obviously work:INF “It is obviously true that working.” In addition, these particles can be paraphrased as impersonal constructions, a combination of the overt copula ser ‘to be’ inflected for tense and default φ-features (third person singular) plus a complementizer que. Compare the adverbial particles in (62) with their corresponding equivalents in (64). The periphrastic adverbial constructions require a finite clause either in indicative or subjunctive mood: (64) a. Es posible que [Mónica llegue tarde]. is possible COMP Mónica arrive:SBJ late "It is possibly true that Mónica will arrive late." 170 b. Es obvio que mi hermana se=divorció. is obvious COMP my sister REFL=divorce:IND;PST "It is obviously true that my sister divorced." c. Es evidente que el cantante está enfermo. is evident COMP the singer is:IND sick "It is evidently true that the singer is sick." Let us call the complementizers associated with these periphrastic particles “assertive complementizers.” Note that assertive complementizers are homophonous with propositional complementizers required by clause selecting verbs (i.e., que in Violeta cree que Víctor es poeta ‘Violeta believes that Víctor is a poet’). However, there are reasons to keep propositional and assertive complementizers as two different realizations in the structure. Not only can they cooccur, but also crucially, they have a different semantics. For instance, propositional complementizers are associated with verbs that induce opaque contexts. Therefore, they disallow inferences in which there is substitution of two names or descriptions that have the same reference. Example (65)c is not a valid inference from the propositions in (65)a and (65)b: (65) a. El oficial del IRS cree que mi novia paga impuestos. the officer of:the IRS believe:3 COMP my girlfriend pay:3 taxes “The IRS officer believes that my girlfriend pays taxes.” 171 b. Mimi es mi novia. Mimi be:3 my girlfriend “Mimi is my girlfriend.” c. El oficial del IRS cree que Mimi paga impuestos. the officer of:the IRS believe:3 COMP Mimi pay:3 taxes “The IRS officer believes that Mimi pays taxes.” In contrast, assertive complementizers actually allow the substitution of two coreferential elements. It is then possible to infer (66)c from the propositions in (66)a and (66)b: (66) a. Es evidente que mi novia paga impuestos. be:3 evident COMP my girlfriend pays taxes “It is evidently true that my girlfriend pays taxes.” b. Mimi es mi novia. Mimi be:3 my girlfriend “Mimi is my girlfriend.” c. Es evidente que Mimi paga impuestos. be:3 evident COMP Mimi pays taxes “It is evidently true that Mimi pays taxes.” Arguably, this difference between complementizers can be explained if one assumes that the particles associated with assertive complementizers predicate about the truth- value of the proposition denoted by the sentence in which they appear. Etxepare and Bellert’s remarks on the nature of modal adverbials (Etxepare 1998, Bellert 1977) 172 confirm this observation. Modal adverbials are sensitive to the propositional content of a sentence. If the truth value of a proposition is questioned, a modal adverb cannot cooccur because one cannot inquiry simultaneously about the truth- value of a proposition and evaluate the truth-value of the same proposition: (67) a. Alfonsina probablemente regresará. Alfonsina probably return:3;FUT “Alfonsina will probably return.” b. *¿Regresará probablemente Alfonsina? return:3;FUT probably Alfonsina “Will Alfonsina probably return?” Note that assertive complementizers cannot be sentence initial. If a sentence begins with a que complementizer, this sentence may be interpreted under the right conditions as a marked clausal type, not as a declarative: (68) a. ¡Que Mónica vaya a la fiesta! COMP Mónica go:SBJV to the party “(I want) Mónica to go to the party!” b. Que el año nuevo sea mejor. COMP the year new be:SBJV better “(I wish) this New Year be better.” The sentences in (68)a and (68)b include a complementizer in the initial position of the clause, but they are not interpreted as declaratives. Instead, they must be taken as imperative or desiderative clauses requiring the presence of irrealis mood 173 (subjunctive). A declarative with an initial complementizer must necessarily be understood as associated with a covert verb as proposed in the performative hypothesis of the 1970’s (Ross 1970). For instance, the sentences in (69) and (70) with a complementizer in first position and the finite verb inflected in indicative may be determined by the pragmatic and discoursive context to the verbs decir ‘to tell’ and creer ‘believe’. Otherwise, they are ungrammatical: (69) a. (Dicen) que Mónica fue a la fiesta. say:3PL COMP Mónica go:3PST;IND to the party “(They say) that Mónica went to the party.” b. *Que Mónica fue a la fiesta. COMP Mónica go:3PST;IND to the party “that Mónica went to the party.” (70) a. (Creen) que el año nuevo es mejor. believe:3PL COMP the year new be:3PST;IND better “(They believe) that the New Year is better.” b. *Que el año nuevo es mejor. COMP the year new be:3PST;IND better “That the New Year is better.” A key remark is that all these sentences can be rescued by the insertion of an overt expletive-like verb, precisely the copula ser that appears with periphrastic adverbials. The paradigm in (71) illustrates how ser licenses the construction and the paradigm in (72) shows that the presence of the copula ser forces indicative mood: 174 (71) a. Es que Mónica fue a la fiesta. be:3 COMP Mónica go:3PST;IND to the party “It is the case that Mónica went to the party.” b. (Creen) que el año nuevo es mejor. believe:3PL COMP the year new be:3PST;IND better “(They believe) that the New Year is better.” (72) a *Es que Mónica vaya a la fiesta. be:3 COMP Mónica go:3SBJV to the party “It is the case that Mónica goes (SBJV) to the party.” b. *Es que el año nuevo sea mejor. be:3 COMP the year new be:3SBJV better “It is the case that the New Year is (SBJV) better.” In other words, the copula ser is incompatible with marked clausal types such as imperatives and desideratives that force subjunctive mood on the finite verb. In conclusion, the existence of adverbial particles and periphrastic adverbial constructions with a modal value suggest that Modern Spanish invoke the presence of an overt element in first position of declarative clauses. This overt element is realized as the copula ser in order to rescue a potential violation of a positional effect where the complementizer appears in first position. Note that the paradigms examined in this section conform to the formulation of the Wackernagel Law in (1). In other words, assertive complementizers being “inherently unstressed” cannot stand in initial absolute position and require the presence of an overt element. In the 175 next section, I will provide additional evidence for the observance of the constraint against first positions from the examination of novel data related to the affirmative marker. 4.4 Assertion markers and the affirmative sí Additional evidence for the existence of a dedicated position in the left edge may be drawn from the existence of assertion markers. These particles constitute a limited class of elements, which encompass very different meanings such as emphasis, affirmation (sí ‘yes’), negation (no ‘no’), and aspectual duration (ya ‘already’): (73) a. La rubia sí enseñó español el semestre pasado. the blonde AFF teach:3;PST Spanish the semester past “The blonde girl indeed taught Spanish last semester.” b. Los lingüistas no estudian literatura. the linguists NEG study:3PL literature “Linguists do not study literature.” c. Alfonsina ya terminó este poema. Alfonsina already finish:3PST this poem “Alfonsina already finished this poem.” Assertion markers can also license absolute verb initial orders. For instance, the paradigms in (73) also permit the finite verb to precede the lexical subject if an assertion marker is inserted: 176 (74) a. Sí enseñó la rubia español el semestre pasado. AFF teach:3;PST the blonde Spanish the semester past “The blonde girl indeed taught Spanish last semester.” b. %Enseñó la rubia español el semestre pasado. teach:3;PST the blonde Spanish the semester past “The blonde girl taught Spanish last semester.” c. No estudian los lingüistas literatura. NEG study:3PL the linguists literature “Linguists do not study literature.” d. %Estudian los lingüistas literatura. study:3PL the linguists literature “Linguists study literature.” e. Ya terminó Alfonsina este poema. already finish:3PST Alfonsina this poem “Alfonsina already finished this poem.” f. %Terminó Alfonsina este poema. finish:3PST Alfonsina this poem “Alfonsina finished this poem.” As pointed out in Section 4.4.1, absolute verb initial orders cannot be understood as declaratives. Thus, sentences (74)b, (74)d and (74)f are ungrammatical unless they are taken as yes/no questions. 177 The properties of assertion markers have not received enough attention and remain to be explored in detail. In this section, I will outline some of these properties, but I will mainly focus on the affirmative marker sí. As mentioned in Chapter 2, Section 2.3.3, the presence of assertion markers gives rise to an unambiguous wide scope interpretation of the preverbal subject, which I have interpreted as an indication of the dislocated nature of this constituent. Thus, by exploring the left edge properties of these assertion markers, it will be possible to understand the syntax of dislocation in the language, a task that will be the topic of Chapter 5. 4.4.1 Selectional properties Let us begin with the peculiar selection requirements displayed by assertion markers. Assertion markers must select for a finite clause regardless of whether or not they are associated with an assertive complementizer. This point will be illustrated with the affirmative marker in (75): (75) a. Sí que [te=quiero]. AFF COMP you=love:1 “I indeed love you.” b. Sí que [querer=te es fácil]. AFF COMP love:INF=you is easy “Indeed, loving you is easy.” 178 c. *Sí que [querer=te]. AFF COMP love:INF=you “Indeed, loving you.” Examples in (75)a and (75)b show that the assertion marker in association with the complementizer must select for a finite clause; otherwise, the outcome is ungrammatical, as in (75)c. In addition, the presence of an assertion marker may also determine the clausal type of a sentence. For instance the affirmative marker can appear in declaratives, contrastive interrogative clauses and reassertive yes/no questions, but not in imperatives, informational questions or non reassertive yes/no questions. This distribution indicates that their presence indeed plays a significant role to determine the clausal typing of a sentence. In fact, the requirement of a finite clause holds even with the bare sí. A similar remark applies for the case of the bare sí. As before, sí must be related to a finite verb or the sentence is ungrammatical: (76) a. Sí [te=quiero]. AFF you=love:1 “I indeed love you.” b. *Sí [querer=te] es fácil. AFF love:INF=you is easy “Indeed, loving you is easy.” 179 Given the existence of examples in which the complementizer que can select for a nonfinite clause, one can conclude that the requirement of a finite clause is due to the influence of the assertive particle. 4.4.2 Association with assertive complementizers As mentioned in passing in the previous section, an interesting property of assertion markers is their ability to be associated with assertive complementizers, a characteristic already observed in periphrastic adverbial constructions (see Section 4.4). Hence, along with the paradigm in (73), it is perfectly possible to find minimal pairs such as those in (77), (78) and (79). In these examples, an assertion marker is needed in pre-complementizer position in order to surpass a violation of a sequence with the complementizer in initial position, a positional effect (see (3)): (77) a. Sí que la rubia enseñó español el semestre pasado. AFF COMP the blonde teach:3;PST Spanish the semester past “Indeed, the blonde girl taught Spanish last semester.” b. *Que la rubia enseñó español el semestre pasado. COMP the blonde teach:3;PST Spanish the semester past “Indeed, the blonde girl taught Spanish last semester.” (78) a. ¿No que los lingüistas no estudian literatura?. NEG COMP the linguists NEG study:3PL literature “Isn’t it that linguists do not study literature?” 180 b. *¿Que los lingüistas no estudian literatura?. COMP the linguists NEG study:3PL literature “Isn’t it that linguists do not study literature?” (79) a. Ya que Alfonsina terminó este poema... already COMP Alfonsina finish:3PST this poem “Because Alfonsina finished this poem...” b. *Que Alfonsina terminó este poema... COMP Alfonsina finish:3PST this poem “Because Alfonsina finished this poem...” In spite of the parallel in the licensing of complementizers that might potentially end up in initial position, the constructions in (73) and their corresponding equivalents in (77), (78) and (79)are far from being synonymous, a difference not found with periphrastic adverbial constructions. For instance, the combination of the affirmative marker and the complementizer in (77)a is intended to strengthen the assertion expressed by the sentence la rubia enseñó español el semestre pasado. In contrast, the association of a negative marker no or the aspectual particle ya with a complementizer as in (78)a and (79)a gives rise to very different meanings. As suggested by the tag question in the English glose for no que (78)a, one can consider the negative marker in this environment to be close to the reinforcement of the assertion found with sí plus que, but with the difference that this parallel does not extend to its clausal type. In other words, whereas sí que in (77)a 181 denotes declarative force, no que in (78)a denotes interrogative force. 46 Finally, the association of ya with the complementizer in (79)a generates an interpretation in which the clause selected by ya que is understood as a consequence of an event denoted by another proposition. 4.4.3 Floating sí and Anchored sí Considering the previous properties of the affirmative marker, I would like to introduce additional data leading to an analysis of the affirmative marker as associated with a c*º [finite] head or a c*º [Q] head. Consider the cases where the affirmative marker is “anchored” to an assertive complementizer. Anchored sí seems to bear an emphatic value related to a presupposed context that the bare sí does not have. Consider the context in (80) and the two possible continuations in (81): (80) Con respecto a las leyes de California, tú... with respect to the laws of California, you “With respect to the laws of California, you...” (81) a. Anchored sí sí que las=conoces bien. AFF COMP acc=know:2 well “Indeed, you know them very well.” 46 This observation amounts to say that the negative marker associated with the position of the complementizer is composed of quantificational features (Q or focus) and features related to modality of the clause. I will leave the exploration of these properties for future research. 182 b. Bare sí sí las=conoces bien. AFF ACC=know:2 well “Indeed, you know them very well.” Only (81)a conveys the meaning in which the lexical subject is contrasted with a previous set of alternatives present in the discourse, whereas (81)b is the reassertion of a simple declaration. According to González (2000) and Dumitrescu (1973), the affirmative particle can have an emphatic or a contrastive interpretation, but not a neutral interpretation. In other words, sí cannot cooccur in out of the blue contexts (as an answer to a wide focus question like what happened?). Interestingly, bare sí can float far away from a verbal host. The example (82) shows that the bare sí in its contrastive interpretation can be separated from the finite verb by modal adverbs, which divide the frontier between the inflectional layer and the left edge: (82) Floating sí SAVANT SÍ probablemente resolvió el teorema de Fermat Savant AFF probably solve:PST the theorem of Fermat (no Sophie Germain). (NEG Sophie Germain) "Savant probably solved Fermat's theorem (and not Sophie Germain)." Interestingly, contrastive sí in (82) must stand in second position when a focused object is fronted to the clause initial position: 183 (83) Floating sí a. AL POLICÍA SÍ probablemente golpeó la activista (y no ACC:the police AFF probably hit:PST the activist and NEG al gerente). ACC:the manager “THE OFFICER, the activist probably hit (and not the manager)." b. *AL POLICÍA probablemente SÍ golpeó la activista (y no ACC:the police probably AFF hit:PST the activist and NEG al gerente). ACC:the manager) "THE OFFICER, the activist probably hit (and not the manager)." In these examples, the affirmative marker bears strong accent and narrow focus on the element to its left al policía "the officer." Taking into account these facts, I would like to propose that the anchored sí and the floating sí correspond to different structures and derivations. In particular, if one assumes the existence of a Polarity Phrase (PolP) below TP, the simultaneous occurrence of an affirmative marker in the anchored construction with a polarity marker (negation or affirmation for instance), suggests that anchored sí is merged in the lower left periphery for a reason different than ph-lexicalization. I would like to suggest that this reason is formally similar to the use of marked particles in Welsh and Breton: anchored sí licenses an assertive complementizer in the way covert propositional verbs would do. Therefore, (85) is the derivation of (84) in which, que 184 ph-lexicalizes the [finite] feature in c*º and sí merges with que leaving the Polº head available: (84) Sí que te=quiero. AFF COMP you=love:1 “I indeed love you.” (85) c*’ 2 c*º TP [finite] 2 que DP T’ 2 Tº PolP te=quiero 2 sí Polº VP 5 In contrast, the floating sí indeed seems to appear in the left edge to save an absolute verb initial sequence and to ph-lexicalize the [Q] feature of c*º. Recall the condition on ph-lexicalization of [Q] stated in (59) and repeated in (86): (86) [Q]-ph-lexicalization The [Q] feature in c*º ph-lexicalizes if the appropriate lexical material associated with c*º does not contain active (“unchecked”) φ-features. The floating sí has a verbal nature due to its selectional properties and qualifies for the condition on [Q]-ph-lexicalization. Due to its association with the polarity head, the features related to the floating sí are inactive. Given a sentence such as (87), I propose that the floating sí adjoins by Long Head Movement to ph-lexicalize the c*º head. This is depicted in the derivation in (88): 185 (87) Sí te=quiero. AFF you=love:1 “I indeed love you.” (88) c*’ 2 c*º Modal Adv [Q] 2 Modal Adv TP 2 DP T’ 2 Tº PolP te quiero 2 Polº VP sí 5 This analysis accounts for the ban on the cooccurrence of a marker with a polarity value in the floating sí construction. Despite the parallelism with Welsh and Breton clausal typing particles, it is important to recall that the constructions exemplified with the affirmative marker in Spanish have a declarative value but with a marked interpretation. In the next section, I will discuss some ideas related to this parallel. 4.4.4 The development of particles in the left edge All the previous properties may have an important diachronic correlate, more evident in the case of the affirmative marker. Recall that in Section 4.4.2, the periphrastic adverbial constructions required the presence of the copula ser. Given the selectional properties of assertion markers, the affirmative sí is likely to have verbal features if certain diachronic facts are taken into account. Unlike Modern Spanish, si in Old Spanish did not stand for an affirmative value. Instead, si had an 186 adverbial value with deictic force. Sí in the example (89) taken from Dumitrescu (1973) can be paraphrased with the expression as you say: (89) si fago, si quiero. ADV do:1 ADV want:1 "I do as you say, I want as you say." In other words, Old Spanish sí did not have the value of affirmation. One can speculate that Old Spanish si underwent a process of fronting to the left edge in a periphrastic structure with the verb ser “to be”, resulting in the emergence of the affirmative value. In Old Spanish, topicalization of si would have occurred in clauses headed by an assertive complementizer: 47 (90) Step 1: que [ ... [ ... es sīc ... Step 2: es sīc i que [ ... [ ... t i (es sīc fronted by topicalization) Step 3: sīc j es t j que [ ... [ ... (sīc is fronted) Step 4: sī (que) [ ... [ ... (ser reanalyzed as a COMP) Accordingly, sí corresponds to the grammaticalization of its predecessor sīc into a marker of finiteness, a process correlated in Old Spanish with the loss of the Long Head Movement construction. This hypothesized evolution may have a correlate in the development of the system of particles of Modern Welsh. Roberts (2005) following Willis (1998) reports that Long Head Movement was possible in Old Welsh during a period in which the 47 Interestingly, a parallel can be found in impersonal constructions with a modal value such as es probable que “it is probable that” and es posible que “it is possible that” and their corresponding adverbials probablemente que "probably" and posiblemente que "possibly." 187 root declarative particles fe/mi of Modern Welsh had not been developed yet. With the progressive declining of Long Head Movement in Welsh, the declarative particles developed from topicalized pronouns, 48 sometimes with an expletive value. In conclusion, Spanish makes use of assertive complementizers to mark finiteness in several environments, including those containing the affirmative marker and the assertive uses of ser. These facts provide strong support for the existence of c*º, a head with verbal properties that selects declarative clauses. Given that c*º needs to be associated with an element of the inflectional layer in order to be properly interpreted, this head can be conceived as a placeholder or dummy position, similar to the position held by a covert light verb or an auxiliary. 49 4.5 Conclusions Within the assumptions of the Narrow Left Edge system, the present chapter has shown that the Spanish lower left edge can be activated in declarative clauses. In particular, constructions involving clitics in a proclitic pattern and constructions involving an assertion marker can correspond to the licensing of c*º by Move. In addition, the construction dubbed “Anchored sí” may correspond to an innovation in which c*º is licensed by Merge. These findings complement the discussion initiated in Chapter 3 with the analysis of wh-interrogatives. In Chapter 3, I argued against the predominant view in the study of Spanish syntax, which considers that verb movement never reaches left 48 As suggested by Roberts (2005; fn. 12), this analysis can also be extended to the development of subject clitics in the Northern Italian Dialects. 49 For parallels in English see the analysis of that by Pesetsky and Torrego (2001), which takes the relation Cº- Tº as one analogous to clitic doubling. 188 peripheral nodes (Suñer 1994, Suñer and Lizardi 1995). I defended an approach in which verb movement does in fact target left edge nodes in clausal marked types such as interrogatives. If the phenomenon of second position can be understood as the result of verb movement to the head of the Assertion Phrase –c*º, then the analysis of interrogative clauses suggested that the Modern Spanish left edge is activated when clausal marked types need licensing. Both the analysis of declaratives in Chapter 4 and interrogatives in Chapter 3 lend support to Roberts typological generalization in (7) repeated as (91): (91) Roberts Generalization a. If a language has V2 in declarative clauses, then it has V2 in nondeclarative clauses. b. If a language has declarative particles, then it has nondeclarative particles. Within a typology of the positional effects and the Narrow Left Edge system, Spanish is different from classical V2 languages and V1 languages because Spanish requires ph-lexicalization of the lower left edge (the Assertion head c*º) by Move or by Merge in the “anchored sí” construction. In contrast, languages such as German need ph-lexicalization of the Cº head within the higher left edge and languages such as Welsh lexicalize the c*º head by Merge. In Chapter 5, I will elaborate on the syntax of the dislocation pattern of subjects and on the type of construal strategy involved in dislocated subjects. 189 Chapter 5 Agreement in the Left Periphery and Dislocated Subjects Lexical preverbal subjects in Spanish have been intensively studied since the pioneering work on inversion by Torrego (1984). In Chapter 2, several pieces of evidence were provided to show that although some researchers have analyzed preverbal subjects in Spanish as base generated constituents within the left periphery (Ordóñez 1997, Ordóñez and Treviño 1999), preverbal lexical subjects do not necessarily behave in this way. Consistent with Suñer (2003) and Zubizarreta (1994)’s insights, I proposed that preverbal lexical subjects display at least two patterns: a dislocated pattern and a canonical pattern. If the analysis of these facts is correct, one question that remains to be answered is how preverbal subjects in a dislocated or a canonical pattern are derived in the same grammar. The purpose of this chapter is to provide a plausible answer to this intriguing question departing from the Narrow Left Edge system (see Chapter 3) and the idea that the head of the Assertion Phrase c*º establishes an Agree relation with the inflectional system (see Chapter 5). I claim that the establishment of an Agree relation between the c*º head and the inflectional system also derives the syntax of dislocation and subjects in the language. Given that it is evident that lexical subjects can be dislocated in the preverbal field and that they can be analyzed in terms of base generation, one question that arises is whether the dislocated subjects detected by the presence of an affirmative marker should receive a similar treatment and be analyzed as base generated constituents in the complementizer system. 190 In what follows, I will claim that preverbal lexical subjects can occupy several positions within the left edge and that one must draw an additional distinction within the dislocation pattern of lexical subjects: preverbal lexical subjects can be analyzed in terms of an adjunction structure or in terms of a movement structure. The adjunction structure for a dislocated constituent has been the favorite analysis for preverbal lexical subjects. Let us call these dislocation structures generated by adjunction higher dislocated subjects (H-subjects). Within the Narrow Left Edge system, H-subjects are adjoined to the maximal projection of the Assertion Phrase, c*P, and are resumed by a null pronominal contained within the inflectional system. This structure occurs when a constituent different than the subject, say XP, occupies the specifier of the c*P: (1) Higher dislocated subjects c*P 2 DP i c*P 2 XP c*’ 2 c*º TP 5 pro i Although the adjunction structure analysis has been the most common alternative for preverbal lexical subjects, I want to claim that a movement structure is also possible. I will refer to the lexical subjects generated by this option as lower dislocated subjects (L-subjects). In the Narrow Left Edge system, L-subjects are generated by movement of the lexical subject to the specifier of the Assertion Phrase: 191 (2) Lower dislocated subjects c*P 2 DP i c*’ 2 c*º TP 5 t i The discussion of empirical paradigms pertaining to constituent ellipsis in Section 5.1, idiomatic expressions in Section 5.2 and restructuring will provide additional evidence to support the idea that L-subjects are fundamentally different from H- subjects. I will claim that L-subjects move to the left edge in order to check Nominative Case and that this movement is possible if feature movement involving the subject agreement occurs to the c*º head. 5.1 Agreement and the COMP-INFL articulation The data and the analysis to be discussed come from constructions involving constituent ellipsis, which have not received enough attention to the best of the author’s knowledge, maybe with the exception of Contreras (2004). We will observe that identity of tense inflection is required under ellipsis, but this is not the case for the specification of subject agreement in the antecedent and the remnant, which may differ: (3) a. [ ANTECEDENT ...αTense] ... [ REMNANT ...βTense] where α and β must be identical in feature specification. b. [ ANTECEDENT ...αAgrS] ... [ REMNANT ...βAgrS] where α and β do not have to be identical in feature specification. 192 These facts indicate that the subject agreement morphology in Spanish is represented syntactically in Tº or a functional projection within the inflectional layer. In contrast, the tense inflection has to be represented lexically on the verbal entry Vº. Given that the lexical subject in the remnant agrees with the unpronounced subject agreement morphology, I will argue that in the structures under discussion, the subject agreement features are displaced to the left edge to create a specifier/head configuration. In this specifier/head configuration, the dislocated subject is in the specifier of the Assertion Phrase (c*P) and the subject agreement features land in the head of c*º. Given the ph-lexicalization requirement on the head c*º must be realized by a polarity or an assertion marker. 5.1.1 Ellipsis and Verbal Morphology in English I will construct the argument for a different location of the tense and subject agreement inflection based on Lasnik’s observations on verbal morphology (Lasnik 1999a). He develops an argument for a hybrid theory of verbal morphology in English where the computational system may represent syntactic units lexically or morphologically. Accordingly, the lexical representation of a syntactic unit takes the verb and its inflection as part of a single lexical item. In this option, syntactic derivations operate with lexical pieces containing the verbal inflection as these lexical pieces are arranged in a numeration, at the very beginning of the derivation. In contrast, the computational system may represent some syntactic units as vocabulary items and through the existence of a morphological component. Thus, the verbal inflection would be represented as an affix and would be inserted by the 193 morphological component as an independent unit. Additional syntactic computations may associate the verbal inflection with the lexical verb in the course of the derivation. A crucial piece of evidence in support of the hybrid approach of the verbal morphology comes from VP-ellipsis. Consider the example (4) from Lasnik (1999a: 108, ex. 39-40) and its two representations in (5): (4) John slept and Mary will too. (5) a. VP ellipsis affecting Vº inflected for past tense John slept and Mary will [slept] too. b. VP ellipsis affecting a bare Vº John PST sleep and Mary will [sleep] too. Example (4) must be represented as in (5)b and not as in (5)a. If ellipsis occurs under identity of the relevant forms in the antecedent and the remnant, the bare form of the verb must undergo deletion in the remnant. As a result, tense can be left stranded by the bare form of a verb in the remnant under ellipsis. As Lasnik explains, (4) suggests that there is no requirement on the identity of tense inflection. In a hybrid approach to morphology this is an indication that the auxiliary and the verb do not form a unit in the lexicon. Otherwise, (5)a should be possible and there would be a clash at LF between the past interpretation to be assigned to the verb and the future interpretation intended by the auxiliary. Now observe what happens in the case of auxiliaries. As reported by Lasnik (1999a: 109-110, ex. 47-48, 56-57), the deletion of be and have in the remnant 194 requires the identity of tense with their antecedent. The ungrammaticality of example (6) illustrates this observation: (6) a. John was here and Mary was too. b. *John was here and Mary will too. Following Lasnik’s reasoning, in order to generate the ungrammaticality of ellipsis involving auxiliaries, the auxiliaries in (6) must be represented as (8)a and not as (8)b. If (8)b were possible, (6) should be parallel the grammaticality of (4), contrary to the facts: (7) a. VP ellipsis affecting AUX inflected for past tense John was here and Mary will [was] here too. b. VP ellipsis affecting a bare AUX John PST be here and Mary will be here. Lasnik founds a similar pattern with the auxiliaries should and have: (8) a. ?John should have left, but Mary shouldn’t. b. *John has left, but Mary shouldn’t. Summarizing, the contrasts between main and auxiliary verbs in the context of ellipsis can be explained if main verbs are bare forms with respect to Inflº in English. According to Lasnik, this means that the grammatical content of Inflº - agreement and tense- is inserted as an affix, independently of the lexical item. By contrast, auxiliaries are fully inflected from the lexicon and presumably, that is why they require identity of the original featural content to be preserved when deletion applies. 195 A related issue is how Inflº ends up as part of the bare verb and especially how this is possible if main verbs do not leave their base positions. For Lasnik, the solution is inherent to the idea that some syntactic units may be affixes and that “bare” verbal forms become associated to their Inflº morphology in a later step in the derivation. Thus, the filter in (9) rules out derivations where the affix has failed to incorporate into the bare form: 50 (9) The “Stranded Affix” Filter A morphologically realized affix must be syntactic dependent of a morphologically realized category, at surface structure (Lasnik 1999a: 98). Although I will not adopt Lasnik’s Stranded Affix Filter in the analysis of Spanish, 51 I will assume his conception of the relation between morphology and syntax. Accordingly, the computational system can manipulate vocabulary items or lexical entries and operate with these units at different stages in the derivation. Syntactic units can be inserted early in the derivation as part of a lexical entry or they can be inserted late and receive their interpretation in the mapping to PF. The choice of how 50 This recalls earlier treatments of the data, in particular, the so-called ‘Affix Hopping’ of the 50s. However, for Lasnik the scope of this operation is a purely morphophonological process occurring at PF in accordance with current conceptions on the syntax-morphology relation. On the contrary, “Affix Hopping” for Chomsky (1957) was part of the syntactic machinery of transformations. The merit of Lasnik’s idea resides precisely in combining an earlier proposal with the idea that lexical items come fully inflected from the lexicon. Both possibilities are thus complementary. 51 The morphophonological nature of the Stranded Affix Filter will suggest an approach based on a condition on ph-lexicalization. See Section 5.1.4 for details. 196 to represent syntactic categories as lexical items or vocabulary items is, however, determined by language specific factors. 5.1.2 Tense is lexically represented Observationally, ellipsis in Spanish is total, namely it cannot leave the verb, the auxiliary or the tense morphology stranded. Under ellipsis, the remnant must be licensed by polarity particles such as si ‘yes’ or no ‘no’ ((10)a-(10)b below) or también ‘too’ and tampoco ‘neither’((10)c-(10)d below): 52 (10) a. Cristina no bailó y su hermano sí bailó. Cristina NEG dance:3PST and her brother AFF dance:3PST “Cristina did not dance and her brother did.” b. Cristina bailó y su hermano no bailó. Cristina dance:3PST and her brother NEG dance:3PST “Cristina danced and her brother did not / too.” c. Cristina bailó y su hermano también bailó. Cristina dance:3PST and her brother too dance:3PST “Cristina danced and her brother too.” d. Cristina no bailó y su hermano tampoco bailó. Cristina NEG dance:3PST and her brother neither dance:3PST “Cristina did not dance and her brother did not either.” 52 A topic that deserves further investigation is the interaction of modal verbs and ellipsis in Spanish. Although auxiliaries and main verbs cannot survive to receive a phonological interpretation in the remnant, modals can be left stranded. Crucially, as reported by López (1999), modals can only receive the deontic reading in this environment and not the epistemic reading. 197 Since there is no do-support in Spanish, if one of these strategies fails as in (11), the sentence is overtly ungrammatical: (11) *Cristina bailó y su hermano bailó. Cristina dance:3PST and her brother dance:3PST “Cristina danced and her brother did not / too” Note that the same can be said with respect to the auxiliaries ser, estar ‘to be’ and haber ‘to have’. This point is illustrated with haber: (12) a. Cristina había bailado y su hermano no había bailado. Cristina have:3IMPF danced and her brother NEG have:3IMPF danced “Cristina had danced and her brother had not.” b. *Cristina había bailado y su hermano había bailado. Cristina have:3IMPF danced and her brother have:3IMPF danced “Cristina had danced and her brother will too.” For instance, in (12)a it is not possible to delete the verb and leave the remnant without a licensor and the same goes for the complex tense form in (12)b. One way to understand this restriction on the possible remnants in Spanish is to say that ellipsis in this language targets a larger constituent than in English. Whereas ellipsis affects VPs in the English examples discussed in Section 5.1.1, this operation affects TPs or IPs in Spanish declaratives. For expository reasons, let us call the Spanish ellipsis style “constituent ellipsis.” A more elaborated argument to understand why constituent ellipsis affects the tense morphology in Spanish can be constructed as follows. Verbal related lexical entries such as main verbs and auxiliaries are constructed with the tense 198 specification in the antecedent and remnant clauses. Assuming that deletion operates under parallel structures and under identity, ellipsis is expected to be possible if and only if the lexical entries related to verbs and auxiliaries are identical in the antecedent and the remnant clause. This is expressed schematically in (13): (13) [ ANTECEDENT ...αTense] ... [ REMNANT ...βTense] where α and β must be identical in feature specification. In turn, the identity of tense specification under ellipsis is only possible under the hypothesis that the Inflº containing the tense inflection is already attached to lexical verbs and auxiliaries from the beginning of the derivation. Thus, following Lasnik’s logic, I postulate the hypothesis in (14): (14) Tense Specification Hypothesis Verbal roots in Spanish are fully inflected for tense morphology from the lexical numeration. The examination of the empirical data confirms the previous logic, giving support to the Tense Specification Hypothesis in (14). Both the auxiliary ser ‘to be’ in (15) and the main verbs in (16) require identity of tense specification in order to undergo ellipsis. For instance, the interpretation of the remnant su hermano no in (15)a is only possible if the structure recovered at LF contains the verb ser “to be” inflected for the present tense as in the antecedent. If a temporal adverb indicating futurity is added, the sentence becomes ungrammatical: 199 (15) Constituent ellipsis and the auxiliary ser a. Cristina es artista y su hermano no es artista. Cristina be:3PRS artist and her brother NEG be:3PRS artist “Cristina is an artist and her brother is not.” b. *Cristina es artista y el próximo año su hermano Cristina is artist and the next year her brother también será artista. too be:3FUT artist “Cristina is an artist and next year her brother will be too.” A similar effect can be observed with main verbs. The remnant su hermano sí “her brother does” in (16)a must be interpreted as inflected for the present tense, the same tense specification of the antecedent. Otherwise, a different specification for tense gives rise to ungrammaticality as suggested by (16)b: (16) Constituent ellipsis and main verbs a. Cristina no escribe libros de sintaxis y su hermano Cristina NEG write:3PRS books of syntax and her brother AFF sí escribe libros de sintaxis. AFF write:3PRS books of syntax “Cristina does not write syntax books, but his brother does” 200 b. *Cristina no escribe libros de sintaxis y el próximo año su Cristina NEG write:3PRS books of syntax and the next year her hermano sí escribirá libros de sintaxis. brother AFF write:FUT books of syntax “Cristina does not write syntax books, but next year her brother will” Following the logic of Lasnik’s hybrid model of the relation between verbal morphology and syntax, these paradigms indicate that main verbs and auxiliaries in Spanish enter the derivation fully inflected for tense as stated in the Tense Specification Hypothesis. This premise explains the requirement on identity of tense inflection between the antecedent and remnant clauses in ellipsis constructions. It also gives an explanation for the lack of do-support in the language. Do-support does not exist because no syntactically independent bare verbal forms exist. 53 5.1.3 The (Mis)matches in Subject Agreement In the previous section, I postulated the Tense Specification Hypothesis, which states that lexical verbs and their tense specification form a lexical unit in Spanish. For expository reasons, I will call the head containing the lexical verbal form and its tense inflection “verbal/tense complex.” In what follows, I will discuss whether or not AGR is present in an abstract way in the verb (namely, represented as a lexical item) or AGR forms an independent unit in the numeration (in other words, represented as a vocabulary item). In order to do this, I will expand on the paradigms 53 Verbal related entries (auxiliaries or main verbs) not specified for tense should be spell-out as infinitives. 201 involving ellipsis to check whether subject agreement requires identity in the remnant with respect to the antecedent. Ellipsis data equivalent to what was already introduced in the previous section allow us to find another surprising contrast. The examples in (17) show that under ellipsis, identity of φ-features is not required: (17) a. Cristina no es doctora y sus hermanos Cristina:3F.SG NEG be:3SG doctor;F.SG and her:PL brother:M;PL sí son doctores. AFF be:3PL doctor;PL “Cristina is not a doctor but her brothers are indeed doctors.” b. Cristina sí escribe libros de sintaxis y sus Cristina:3F.SG AFF write:3PRS books of syntax and her:PL hermanos no escriben libros de sintaxis. brother:3M;PL NEG write:3PRS books of syntax “Cristina indeed writes syntax books, but her brothers do not.” As long as there is identity of tense specification, Spanish shows that the remnant subject does not need to match in number or gender with their antecedent. Similarly, the person feature does not need to be identical in the antecedent and the remnant: (18) a. Cristina no es doctora y yo sí soy doctor. Cristina:3F.SG NEG be:3SG doctor;F.SG and I AFF be:1SG doctor;PL “Cristina is not a doctor but indeed I am a doctor.” 202 b. Cristina sí escribe libros de sintaxis y tú no Cristina:3F.SG AFF write:3PRS books of syntax and you NEG escribes libros de sintaxis. write:2PRS books of syntax What does this difference in the ellipsis construction indicate? I believe that this contrast indicates that Spanish φ-features related to subject agreement are located in a projection different from the one that includes the verbal/tense complex. I would like to hypothesize that as same as Inflº is actually an affix with respect to main verbs in English, subject agreement in Spanish is also an affix. Main verbs and auxiliaries are in this sense ‘bare’ with respect to subject agreement in Spanish. 5.1.4 Proxies and Subject Agreement Movement In the previous section, I have claimed that Spanish Inflº differ from the English one with respect to where agreement and tense inflection is inserted. On the one hand, either auxiliaries or main verbs in Spanish, irrespectively of their status, require to be inflected for tense from the very beginning of the derivation. On the other hand, subject agreement φ-features seem to be inserted in a projection different than the one formed by the verbal/tense complex: (19) Spanish a. Verbal roots enter the derivation inflected for tense. b. Verbal roots enter the derivation uninflected for subject agreement morphology. 203 In other words, Spanish and English operate with different units in the numeration. The following diagrams represent the different association with respect to tense, verbal items and agreement in the two languages: (20) a. Spanish 1 2 [AGR] [T]====[V] b. English 2 1 [AGR]===[T] [V] The two clusters in (20) represent how Spanish associates [AGR] in a distinct unit from the complex [T/V]. In this grammar, [AGR] is instantiated as an affix with respect to [T/V]. On the contrary, in English, subject agreement is inactive or parasitic with respect to the projection containing tense specification. This dependency is expressed by the formation of a complex unit [AGR/T]. Giorgi and Pianesi (1997) corroborate this observation when they note that tense inflection must be specified as present if the third person singular morphology is present, but not vice versa. Crucially, López (1999) has shown that the remnant in Spanish constituent ellipsis constructions displays several properties attributed to left dislocated constituents. It is uncontroversial that lexical preverbal subjects can appear dislocated in the left periphery and some of these subjects can be analyzed as base generated constituents. For example, consider the cases of the so called “agreement ad sensum”, which allows agreement between the plural inflection of the verb and a 204 morphological singular nominal with a collective interpretation. It is not surprising to find that a context that favors agreement ad sensum is across syntactic islands: (21) Me=preguntaron que el jurado quién creía que ask:3PL;PST COMP the jury who believe:3IPFV COMP estaban presionados. be:3PL;IPFV pressured “As for the jury, they asked me who used to believe that they felt pressured.” (intended) In (21), the dislocated subject el jurado “the jury” must be base generated in their surface position. A base generation analysis of the remnant in ellipsis constructions also fails to account for the parallelism between canonical subjects and dislocated subjects involving assertion markers. Only canonical subjects (i.e., (22)a) and dislocated subjects diagnosed by an assertion marker (i.e., (22)b or (22)c) disallow agreement ad sensum. In contrast, lexical subjects that are clearly base generated in the periphery (i.e., (22)d) allow agreement ad sensum (see Brucart 1997): (22) a. *La escuadra atacaron al enemigo. the squad:3SG attack:3PL;PST to:the enemy “The squad attacked the enemy.” 205 b. *La escuadra sí atacaron a los enemigos. the squad:3SG AFF attack:3PL;PST to:the enemy “The squad did attack the enemy.” c. *La escuadra sí que atacaron a los enemigos. the squad:3SG AFF COMP attack:3PL;PST to:the enemy “The squad did attack the enemy.” d. La escuadra, el general explicó que atacaron the squad:3SG the general explain:3;PST COMP attack:3PL;PST al enemigo. to:the enemy “The squad, the general explained that they attacked the enemy” These reasons lead us to reject a base generation analysis of the remnant resumed by a null resumptive pro, as suggested by López. 54 I would like to propose that the asymmetry between tense and subject specification detected in ellipsis constructions can be explained by assuming that the subject agreement features are base generated in the Tº head (see Section 5.1.3), but they need to be displaced to a different projection. This displacement of features is motivated by the Single Licensing Condition (Nash and Rouveret 2002): (23) Single Licensing Condition (SLC) 54 It is hard to explain the asymmetry in obligatory identity of tense versus optional identity of subject agreement displayed by the Spanish ellipsis construction in a base generation analysis. Moreover, an analysis in terms of base generation and resumption by a null pro would entail that strong personal pronouns would be ‘duplicated’ by this null pro when they appear as the remnant, a problematic assumption from a conceptual point of view. 206 A functional category can enter into a licensing relation with the feature content of only one terminal node in its checking domain. As suggested by the definition in (23), the Single Licensing Condition ensures that a head can only support a single licensing operation. Therefore, if a licensing relation L applies to a functional head F, further licensing operations will be blocked to apply to F. Suppose that the head F has f 1 and f 2 as features that require licensing to comply with the Principle of Full Interpretation. If the licensing relation L applied to f 1 , L’ cannot applied to f 2 by the operation of the Single Licensing Condition. In Nash and Rouveret’s theory, the grammar resorts to the projection of the offending feature, in our example f 2 , into a dedicated projection that can be present or absent in the numeration. When these dedicated projections are not included in the numeration, they are called “proxies”. A proxy is a copy of a functional projection without any semantic content, created to satisfy a formal property, not legible at the PF and LF components. The Single Licensing Condition allows us to explain why Spanish displays a canonical pattern in the inflectional system and a dislocation pattern in the left edge. Consider the derivation of the sentence (24)a. The structure in (24)b represents the stage where V-to-T has applied: (24) a. Mimi memorizó las leyes de California. Mimi:3F.SG memorize:3PST the laws of California “Yesterday, Mimi memorized the laws in California.” 207 b. T’ 2 Tº VP memorizó 5 Mimi t i las leyes California After the lexical verb has been inserted into the derivation with its tense specification, V-to-T raising has to occur. Given that tense is an operator and needs to take scope over the predicated event, I will assume with Contreras (2004) and Koeneman (2000) that the inflected verb raises to the Tº head to take scope. Note that Tº is the locus where subject agreement features are inserted. Subject agreement features are uninterpretable but they cannot be licensed in the local specifier of Tº because of the Single Licensing Condition. The grammar then displaces the subject agreement features into a proxy of TP or into c*º. In the first situation, the proxy of TP is a functional head without semantic content. In the second case, c*º is a functional head with full content. I claim that the generation of a proxy of TP can be correlated with the canonical pattern of preverbal lexical subjects whereas the displacement of the subject agreement features to c*º correlates with the dislocation pattern of preverbal subjects. Both the Canonical and Dislocation Pattern are depicted in (25): 208 (25) a. Canonical Pattern Projection of the subject agreement features into a proxy of TP ProxyTP 2 SPEC ProxyT’ 2 ProxyTº TP [AGR] 2 Tº VP memorizó5 Mimi t i las leyes California b. Dislocation Pattern Projection of the subject agreement features into c*º c*P 2 SPEC c*’ 2 c*º TP [AGR] 2 Tº VP memorizó5 Mimi t i las leyes California Consider the derivation of lexical preverbal subjects in the canonical pattern in sentences such as (26): (26) Mimi ayer memorizó las leyes de California. Mimi:3F.SG yesterday memorize:3PST the laws of California “Yesterday, Mimi memorized the laws in California.” As suggested in Section 5.1.2, inflected verbs and auxiliaries enter into the derivation inflected for tense. The subject agreement features are merged into the Tº head, which merges with the VP memorizó las leyes de California “memorized the laws of California.” At this point, the derivation can follow by inserting the temporal adverb 209 in the specifier of TP without violating the Single Licensing Condition because adverb insertion does not involve any feature checking relation: (27) TP 2 AYER T’ 2 Tº VP memorizó 5 Mimi t i las leyes de California However, the uninterpretable φ-features related to the subject and the Case features of the subject remain active and unchecked. An Agree relation between Tº and the subject in the [Spec, VP] cannot be established because it would violate the Single Licensing Condition. The lexical subject could not have raised to the inner specifier of TP either because of the Single Licensing Condition. Therefore, the uninterpretable φ-features of Tº move into a proxy of TP. Next, the lexical subject moves to the specifier of the Proxy TP. Through the projection of a proxy TP, the φ- features of the proxy TP and the Nominative Case of the lexical subject are licensed. Now consider the dislocation pattern of preverbal subjects diagnosed by the presence of an assertion marker. The derivation of sentences such as (28) is different from the derivation of the anchored sí and the floating sí constructions discussed in Chapter 4. In particular, (28) corresponds to cases where the material preceding the affirmative marker does not receive a strong contrastive intonation: 210 (28) Mimi sí entiende las leyes de California y sus Mimi:3F.SG AFF understand:3PRS the laws of California and her:PL amigos no entienden las leyes de California. friends:3M;PL NEG understand:3PRS the laws of California “Mimi actually understands the laws in California and her friends do not understand the laws in California.” Consider the derivation of the sentence (28) given in (29) after verb movement to the inflectional domain and before the agreement subject moves to host the lexical subject in the canonical pattern: (29) TP 2 SPEC T*’ 2 Tº PolP memorizó 2 Polº VP sí 5 Mimi t i las leyes de California The ellipsis data suggest that the finite verb does not move along with the affirmative marker. In other words, the affirmative marker does not incorporate into the verb in Tº because finite verbs or auxiliaries are not part of the remnant in constituent ellipsis. Therefore, I am forced to assume that the affirmative marker moves from the head of the Polarity Phrase to the head c*º, the head of the Assertion Phrase: 211 (30) c*P 2 SPEC c*’ 2 c*º TP [finite] 2 sí SPEC T*’ 2 [AGR] Tº PolP memorizó 2 Polº VP t j 5 Mimi t i las leyes de California Given that the lexical subject requires Nominative Case and the Tº head contains unchecked uninterpretable features, the derivation cannot proceed with the creation of a proxy as in the canonical pattern. This is because a proxy is typically contentless and the affirmative marker would not be interpreted properly in situ or attaching to the proxy of TP, an effect of the Principle of Full Interpretation. Therefore the most economical derivation is the one which fissions the subject agreement features with a content head present in the numeration: the c*º [finite] head. The subject agreement movement allows the lexical subject to surpass the Case position in the specifier of TP, which would have created too long a chain. Does the derivation of examples such as (28) end here? No, the subject agreement present in the c*º head still requires lexical support. Recall that the licensing conditions of the c* [finite] stated in Chapter 3 required both φ-feature visibility and lexical support: 212 (31) [finite]-ph-lexicalization The [finite] feature in c*º ph-lexicalizes if the appropriate licensing material contains visible φ-features, unchecked at the point of the spell out. Subject agreement movement cannot pied pipe the phonological matrix of the verbal head as verbal heads do not appear in the left edge in the absence of clitic morphology. Movement of the affirmative marker sí from its base position in Polº is required to provide ph-lexicalization of c*º [finite]. Once the assertion head has been licensed, the lexical subject can raise to the specifier of c*P to check the φ-features that underwent fission and get Nominative Case assigned: (32) c*P 2 Mimi K c*’ 2 c*º TP [finite] 2 sí_[AGR] SPEC T*’ 2 Tº PolP memorizó 2 Polº VP t j 5 t k t i las leyes de California In the next two sections, I will provide additional support for this analysis. In Section 5.2, I will provide empirical evidence from the examination of idiomatic expressions to argue for the idea that L-subjects are derived by movement. In Section 5.3, I will discuss the idea that Nominative Case is assigned to L-subjects in their left edge position and the consequences of this outcome. 213 5.2 L-subjects and Movement to the Lower Left Edge In this section, I will draw evidence for the availability of movement in dislocated subjects from the examination of idiomatic expressions with a sentential value. Based on the distinction between sentential idioms including functional information such as tense and modality (“TP-idioms”) and those that do not include such functional information (“vP-idioms”), I argue that [Spec, vP] is the extraction point of dislocated subjects generated by Move. Idiom chunks constitute a useful test to probe into the structure and operations available to a certain grammar. In particular, I will consider a class of idioms called phrasal idioms in Nunberg, Sag and Wasow (1994)'s terminology. A phrasal idiom is a non-compositional expression, namely, one that does not obtain its meaning from the semantic calculation of the meaning of its components. Phrasal idioms contrast with those idioms that have been argued to be somewhat compositional or idiomatically combining expressions in Nunberg, Sag and Wasow's classification (i.e., to take advantage of). The crucial property associated with phrasal idioms is one that I call Idiomatic Constituenthood Condition: (33) Idiomatic Constituenthood Condition (ICC) A phrasal idiom must correspond to a syntactic constituent. According to the Idiomatic Constituenthood Condition, phrasal idioms overlap only with syntactic blocks or constituents allowed by the grammar. The Idiomatic Constituenthood Condition ensures that a verbal predicate and its complement may 214 constitute an idiom chunk and predicts that subject and verb cannot compose an idiom chunk in exclusion of the object, an old observation (see the discussion in Marantz 1984: 27-31). In addition, (33) also suggests that compositional idioms do not need to correspond to syntactic constituents. Based on the observation that dislocation of subjects and non subjects (direct and indirect objects) can be analyzed as adjunction of the relevant constituent to the lower left edge, I will simply assume the structure in (34), for what I dub Higher Dislocated Subjects (H-subjects): (34) Higher dislocated subjects c*P 2 DP i c*P 2 XP c*’ 2 c*º TP 5 pro i Although I will not question the existence of (34), I want to claim that some idiomatic constructions provide evidence for the generation of dislocated subjects by movement. Thus, in addition to (34), I claim that idiomatic expressions, and more precisely vP-idioms provide support for a structure like (35), where the DP subject has been dislocated by movement: 215 (35) Lower dislocated subjects c*P 2 DP i c*’ 2 c*º TP 5 t i Thus, lower dislocated subjects (L-subjects) are generated by movement of the lexical subject to the specifier of c*P. 5.2.1 TP and vP-idioms I will start the discussion of dislocation and preverbal lexical subjects by proposing a classification of the different types of sentential idioms. Sentential idioms are phrasal idioms including a subject and a verbal predicate. Considering (33), sentential idioms can be divided into two broad categories: TP-idioms and vP- idioms: (36) Classification of idiom chunks Type of idiom Defining property vP-idioms Not sensitive to functional information (tense and/or modality) TP-idioms Sensitive to functional information (tense and/or modality) The logic of the division proposed in (36) relies on the ability of idioms to include grammatical information. If an idiom is not affected by tense or modality, it must be the case that the idiom does not contain such information. I will assume that this is what happens in vP-idioms -frozen expressions including a verbal phrase and its subject specifier. Conversely, one can interpret the sensitivity of some idioms to 216 functional elements as a manifestation of the idiomaticity of the grammatical features involved. Let us illustrate the proposed typology with the following paradigm: (37) vP-idioms a. Hierba mala nunca muere herb bad never dies “Bad seed never dies” IDIOMATIC "Weed never dies” LITERAL b. Hierba mala nunca morirá herb bad never dies:FUT “Bad seed will never die” IDIOMATIC "Weed will never die” LITERAL c. Hierba mala nunca debe morir herb bad never must die “Bad seed must never die” IDIOMATIC "Weed never dies” LITERAL Example (37)a illustrates the conventional form of the idiom. As observed in (37)b and (37)c, the idiomatic interpretation remains constant even if one uses future tense or the modal deber "must". This constitutes strong evidence that grammatical information is merged with the idiomatic expression contained in vP. Another notable example is in (38) below: 217 (38) El buey suelto bien se=lame solo the ox loose well REFL=licks alone “A sound man needs no physician” IDIOMATIC “The ox that is loose licks himself well" LITERAL As in (37)a, this vP-idiom contains a low adverb (in this case bien "well"), which cannot be replaced by an adverbial of a different or of a similar class (for instance rápidamente "quickly") without losing its idiomatic reading. Moreover, the expression cannot keep its idiomatic reading either if the verb precedes the adverbial (by vº-to-Tº movement) as in regular non-idiomatic cases. By contrast, TP-idioms display frozen information with respect to tense and modality. Accordingly, their meaning is affected if this information is altered: (39) TP-idioms a. Otro gallo nos=cantará another rooster us=sing:FUT “That's a horse of a different color” IDIOMATIC “Another rooster will sing for us" LITERAL b. Otro gallo nos=cantaba another rooster us=sing:PST “That's a horse of a different color” *IDIOMATIC “Another rooster used to sing for us" LITERAL 218 c. Otro gallo nos=debe cantar another rooster us=must sing:INF “That's a horse of a different color” *IDIOMATIC “Another rooster must sing for us" LITERAL Sentence (39)a loses its idiomatic interpretation when present tense is replaced by the imperfect form of the preterit -(39)b. Something similar occurs in (39)c. If one adds a modal to the original verbal form of the idiom, the idiomatic reading is absent. 5.2.2 Clitic Left Dislocation and Idioms Sentential idioms are especially useful to investigate the syntax of subjects in interaction with dislocation. Typically, an idiomatic construal does not support clitic left dislocation (CLLD) of objects (cfr. Vat 1981/1997, Zaenen 1997). However, the examination of additional data and the distinction between TP-idioms and vP-idioms suggests that the relationship between dislocation and idiomaticity responds to the following generalization: (40) In a language allowing CLLD, if a frozen expression includes tense as part of its idiomatic meaning (i.e., if it is a TP-idiom), this expression does support an idiomatic CLLD. The statement in (40) predicts an interesting asymmetry between idioms involving direct objects and idioms involving subjects/indirect objects. As argued in Chapter 5, dislocated direct objects are correlated with agreement with Tº and dislocated indirect objects are instead linked to a lower projection, presumably vP. In addition, the generalization in (40) explains the contrast between the paradigms (41) and (42): 219 (41) Camarón que se=duerme se=lo=lleva la corriente. shrimp that REFL=falls_asleep REFL=it=carries_away the current “You snooze, you lose.” (42) a. Cristina [ VP metió la pata] Cristina stuck the foot “Cristina stuck his foot in (location)" LITERAL “Cristina stuck his foot in his mouth" IDIOMATIC b. Cristina [ VP tiró la toalla] “Cristina threw the towel” LITERAL “Cristina threw in the towel” IDIOMATIC Example (41) is an instance of a CLLD object camarón “shrimp”, which has an idiomatic interpretation. Although it is true that idioms resist clitic left dislocation, this observation does not entail the impossibility of the existence of a CLLD idiom. In particular, idioms such as (41) display properties related to TP-idioms. For instance, they do not need to be specific (the relative clause can be in the subjunctive que se duerma); their associated tense information is frozen and they do not have a non dislocated idiomatic counterpart. These examples permit a literal and an idiomatic interpretation. However, if the object in the above examples is clitic left dislocated, the expression loses its idiomatic reading and can only be interpreted literally: 220 (43) CLLD and idioms a. La pata la=metió Cristina (LITERAL, * IDIOMATIC) the foot it=stuck Cristina "Her foot, Cristina stuck in her mouth." b. La toalla la=tiró Cristina (LITERAL, * IDIOMATIC) the towel it=threw Cristina "The towel, Cristina threw it" In simple words, certain types of idioms in Spanish resist dislocation by CLLD. If CLLD is analyzed in terms of base generation, one can easily explain the above contrasts in terms of θ-role assignment. Given that idioms are not associated with any specific thematic role, one can use Rizzi's terminology, and claim that they are not referential. If this is so, they are expected not to appear in contexts in which a referential θ-role is assigned or there is no way for the idiom components to form a unit at some point of the derivation. Consequently, one would expect that if the Dislocation Analysis is correct (i.e. if Spanish is a language that resorts to dislocation/base-generation for preverbal lexical subjects), no idioms including the subject should be possible. However, as I will demonstrate below, this is not supported by the empirical data and represents strong evidence against the Dislocation Analysis. 5.2.3 Evidence for movement Interestingly, TP- and vP-idioms are correlated with subject dislocation in not so obvious ways. From the examination of the data, and given the distinction 221 between H-subjects and L-subjects, I conclude that vP-idioms provide evidence for the extraction site of L-subjects and for the type of relation: (44) L-subjects are extracted from the specifier of vP and displaced by movement to their surface position. We will observe that in contexts where movement is allowed, L-subjects can be displaced far away from their initial position inside the idiom chunk and retain its non-literal interpretation. In contrast, in contexts where syntactic movement is forbidden or not available, the idiomatic reading disappears or the sentence is ungrammatical. Crucially, subjects of TP-idioms are ungrammatical both in environments where movement is possible and in environments where it is not. First, the subject of vP-idioms must be specific for the idiomatic interpretation to hold, typical of dislocated constituents. Accordingly, non-specific subjects of vP-idioms detected by the addition of a relative clause in subjunctive only tolerate the non-idiomatic/literal interpretation: (45) a. Hierba mala [que crezca] nunca muere 55 herb bad that grows:SBJV never dies “Bad seed never dies” *IDIOMATIC “Weed never dies” LITERAL 55 There is a strong preference to have a postverbal subject with this sentence in the literal meaning. This is again another indication of the correlation between dislocation and the idiomatic interpretation in these idioms. 222 b. Hierba mala [que crece] nunca muere herb bad that grows:IND never dies “Bad seed never dies” IDIOMATIC “Weed never dies” LITERAL In (45)a the idiomatic subject is modified by a relative clause in subjunctive and only the literal interpretation is possible. In contrast, the modification by a relative clause in the indicative mood as in (45)b still allows both idiomatic and non-idiomatic readings. This requirement does not hold for subjects of TP-idioms, which can appear in irrealis contexts (i.e. with future tense in Otro gallo nos cantará Lit. "Another rooster will sing for us"). 56 Second, Rizzi (1990: 86) following Chomsky (1981: 325) notes that idiom chunks have the status of quasi-arguments (i.e., they do not receive a θ-role related to event participants). This immediately suggests two predictions: (46) a. Idiomatic subjects should be possible as specifiers of verbs that do not assign a (referential) θ-role to its external argument (for instance, raising verbs). b. Idiomatic subjects cannot stand as specifiers of verbs that require a referential θ-role (for instance, control verbs). Interestingly, the Spanish data confirms only the second of these predictions ((46)b) and partially disconfirms the first ((46)a). 56 Rizzi (1990: 79, ex. (15)) also discusses the observation that compositional idioms can be relativized. Interestingly, he reports that compositional idioms only allow relativization of the nominal, but not the compositional component. 223 To begin with, consider the case of raising verbs. Interestingly, only subjects of vP-idioms can appear as specifiers of raising verbs and retain its idiomatic interpretation, TP-idioms are ruled in only in the literal interpretation: (47) a. vP-idioms and raising verbs Hierba mala parece que nunca muere herb bad seems that never dies “(It seems that) bad seed never dies” IDIOMATIC “(It seems that) weed never dies” LITERAL b. TP-idioms and raising verbs Otro gallo parece que nos=cantará another rooster seems that us=sing:FUT “(It seems that) that's a horse of a different color” *IDIOMATIC “(It seems that) another rooster will sing for us" LITERAL Despite the fact the subject of the vP-idiom in (47)a is separated from its idiomatic predicate by the verb parecer "to seem", literal and non-literal interpretations are available. This is not the case with a TP-idiom subject in (47)b: the non-literal reading is lost, which challenges prediction (46).. Recall from Chapter 2 that subject dislocation was detected by the presence of assertion markers such as the affirmative particle sí "yes". The subject of a vP- idiom can be separated from its predicate by an affirmative marker as in (48): 224 (48) vP-idioms a. El buey suelto sí que bien se=lame solo the ox loose yes that well REFL=licks alone “A sound man needs no physician” IDIOMATIC “The ox that is loose licks himself well” LITERAL b. Hierba mala sí que nunca muere herb bad yes that never dies “Bad seed never dies” IDIOMATIC “Weed never dies” LITERAL In contrast, if the affirmative marker separates the subject of a TP-idiom and the verbal predicate, the idiomatic reading vanishes: (49) TP-idioms Otro gallo sí que nos=cantará another rooster yes that us=sing:FUT “That's a horse of a different color” *IDIOMATIC “Another rooster will sing for us" LITERAL Next, consider the case of control predicates analyzed as generated by movement (Hornstein 1999) 57 . Given that control predicates require their external argument to be associated with a participant role, it is not surprising that both vP- and TP-idioms lose their non-literal meaning (idiom chunk domain in brackets): 57 In a non-movement based analysis of control, it would be hard to tease apart the reasons why idiomatic subjects are disallowed with control predicates. This is because one alternatively might argue that the subject and its associated idiom predicate do not form a constituent at any level of the derivation. 225 (50) Control predicates and vP-/TP-idioms a. vP-idiom chunks El buey suelto quiere [bien lamer=se solo] the ox loose wants well licks:INF=REFL alone “A sound man does not want to have a physician” *IDIOMATIC “The ox that is loose wants to lick himself” LITERAL b. TP-idiom chunks Otro gallo querrá [PRO cantar=nos] another rooster want:FUT sing=us “That's a horse of a different color” *IDIOMATIC “Another rooster will want to sing for us" LITERAL Examples (50)a with a vP-idiom and (50)b with a TP-idiom are ruled out under the non-literal reading. This follows from the selectional restrictions that the verb querer “to want” impose on its external argument, absent with raising verbs: the subject of querer “to want” requires an agent to perform the action. Similarly with other vP- idioms (idiom chunk domain in brackets): (51) Hierba mala no quiere [nunca morir] herb bad not wants never die:INF “Bad seed does not want to never die” *IDIOMATIC "Weed does not want to never die” LITERAL Something that I do not have a good explanation for is that for some speakers example (51) above retains the idiomatic reading if the sentence is constructed as 226 Hierba mala nunca quiere morir "(lit.) Weed never wants to die". Probably, the availability of this ambiguity is due to the well-known observation that Spanish allows restructuring of verbal complexes. If so, one would be forced to assume that the referential θ-role of the restructured complex is also reanalyzed. One piece of evidence for this comes from the observation that examples like (50)a (repeated as (52)) under a restructuring analysis the reflexive se can also appear as enclitic in the idiomatic reading: (52) a. El buey suelto bien quiere [lamer=se solo] the ox loose well wants licks:INF=REFL alone b. El buey suelto bien se=quiere [lamer solo] the ox loose well REFL=wants licks:INF alone “A sound man does not want to have a physician” IDIOMATIC “The ox that is loose wants to lick himself” LITERAL If restructuring were impossible, one would expect the idiomatic reading to be correlated with a proclitic placement of the reflexive. 5.2.4 Evidence against base generation As discussed in Rizzi (1990: 78-80), idiomatic expressions are barred from extraction from weak islands due to its intrinsic non referential character. Thus, the examples in (53)b based on (53)a illustrate this observation with a vP-idiom chunk: 227 (53) vP-idioms and wh-islands a. Me=preguntaba porqué [hierba mala nunca muere] me=wondered for:what herb bad never dies “I was wondering why bad seed never dies” IDIOMATIC b. Hierba mala i me=preguntaba porqué [t i nunca muere] herb bad me=wondered for:what never dies “Bad seed I was wondering why bad seed never dies” *IDIOMATIC As indicated by (53)b, the idiomatic subject hierba mala "bad seed" resists extraction out of a weak islands such as a wh-island. Additional evidence for a movement analysis of idiomatic subjects comes from the examination of constructions such as hanging topic left dislocations. As well known, hanging topic left dislocations are base generated and display absence of "case connectivity". Therefore, idiomatic subjects in the examples discussed above are expected to lose their idiomatic interpretation if placed in these environments. This prediction is borne out: (54) Hanging topics a. Hablando de hierba mala, esta nunca muere speaking of herb bad this:F never dies “As for bad seed, bad seed never dies” *IDIOMATIC b. En cuanto a otro gallo, este nos=cantará in how_much ACC another rooster this:M us=sing:FUT “As for this, that's a horse of a different color” *IDIOMATIC 228 In (54)a and (54)b, the idiomatic subject receives prepositional case and accusative case respectively and the resumptive element este "this (M)" or esta "this (F)" is in nominative. As predicted, this context does not tolerate an idiomatic interpretation of the idiomatic subject. This is confirmed by the fact that under the literal or referential reading, the idiomatic subject is understood as a (topic)-theme or a definite participant of the event denoted by the predicate. Finally, one can establish a parallel between raising verbs and the syntax of sí que "yes that." Accordingly, displacement of the dislocated subject must also be related to Case. 5.3 The Makeup of the Lower Left Edge In this section, I want to propose that the Agree relation is the inner mechanism at work in the left edge in order to explain why the c*º head needs to be licensed by movement of subjects. Agree between the c*º head and the inflectional layer applies to license c*º’s uninterpretable subject agreement features, attracting a subject in [Spec, v*P] position. Thus, the subject agreement movement discussed in section 5.1 is best represented in terms of the Agree operation. The key point is based on Boeckx’s typology of complementizers (Boeckx 2003). According to him, complementizers across languages display different possibilities to induce extraction in syntactic islands. Under resumption, some languages such as Hebrew allow extraction across strong and weak islands. Other languages such as Serbo-Croatian and Vata display island effects in both strong and weak islands. A third group including Greek, Romanian, and Scottish Gaelic 229 displays selective island effects: only strong islands but not weak islands are sensitive to extraction under resumption. Boeckx proposes that these patterns follow from the fact that some complementizers induce agreement and others do not. A language with a complementizer establishing an Agree relation disallows extraction from non selected domains such as strong islands, but allows extraction from selected domains such as weak islands. On the other hand, if a language does not have an agreeing complementizer, extraction from all island types is permitted under Match, a relationship which is less strict than that entailed by Agree. Note that CLLD of direct objects in Spanish has been reported to display selective island sensitivity (Suñer 2006), indicating the availability of an agreeing complementizer in the left edge. Examples in (55) illustrate this point with wh-words that have been CLLDed out of a strong island such as the Relative Clause Island and out of a strong island such as a wh-island. Island boundaries are represented by angle brackets: (55) a. Relative Clause Island *¿Qué auto i llamó Sandra <al muchacho al que Joanna what car called Sandra ACC:the guy DAT:the COMP Joanna lo i =compró e i >? it=bought “Lit. What car did Sandra call the guy who Joanna bought it...?” 230 b. Wh-Island (Suñer 1992: 237, ex. 9) ¿A cuántos i de los niños no sabían <a qué familia ACC how.many of the children not knew DAT what family se=los i =había encomendado e i el director del orfanato>? DAT=ACC=had entrusted the director of:the orphanage "How many of the children didn't they know to what family the director of the orphanage has entrusted?" Similarly, extraction of subjects displays sensitivity only to strong islands, but not to weak islands: (56) a. Relative Clause Island *Me=dices que esas mujeres <conociste a los sintactistas me=say:2 COMP those women:3PL met:2 ACC the syntacticians que saludaron>. COMP met:3PL Lit. “You said that those women you met the syntacticians that they want to greet.” b. Wh-Island Te=preguntabas que esa mujer <a quiénes saludó ayer>. you=wonder:2PST COMP that woman:3 ACC:who greet:3PST yesterday Lit. “You were wondering this woman who she greeted at home.” If a wh-subject is extracted from a wh-island, sensitivity to wh-islands is suspended: 231 (57) ¿Qué lingüista te=preguntabas que <por qué saludó a which linguist you=wondered:2 COMP for:what greeted:3 Melissa?> ACC:Melissa “Who is such that you were wondering why that person greeted Melissa?” If Boeckx's theory is correct, the selective island sensitivity of direct objects and subjects discussed above finds a potential explanation. The Selective Island Sensitivity indicates that there is a probe position in the complementizer system that triggers an Agree relation. Given that Agree is constrained to selected domains, it cannot penetrate and search for a matching goal inside adjuncts, which become strong islands for extraction. 5.3.1 The ReconstructionPuzzle A diagnostic to test the availability of movement concerns the presence or absence of reconstruction effects. If reconstruction is a property of displaced expressions, and dislocated subjects can be generated by movement as suggested in Section 5.2, then reconstruction effects should be expected. However, the data from bound variable interpretation and scope interactions seem to contradict this expectation. The reconstruction Puzzle is illustrated below. In (58), a pronoun contained within a dislocated subject cannot be interpreted as a bound variable by a structurally lower quantified phrase, even in the presence of the affirmative marker: 232 (58) BOUND READING IMPOSSIBLE *Creo que su i abogado sí que a cada estudiante i believe:1SG COMP her lawyer AFF COMP ACC each student lo=defraudó de la peor manera posible. him=defrauded of the worst way possible “I believe that his lawyer defrauded each student in the worst possible way.” Similarly, scope interactions between an indefinite subject in dislocated position and a quantified phrase adjoined to TP only favor the surface reading, displaying absence of reconstruction as (59) shows: (59) UNAMBIGUOUS: un > cada (collective); *cada > un (distributive) Creo que un abogado sí que a cada estudiante believe:1SG COMP a lawyer AFF COMP ACC each student lo=defraudó de la peor manera posible. him=defrauded of the worst way possible “I believe that a lawyer defrauded each student in the worst possible way.” Example (59) favors only the collective interpretation corresponding to the surface scope. In other words, it can only mean that there is one single lawyer who, for any number of students, defrauded all of them. In contrast, example (59) lacks a distributive reading of cada estudiante; it does not mean that for each of the relevant students in context, there is one different lawyer such that each of the students was defrauded by a different lawyer. 233 In conclusion, if reconstruction is taken as a diagnostic of movement, the absence of reconstruction in bound variable interpretations and scope interactions should lead us to conclude that movement is also absent. This is in clear contradiction to the findings obtained with sentential idioms in Section 5.2. 5.3.2 Nominative Case in the Left Edge Even though the evidence for movement in sentential idioms and the evidence for the lack of reconstruction effects seem inconsistent, I want to suggest a possible way to solve the Reconstruction Puzzle. Dislocated subjects detected by the presence of the affirmative marker display absence of reconstruction effects because they check Nominative Case in the left edge, taking clausal scope. Thus, the type of chain created is similar to the one obtained by A-movement or "EPP checking," which, it has been argued not to display reconstruction effects. Chomsky (1995: 326) claims that reconstruction is contingent on the formation of operator/variable sequences, which correspond to typical cases of Ā- chains such as those formed by wh-questions, but not to those instances involving A- chains. Consider the examples in (60) from Chomsky (1995: 327, ex. 135), where the pronoun him and its trace form an A-chain: 58 (60) a. John expected [him to seem to me [ α t i to be intelligent]] b. *John i expected [him i to seem to me [ α t i to be intelligent]] c. John i expected [e to seem to me [ α him i to be intelligent]] 58 The star in (60)a was added by me for the sake of the exposition. 234 In (60)a, him must move out of the minimal domain α in order to receive Case. Let Assuming that binding conditions apply at LF, Condition B must operate to rule out the ungrammatical interpretation of (60)a, expressed as (60)b, where the pronoun him and the matrix subject John have the same reference. Crucially, the rejection of (60)b by Condition B is only achieved if the pronoun is interpreted outside α, namely, within the same minimal domain containing the binder. If reconstruction of ‘him’ were to operate as in (60)c, there would be no way to exclude the intended reading of (60)b because Condition B is expected to be respected and allow the coreferent reading. Thus, reconstruction should not apply to A-chains such as those created by him and its trace. An additional argument is provided in (61), also from Chomsky (1995: 327, ex. 137): (61) a. (It seems that) everyone isn’t there yet b. I expected [everyone not to be there yet] c. Everyone i seems [t i not to be there] Chomsky notes that negation can involve wide scope in (61)a and (61)b, but not in (61)c. Whereas the Case position of everyone and the negation are within the same minimal domain in (61)a and (61)b, the same does not occur in (61)c. Again, if reconstruction were possible in A-chains, no contrast in the scope interpretations should be expected. Finally, the most suggestive piece of evidence for the idea that traces of A- chains are invisible to the computational system is found in Jaeggli (1980). He notes 235 that Ā-traces block contraction, but A-traces do not. For example, Baltin (2003: 227, ex. 5-6) reports that a flap indicating contraction is possible in (62)a, which constitutes an A-chain environment. Therefore, the trace of an A-chain in (62)a allows the pronunciation in (62)b: (62) a. Does there i really need to t i be a separate constraint? b. Does there really [niyDa] be a separate constraint? In contrast, the same effect does not occur in (63)a-(63)b where there is an intervening referential element or in (64)a-(64)b where Ā-movement has applied: (63) Blocking of contraction with a referential DP a. I need Sally to be there b. *I niyDa Sally be there (64) Blocking of contraction with wh-movement a. Who do you need to be there b. *Who do you niyDa be there? Although some researchers have considered reconstruction to be possible in A-chains created by Case theoretical reasons (Hornstein (1999: 65, ex. 63; Lebeaux 1995; Hornstein 1995; Romero 1997 and Sportiche 1996, Bobaljik and wurmbrand 1999 among others), I will take the position that “movement” in A-dependencies does not leave a trace as in cases of Ā-chains as in Lasnik (1999b: 206-208). The adoption of this assumption is motivated by the observation that the traces created by A-movement are fundamentally different from those created by Ā-chains in terms of the Binding Theory. A-traces behave are nonreferential and behave like anaphors 236 whereas Ā-traces are referential and behave like variables. Thus, the deletion of A-traces do not have consequences for the semantic component, but the deletion of Ā-traces would create a situation where the wh-operator does not range over the set of individuals expressed by a variable. Therefore, I claim that the dislocated subjects associated with an assertion marker are displaced by movement to the specifier of c*º and that this movement is akin to one creating A-chains. Let us assume that these subjects are L-subjects. below I provide three pieces of evidence for the parallel between constructions involving L-subjects and structures involving raising. First, the specifier of c*º in the sí que construction is a nonthematic position. Thus, if an L-subject occupies this specifier, the L-subject is the only constituent that can be related to the affirmative marker sí and its associated verbal predicate: (65) a. Melissa sí que escribió esta tesis Melissa AFF COMP wrote this thesis "Melissa indeed wrote this thesis" b. *Melissa sí que Arantxa escribió esta tesis Melissa AFF COMP Arantxa wrote this thesis "Melissa indeed Arantxa wrote this thesis" c. Melissa, sí que Arantxa escribió esta tesis Melissa AFF COMP Arantxa wrote this thesis "Melissa, Arantxa indeed wrote this thesis" 237 The subject Melissa in the above examples must be associated with the sí que construction and the main verb as in (65)a. Otherwise, the subject Melissa would be interpreted as a hanging topic as the contrast (65)b-(65)c shows, and would be disconnected from the inflectional layer. A second piece of evidence comes from restructuring processes analyzed as involving raising (Solà 2002). Note that verbal complexes allowing optional restructuring require obligatory restructuring in the presence of an affirmative marker, presumably indicating some form of connectivity similar to "A-movement." For instance, the subject Melissa can be interpreted as the subject of decidir and escribir in (66) (with restructuring) or it can be interpreted as only the subject of decidir (without restructuring). In the latter case, the subject of escribir is unspecified and different from Melissa: (66) Melissa decidió escribir esta tesis. Melissa decided write:INF this thesis “Melissa decided to write this thesis.” (with restructuring) "Melissa decided the writing of this thesis." (with no restructuring) Interestingly, if the affirmative marker sí is added to the conjugated verb, restructuring must take place: (67) Melissa sí decidió escribir esta tesis. Melissa AFF decided write:INF this thesis “Melissa decided to write this thesis.” (only with restructuring) 238 In contrast, if restructuring is obligatory, sí cannot be associated with the non finite verb: (68) a. Melissa sí necesita escribir esta tesis. Melissa AFF needs write:INF this thesis “Melissa needs to write this thesis.” b. *Melissa necesita sí escribir esta tesis. Melissa needs AFF write:INF this thesis “Melissa needs to write this thesis.” Verbs such as necesitar “to need” must undergo restructuring to share a subject. This is possible only if sí is attached to necesitar, as in (68)a, but it is impossible if sí is attached to escribir, as in (68)b. Furthermore, when restructuring is impossible, as with verbs allowing either subject or object control, the addition of emphatic sí to the non finite verb forces control by a subject: (69) Melissa pidió a Arantxa sí firmar su propia confesión. Melissa asked ACC Arantxa AFF sign:INF her own confesion “Melissa asked Arantxa to sign her own confesion.” The previous paradigms indicate that restructuring is correlated with the presence of the affirmative marker heading the entire verbal complex. External Merge of these preverbal subjects would leave these contrasts unexplained. The third and final piece of evidence contemplates the availability of weak crossover effects, considered a detector of an operator/variable sequences. Given that operator/variable sequences are a property of Ā-chains, the constructions involving 239 L-subjects are predicted to lack this effect if their constitution is akin to A- chains. This expectation is confirmed by the empirical data: (70) ¿Quién i sí que a su i novia la=vió bailar who AFF COMP ACC his girlfriend ACC:her=saw dance:INF en la fiesta? in the party “Who saw his girlfriend dancing in this party?” (intended) Given the discussion in the present section, I conclude that L-subjects are licensed by movement of the lexical subject to the specifier of c*P. In formal terms, the displacement of subjects to the lower left edge is motivated to check the subject agreement features of c*º. As discussed in Section 5.1.4, L-subjects, namely dislocated subjects displaced by movement to the specifier of c*P, check the subject agreement features that were displaced into c*º. As a result L-subjects receive Nominative Case in the left edge, an idea already suggested in Pesetsky and Torrego’s analysis of wh-questions in English (Pesetsky and Torrego 2001) 5.3.3 Multiple dislocations in the Lower Left Edge An immediate consequence of the present approach is that once the head c*º has met its licensing requirements; an intervention effect is created in relation to any further application of Move: 240 (71) Intervention effect If an XP has checked the F-feature in c*º, a head W that c- commands c*º cannot attract a lower YP inside c*º to check W’s own F-feature: *YP [Wº [ c*P XP [c*º ... t i [F] [F] The intervention effect in the lower left edge is a byproduct of the application of Agree. Once c*ºº has established an Agree relation with the inflectional system and an XP has been displaced by Move, an additional Agree relation can only target XP, but cannot target any constituent below XP. The existence of an intervention effect finds support in the availability of a phenomenon called by Rizzi (1997) recursivity, namely, the property of some languages to display free ordering of multiple dislocations. Example (72) is reported in Suñer (2003b: ex. 11): (72) a. Mi hermana, las empanadas, los domingos . my sister:3SG the turnovers:3PL the Sundays las=hac-e de pollo them=make-3SG of chicken "My sister, the turnovers, on Sundays, she fills them with chicken." 241 b. Las empanadas, mi hermana, los domingos. the turnovers:3PL my sister:3SG the Sundays las=hac-e de pollo. them=make-3SG of chicken "The turnovers, my sister, on Sundays, she fills them with chicken." c. Los domingos, mi hermana, las empanadas the Sundays my sister:3SG the turnovers:3PL las=hac-e de pollo. them=make-3SG of chicken "On Sundays, my sister, the turnovers, she fills them with chicken." In (72), free recursion of a direct object, a subject, and an adverbial is possible; they can appear displaced in the left edge in any order. If recursivity is an exclusive property of TOP* positions as argued by Rizzi (1997), then recursivity does not have to follow from anything else in the grammar. Thus, it becomes a mysterious property, present in some languages (Spanish and Italian) but absent in others (German, Kinande and English). The proposal offered here explains these facts if c*º is a probe establishing an Agree relation. If c*º has fulfilled its formal licensing requirements, an Intervention Effect is automatically created for additional probing, barring additional instances of Move. Therefore, the dislocated constituents in (72) cannot be generated by Move, but can be generated by Merge. 242 5.4 Conclusions In this chapter, I have analyzed the dislocation pattern of subjects. The main empirical finding suggests that the dislocation pattern should be understood as generated by the application of Move or Merge. These two core operations target the left periphery under different conditions. Whereas Move generates L-subjects in the specifier of c*º, Merge generates H-subjects when the specifier of c*º is targeted by another constituent. Crucially, L-subjects arise in the left edge as the result of overt raising of the subject agreement features when c*º is specified as [finite] 243 References Alexiadou, Artemis. 2005. “On Nominative Case Features and Split Agreement.” In: Brandner, Ellen and Heike Zinsmeister (ed). New Perspectives on Case Theory. Stanford: CSLI Publications. Alexiadou, Artemis and Elena Anagnostopoulou. 1998. “Parametrizing Agr: Word Order, V-movement and the EPP-checking.” In: Natural Language and Linguistic Theory 16: 491-539. Alvar, Manuel.1996. Manual de Dialectología Hispánica: el español de América. Barcelona: Ariel. Aoun, Joseph, Elabbas Benmamoun and Dominique Sportiche. 1994. “Agreement, Word Order and Conjunction in Some Varieties of Arabic.” Linguistic Inquiry 25: 195-220. Aoun, Joseph and Elabbas Benmamoun. 1998. “Minimality, Reconstruction and PF- Movement.” In: Linguistic Inquiry 29: 569-597. Aoun, Joseph; Lina Choueiri and Norbert Hornstein. 2001. “Resumption, Movement, and Derivational Economy.” In: Linguistic Inquiry 32: 371-403. Aoun, Joseph and Lina Choueiri. 2000. “Epithets.” In: Natural Language and Linguistic Theory 18: 1-39. Aoun, Joseph and Audrey Li. 2004. Essays on the Representational and Derivational Nature of the Grammar. Cambridge, Massachusetts: MIT Press. Aoun, Joseph and Audrey Li. 1993. Syntax of Scope. Cambridge, Massachusetts: MIT Press. Aoun, Joseph and Audrey Li. 1990. “Minimal Disjointness.” In: Linguistics 28: 189- 204. Aoun, Joseph, and Audrey Li. 1989. “Constituency and Scope.” In: Linguistic Inquiry, 20: 124-172. Arnaiz, Alfredo. 1996. N-Words and wh-in situ. PhD dissertation. University of Southern California, Los Angeles. Baauw, Sergio. 1998. “Subject-Verb Inversion in Spanish Wh-questions: Movement as Symmetry Breaker.” In: Bezooijen, Renée van and René Kager (eds). Linguistics in the Netherlands. 244 Baker, Mark. 1996. The Polysynthesis Parameter. New York: Oxford University Press. Baltin, Mark. 2001. “A-Movements.” In: Baltin, Mark and Chris Collins (eds). The Handbook of Contemporary Syntactic Theory. Blackwell: Oxford, 226-254. Barbiers, Sjef. 2002. “Current Issues in Modality.” In: Barbiers, Sjef (ed). Modality and Its Interaction with the Verbal System. Philadelphia: Benjamins, 11-27. Beas, Omar. 2007. “The Narrow Structure of the Left Periphery.” Unpublished manuscript. University of Southern California. Beas, Omar. 2005a. “Verbal Morphology, Ellipsis and Phrase Structure in Spanish.” Talk presented at the USC Open House Student Workshop. Bellert, Irena. 1977. “On Semantic and Distributional Properties of Adverbs.” In: Linguistic Inquiry 8: 337-351. Besten den, H. 1983. “On the Interaction of Root Transformations and Lexical Deletive Rules.” In: Abraham, W. (ed). On the Formal Syntax of Westgermania. Benjamins, Amsterdam, 47-131. Bobaljik, Jonathan David and Susi Wurmbrand.1999. “Modals, Raising and A- Reconstruction.” Presented at Leiden and Salzburg, (http://bobaljik.uconn.edu/files.html) Boeckx, Cedric. 2003. Islands and Chains: Resumption as Stranding. Philadelphia: Benjamins. Bok Bennema, Reineke. 2001. “Evidence for an Aspectual Functional Head in French and Spanish.” Ms. Borsley, Robert and Janig Stephens. 1989. Agreement and the Position of Subjects in Breton. In: Natural Language and Linguistic Theory 7: 407-427. Borsley, Robert, Maria Luisa Rivero and Janig Stephens. 1996. “Long Head Movement in Breton.” In: Borsley, Robert and Ian Roberts (eds). The Syntax of the Celtic Languages. Cambridge University Press, Cambridge. Brucart, Jose Maria. 1997. “Concordancia ad sensum y partitividad en español.” In: Almeida, Manuel and Josefa Dorta Luis (eds). Contribuciones al estudio de la 245 linguistica hispánica: Homenaje al profesor Ramón Trujillo. Tenerife: Montesinos Editor, 157-184 Bury, Dirk. 2003. Phrase structure and derived heads. PhD dissertation, University College London. Bury, Dirk. 2000. “Effects of self attachment and the status of functional projections.” In: Snippets 2 (http://www.lededizioni.it/ledonline/snippets.html). Camacho, José. 2003. Is there a Place for Subjects in Spanish? Ms. (http://www.rci.rutgers.edu/~jcamacho/publications/publications.htm). Campos, Héctor. 1999. “Spanish as a CP/IP Absorption Language.” In: Gutiérrez- Rexach, Javier and Fernando Martínez Gil (eds) Advances in Hispanic Linguistics. Papers from the 2 nd Hispanic Linguistics Symposium. Cascadilla Press: Somerville,.332-345. Campos, Héctor. 1997.”Three Types of Subject Raising in Spanish.” Unpublished manuscript, Georgetown University. Campos, Héctor. 1991. “Silent Objects and Subjects in Spanish.” In: Campos, Héctor and Fernando Martínez Gil (eds) Current Studies in Spanish Linguistics. Washington: Georgetown University Press, 117-141. Campos, Héctor, and Mary Zampini. 1990. “Focalization Strategies in Spanish.” Probus 2: 47-64. Cardinaletti, Anna. 1997. “Subjects and Clause Structure.” In Haegeman, Liliane. (ed.). The New Comparative Syntax. London: Longman. Cardinaletti, Anna and Ian Roberts. 1990. Clause Structure and X-Second, ms. University of Venice and University of Geneva. Cecchetto, Carlo. 2000. “Doubling Structures and Reconstruction.” In: Probus 12: 94-126. Cheng, Lisa L.-S. 1991. On the typology of wh-questions. PhD dissertation, MIT, Cambridge, Mass. Chomsky, Noam. 2001. “Derivation by Phase.” In: Michael Kenstowicz (ed) Ken Hale: A Life in Language. Cambridge, Mass.: MIT Press, 1-52. 246 Chomsky, Noam. 2000. “Minimalist Inquiries: the framework.” In Martin, Roger; David Michaels and Juan Uriagereka (eds). Step by step: essays on minimalist syntax in honor of Howard Lasnik. Cambridge, Mass.: MIT Press, 89- 155. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam. 1993. “A Minimalist Program for Linguistic Theory.” In: Hale, Kenneth and Samuel Kayser (ed). The View from the Building. Cambridge, Mass.: MIT Press, 1-52. Chomsky, Noam. 1986. Knowledge of Language. Praeger, New York. Chomsky, Noam. 1982. Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, Massachusetts: MIT Press. Chomsky, Noam. 1981. Lectures on Government and Binding. Foris: Dordrecht. Chomsky, Noam. 1977. “On Wh-movement.” In: Cullicover, Peter; Thomas Wasow and Adrian Akmajian (eds). Formal Syntax. New York: Academic Press, 71-132. Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton. Cinque, Guglielmo. 1999. Adverbs and Functional Heads. Oxford: Oxford University Press. Cinque, Guglielmo. 1995. “Adverbs and the Universal Hierarchy of Functional Projections.” GLOW Newsletter 34, 14-15. Cinque, Guglielmo. 1990. Types of Ā-Dependencies. MIT Press, Cambridge, Massachusetts: MIT Press. Cinque, Guglielmo. 1977. “The Movement Nature of Left Dislocation.” In: Linguistic Inquiry 8: 387-412. Contreras, Heles. 2004. “A Restricted View of Head Movement.” In: Bok Bennema, Reineke (ed). Romance Languages and Linguistic Theory 2002: Selected Papers from ‘Going Romance’-Groningen, 28-30 November 2002. Philadelphia: John Benjamins Publishing Company. Cullicover, Peter and Paul Postal. 2001. Parasitic Gaps. Cambridge: MIT Press. Diesing, Molly. 1992. Indefinites. Cambridge: MIT Press. 247 Dumitrescu, Dominita. 1973. "Apuntes sobre el uso enfático de sí (adv.) en el español contemporáneo", Revue Roumaine de Linguistique 5: 407-413. Emonds, Joseph. 1978. The verbal complex V’-V in French. Linguistic Inquiry 9, 151–175. Emonds, Joseph. 1970. Root and Structure Preserving Transformations. PhD dissertation. MIT, Cambridge Massachusetts. Enς, Murvet. 1991. “The Semantics of Specificity.” Linguistic Inquiry 22: 1- 25. Engdahl, Elisabet. 1985. “Parasitic Gaps, Resumptive Pronouns and Subject Extraction.” In: Linguistics 23, 3-44. Engdahl, Elisabet. 1983. “Parasitic Gaps.” In: Linguistics and Philosophy 6: 5-34. Epstein, Samuel David. 1998. “Overt scope marking and covert verb second.” In: Linguistic Inquiry 29: 181-227. Escobar, Linda. 1997. “Clitic Left Dislocation and Other Relatives.” In: Anagnostopoulou, Elena, Henk van Riemsdijk and Frans Zwarts (eds). Materials on Left Dislocation. Amsterdam: Benjamins, 233-274. Etxepare, Ricardo. 1998. “On the Structure of Declarative Clauses.”In: Schwegler, Armin, Bernard Tranel and Myriam Uribe Etxebarria (eds). Romance Linguistics: Theoretical Perspectives, Amsterdam: Benjamins. Evans, Gareth. 1980. “On Pronouns.” In: Linguistic Inquiry 11: 337-362. Fontana, Josep. 1993. Phrase Structure and the Syntax of Clitics in the History of Spanish. PhD dissertation, University of Pennsylvania. Fox, Danny. 2000. Economy and Semantic Interpretation. Cambridge, Massachusetts: MIT Press. Fox, Danny. 1993. “Relative Clauses and Resumptive Pronouns in Hebrew.” Ms. García Mayo, Pilar. 1992. “Pro as a parasitic gap licenser.” In: Smith, Evan and Flore Zephir (eds). Proceedings of the 1992 Mid-America Linguistics Conference and Conference on Siouan-Caddoan Languages. University of Missouri, 15-24. 248 González, María José. 2000. “Emphatic Affirmation in the Spanish of the Basque Country: The Equivalence between ya and sí (que).” In: Campos, Héctor, Elena Herburger, Alfonso Morales Front and Thomas Walsh (eds). Hispanic Linguistics at the Turn of the Millennium: Papers from the 3rd Hispanic Linguistics Symposium. Cascadilla Press: Somerville, 308-321. Goodall, Grant. 2004. “On the Syntax and Processing of Wh-questions in Spanish.” In: Schmeiser, Benjamin, Vineeta Chand, Ann Kelleher and Angelo Rodriguez (eds)..WCCFL 23 Proceedings. Cascadilla Press: Somerville, 101-114. Goodall, Grant. 1991. “Spec of IP and Spec of CP in Spanish wh-questions”. Paper read at the Linguistic Symposium of Romance Languages, 21. Giorgi, Alessandra and Fabio Pianesi. 1997. Tense and Aspect: From Semantics to Morphosyntax. Oxford University Press, New York. Grimshaw, Jane. 1991. "Extended projections and locality." In P. Coopmans, M. Everaert, and Jane Grimshaw (eds). Lexical structure. New Jersey: Erlbaum. Gutiérrez Bravo, Rodrigo. 2007. “Prominence Scales abd Unmarked Word Order in Spanish.” In: Natural Language and Linguistic Theory 25: 235-271. Gutiérrez Bravo, Rodrigo. 2002. Structural Markedness and Syntactic Structure: A Study of Word Order and the Left Periphery in Mexican Spanish. PhD dissertation, University of California, Santa Cruz. Haverkort, Marco. 1993. “Romance and Germanic Clitics: A Comparison of their Syntactic Behavior.” In: Shannon, Thomas and Johan Snapper. The Berkeley Conference on Dutch Linguistics 1993. Lanham: University Press of America, 131-150. Hernanz, Maria Luisa and José Maria Brucart. 1987. La sintaxis. Barcelona: Crítica. Hornstein, Norbert. 1999. “Minimalism and Quantifier Raising”. In: Epstein, Samuel and Norbert Hornstein. Working Minimalism. Cambridge, MA: MIT Press. Hornstein, Norbert. 1995. Logical Form: From GB to Minimalism. Oxford: Blackwell Publishers. Hornstein, Norbert and Amy Weinberg. 1990. “On the Necessity of LF.” In: The Linguistic Review 7: 129-167. 249 Higginbotham, James. 1992. “Anaphoric Reference and Common Reference.” Unpublished manuscript, MIT. Iatridou, Sabine. 1991. “Clitics and Island Effects.” In: University of Pennsylvania Working Papers in Linguistics, 2. Inclán Nichol, Sara. 1997. Absence of Verb Inversion and Specificity in Peninsular Spanish Wh-questions. PhD dissertation, UCLA, Los Angeles. Jackendoff, Ray. 1971. “Modal Structure in Semantic Representation.” In: Linguistic Inquiry 2: 479-454. Jaeggli, Osvaldo. 1980. “Remarks on to-Contraction.” Linguistic Inquiry 11: 239- 245. Jelinek, Eloise. 1984. "Empty Categories, Case, and Configurationality." In: Natural Language and Linguistic Theory 2: 39-76. Kiss, Katalin. 1993. “Wh-movement and Specificity.” In: Natural Language and Linguistic Theory 11: 85-120. Koeneman, Olaf 2000. The Flexible Nature of Verb Movement. PhD dissertation, Utrecht University. Kuno, Susumu and Etsuko Kaburaki. 1977. “Empathy and Syntax.” In Linguistic Inquiry 8: 627-672. Laka, Itziar. 1990. Negation in Syntax: On the Nature of Functional Categories and Projections. PhD dissertation, MIT, Massachusetts. Larson, Richard. 1988. “On the Double Object Construction.” In: Linguistic Inquiry, 19: 335-391. Lasnik, Howard. 1999a. Minimalist Analysis. Oxford Blackwell. Lasnik, Howard. 1999b. “Chains of Arguments.” In: Epstein, Samuel and Norbert Hornstein. Working Minimalism. Cambridge, MA: MIT Press. Lebeaux, David 1995. “Where does Binding Theory Apply?” In: University of Maryland Working Papers in Linguistics 3, 63-88. Leonetti, Manuel. 2004. "Specificity and Differential Object Marking in Spanish", In: Catalan Journal of Linguistics 3: 75-114. (http://www2.uah.es/leonetti/papers/Specif&DOM.pdf). 250 Leonetti, Manuel. 2003. “Specificity and Object Marking: the Case of Spanish a”, In: von Heusinger, Klaus and Georg Kaiser (eds.). Arbeitspapier Nr. 113. Proceedings of the Workshop “Semantic and Syntactic Aspects of Specificity in Romance Languages., Fachbereich Sprachwissenschaft, Universität Konstanz, 67-101. (http://www2.uah.es/leonetti/papers/Leonetti- SpecificityandObjectMarking.pdf). Leonetti, Manuel and A. Ahem. 2004. “The Spanish Subjunctive: Procedural Semantics and Pragmatic Inference” In: Placencia, Maria Elena and Rossina Márquez-Reiter (eds.). Current Trends in the Pragmatics of Spanish, Amsterdam, John Benjamins, 34-56. Lipsky, John. 2002. El español de América. Madrid: Cátedra. Lois, Ximena. 1989. Aspects de la syntaxe de l’espagnol et théorie de la grammaire. PhD dissertation, University of Paris-VIII. Longobardi, Giuseppe. 2000. “Postverbal Subjects and the Mapping Hypothesis.” In: Linguistic Inquiry 31: 691-702. Longobardi, Giuseppe. 1994. “Reference and Proper Names: A Theory of N- Movement in syntax and logical form.” In: Linguistic Inquiry 25: 609-665. López, Luis. 1999. “The Syntax of Contrastive Focus: Evidence from VP-Ellipsis.” In: Gutiérrez Rexach, Javier and Fernando Martínez Gil (eds). Advances in Hispanic Linguistics. Somerville, MA: Cascadilla Press, 412-427. Luján, Marta. 1986. “Los pronombres implícitos y explícitos del español.” Revista Argentina de Linguística, xx. Mallén, Enrique. 1993. “Verb Movement in Spanish and the Licensing Condition on Variables.” In: Linguistiche Beritche 148. Mejías Bikandi, Errapel. 1995. “The VP internal Subject Hypothesis and Spanish sentence structure.” In: Amastae, Jon, Grant Goodall, Mario Montalbetti and Marianne Phinney (eds.) Contemporary Research in Romance Linguistics: Papers from the 22 nd Linguistic Symposium on Romance Languages El Paso- Ciudad Juarez 1992. Amsterdam: John Benjamins Publishing Company. McCloskey, James. 1990. “Resumptive Pronouns, Ā-Binding and Levels of Representation in Irish. In: Hendrick, Randall (ed). The Syntax and Semantics of Modern Celtic Languages. New York: Academic Press, 199-248. 251 McDowell, Joyce. 1987. Assertion and Modality. PhD dissertation, USC, Los Angeles, California. Montalbetti, Mario. 1984. After Binding: On the Interpretation of Pronouns. PhD dissertation, MIT, Cambridge Massachusetts. Moreno Quibén, Norberto. 2000. “La especificidad y la estructura del sintagma verbal: La generalización de Leonetti.” Paper presented at the 10º Coloquio de Gramática Generativa, Universidad de Alcalá (Alcalá de Henares, abril del 2000). (http://www.uclm.es/profesorado/nmoreno/investigacion/Alcal2000Leonetti.pdf). Nash, Lea and Alain Rouveret. 2002. “Cliticization as Unselective Attract.” Catalan Journal of Linguistics 1: 145-187. Nunberg, Geoffrey, Ivan Sag and Thomas Wasow. 1994. “Idioms”. Language 70: 491-538. Olarrea, Antxon. 1996. Pre and Post-verbal Subject Position in Spanish: A Minimalist Account. PhD dissertation, University of Washington. Olsen, S. 1985. “On Deriving V1 and V2 Structures in German.” In: Toman, J. (ed). Studies in German Grammar. Dordrecht: Foris, 133-163. Ordóñez, Francisco. 2000. The Clausal Architecture of Spanish: A Comparative Study. New York: Garland Publishing Company. (This is the published version of Ordóñez 1997’s doctoral dissertation). Ordóñez, Francisco. 1998. “Postverbal Asymmetries in Spanish.” In: Natural Language and Linguistic Theory 16: 313-346. Ordóñez, Francisco. 1997. Word Order and Clausal Structure in Spanish and Other Romance Languages. PhD dissertation, Cuny, Graduate Center. Ordóñez, Francisco and Esthela Treviño. 1999. “Left Dislocated Subjects and the Prodrop Parameter: A Case Study of Spanish.” In: Lingua 107: 39-68. Perlmutter, David. 1970. “The Two Verbs begin.” In: Jacobs, R. and P. Rosenbaum (eds). Readings in Transformational Grammar. Boston: Ginn. Pesetsky, David. 1987. “Wh-in situ: Movement and Unselective Binding.” In: Reuland, Eric and Alice Meulen (eds). The Representation of (In)definiteness. Cambridge, Massachussets: MIT Press, 98-129. 252 Pesetsky, David and Esther Torrego. 2001. “Tº to Cº movement; Causes and Consequences.” In: Michael Kenstowicz (ed) Ken Hale: A Life in Language. Cambridge, Massachussets: MIT Press, 1-52. Piera, Carlos. 1987. “Sobre la estructura de las cláusulas de infinitivo”. In: Violeta Demonte and Marina Fernández Lagunilla (eds). Sintaxis de las lenguas románicas. Madrid: El Arquero. Poletto, Cecilia. 2000. The Higher Functional Field in the Northern Italian Dialects. New York: Oxford University Press. Pollock, Jean Ives. 1989. “Verb Movement, UG and the Structure of IP.” Linguistic Inquiry 20: 365-424. Platzack, Christer. 1983. “Existential Sentences in English, German, Icelandic and Swedish.” In: Karlsson, Fred (ed). Papers from the Seventh Scandinavian Conference of Linguistics: Hanasari, Finland. Helsinsky: University of Helsinski, 80-100. Riemsdijk, Henk van. 1997. “Left Dislocation.” In: Anagnostopoulou, Elena, Henk van Riemsdijk and Franz Zwarts (eds). 1997. Materials on Left Dislocation. Amsterdam: Benjamins, 233-274. Riemsdijk, Henk van and Frank Zwarts. 1974. “Left Dislocation in Dutch and the Status of Copying Rules.” In: Anagnostopoulou, Elena, Henk van Riemsdijk and Franz Zwarts (eds). 1997. Materials on Left Dislocation. Amsterdam: Benjamins, 13-29. Rivero, María Luisa. 1980. “On Left Dislocation and Topicalization in Spanish.” In: Linguistic Inquiry 11: 363-393. Rizzi, Luigi and Ur Shlonsky. 2005. “Strategies of Subject Extraction.” (http://www.ciscl.unisi.it/doc/doc_pub/rizzi_shlonsky2005.pdf). Rizzi, Luigi. 1997. “The Fine Structure of the Left Periphery.” In: Haegeman, Lilianne (ed.). Elements of Grammar: A Handbook of Generative Syntax. Kluwer: Dordrecht, 281-337. Rizzi, Luigi. 1996. “Residual Verb Second and the Wh-Criterion.” In: Belleti, Adriana and Luigi Rizzi (eds.). Parameters and Functional Heads. New York: Oxford University Press. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Massachusetts: MIT Press. 253 Rizzi, Luigi. 1986. “Null Objects in Italian and the Theory of pro.” In: Linguistic Inquiry, 17: 501-557. Roberts, Ian. 2005. Principles and Parameters in a VSO Language. A Case Study in Welsh. New York: Oxford University Press. Romero, Maribel. 1997. “The Correlation between Scope Reconstruction and Connectivity Effects”. In: E. Curtis, J. Lyle and G. Webster (eds). Proceedings of the XVI West Coast Conference in Linguistics, 351-365. Ron, María Pilar. 1998. The Position of the Subject in Spanish and Clausal Structure: Evidence from Dialectal Variation. PhD dissertation, Northwestern University, Illinois. Ross, John Robert. 1970. ”On Declarative Sentences” In: Jacobs, A. and Peter Rosenbaum. (eds). Readings in English Transformational Grammar. Waltham, Massachusetts: Ginn. Ross, John Robert. 1969. “Auxiliaries as main verbs.” In: Todd, W. (ed). Studies in Philosophical Linguistics, Series 1, Evanston: Great Expectations, 77-102. Ross, John Robert. 1967. Constraints on Variables in Syntax. PhD dissertation. MIT. Published in 1986 as Infinite Syntax. Ablex Publishing Corporation: Norwood. Rosselló, Joana. 2000. “A Minimalist Approach to the Null Subject Parameter” In: Catalan Working Papers in Linguistics 8: 97-128. Sauerland, Uli and Paul Elbourne. 2002. “Total Reconstruction, PF-movement and Derivational Order.” In: Linguistic Inquiry 33: 283-319. Schneider Zioga, Patricia. 2007. “Antiagreement, Antilocality and Minimality: The Syntax of Dislocated Subjects.” In: Natural Language and Linguistic Theory 25: 403-446. Schneider Zioga, Patricia. 2002. “The Case of Antiagreement.” In: Rackowski, A and Norvin Richards (eds). Proceedings of AFLA 8: the Eighth Meeting of the Austronesian Formal Linguistics Association, MIT, MITWPL 44, Cambridge, MA, 325-339. Schneider Zioga, Patricia. 2000. "Antiagreement and the Fine Structure of the Left Periphery," University of California Irvine Working Papers in Linguistics, Vol. 6 (2000) (http://aris.ss.uci.edu/ling/people/lecturers/schneider-zioga/schneider- zioga.html). 254 Schneider Zioga, Patricia. 1993. The Syntax of Clitic Doubling in Modern Greek. PhD dissertation, USC, Los Angeles, California. Simpson, Andrew and Zoe Wu. 2002. “Agreement, Shells and Focus.” In: Language, 78: 287-313. Solà, Jaume. 2002. “Clitic climbing and null subject languages.” Catalan Journal of Linguistics 1: 225-255. Sportiche, Dominique. 1996. “A-Reconstruction and Constituent Structure”. In: Colloquium Université of Montréal (October, 1996). Stephens, Janig. 1990. “Nonfinite clauses in Breton”. In: Ball, Martin J., James Fife, Erich Poppe, and Jenny Rowland (eds.). Celtic Linguistics. Amsterdam: John Benjamins, 151-167. Suñer, Margarita. 2006. “Left Dislocations with and without Epithets.” In: Probus 18: 127-158. Suñer, Margarita. 2003. “The Lexical Pre-verbal Subject in Spanish: Where are thou?” In: Núñez Cedeño, Rafael; Luis López and Richard Cameron (eds). A Romance Perspective on Language Knowledge and Use. Amsterdam: Benjamins, 95-109. Suñer, Margarita. 2000a. "Clitic doubling of Strong Pronouns in Spanish: An Instance of Object-shift." In: Franco, Jon; Alazne Landa, and Juan Martín (eds.), Grammatical Analyses in Basque and Romance Linguistics. Amsterdam: Benjamins, 233-256. Suñer, Margarita. 2000b. "Object shift: Comparing a Romance Language to Germanic." Probus 12: 261-289. Suñer, Margarita. 1994. “V-Movement and the Licensing of Argumental Wh-phrases in Spanish.” In: Natural Language and Linguistic Theory 12: 335-372. Suñer, Margarita. 1992. “Two Properties of Clitics in Clitic Doubled Constructions.” In: Huang, James and Robert May (eds). Logical Structure and Linguistic Structure: Cross Linguistic Perspectives. Reidel. Suñer, Margarita. 1988. “The Role of Agreement in Clitic Doubled Constructions.” In: Natural Language and Linguistic Theory 6: 391-434. Suñer, Margarita and Carmen Lizardi. 1995. “Dialectal Variation in an Argumental/Non-Argumental Asymmetry in Spanish.” In: Amastae, Jon, Grant 255 Goodall, Mario Montalbetti and M. Phinney. Contemporary Research in Romance Languages, El Paso/Cd Juárez 1992, Amsterdam: John Benjamins. Tellier, Christine. 1991. Licensing Theory and French Parasitic Gaps. Dordrecht: Kluwer. Timm, Lenora. 1988. “Relative clause formation in Breton.” In: Word 39: 79- 107. Toribio, Jacqueline Almeida. 2002. “Setting Parametric Limits on Dialectal Variation in Spanish.” In: Lingua 10: 315-341. Torrego, Esther. 1995. “On the Nature of Clitic Doubling.” In: Campos, Héctor and Paula Kempchinsky (eds.), Evolution and Revolution in Linguistic Theory, Washington, DC: Georgetown University Press, 300-418. Torrego, Esther. 1984. “On Inversion in Spanish and Some of Its Effects.” In: Linguistic Inquiry 15: 103-129. Travis, Lisa. 1991. “Parameters of Phrase Structure and the Verb second Phenomena.” In: Freidin, Robert (ed). Principles and Parameters in Comparative Grammar. Cambridge, Massachusetts: MIT Press, 339-364. Uriagereka, Juan. 1995. “Aspects of the Syntax of Clitic Placement in Western Romance.” Linguistic Inquiry 26: 79-125. Uribe Etxebarria, Maria. 1995. “On the Nature of SPEC/IP and Its Relevance for Scope Asymmetries in Spanish and English.” In: Amastae, J., Grant Goodall, Mario Montalbetti and M. Phinney (eds.) Contemporary Research in Romance Linguistics: Papers from the 22 nd Linguistic Symposium on Romance Languages El Paso-Ciudad Juarez 1992. Amsterdam: John Benjamins Publishing Company, 355-366. Vat, Jan. 1981. “Left Dislocation, Connectedness and Reconstruction.” In: Groninger Arbeiten zur Germanistischen Linguistik 20. Vikner, S. 1995. Verb Movement and Expletive Subjects in the Germanic Languages. Oxford University Press, New York. Villalba, Xavier. 2000. The Syntax of Sentence Periphery. PhD dissertation, Universitat Autònoma de Barcelona. Villalba, Xavier. 1999. “Nihil est in LF quod prius non fuerit in SS.” In: Catalan Working Papers in Linguistics 7: 239-252. 256 Webelhuth, Gert. 1985. “German is Configurational.” The Linguistic Review 4: 203- 246. Weinberg, Amy, 1999. “A Minimalist Theory of Sentence Processing” In: Epstein, Samuel and Norbert Hornstein. Working Minimalism. Cambridge, MA: MIT Press. Willis, David. 1998. Syntactic Change in Welsh: A Study of the Loss of Verb Second. Oxford: Clarendon Press. Wurmbrand, Susi. 1999. “Modal Verbs Must Be Raising Verbs.” In: Bird, S.; Andrew Carnie, J. Haugen and P. Norquest (eds). WCCFL 18 Proceedings. Somerville: Cascadilla Press, 599-612. Zubizarreta, María Luisa. 2007. “Aspects of the Left Periphery in Spanish.” Paper presented at the Linguistic Symposium of Romance Languages. Zubizarreta, María Luisa. 2001. “The Constraint on Preverbal Subjects in Romance Interrogatives. A Minimality Effect.” In: Pollock, Jean Ives and A. Hulk (eds). Subject Inversion in Romance and the theory of universal grammar. Oxford: Oxford University Press. Zubizarreta, María Luisa. 1999. “Word Order in Spanish and the Nature of Nominative Case.” In: Johnson, Kyle and Ian Roberts (eds.). Beyond Principles and Parameters, Kluwer Academic Publishers, 223-250 (This is an English version of Zubizarreta, 1994). Zubizarreta, María Luisa. 1998a. Prosody, Focus and Word Order. Cambridge, Massachusetts: MIT Press. Zubizarreta, María Luisa. 1998b. "The Structure of the Higher Middle Field: the Position of the Verb and the Subject" (Handout) Colloquium on Romance Inversion sponsored by the University of Amsterdam, May 1998. Zubizarreta, María Luisa. 1994. "El orden de palabras en español y el caso nominativo.” In: Demonte, Violeta (ed). Nueva Revista de Filología Hispánica VI: 29-41.

Abstract (if available)

Abstract This dissertation is an investigation of two interrelated clausal left edge properties in the Spanish syntax: the positional effects of left edge elements (akin to the Tobler Mussafia and Wackernagel effects) and the availability of constituent dislocation. I focus on the syntax of preverbal lexical subjects and their relation to the several positions in the clause. Based on the projectionist approaches developed by Rizzi (1997), Poletto (2000), among others, this dissertation sketches a more reduced version of the complementizer system dubbed the Narrow Left Edge. Accordingly, the left periphery is composed by two core projections: the Point of View Projection (CP) and the Assertion Projection (c*P). In particular, I claim that the phenomenon of second position placement is manifested in Modern Spanish (cfr. Suñer 1994 and Fontana 1993) and must be related to the licensing of the c*P projection in declaratives and nondeclaratives. In addition, the mechanics of word order dislocation shows that constituents can be displaced to the left edge by movement or by base generation. These two effects can be explained if an Agree relation is postulated between the Assertion Phrase c*º in the lower left edge, and Tº in the inflectional layer.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The Spanish feminine el at the syntax-phonology interface

PDF

Copy theory of movement and PF conditions on spell-out

PDF

A reduplicative analysis of sentence modal adverbs in Spanish

PDF

Perspective in Turkish complementation

PDF

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

PDF

Machos y malinchistas: Chicano/Latino gang narratives, masculinity, & affect

Asset Metadata

Creator Beas, Omar (author)

Core Title Agreement on the left edge: the syntax of left dislocation in Spanish

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Linguistics

Publication Date 07/30/2009

Defense Date 04/19/2007

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

Tag dislocation,first/second position,left edge,OAI-PMH Harvest,preverbal lexical subjects,Spanish declaratives,Spanish interrogatives

Language English

Advisor Zubizarreta, Maria Luisa (committee chair), Hoji, Hajime (committee member), Saltarelli, Mario (committee member), Vergnaud, Jean-Roger (committee member), Zioga, Patricia Schneider (committee member)

Creator Email obeas@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m722

Unique identifier UC1203870

Identifier etd-Beas-20070730 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-535629 (legacy record id),usctheses-m722 (legacy record id)

Legacy Identifier etd-Beas-20070730.pdf

Dmrecord 535629

Document Type Dissertation

Rights Beas, Omar

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu