Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
A reduplicative analysis of sentence modal adverbs in Spanish
(USC Thesis Other)
A reduplicative analysis of sentence modal adverbs in Spanish
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
A REDUPLICATIVE ANALYSIS OF SENTENCE MODAL ADVERBS IN SPANISH by Héctor Velásquez Chafloque ______________________________________________________________________________ A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA in Partial Fulfillment of the Requirements for the Degree of DOCTOR OF PHILOSOPHY (HISPANIC LINGUISTICS) Los Angeles, U.S.A. August 2014 © 2014 Héctor Velásquez Chafloque All rights reserved ii To my parents / A mis padres, Luis y Martha, con amor y gratitud. iii Acknowledgements This dissertation exists because of the felicitous confluence of good intentions and concrete help coming from many people and several places at different times. While I am aware that this section may seem long to the reader—obviously, you can skip it—, I also know that I just want to express my thankfulness as sincerely and extensively as I can. First and foremost, this work is the result of the inspiration and guidance I received during my years at USC from my advisor, Maria Luisa Zubizarreta. Her classes, her published research, and her personal advice have motivated the main ideas for this project from the very beginning to the very end. Beyond that, Maria Luisa’s patience has been put to the test more than once, particularly every time I got disheartened—not an uncommon case, I must say—by obstacles of varied nature. I deeply thank Maria Luisa for her constant support, both in academic and human terms. I also thank Roumyana Pancheva and Mario Saltarelli, also members of my dissertation committee. Roumy and Mario have offered ideas and inspiration to this work, and have leaded me to get something valuable from some apparently obscure initial intuitions. Their lucid advice has helped me to sharpen many proposals in this work. I definitely appreciate their friendly and always supportive way of providing assistance. My gratitude goes to the faculty of the Department of Linguistics of USC. First, I am grateful to James Higginbotham and Jean-Roger Vergnaud, who will always be in my thankful memories. I also thank Hagit Borer, Ed Finnegan, Louis Goldstein, Elena Guerzoni, Hajime Hoji, Elsi Kayser, Ania Lubowicz, Barry Schein, Andrew Simpson, and Rachel Walker. I give special thanks to the gran maestra Carmen Silva-Corvalán, who kindly offered me her help and friendship from the very first day at USC. Each and all of my professors have taught me valuable lessons, not only about their respective fields, but also—via their direct example—about linguistic research and teaching at the highest level. iv Thanks also go to my own friendly cohort, formed by Mary Byram-Washburn, Xiao He, Canan Ipek, Lucy Kyoungsook Kim, David Li, Kathy McKinney-Bock, Sarah Ouwayda, Ben Parrell, and Erika Varis-Doggett. It has been my privilege to be part of this fantastic group. Additionally, I want to thank some friends of mine from other Linguistic cohorts: Janet Anderson, Andrés Benítez-Pozo, Arunima Choudhury, Alfredo García-Pardo, Mythili Menon, Iris Ouyang, Magdalena Pire-Schmidt, Sergio Robles-Puente, Michael Rushforth, Laura Tejada, Barbara Tomaszewicz, and Xin Zhao. Their company and joyful friendship has always motivated me, especially during the hardest times. I am also thankful to the Peruvian USC Linguistics alumni Alfredo Arnaiz, Omar Beas, Mónica Cabrera, Álvaro Cerrón-Palomino, and Liliana Sánchez. They have generously shared their experiences, and have been available for me whenever I need their advice or support—in the case of Monica, even when I needed some housing. I want to extend these thanks to José Camacho, also a USC Linguistics alumnus who, together with Liliana Sánchez (both current faculty members at Rutgers University), helped and encouraged me during these study years in the U.S.—and even before, indeed. I have also received valuable help from the staff of the USC Department of Linguistics. In particular, I want to thank Joyce Perez, who has been a really great student advisor during most of the time I have been a student in the Department. I am direct witness that Joyce has always been willing to walk the extra mile for whoever asked for it. I also thank Brandon Washington, who assumed Joyce’s position in my final year as a student at USC. Talking about my final year, I am deeply grateful to Dr. Jane Cody, Associate Dean for Academic Programs at Dornsife College, for offering me her support and trust in moments when I was particularly in need for them. The same deep level of gratitude, and for the same reasons, goes to Gayle Vierma, director of the USC Spanish Language Program during most of the years I have taught there, and to Dan Bayer, executive director of the USC Language Center. It is just fair to say that, without Gayle’s and Dan’s generous support, this dissertation would have not been possible. Many thanks to both! v During my years as instructor at the USC Department of Spanish, I have shared many experiences with senior lecturers and other instructors from different Departments. With an open heart, they offered me their help and friendship. For that, I thank Gloria Arjona, Vianey Cabrera, Lorena Gallego, Anahit Hakoupian, Michelle Har-Kim, Ana Paulina Lee, Verónica Medda, Alex Montes, Andrea Parra, Charles Paus, Payton Phillips, Yubane Sánchez, Consuelo Sigüenza- Ortiz, Katrina Spencer, Liana Stepanyan, and David Zarazúa. Additionally, I extend my thanks to the Department staff: Amelia Acosta, Sandra Hoof, Debbie Romero, and Jesus Solano. Sincere thanks also go to Martha Galván, whose smiling kindness will always be remembered. For her friendly help and encouragement during all these years, I also want to thank Bertha Arce, from the Business Office of Dornsife College. Talking about business, I am grateful to the Del Amo Foundation for their financial support during the academic year 2011- 2012, and also for funding a study trip to Spain during the summer of 2011. I also thank Chad Vicenik, from UCLA, who introduced me to the ToBI prosody annotation system in the summer of 2010. Also from UCLA, I am grateful to Dr. Sun-Ah Jun, who accepted me as a free attendant to her course of Prosody in the spring of 2012. I extend this thanks to her assistants at that time, Jason Bishop and Marc Garellek. I must say that my currently knowledge about the prosody of focus—which allowed me to successfully prepare my qualifying paper—is due, in a significant part, to the generosity of UCLA researchers and instructors. Many thanks to all of them! In previous stages of the present dissertation, I have benefited from comments and observations from the audiences at the LSA Annual Meeting in 2011, the Hispanic Linguistics Symposium in 2012, and various meetings of Syntax-Plus held between 2012 and 2014 at USC. I give thanks to all of them. Also I am grateful to Violeta Demonte, Jairo Nunes, Pilar Prieto, Tim Stowell, and Henk van Williams, for their advice at different stages of preparation of this work. And although we have never met in person, I am thankful to Chih-hsiang Shu, with whom I shared the interest about sentence adverbs, and who freely shared with me his dissertation on that topic (Shu 2011)—certainly, a careful visit to the Kingdom of Agree remains pending in the present work. Naturally, I assume the full responsibility for all mistakes and shortcomings. vi I am also deeply grateful to my professors, family and friends in Peru. First, I want to thank my professors Luis Jaime Cisneros, José Luis Rivarola, and Enrique Carrión Ordóñez. Although already departed, they will always live in my memory and will always be present my teaching activity. Sincere thanks also go to Rodolfo Cerrón-Palomino and Mario Montalbetti. Coming from different interests and perspectives, they both have stimulated my initial curiosity about Linguistics and have guided my first steps in teaching and research in the field. I am also thankful to the Department of Humanities of the Pontifical Catholic University of Peru for their support, both in financial and moral terms. In particular, I want to thank Pepi Patrón and Miguel Giusti, who occupied consecutively the position of chair of the Department during my saty in the U.S. Thanks also go to the staff of the Department, in particular, to my good friends Dolly Trujillo and Myrian Maldonado (¡Gracias, amigas!). I give special thanks to Jorge Iván Pérez Silva, my good friend since the very early ages. At this point in my life, Jorge’s generosity towards me is already a true legend. Now again, just like in those days far ago, his unlimited friendship has a crucial role in the success of my most important projects. Also I want to thank the support—of several kinds—that I have received, at different times and places, from my Peruvian friends Alexandra Barrionuevo, Rafael Camacho, Paola Cépeda, Karen Coral, Claudia Crespo, Ernesto Cuba, Roberto García, Kelly Gómez, Daniel Holgado, Carlo Linares, Natalia Matta, Maritza Menéndez, Andrea Patriau, Luisa Ramírez, Leticia Robles, Miguel Rodríguez-Mondoñedo, Daniel Salas, Elizabeth Tavera, and Teresa Torres Bustamante. I am truly happy to have been able to count on their friendship. I express my deep gratitude to Elizabeth Reyes, from USC Counseling Services, and to Cecily Roberts, from USC Occupational Therapy Faculty Practice. At different crucial moments of my life as a student—as a human being, indeed—both of them have helped me decisively to overcome a certain kind of invisible but real obstacles. vii I also thank the doctors from the the USC Ostrow School of Dentistry, for their dedication to my complicated case: Manmeet Arora, Johnny Avestian, Mohammed Mashyakhy, Vi Nguyen, and Sowmya Rajagopal. I extend my gratitude to the faculty of the school that oversaw my process. I am grateful to the USC Thornton School of Music for all I learned while singing in the USC Apollo Choir. Thanks to the former conductors Dominic Gregorio, Christopher Eanes, Christopher Haygood, Brandon Brack, Christopher Gravis, E. Jason Armstrong, Yohan Partan, and Tammy Alderman. Thanks also to all singers and accompanists. The fun and the excitement of rehearsals and presentations are already part of my most beloved memories. I have been fortunate to have had the opportunity to practice and share my knowledge about Spanish language with four tutored adult students, namely Christina Faegri, David Lloyd, Robby Mockler, and Gillian Pears. I am thankful for their warm, friendly attitude and, certainly, for all I have learned from each of them. I want to thank the 8 a.m. mass choir at Saint Agnes church, conducted by my good friend Orinio Opinaldo, and also formed, at different moments, by Juan Acuña, Lita Alfon, the Alvarado family—Mayra, Carlos, and their son Henry—, Louie and Ching Arzadon, Nancy Davis, the Hernández family—Elvira and her children Jessica and Álvaro—, Miranta Louis, Flor Nocum, Chito and Josie Troncoso, and Armando and Luz Villanueva. I thank each and all of them. Very special thanks go to Nema Damaso and her children Diana and Charlton James (a.k.a. ‘C.J.’). Thanks also to the OneLA group. For all that I learned and all that we shared, I thank Carmen Amaya, Norma Castillo, Fr. William Delaney, Carolina García, Tom Holler, Robert Hoo, Alejandro Huerta, Mary Jackson, Sr. Maribeth Larkin, Clementina López, Marcelo Martínez, Rocy Reynoso, LAPD Senior Lead Officer Geraldine Vazquez, Lambreni Waddell, and Ethan Weiss. Special thanks go to doña María Paz, whose example of generous dedication to the cause of social justice will always be alive among the members of St. Agnes community. viii I also thank my friends Kirk Bruner and James Carey for their generous hospitality, particularly during Thanksgiving Day—wherever I go, I will remember them every fourth Thursday in November. Thanks also to All Peoples Church, especially to my good friends Robert and Aeros Pierce, excellent both as musicians and as human beings. Thanks also to Santos Avilés, my landlord during all these long-but-short years. Now I want to thank all my family. My parents, Luis and Martha, have been patiently waiting for this to happen for decades so far. Many thanks go to them for their love and their continue support, which has literally lasted all my life. For these, and a million reasons more, this dissertation is lovingly dedicated to them. Thanks also to my son Alejandro José for his love, care and understanding. I also sincerely thank Carla Barrionuevo, Alejandro’s mom. I am grateful to my brother Luis Martín and to my sister Rosa María, just for everything. Warm thanks also go to Paco Chafloque, more a soul brother than an uncle. In Florida, I send my total gratitude to my primísima Natalý Bohn and to her mom, my madrina Socorro Fernández. They have done all in their hands to welcome me in the U.S. and have joyfully offered me their support at all times. My family’s help, in every possible aspect, has been decisive for me to arrive to this final point on the road. Thanks to the good man at the corner of Willow St. and Long Beach Blvd., one Saturday morning by the end of 2013. Somehow, I know that his true blessing has a lot to do with the completion of this work. Finally, I give thanks to all my former, current, and future students, here and there, now and beyond. I am grateful for their time and patience. For me, being a teacher has been—and, I hope, it will always be—a pleasure, a responsibility, and a true privilege. Queridos Cola y Valicha: estas primeras palabras en castellano son para ustedes. Gracias por todo lo que me han dado, que es simplemente inmenso, tanto en lo material como en lo intangible. Ustedes dos son mis maestros en el Amor, y el Amor, al fin y al cabo, es lo único que verdaderamente importa. ix Querido Choche: muchas gracias por todo. Espero que ahora podamos pasar juntos mucho tiempo para compensar todos estos años de separación obligada entre un viejo piapiá y un su querido hijito. Estimada Elvira: muchas gracias por su insistente pregunta dominical, que siempre me hacía consciente del paso irreversible de las semanas. Y, pues, ya vio usted que mi “ya merito” resultó siendo cierto. Muchas gracias también por asistir a mi defensa. Mis queridos compañeros Magdalena y Sergio: creo que, a estas alturas, no necesito decirles que, sin ustedes, simplemente no la habría hecho. Así nomás, tal cual. Y sea que lo diga o no, ustedes saben que es cierto. Por eso, amigos, les agradezco desde el fondo del corazón. Seguramente nos volveremos a ver algún día, quizá pronto. Entonces, nos juntaremos para reírnos en buena onda de lo que, en otro tiempo, tanto nos estresó. ¡Un abrazo grande! x Table of contents Dedication ii Acknowledgements iii Abbreviations xii Abstract xvii Chapter 1. Fundamental issues 1 1. Sentence adverbs 1 2. Sentence modal adverbs 8 3. Modal adverbs in Cinque’s Universal Hierarchy (1999) 10 4. A semantic-based proposal for the ordering of adverbs (Nilsen 2003) 15 5. Truth conditions in modal adverbs 20 6. A distributional challenge for sentence modal adverbs 22 7. Information structure: focus and presupposition 29 7.1. Definitions 29 7.2. Assertion structure 30 7.3. Focus and question-answer congruence 32 7.4. Set of alternative focus values vs. set of alternative propositions 34 7.5. Informative focus vs. contrastive focus 35 8. Summary 38 Chapter 2. Theoretical framework 42 1. Displacement 43 2. The GB view of movement: traces, ECP, remnant movement 45 2.1. The GB theory 45 2.2. Traces 48 2.3. Remnant movement 50 3. A remnant-movement analysis of association with focus 58 xi 4. Movement as Copy+Merge. Scattered complementary deletion 61 4.1. The minimalist program 61 4.2. The Copy Theory of Movement (CTM) 66 4.3. Some advantages of the CTM over traces 67 4.4. Linearization of chains and the LCA 73 4.5. Chain Reduction and Formal Feature Elimination 74 4.6. Scattered deletion 78 4.6.1. Order of clitics in Bulgarian (Bošković 2001) 78 4.6.2. English stress assignment and tone-sandhi in Taiwanese (Vergnaud 2014) 80 4.6.3. English extrapositions (Wilder 1995) 82 4.6.4. XP-split constructions in German and Croatian (Fanselow & Ćavar 2002) 84 4.6.5. R- and Ha-constructions in Korean (Jo 2013) 90 5. Summary 94 Chapter 3. Analysis. Reduplication and ellipsis in sentence modal adverbs in Spanish 97 1. Reduplication and ellipsis in Spanish 98 2. Main proposal. The [+foc] feature 105 2.1. An inventory of resources 105 2.2. ‘Transportability’ as focus sensitivity 106 2.3. The bi-clausal derivation. Basic analytical remarks 111 2.4. Association with focus-containing phrase (fcP) 114 2.5. Linearization and the Inclusiveness Condition 120 3. Preliminary applications of the bi-clausal reduplicative analysis 125 3.1. Short answers 125 3.2. Focus association at a distance 128 3.3. Setence modal adverbs in different functional positions 131 3.4. The problem with F-marking of subjects in situ 137 3.5. F-marked subjects with prosodic prominence: the structural view 139 xii 3.6. The reduplicative structure, HT and CLLD constructions 141 3.7. Sentence modal adverbs in sentence-final position 148 4. Summary 152 Chapter 4. Possible extensions of the main proposal I: Reduplication with other focus sensitive units 154 1. The reduplicative structure (RS) 154 2. The RS and other focus sensitive units 157 2.1. Sp. sólo ‘only’ 157 2.2. Sp. también ‘also’ and incluso ‘even’ 165 2.3. Sp sólo, también, incluso: Some similarities and differences 168 2.4. Sp. probablemente ‘probably’ and sólo ‘only’ in the same sentence 175 3. Other –mente ‘-ly’ sentence adverbs, and negation in Spanish 184 3.1. Other –mente ‘-ly’ sentence adverbs 185 3.2. Focus and negation 198 4. Marked word order in Spanish. The RS approach 204 5. Summary 225 Chapter 5. Possible extensions of the main proposal II: Copy+Merge in certain focal constructions 228 1. Focalizing-SER (‘to be’) in Caribbean Spanish. The RS approach 229 1.1. The phenomenon 229 1.2. The RS analysis of FS 232 2. Considerations on copular structures in Spanish. The ‘Hausa-Spanish connection’ 239 3. Revisiting pseudo-clefts 249 3.1. PsCls: a CTM-inspired proposal 249 3.2 Copies without chains? 264 3.3 Another possibility: a total copy of CP 273 4. It-clefts: just a tentative proposal 275 5. Pseudo-cleftsin colloquial Peninsular Spanish (CPS) 292 xiii 6. Summary 305 Chapter 6. Conclusions and some projections for future research 309 References 311 xiv Abbreviations 1 First person 2 Second person 3 Third person ACC Accusative [Case] Alt-fv Set of alternative focus-values Alt-p Set of alternative propositions AS Assertion Structure C-NSR Constituent-driven Nuclear Stress Rule ChR Chain Reduction CISD Chain Internal Selective Deletion Cl Clitic CLLD Clitic Left Dislocation [construction] COP Copula CPS Colloquial Peninsular Spanish CTM Copy Theory of Movement DO Direct Object e-COP Empty Copula ECP Empty Category Principle Eq Equative GB Government and Binding [Theory] DAT Dative [Case] DECL Declarative ECP Empty Category Principle EPP Extended Projection Principle F Focus F-marking Focus Marking fcP Focus-containing Phrase FDC Focus Deletion Constraint xv FFE Formal Feature Elimination FPR Focus Prominence Rule FocP Focus Phrase FS Focalizing SER [construction] FSU Focus Sensitive Unit Ft Feature(s) FUT Future tense HON Honorary [morpheme] HT Hanging Topic [construction] id-COP Identity/Identificational Copula IND Indicative [mood] IO Indirect Object ip Intermediate Phrase IP Intonational Phrase [in prosody] / Inflexional Phrase [in syntax] IS Information Structure L1 First Language/Mother Language LCA Linear Correspondence Axiom LF Logical Form (or semantic component) LOC Locative N Neuter NEG Negation NOM Nominative [Case] NPI Negative Polarity Item NS Nuclear Stress NSR Nuclear Stress Rule P&P Principles and Parameters [Theory] PF Phonetic Form (or phonetic component) PG Parasitic Gap PISH Predicate-Internal Subject Hypothesis PL Plural PLD Primary Linguistic Data xvi PMC Parallel Movement Constraint PPI Positive Polarity Item PRS Present tense PsCl Pseudo-Cleft [construction] PST Past tense RS Reduplicative Structure S Sentence S-NSR Selection-driven Nuclear Stress Rule Sbj Subject SBJV Subjunctive [mood] SG Singular SM Structurated Meaning TAM Tense, Aspect, Mood (bundle of verbal-related features) TOP Topic [morpheme] TS Tone Sandhi UG Universal Grammar Unit X Unit X is deleted at PF (single strikethrough) Unit X Unit X is deleted at LF (double strikethrough) xvii Abstract The main problem that this work deals with is the question about the syntactic representation one must attribute to sentence modal adverbs in Spanish (for instance, probablemente ‘probably’) when they surface in low sentence positions. Starting by the verification of the propositional scope of this kind of adverbs, the fact that they can surface adjacent to sub-clausal units, as vP, DP, AdjP, etc. seems puzzling. Whereas the adverb probablemente is visibly located in a syntactic position with clausal scope in the sentence Probablemente, Juan comprará un carro ‘Probably, J. will-buy a car’, this is not the case in the sentence Juan comprará un carro probablemente rojo ‘J. will buy a car such that its color will probably be red’. First, I observe that the adjacency between the adverb and the sub-clausal unit is triggered by focus sensitivity: the adverb appears immediately adjacent to the focus marked unit. This adjacency may be a mechanism in virtue of which the left edge of the unit in focus gets unambiguously marked. I propose that the adverb is always base generated in a high sentence position, but focus sensitivity on the adverb may trigger later reduplication whenever a unit is focus-marked within its c-command domain. I claim that the unit to be copied is the clause c- commanded by the sentence modal adverb. This clause is then copied and re-merged in the Spec position of the maximal projection of the adverb. I anchor my proposal in the Copy Theory of Movement (Nunes 2001, 2004). This theory interprets movement as Copy+Merge. Crucially, the ‘original’ clause and its clone form a chain. Linearization then enforces a process of chain reduction, involving deletion. In this case, ellipsis applies on the lower copy, but cannot delete the unit in focus, since an information relevant for interpretation would be lost (Copy cannot reproduce F-marking). As a result, the modal adverb ends up by surfacing adjacent to the F-marked unit. On the basis of this proposal, I extend the reduplicative structure to other cases of focus-related words and constructions. KEY WORDS: Copy Theory of Movement, Ellipsis, Informative focus, Modal adverbs, Reduplication, Sentence adverbs 1 Chapter 1 Fundamental issues This chapter is intended to present the fundamental issues that this work deals with. In section 1, I introduce sentence adverbs and VP-adverbs as two different kinds of syntactic units with regards to their respective scopes (Jackendoff 1972). To establish this difference, some criteria are offered. In section 2, sentence modal adverbs are introduced. These are presented as part of a more general class of adverbs, namely speaker oriented adverbs. As such, sentence modal adverbs are ordered with respect to subject oriented adverbs, manner adverbs, and other adverbial sub-classes (Alexiadou 1997). Section 3 deals with the position of sentence modal adverbs in the context of a richly articulated hierarchy of adverbs that has been proposed as universal (Cinque 1999). In section 4, an alternative (or complementary) way for ordering of adverbs is offered. This alternative takes into account the semantic content of sentence adverbs, and capitalizes on the fact that they can be analyzed as polarity items (Nilsen 2003). Section 5, in turn, presents a brief consideration about the semantics of modals, and establishes that sentence modal adverbs are always propositional in scope (Chierchia & McConnell-Ginet 2000). In section 6, I present the main problem for this work, namely the fact that, being propositional in scope, sentence modal adverbs can also appear grammatically in low sentence positions, adjacent to sub-clausal units. Preliminary, I outline an analysis of these cases that resorts to the notion of focus (Krifka 2008a) and movement as copy and re-merge. Section 7 presents the information notions of focus and presupposition, as they are to be understood in this work. Section 8 summarizes the chapter. 1. Sentence adverbs Adverbs have been known since old Greek grammars as modifiers of verbs. Indeed, the word adverb itself refers to this possibility. It comes from Latin ad verbum ‘next to the verb’. Flavius Charisius, a Latin grammarian from the 4th century, wrote, Adverbium est pars orationis quae adiecta verbo significationem eius inplet atque explanat ‘The adverb is the part of a sentence such that, when located next to a verb, completes and expands its meaning’ (Chuaqui Farrú 2 2006: 128) 1 . Taken as a class, adverbs were also known by traditional grammars to be modifiers of adjectives or other adverbs. Let’s illustrate these three possibilities for Spanish. (1) a. María canta bonito. ‘M. sings well.’ b. Es una muy buena película. ‘It is a very good movie.’ c. Pedro baila realmente bien. ‘P. dances really well.’ It has also been observed that some adverbs can modify sentences. (2) a. Usualmente, Juan canta bien. ‘Usually, J. sings well.’ b. Afortunadamente, es una muy buena película. ‘Fortunately, it is a very good movie.’ c. Sinceramente, Pedro dances muy bien. ‘Sincerely, P. baila very well.’ In (2a-c), the words usualmente, afortunadamente, and sinceramente can be considered, intuitively, as modifiers of the whole sentence that follows, hence the label sentence adverbs. 2 In this vein, sentence adverbs can be seen as lexical items that take syntactic scope over a constituent with propositional content, as opposed to other adverbs that may take scope over a 1 The Latin word adverbium, in turn, is a direct translation from the Greek word epírrēma (έπίρρημα), with identical meaning. 2 For a formalization of this intuition within the generative grammar framework, see Shu (2011: 21 ff). On the other hand, the terms sentential adverbs and sentence adverbs have been used as synonyms. I will mostly use the latter from now on. The term adverbial, in turn, will be used here to refer to any phrase (for instance, PPs) with the same distribution as adverbs or AdvPs. 3 sub-clausal constituent, tipically VP. 3 For instance, the adverb evidently in (3a) is considered a sentence adverb, while the adverb easily in (3b) is labeled as a (manner) VP-adverb. (3) a. Evidently Horatio has lost his mind. b. Stanley easily ate his Wheaties. (Jackendoff 1972: 50). On syntactic grounds, the difference of scope is related to the structural position. In (3a), evidently appears in a high position, presumably within the CP domain, from which it c- commands the following sentence. In (3b), the adverb easily appears in a sentence-internal position, from which it cannot take scope over the whole sentence, but from which it is able to c- command the VP ate his Wheaties. The basic structural descriptions for (3a,b) are given in (4a,b), respectively. (4) a. [ CP Evidently [ TP Horatio has lost his mind]] (Sentence adverb) b. [ TP Stanley [ VP easily [ VP ate his Wheaties]]] (VP-adverb) So far thus, the difference between sentence adverbs and VP-adverbs seems relatively straightforward. Up to this point, it looks like the difference between these two types of adverbs is due to some selectional feature on the adverb, which would come already specified from the lexicon either as sentential (high) or as verbal (low). In this vein, for instance, evidently cannot be used as VP-adverb, as the example in (5a) shows. Although the adverb appears after the verb in (5b), the comma intonation is an index of the external position of that adverb with respect to the rest of the sentence. (5) a. * Horatio has lost his mind evidently. b. Horatio has lost his mind, evidently. On the other hand, as (6a) shows, easily has no sentential scope. The basic structure in (6b) shows that the adverb has been left-dislocated, presumably to a topic position (or similar). 3 For the time being, I make no difference between VP and vP. 4 These cases suggest that the adverb may be lexically marked with a feature that identifies it either as sentence adverb or as VP-adverb. (6) a. ? Easily, Stanley ate his Wheaties. b. [ TopP Easily i [ TP Stanley [ VP ate his Wheaties t i ]]] There are some adverbs, however, that seem ‘unmarked’, since they can grammatically appear in either as sentential or as verbal modifiers. Take for instance the speech-act adverb frankly. As examples (7a,b) show, this adverb can appear either as sentence adverb or as manner adverb. The difference in scope can be attributed to the different position the lexical item occupies in the syntactic structure. (7) a. Frankly, I didn’t like the movie. (sentence adverb) 4 b. John spoke frankly with his wife. (VP-adverb) Another subclass of sentence adverbs known as subject-oriented adverbs, like cleverly, intelligently, clumsily, stupidly, etc. offers and interesting illustration of the aforementioned lexical ‘unmarkedness’ of some sentence adverbs (Jackendoff 1972: 57). (8) a. Cleverly, John answered the questions. b. John answered the questions foolishly. c. Cleverly, John answered the questions foolishly. Example (8a) can be paraphrased as ‘It was clever of John to answer the questions’. This interpretation is consistent with the adverb cleverly taking scope over the sentence that follows; thus cleverly is used as a sentence adverb. In (8b), in turn, the adverb foolishly expresses the manner in which John answered some questions; its paraphrase can be ‘John answered the questions in a fool manner’. Here, the adverb takes scope over the VP only; thus foolishly is used as a VP-adverb. Interestingly, (8c) combines both adverbs in a single sentence without 4 It’s been suggested that, in its sentential use, frankly is related to a ‘high’ speech-verb (cf. frankly speaking). This speech-verb can be later deleted at PF: Frankly speaking, I didn’t like the movie (Velásquez 2011a). 5 contradiction, despite the fact that the lexical meanings of cleverly and foolishly are mutually opposed. It is possible to imagine a context where the clever thing for John to do is to answer some questions in a fool mode (maybe John is a spy, and he is being interrogated by the enemy). It is also possible to invert the positions for the adverbs in (8c), namely Foolishly, John answered the questions cleverly, and again no contradiction would arise. This fact indicates that, in spite of belonging to the same semantic subclass, and in spite of being antonyms, both adverbs cleverly and foolishly can co-occur grammatically in the same sentence since they belong to two different syntactic domains: (i) a high domain, as a subject-oriented adverb; and (ii) a low domain, as a manner adverb. This sub-class of adverbs then would not be specified as belonging to just one of two domains. Coordination of adverbs in –mente in Spanish provides further evidence of the existence of those two separate syntactic domains. Morphologically, the adverbs ending in –mente in Spanish form a class of adverbs that can be taken as equivalent to the –ly adverbs of English. 5 Similarly to English, these adverbs are built by adding an affix to an adjective, which functions as its morphological base. 6 Just like English too, some -mente adverbs appear as sentence adverbs (afortunadamente ‘fortunately’), some appear as VP-adverbs (completamente ‘completely’), and some others may appear alternatively either as sentence adverbs or VP- adverbs (francamente ‘frankly’). See examples of the former two classes in (9) and (10): (9) Afortunadamente, Juan limpió la casa. (Sentence adverb) Fortunately, J. cleaned the house. (10) Juan limpió la casa completamente. (VP-adverb) J. cleaned the house completely. 5 Those suffixes are equivalent to –a/ -os in Greek and –wiese in German (Alexiadou 1997: 3). 6 It is debatable, however, whether –mente is a real suffix or instead it is a “bound-word” in Spanish, a sort of intermediate step between a full lexical unit (the word mente ‘mind’, which is still in use in Modern Spanish) and a pure functional affix. Interestingly, -mente adverbs are the only words that allegedly bear two main stresses—the one on –mente and the one on the adjectival base; cfr. últimamente ‘lately’, with stress on syllables [ul] and [men]. 6 These two kinds of adverbs may co-occur in the same sentence as seen in (11a), but they cannot be coordinated. 7 This is seen in (11b,c). This latter fact clearly indicates that sentence adverbs and VP-adverbs belong to different domains. (11) a. Afortunadamente, Juan limpió la casa completamente Fortunately, Juan cleaned the house completely. b. * Afortunada y completamente, Juan limpió la casa. c. * Juan limpió la casa afortunada y completamente. Cfr. Juan limpió la casa rápida y completamente ‘J. cleaned the house quick(-ly) and completely’ In some languages, the different scope for the same adverb may be morphologically marked. Alexiadou (1997) reports the following contrast with the adverbial endings –a and –os in Modern Greek. (12) a. Zi ikonomika. (VP-adverb) lives economically ‘He/she lives economically.’ b. Ikonomikos, den pame kala. (Sentence adverb) Financially NEG go.1PL well ‘Financially speaking, we are not doing well.’ (Alexiadou 1997: 3) 7 When coordinated, these adverbs ‘share’ one common suffix –mente, which is pronounced only in the last adverb of the list: Juan limpió la casa fácil, efectiva y rápidamente ‘J. cleaned the house easy(-ly), effective(-ly), and quickly’. On this issue, see Alarcos (1999), and Kovacci (1999). 7 With adverbs that can appear in the two domains, it is possible in Spanish to build a ‘minimal pair’ to distinguish sentence adverbs and VP-adverbs. This is shown in (13a,b). On the other hand, (13c) combines the two possibilities in one single sentence. (13) a. No quise hablar francamente NEG I.wanted to.speak frankly ‘I didn’t want to speak openly’ b. No quise hablar, francamente ‘I didn’t want to speak, frankly/honestly’ c. Francamente, no quise hablar francamente ‘Frankly, I didn’t want to speak honestly’ Negation can also be used to distinguish sentence adverbs from verbal adverbs. In (14), the sentence adverb is not affected by NEG, whereas the VP-adverb in (15) is affected. (14) a. El doctor viajó, afortunadamente ‘The doctor traveled, fortunately’ b. El doctor no viajó, afortunadamente ‘The doctor didn’t travel, fortunately’ (15) a. El estudiante lee cuidadosamente ‘The student reads carefully’ b. El estudiante no lee cuidadosamente ‘The student doesn’t read carefully’ The syntactic node ‘polarity’, which encodes the opposition between affirmation and negation, has been seen as occupying the rightmost edge of the C-Domain for Modern Spanish 8 (Zubizarreta 2009). Polarity is at the “border line” separating the two big domains within the clausal structure, namely C-Domain and I-Domain. See the following phrase-structure schema, from Zubizarreta (2009). (16) [ Force [ Wh [ Top [ Pol [phi-P [ T [ Asp [ V … C-Domain I-Domain (Zubizarreta 2009: 343) The C-Domain is the domain “where sentence grammar meets discourse” (Zubizarreta 2009: 342). From left to right, this domain (i) encodes sentential force (declarative, interrogative, exclamative), (ii) attracts wh-phrases, (iii) hosts topics, point-of-view constituents, and discourse-linked focus, and (iv) includes polarity itself. On the other hand, I-Domain is “the inflectional domain of the verb”. It consists, from left to right, of (i) a bundle of phi-features, like person, number, and Det; and (ii) temporal elements, such as Tense and Aspect. If NEG is a reliable criterion to distinguish scopes, we can conclude that, syntactically, adverbs with sentential scope belong to the C-Domain, whereas adverbs with VP scope belong to the I-Domain. This idea is semantically supported for sentence adverbs, since their interpretation is related to sentential force (presumably), point of view (fortunately), or topic (economically). 8 2. Sentence modal adverbs I review now a specific class of sentence adverbs, namely sentence modal adverbs. These adverbs express “the speaker’s degree of confidence about the truth of the proposition (based on the kind of information he/she has)” (Cinque 1999: 86). Being related to truth, these adverbs have been further specified as epistemic modals. Von Fintel (2006: 2) defines epistemic modality as follows, “Epistemic modality (Greek episteme, meaning ‘knowledge’) concerns what is possible or necessary given what is known and what the available evidence is”. In turn, 8 Note that the PS schema in (16) predicts that NEG cannot take scope over force-related items, as for instance the modal adverb probablemente. This prediction is borne out. *No probablemente viajará vs. Probablemente no viajará ‘Probably (he) will not travel’. This can be related to the consideration of modal adverbs as PPIs, that is to say, units that require an affirmative environment to surface grammatically (see below). 9 Portner (2009: 4) considers adverbs like maybe, probably, and possibly as expressions of sentential modality. Some other adverbs belonging to this subclass in English are likely, presumably, and supposedly. As for Spanish, the adverbs probablemente ‘probably’, presumiblemente ‘presumably’, and supuestamente ‘supposedly’ are examples of sentence modal adverbs. In turn, Kovacci (1999: 755) includes in this class quizá(s), tal vez ‘maybe’ and posiblemente ‘possibly’. 9 Some of these adverbs are sensitive to quantitative modification (muy probablemente ‘very probably’), but not all of them (*muy supuestamente lit. ‘very supposedly’). In any case, since epistemic modality, by definition, expresses a certain “degree of confidence” hold by the speaker, those different degrees of confidence form a scale. 10 According to Kovacci (1999: 755), for Spanish, the adverb seguramente ‘almost for sure’ occupies the higher degree of confidence in that scale, while the adverb difícilmente ‘hardly’ (which can be quantitatively modified) corresponds to the lower degree. At this point, it becomes important to consider the syntactic position occupied by modal adverbs. While the traditional approach considered adverbs in general as non-selected adjuncts that can freely appear in different positions within the sentence, this view has been refined in more recent research, in particular with respect to their co-occurrence in the same sentence. Alexiadou (1997), for instance, observes that the adverbs co-occurring in (17), which belong to different sub-classes, “appear in specific scope relations which, when reversed, lead to ungrammaticality” (Alexiadou 1997: 9). (17) Probably, John cleverly frequently avoided Mary carefully. On the basis of previous work by Jakendoff (1972), Alexiadou proposes the following hierarchy for classes of adverbs. This order has been cross-linguistically attested both in terms of scope and linearization. 9 Additionally, Kovacci (1999) includes seguramente ‘almost for sure’, acaso ‘perhaps’, and difícilmente ‘hardly’. 10 For English, Higashimori (1978: 18) proposes a [0-1] continuum for the degree of confidence. Adverbs possibly and conceivably would be closer to the extreme 0, while probably and presumably would be closer to 1. Necessarily would correspond to the value 1. 10 (18) Speaker-oriented adverbs > Subject-oriented adverbs > Manner adverbs This general idea of a fixed order for the sub-classes of adverbs has been lately even more refined. In that vein, a very fine-grained universal hierarchy for adverbs has been proposed in Cinque (1999). 3. Modal adverbs in Cinque’s Universal Hierarchy (1999) Cinque (1999) presents a proposal according to which there is a fixed order of adverbs, applicable in all languages. Adverbs are supposed to occupy the Spec position of a universally ordered series of functional heads. Cinque presents cross-linguistic evidence in support of this proposal. This author starts by considering adverbs in Italian and French. The adverbs are divided in two groups: lower pre-verbal adverbs, and higher sentence AdvPs (Cinque 1999: 4 ff). 11 Cinque’s method consists of comparing pair of sentences in Italian and in French to show the relative fixed order of the adverbs therein included. Examples (19)-(23) illustrate three of these comparisons for lower adverbs, preceded by the conclusion demonstrated. Negative adverbs mica and pas precede già and déjà ‘already’, respectively. (19) a. Non hanno mica già chiamato, che io sappia. ‘They have not already telephoned, that I know.’ b. *Non hanno già mica chiamato, che io sappia. ‘They have already not telephoned, that I know.’ (20) a. Si tu n’as pas déjà mangé, tu peux le prendre. ‘If you have not already eaten, you can take it.’ b. *Si tu n’as déjà pas mangé, tu peux le prendre. ‘If you have already not eaten, you can take it.’ 11 A third group (lower pre-verbal adverbs in final position) is later reviewed, but it is finally subsumed into the first group as a mere ‘illusion’ attributed to focus-related movement. 11 Adverbs già and déjà precede the adverbs più and plus ‘any longer’, respectively. (21) a. All’epoca non possedeva già più nulla. ‘At the time, he did not possess already any longer anything.’ b. *All’epoca non possedeva più già nulla. ‘At the time, he did not possess any longer already anything.’ (22) a. A l’époque, il ne possédait déjà plus rien. ‘At the time, he did not possess already any longer anything.’ b. *A l’époque, il ne possédait plus déjà rien. ‘At the time, he did not possess any longer already anything.’ In all these cases, transitivity applies to precedence: if mica precedes già and già precedes più, then mica precedes più. (23) a. Non hanno chiamato mica più, da allora. ‘They haven’t telephoned not any longer, since then.’ b. *Non hanno chiamato più mica, da allora. ‘They haven’t telephoned any longer not, since then.’ The method is applied to the remaining lower adverbs, and also to the higher sentence adverbs to set their relative order, based on transitivity. That order is illustrated in (24)-(25); the first line shows the Italian adverbs, and the second line shows the equivalents in French. 12 (24) For the lower (pre-VP) AdvPs solitamente > mica > già > più > sempre > completamente > généralement > pas > déjà > plus > toujours > complètement > ‘usually’ NEG ‘already’ ‘more’ ‘always’ ‘completely’ tutto > bene tout > bien ‘all’ ‘well’ (25) For the higher (sentential) AdvPs francamente > fortunatamente > evidentemente > probabilmente > franchement > heuresement > évidemment > probablement > ‘frankly’ ‘fortunately’ ‘evidently’ ‘probably’ ora > forse > intelligentemente maintenant > peutêtre > intelligentement ‘now’ ‘maybe’ ‘inteligently’ These two orders can be replied in Spanish, with the exception of the particle NEG. (26) For the Spanish “lower” (pre-VP) AdvPs usualmente > ya > más > siempre > completamente > todo > bien ‘usually’ ‘already’ ‘more’ ‘always’ ‘completely’ ‘all’ ‘well’ 13 (27) For the Spanish “higher” (sentential) AdvPs francamente > afortunadamente > evidentemente > probablemente > ahora > ‘frankly’ > ‘fortunately’ > ‘evidently’> ‘probably’ > ‘now’ > tal vez > inteligentemente ‘maybe’ > ‘intelligently’ Cinque presents evidence of a similar ordering for adverbs in other languages with independent adverbs. Some languages show specific morphemes located in the head of the proposed functional projections. The author establishes a parallel fixed order for those morphemes. Finally, this author matches the hierarchy found for AdvPs with the hierarchy found for functional heads. The result of this is the following hierarchy of clausal functional projections, which is postulated as universal (Cinque 1999). (28) [ frankly Mood speech act [ fortunately Mood evaluative [ allegedly Mood evidential [ probably Mod epistemic [ once T(Past) [ then T(Future) [ perhaps Mood irrealis [ necessarily Mod necessity [ possibly Mod possibility [ usually Asp habitual [ again Asp repetitive(I) [ often Asp frequentative(I) [ intentionally Mod volitional [ quickly Asp celerative(I) [ already T(anterior) [ no longer Asp terminative [ still Asp continuative [ always Asp perfect(?) [ just Asp retrospective [ soon Asp proximative [ briefly Asp durative [ characteristically(?) Asp generic/progressive [almost Asp prospective [ completely Asp SgCompletive(I) [ tutto Asp PlCompletive [ well Voice [ fast/early Asp celerative(II) [ again Asp repetitive(II) [ often Asp frequentative(II) [ completely Asp SgCompletive(II) This sequence is, at the same time, hierarchical (since each functional head s-select its complements) and linear (because it imposes a fixed order for the AdvPs located in the specifiers). Three important characteristics must be attributed to this sequence: (i) transitivity, (ii) asymmetry, and (iii) connectedness. With respect to this issue, Nilsen (2003: 8) notes, 14 Suppose that we denote the relevant (descriptive) ordering relation of functional material as “≺” where x ≺ y iff x can precede y. Cinque’s approach leads one to expect ≺ to be a linear ordering, i.e. a transitive (∀x, y, z (x ≺ y ∧ y ≺ z) → x ≺ z)), asymmetric (∀x, y ((x ≺ y ∧ y ≺ x) → x = y)) and connected (∀x, y (x ≺ y ∨ y ≺ x)) ordering of classes of functional material. The hierarchy in (28) considers the modals probably, perhaps, and possibly as belonging to different functional projections, where the syntactic order probably > perhaps > possibly holds. 12 The adverb probably is labeled as epistemic modal, while perhaps is taken as modal irrealis-mood. Cinque indicates that, in Italian, forse ‘perhaps’ normally follows ora ‘now’ and allora ‘then’; this latter, in turn, follows probabilemente ‘probably’. Quoting Bellert (1977: 344), Cinque notes that, in English, probably cannot appear grammatically in a question like (29a), but perhaps can, as seen in (29b). (29) a. * Has Gianni probably have been there before? b. Has Gianni perhaps have been there before? To my own grammaticality judgment, Bellert’s observation applies to Spanish. The translations of (29) into Spanish are shown in (30). (30) a. * Ha estado Juan probablemente allí antes? 13 b. Ha estado Juan quizás allí antes? Similarly, Cinque (1999: 89) considers possibly as alethic modal. This author considers that alethic modality appears between epistemic modality and root modality, but no further 12 Interestingly, in (28), necessarily intervenes between probably and possibly. If this is the case, the syntactic hierarchy proposed does not parallel the scale of speaker’s degree of confidence, in which necessarily is located at the extreme of total epistemic certainty, i.e. possible < probably < necessarily. 13 In this work, I will not write the opening question mark mandated by Spanish orthography to avoid confusions with the question mark used to indicate uncertainty with respect to grammaticality. 15 explanation is offered. 14 Cinque further shows that possibly can appear grammatically in questions, as seen in (31b), whereas probably, which is an epistemic modal, cannot, as seen in (31a). (31) a. * Can he probably have said that? b. Can he possibly have said that? Cinque considers probably and its synonyms likely, presumably, and supposedly as a clear-cut epistemic modals. According to the definition cited above, these adverbs manifest a certain degree of confidence that some speaker has about the truth of the proposition he/she utters when the speech act takes place. 15 As seen in (28), Cinque’s hierarchy distinguishes around thirty different functional heads. It can be questioned whether this richness may be at odds with the task of acquisition of L1 by the child. From semantic grounds, it has been proposed that the rigid syntactic order imposed by Cinque’s hierarchy may be derived from a different source. This is the approach taken by Nilsen (2003), which I briefly explore in the following section. 4. A semantic-based proposal for the ordering of adverbs (Nilsen 2003) Nilsen (2003) proposes an alternative view to Cinque’s syntactic hierarchy such that, if tenable, would dispense with the long list of functional projections illustrated in (28), at least in part. The basic idea is that sentence adverbs are positive polarity items (PPIs), and this fact may be used to compute precedence relations between any given pair of sentence adverbs. 14 With respect to alethic modality, Von Fintel (2006: 2) observes, “Alethic modality (Greek: aletheia, meaning ‘truth’), sometimes logical or metaphysical modality, concerns what is possible or necessary in the widest sense. It is in fact hard to find convincing examples of alethic modality in natural language”. 15 The expression Es muy posible que S, lit. ‘It is very possible that S’ is not uncommon in Spanish. To my own judgment, this expression has a higher (or equal) degree of certainty than Es probable que S, ‘It is probable that S’. The negated adjectival version Es improbable que S ‘It is improbable that’ still has a minimal degree of certainty for S to occur, whereas Es imposible que S ‘It is impossible that S’ has 0 degrees of certainty. This indicates that, in Higashimori’s (1978) terms, probability will always be bigger than 0 (no matter how small), while possibility may eventually take the value 0 at its lower extreme. In practical terms, something improbable—even something very improbable—can still occur, whereas something impossible simply cannot happen. 16 Giannakidou (2008) defines PPIs as “expressions that are ‘repelled’ by negation and tend to escape its scope […] Expressions like some, already, would rather, and speaker oriented adverbs have been identified as PPIs in the literature” (Giannakidou 2008: 4). Just like negative polarity items (NPIs) demand a negative environment to appear felicitously, PPIs require an affirmative environment to surface grammatically. In other words, PPIs are infelicitous in non- affirmative contexts—like, for instance, those in (32). The example in (32a) is offered by Progovac (1994: 2), while the sentences in (32b-d) are taken from Giannakidou (2008: 4). 16 PPIs are transcribed in italics. (32) a. # Mary didn’t insult someone. b. # Bill wouldn’t rather be in Montpellier. c. # John isn’t here already. d. # John didn’t unfortunately die. In this vein, Bellert (1977) observed that speaker oriented adverbs (SpOAs) in English are degraded in environments that license NPIs (negative environments), that is to say in questions (33a), in antecedents of conditionals, (33b) in imperatives (33c), under negation (33d) under clause-embedding predicates like hope (33e), within the scope of monotone decreasing subject quantifiers like No N (33f). The common characteristic of these environments is that they are non-affirmative. (33) SpOA = {fortunately, evidently, honestly…} a. Did Stanley (*SpOA) eat the wheaties? b. If Stanley (*SpOA) eat the wheaties… c. (*SpOA) Eat (*SpOA) the wheaties! d. Stanley (SpOA) didn’t (*SpOA) eat the wheaties. e. I hope Stanley (*SpOA) ate the wheaties. f. No students (*SpOA) ate the wheaties. 16 See also Israel (2011) for a pragmatic approach to polarity items. 17 On the other hand, according to Bellert, SpOAs can appear in degree clauses (34a), in clause-embedders like think (34b), and ordinary declaratives (34c). All these imply affirmative environments. (34) a. Stanley was so hungry that he ( SpOA) ate wheaties. b. I think Stanley ( SpOA) ate the wheaties. c. Stanley ate ( SpOA) wheaties. Spanish SpOAs also behave as PPIs. See (35). 17 (35) SpOA = {afortunadamente ‘fortunately’, evidentemente ‘evidently’, honestamente ‘honestly’…} a. Comió (*SpOA) Esteban el cereal? ‘Did Stanley (*SpOA) eat the wheaties?’ b. Si Esteban (* SpOA) comiera el cereal… ‘If Stanley (*SpOA) ate the wheaties…’ c. (*SpOA) Come (*SpOA) el cereal! ‘(*SpOA) Eat (*SpOA) the wheaties!’ d. Esteban (SpOA) no (*SpOA) comió el cereal ‘Stanley (SpOA) didn’t (*SpOA) eat the wheaties’ e. Espero que Esteban (*SpOA) coma el cereal ‘I hope Stanley (*SpOA) ate the wheaties’ f. Ningún estudiante (*SpOA) comió el cereal ‘No students (*SpOA) ate the wheaties’ 17 In all these cases, a parenthetical reading for the SpOA should be avoided. About parentheticals, see below. 18 In (35) it is visible a basic parallelism between English and Spanish. The common idea is that, in both languages, PPIs cannot surface felicitously in non-affirmative environments. According to Nilsen (2003), from a semantic point of view, a high sentential expression as I think S creates an affirmative environment for S. This may seem counterintuitive at first sight. The idea is that if someone thinks S, he or she is taking S for real in his/her cognitive horizon (i.e. the proposition encoded by S is considered true). In contrast, the expression I hope S does not commit to the truth of S, so the expression creates a non-affirmative environment. Now if SpOAs are PPIs, the prediction would be that they can appear grammatically within the domain of I think but not in the domain of I hope. This prediction is confirmed in the examples in (36). (36) a. I think that fortunately John will come back today. b. # I hope that fortunately John will come back today. In (37), Spanish equivalent expressions show the same behavior, with a revealing additional change in the verbal inflection. The subordinate verb with Creo que S ‘I think (that) S’ is inflected in indicative mood, while with Espero que S ‘I hope (that) S’, the equivalent verb is inflected in subjunctive mood (typically associated with the irrealis mood). (37) a. Creo que ( afortunadamente) Juan regresará hoy. I.believe that fortunately J. will.come.back.IND today ‘I believe that fortunately Juan will come back today’ b. Espero que (* afortunadamente) Juan regrese hoy. I.hope that fortunately J. will.come.back.SBJV today Applying the same logic to sentence adverbs, Nilsen chooses, as an illustration of his proposal, the evaluative fortunately and the modal probably. Being sentence adverbs, both of them are PPIs. Nilsen notes that fortunately creates an affirmative environment (fortunately S 19 entails S) for its domain, while probably does not create an affirmative environment (probably S does not entail S) for its c-commanding domain. Things being so, the prediction is twofold: (i) fortunately can precede probably, since this latter would surface in an affirmative environment, and (ii) probably cannot precede fortunately, because this latter would be then in a non- affirmative environment, which is infelicitous for PPIs. As shown in (38a,b), both predictions are borne out. (38) a. Fortunately, John will probably come back tomorrow. b. * Probably, John will fortunately come back tomorrow. The equivalent sentence adverbs in Spanish further confirm the predictions. Now, since Spanish rejects two –mente adverbs appearing either consecutively or ‘too close’ from each other, an adverbial may be used to replace one of the adverbs (por fortuna ‘fortunately’). (39) a. Por fortuna, Juan probablemente regresará mañana. Fortunately J. probably will.come.back tomorrow ‘Fortunately, Juan will probably come back tomorrow’ b. * Probablemente, Juan por fortuna regresará mañana. Probably J. fortunately will.come.back tomorrow Nilsen’s idea at this point is that the kind of test applied in (38) for the pair {fortunately, probably} can be generalized for any given pair of adverbs belonging to Cinque’s hierarchy, so at least some of the ordering restrictions that Cinque attributes to different positions in the syntactic structure can be derived instead from the semantic value of the adverbs considered. Throughout this work, I will take Cinque’s hierarchy as a point reference, without necessarily assuming all its theoretical presuppositions. In any case, I consider that semantic scope should also be taken into account when deriving the relative order for adverbs. 20 As for sentence modal adverbs—the central object in this work—one aspect of its semantics is crucial: they express the speaker’s confidence with respect to the truth of a proposition. This means that, semantically, sentence modal adverbs have always propositional scope. This fact calls for a brief consideration of the way the truth conditions of modals are to be understood. 5. Truth conditions in modal adverbs Chierchia & McConnell-Ginet (2000: 257 ff) indicate that propositions can be understood as functions from circumstances into truth values. Circumstances, in turn, are presented as pairs <w, i>, where <w> stands for “possible world” and <i> represents “time”. These authors discuss the truth conditions applicable to (40)-(41). (40) It is probably Pavarotti that wrote the letter. (41) There is a slight chance that Pavarotti wrote the letter. As a first approximation, Chierchia & McConnell-Ginet (2000) note that there’s a difference between (40) and (41) with respect to the degree of confidence of the speaker: a speaker uttering (40) considers Pavarotti more likely to have written the letter than a speaker uttering (41). In other words, the degree of confidence of the speaker in (40) is higher than the degree of confidence of the speaker in (41) with respect to the truth of the proposition conveyed by the sentence Pavarotti wrote the letter. As the authors observe, What [(40)] and [(41)] seem to call for is a way of ranking the relevant possibilities. Some of them are more likely to occur than others. The criteria we use to rank the relevant situations can vary from context to context. In the case at hand we have a modal base determined by what we know about the circumstances in which the letter was written. But beyond that, we also have ideas about what normally is the case. What normally is the case can also be regarded as a conversational background, a stereotypical one. It describes worlds where only what is normal happens […]. A stereotypical conversational background will also 21 be a set of propositions. So we have a first conversational background that determines the relevant set of worlds and a second conversational background that provides us with standards that we can use to rank or impose an ordering on the relevant set of worlds. Intuitively, we can say that a world w in the modal base will be closer to our standards of normality than a world w’ iff w makes true more propositions in the stereotypical background than w’. (Chierchia & McConnell- Ginet 2000: 300) On the basis of these considerations, the authors present the truth conditions for (40) and (41) in (42) and (43), respectively. (42) Sentence [(40)] is true in <w, i> iff Pavarotti is the author of the letter in all the relevant situations (those compatible with what we know) that comes closest to our standards of normality. (43) Sentence [(41)] is true at <w, i> iff Pavarotti is the author of the letter in a situation that is relevant (or compatible with what we know) but not close enough to the standards of normality. Now considering that It is probable that Pavarotti wrote the letter is interpretively synonym of Probably, Pavarotti wrote the letter, the truth conditions for this latter sentence are also expressed in (42). Note that modal adverbs do not entail the truth of the proposition expressed by the sentence to which they modify. In interpretive terms, whenever an epistemic modal high adverb like probably merges with a clause, the truth conditions of the resultant construction is necessarily affected. Crucially, truth values can only be declared of propositions. Semantically, epistemic modal adverbs have an interpretive effect over propositions. In syntax, only sentences are able to convey propositional content. Thus we would expect that epistemic modal adverbs appear in syntax related to sentences only. In other words, from a semantic point of view, we can conclude that sentence modal adverbs are always clausal in nature. 22 6. A distributional challenge for sentence modal adverbs Keyser (1968: 367 ff) observes that the adverb immediately can validly appear in different positions in the context of a sentence. See (44). (44) a. Immediately John will send back the money to the girl. b. John immediately will send back the money to the girl. c. John will immediately send back the money to the girl. d. John will send back the money to the girl immediately. Working within the generative transformational framework, Keyser establishes that up to three transformational rules would be necessary to account for the set of possibilities in (44). However, this author states that “postulating three separate rules is undesirable”. Moreover, those rules would not be able to predict the ungrammaticality of some other sentence internal positions, like those in (45). (45) a. * John sent immediately back the money to the girl. b. * John sent back immediately the money to the girl. c. * John sent back the money immediately to the girl. In order to account for the observed facts, Keyser adds to his analysis a single rule that he names Transportability Convention. In Keyser’s words, “[t]his convention permits a particular constituent to occupy any position in a derived tree so long as the sister relationships with all other nodes in the tree are maintained”. In the following tree, taken from Keyser (1968: 367), “angle brackets indicate that if one of the possible structures is chosen, then all the others are automatically suppressed”. 23 (46) Sentence Adj Subj Adj Modal Adj Predicate Adj VP <immediate#ly> John <immediate#ly> will <immediate#ly> sent back the money to the girl <immediate#ly> The syntactic tree in (46) contains a n-ary node at the top. It relies on a transformational rule known then as adverb incorporation, along with a feature [+transportable] “which indicates that, at some point in the derivation, not necessarily at the point where the adverb is attached, the node must be interpreted under the transportability convention” (Keyser 1968: 368). In current versions of generative grammar, none of Keyser’s conventions are at use anymore. However, the main observation behind transportability, namely that some adverbs may appear grammatically at different points within a sentence, is still valid and worth exploring. Let’s take, for instance, the modal adverb probably. As examples in (47) show, this sentence adverb does surface in high sentence position (see 47a), but it can also appear in different low sentence positions (see 47b-d). (47) a. Probably, John will buy a new car next month. b. John probably will buy a new car next month. c. John will probably buy a new car next month. d. John will buy a new car probably next month. At the end of the previous section, we concluded that epistemic modal adverbs such as probably are clausal in nature. Following Chierchia & McConnell-Ginet (2000), we saw that these adverbs modify the truth conditions of a proposition. Being the case that, in syntax, 24 propositions can be expressed only by sentences, we would expect that epistemic modal adverbs appear adjacent only to clauses, that is to say, in a high position at the CP domain, as seen in (47a). Form these assumptions, sentences (47b-d) are puzzling. We see a sentence modal adverb appearing adjacent to T’ (47b), to VP (47c), and to an adverbial DP (47d). At least two facts are now in demand for explanation. On the one hand, we have lexical units with semantic propositional scope adjacent to sub-clausal constituents in syntax. On the other hand, we see ‘odd’ syntactic units like probably next month in (47d), which appear to be formed by an adverb plus a DP. More generally, it looks like sentence modal adverbs can freely merge with any kind of sub-clausal unit. Following Chomsky (1995), Larson & Segal (1995), Heim & Kratzer (1998), Chierchia & McConnell-Ginet (2000), among many others, I assume in this work that syntax provides well-formed structures for the semantic component to interpret. From this point of view, it is important to establish how come this component can assign an interpretation to these sentences, where a mismatch between semantics and syntax seems to be at stake. Syntax-semantics mismatches are, in principle, undesirable, considering the facts that (i) both aspects of knowledge are part of a broader unified cognitive system, (ii) they should be suitable for the child to learn, and (iii) they show many empirical convergent results (Larson & Segal 1995: 72 ff). Shu (2011) describes these mismatches for the more general case of mood-related expressions (a set that explicitly include sentence adverbs) as follows, Assuming structures to be uniform, there should be rigid correspondence between the semantic function of a linguistic expression and its syntactic structure, so all mood-related expressions should be encoded at C, but in the mainstream theories of a subset of mood-related expressions, including negation, sentential coordinators, and sentence adverbs, it is frequently proposed that these expressions are encoded on some lower categories. (Shu 2011: 4). 25 Now facing this challenge, let’s review, in broad terms, some possible lines of solution. One first line of explanation may be that, against our previous conclusion, sentence modals are not always propositional. One and the same adverb may be propositional at times and non-propositional at some other times, so these units would be underspecified with respect to their semantic type. This fact would license them to appear in unexpected positions. I consider this line of solution inadequate, since it would compromise the interpretive component. In this case, this component would have the burden of evaluating two things, without much aid from syntax: first, whether the adverb has propositional or non-propositional value, and, second, in this latter case, which type (in the sense of the semantic theory of types), among various available, corresponds to any specific position for the modal adverb. This would make more difficult the task of learning a language for the child. From this point of view, maintaining the idea that sentence modal adverbs are always clausal is more economical. If we don’t want to overload the semantic component, then a second line of explanation is available. This one would appeal to syntactic mechanisms. According to this idea, the modal adverb would be generated in a high CP position, but then it would move downwards to adjoin a sub-clausal unit. Now, from a derivational point of view, downward movement is not a valid option. Another problem would be the justification for this movement, either overt or covert, in terms of feature checking. In case we postulate that the adverb is generated in a low position and then moves upwards to the CP field, the problem would be that the adverb in low position would be base-merged with different types of constituents, with which the original problem of multi- type merge would remain unsolved. There’s still a third line of solution. This one takes into account information structure. More specifically, it takes into the picture sensitivity to focus. In this line, sentence adverbs are considered as focus-sensitive units (FSUs). In case a syntactic unit within the domain of the adverb is marked with focus, the sentence adverb associates with that unit, in a fashion somehow similar to what occurs with better-known focus-sensitive units only and even. 26 Krifka (2008a) illustrate the sensitivity to focus with the sentence adverb fortunately. In this case, the adverb associates with a focused adjective (marked with ‘F’) in the c-commanded sentence. (48) Fortunately, Bill spilled [white] F wine on the carpet. Krifka goes on saying, […] among the two alternatives, JOHN SPILLED RED WINE and JOHN SPILLED WHITE WINE, the latter one was more fortunate […]. Hence, fortunate expresses a comparison between the denotation of the focus and the denotation of the alternatives, a genuinely semantic use of focus (Krifka 2008a: 254). In turn, Shu (2011: 136) exemplifies this idea with the adverb probably: (49) a. Probably, John [likes] F Mary. b. Probably, [John] F likes Mary. In general terms, the interpretation of focus implies the evaluation of alternatives. In this case, sentences (49a,b) can be understood with respect to the alternatives in (50a,b), respectively. (50) a. Among alternatives such as LIKING, HATING, DESPISING, NOT CARING, etc., the first one is the more probable attitude John has of Mary. b. Among alternatives such as JOHN, PETER, JENNY, etc., the first one is the more probable person who likes Mary. Note that, although the linear string is exactly the same in (49a) and (49b), both sentences are different with respect to their respective informational structure. This is shown in (51), which makes explicit the kind of questions for which (49a) and (49b) are possible answers. 27 (51) a. Question: Well-formed answer: Ill-formed answer: What does John feel for Mary? Probably, John [likes] F Mary. * Probably, [John] F likes Mary. b. Question: Ill-formed answer: Well-formed answer: Who likes Mary? * Probably, John [likes] F Mary. Probably, [John] F likes Mary. In (52a,b), the informational structure in (47c,d) is made explicit. Note that the sentence modal adverb can surface immediately adjacent to the unit in focus. (52) a. Question: Non-modalized answer: Modalized answer: What will John do? John will [buy a new car] F John will probably [buy a new car] F b. Question: Non-modalized answer: Modalized answer: When will John buy a new car? John will buy a new car [next month] F John will buy a new car probably [next month] F The examples in (52) lead us to formulate the main hypothesis behind the third line of solution, namely that sentence adverbs, by virtue of their sensitivity to focus, can appear immediately adjacent to the unit in focus. This hypothesis relates focus and ‘transportability’ in a straightforward fashion. Being FSUs, sentence modal adverbs are able to appear immediately adjacent to the unit in focus, whatever its label may be (T’, VP, DP, AdjP, PP, etc.). This line of solution considers sentence modals as strictly propositional. They are generated in the CP field, from which they c-command the unit in focus. If we consider that the adverbs don’t move, as shown in the previous alternative solution, this proposal must respond to the challenge of providing a syntactic account for the adjacency to the focused unit. In this work, I consider that the third line of solution is the best option. In this sense, I assume the syntactic challenge mentioned in the previous paragraph. Thus in the following chapters, I attempt to provide a syntactic account for ‘transportability’ for sentence modal adverbs, particularly in Spanish. I take as my theoretical base some ideas put forth in the context 28 of the Copy Theory of Movement (Nunes 2001, 2004). The basic suggestion of the CTM is that movement can (and must) be understood as copy and re-merge of a given syntactic unit. For the case of sentence adverbs, I propose that there is a process of copy, triggered by focus marking, which duplicates the clause c-commanded by the sentence adverb and re-merges it in a high position in such a way that the copied clause and the ‘original’ clause form a chain within a bi- clausal structure: one copy is in [Spec,AdvP] and the other one is in [Compl,AdvP]. On this structure, deletion applies driven by linearization reasons. Crucially, focus marking cannot be copied, so the F-marked unit in [Compl,AdvP] cannot be deleted. This process creates the observed immediate adjacency between the modal adverb and the unit in focus in the linear string. As an initial example, consider the sentence in (53), whose rough analysis is shown in (54). (53) John will come back probably on Monday. (54) a. Probably, John will come back on Monday (F-marking →) b. Probably, John will come back [on Monday] F (clausal reduplication →) c. [John will come back on Monday] probably, [John will come back [on Monday] F ] (deletion on the lower copy →) d. [John will come back on Monday] probably, [John will come back [on Monday] F ] (complementary deletion on the higher copy →) e. [John will come back on Monday] probably, [John will come back [on Monday] F ] = John will come back probably on Monday. In the linear string in (54e), we see that the sentence modal adverb appears immediately adjacent to the unit in focus; at the same time, it keeps its sentential scope. In this work, after presenting this proposal for sentence modal adverbs, I further evaluate the possibility of extending the idea to different cases of association with focus. 29 7. Information structure: focus and presupposition 7.1. Definitions The information structure (IS) divides the content of a sentence in two parts, namely focus and presupposition. Although there is no consensus about what exactly one must understand by those terms, I consider here that their definitions are related to the relevance of the information that speakers share from the beginning of the conversation and the information they go ‘building together’ during the discourse. 18 In that sense, I stick to the definitions presented on (55) and (56), which follow Jackendoff (1972), Chomsky (1976), and Zubizarreta (1998), among others. (55) The focus is the non presupposed part of the sentence. (56) The presupposed part of the sentence is what the speaker and the hearer assume to be the case (i.e., the shared assumptions or the common ground) at the point at which the sentence is uttered in a discourse. In this approach, a useful way to identify the focused part—and, consequently, the presupposed part—of the sentence is to check the question to which that sentence could be a possible answer. An English sentence like (57), for instance, is a possible answer for different questions in (58) (Zubizarreta 1998: 2-7). (57) Mary bought the car. (58) a. What happened? b. What did Mary do? c. What did Mary buy? d. Who bought the car? e. What happened to the car? f. What did Mary do with the car? 18 It is convenient to see this as a dynamic process. When the conversation starts, the speakers already share some common content (the common ground). As the conversation proceeds, the common ground gets incremented with each intervention. Once successfully transmitted, new information becomes part of the common ground. 30 The basic idea is that congruent questions and answers share the same presuppositions. Thus, any part of the answer that is not shared with the question will be considered the focus of the sentence. In order to make explicit the unit in focus, this unit is marked with a label “F”. A syntactic description of a sentence S annotated with F labels is called the F-structure of S. A given string of lexical units, as the one presented in (57), may be associated with different F- structures. For instance, the F-structures correspondents this latter sentence are those presented in (59) following the question to which they are intended to be an answer. (59) a. What happened? [Mary bought the car] F b. What did Mary do? Mary [bought the car] F c. What did Mary buy? Mary bought [the car] F d. Who bought that car? [Mary] F bought the car e. What happened to that car? [Mary bought] F the car f. What did Mary do with the car? Mary [bought] F the car 7.2. Assertion structure Note that the F-structure in (59e) shows that the part in focus can be a non-constituent (“Mary bought”). On that issue, Zubizarreta (1998) proposes that the relevant representation of the divide focus / presupposition does not correspond to LF—as previously suggested by Chomsky (1976)—, but instead corresponds to “a more abstract representation derived from LF via some interpretive mechanisms.” (p. 4). That representation is called Assertion Structure (AS). The AS of a sentence consists of two ordered assertions. The first one expresses the existential presupposition that the question provides. The second one, known as main assertion, is presented as “an equative relation between a definite 19 variable (the restriction of which is the presupposition provided by the context question) and a value” (Ibidem). In this sense, each F- structure is associated with one particular AS. In (60), each AS follows its correspondent F- structure (the symbol “=” stands for the aforementioned equative relation). 19 In some cases, weak determiners (few, many) can be associated with the variable, which then would not be definite. A more general approach in terms of λ-functions will be proposed below to include these cases. 31 (60) a. (What happened? →) [Mary bought the car] F A 1 : There is an x, such that x happened. A 2 : The x, such that x happened = [Mary bought the car] b. (What did Mary do? →) [Mary [bought the car] F ] A 1 : There is an x, such that Mary did x. A 2 : The x, such that Mary i did x = [She i bought the car] c. (What did Mary buy? →) [Mary bought [the car] F ] A 1 : There is an x, such that Mary bought x. A 2 : The x, such that Mary bought x = [the car] d. (Who bought the car? →) [[Mary] F bought the car] A 1 : There is an x, such that x bought the car. A 2 : The x, such that x bought the car = [Mary] e. (What happened to the car? →) [[Mary bought] F the car] A 1 : There is an x, such that x happened to the car. A 2 : The x, such that x happened to the car i = [Mary bought it i ] f. (What did Mary do with the car? →) [Mary [bought] F the car] A 1 : There is an x, such that Mary did x to the car. A 2 : The x, such that Mary i did x to the car j = [She i bought it j ] The ASs in (60) show that the unit in focus expresses the value assigned to a variable x (A 2 ) while the presupposed part (A 1 ) acts as the restrictor for that variable. About the relation between variables in the AS, Zubizarreta (1998) suggests that the relation between the indefinite variable in A 1 and the definite description in A 2 is analogous to the relation between an E-type pronoun and its antecedent (Evans 1980). In this kind of relation, a definite description A 2 picks up a referent introduced by a previous assertion A 1 . See an example of a E-type pronoun in (61). 32 (61) a. Some sailor walked into the room. He was wearing a red shirt. b. A 1 : There is an x (x = a sailor) such that x walked into the room. A 2 : The x (such that x = a sailor & x walked into the room) was wearing a red shirt. (Zubizarreta 1998: 5) 7.3. Focus and question-answer congruence In (60), in order to illustrate how one and the same string of lexical items can bear different information structures, I made explicit the question for which each sentence, associated with a different focus marking, could be a valid answer. On this issue, Krifka (2001, 2008b) analyzes the relation between questions (Q) and answers (A) from the point of view of the semantic approach known as structured meaning. The basic idea of this proposal is that the meaning of a question is a function such that this function yields a proposition when applied to the meaning of an answer (Krifka 2001: 288). Functions are represented using the lambda notation, as usual in formal semantics. Now, in order to achieve congruence 20 between Q and A, their respective backgrounds (B), expressed as λ-functions, must coincide, and the unit in focus (F) in the answer must belong to a set of alternatives presupposed in the background of Q; this set is called the restriction set (R). The meaning of a question ( 〚Q 〛) and the meaning of the answer ( 〚A 〛) are understood as order pairs <Background, Restriction> and <Background, Focus>, respectively. This idea is summarized as follows: (62) Criterion for congruent Q-A pairs, where 〚Q 〛 = <B, R>, and 〚A 〛 = <B’, F>: B = B’ and F ∈ R (Krifka 2001: 296) Informally: A pair question-answer is congruent if their respective background functions are identical, and the focused element in the answer belongs to the restriction set presupposed in the question. 20 Intuitively, a congruent Q-A relation is the relation established between a question and (what we understand as) a valid or felicitous answer to it. 33 Krifka (2001) illustrates the congruence criterion with the question in (63), followed by three possible answers, shown in (64). It is shown that only the answer (A 1 ) is a congruent answer for the question. The analysis in terms of pairs <B, R> and <B’, F> would be as follows. (63) 〚Q 〛 = 〚Who did Mary see? 〛 = <λx[SEE (x) (Mary)], PERSON> (64) a. 〚A 1 〛 = 〚 Mary saw [John] F 〛 = <λx[SEE (x) (Mary)], John> where John ∈ PERSON b. 〚A 2 〛 = 〚 [Mary] F saw John 〛 = <λx[SEE (John) (x)], Mary> wh. Mary ∈ PERSON c. 〚A 3 〛 = 〚 Mary saw [the car] F 〛 = <λx[SEE (x) (Mary)], the car> wh. car ∉ PERSON The congruence criterion, on the one hand, validates the answer A 1 as felicitous, since the correspondent functions in the backgrounds of Q and A are identical (i.e., λx[SEE (x) (Mary)]), and the focused element (John) in A 1 belongs to the restriction set PERSON. On the other hand, the same criterion rules out the answers A 2 and A 3 . In A 2 , the focused element (Mary) belongs to the restriction set PERSON, but crucially its background function (λx[SEE (John) (x)]) is different from the background function in Q (λx[SEE (x) (Mary)]). Finally, in A 3 the background function is the same as that in Q, but the element in focus (the car) does not belong to the restriction set. The congruence criterion has been further refined, particularly in Krifka (2008b), in order to deal with some complexities posited by focus in multiple-wh questions (Which student read which novel?) and by cases of under- and overfocusing. 21 However, for the purposes of this work, the important idea is that we can represent the meaning of a sentence S containing a focused unit by considering the sentence S as a (potential) congruent answer to a certain question; in turn, this indicates that the meaning of S can be represented by an ordered pair <B, F>, where B is expressed as a λ-function. 21 In Krifka (2008b), the revised extended congruence criterion looks like this: A question-answer pair Q-A with meanings 〚Q 〛 and 〚A 〛(where 〚A 〛 = <B, F>) is congruent iff: i. 〚Q 〛 ⊆ B ii. F ∈ DOM ( 〚Q 〛 ) ( “DOM” = domain) 34 Let’s exemplify this with the sentences already presented in (60), repeated here as (65) next to their correspondent ordered pairs <B, F>: S = potential answer 〚S 〛 = 〚Answer 〛 = <Background, Focus> (65) a. [Mary bought the car] F < λx[HAPPEN (x)], Mary bought the car > b. Mary [bought the car] F < λx[DO (x) (Mary)], _bought the car > c. Mary bought [the car] F < λx[BUY (x) (Mary)], the car > d. [Mary] F bought the car < λx[BUY (the car) (x)], Mary > e. [Mary bought] F the car < λx[HAPPEN TO (the car) (x)], Mary bought_ > f. Mary [bought] F the car < λx[DO (x) WITH (the car) (Mary)] , _bought_ > Now let’s consider what was previously said about the assertion structure (AS). We declared that the AS consists of two ordered assertions. The first one encodes the presupposition, which always contains a variable. The second one (an equative expression) assigns a value to the variable introduced in the presupposition, and that value is, precisely, the denotation of the focused unit. This is clearly similar to the ordered pairs we see in the second column of (65). The Background (B) is what is presupposed (shared identically by Q and A) and it always contains a variable. The value assigned by the second assertion to that variable is precisely what appears as the second element of the order pair, namely the unit in focus (F). We conclude then that the AS of a sentence can also be represented as (some version of) an ordered pair <B, F>. 22 7.4. Set of alternative focus values vs. set of alternative propositions As it is known, the interpretation of a simple sentence (i.e. a sentence non-marked for focus) is understood in terms of its truth conditions. Now, when that sentence or any of its sub-units has been F-marked, its interpretation is additionally linked to a set of alternative propositions (Rooth 1985, Krifka 2008a). These propositions shared the same presuppositions, and are different among them according to the different values assigned to the focus variable. 22 This is particularly useful considering that B (i.e. the first assertion) is represented as a general λ-function, which not necessarily indicates a definite variable. 35 Let’s take, for instance, the simple question-answer pair What did Mary buy? Mary bought a car. The unit in focus is a car. The interpretation of this answer indicates that, the idea that Mary bought something (already in the common ground), must be suplemented with the idea that Mary bought a car, and this thing has been chosen from a group of some other things that could have been alternatively chosen as object of the purchase. The expressions that denote these alternative objects form a set. In this work, I call this Set of alternative focus values, or Alt- fv. In this case, Alt-fv may look like the set in (66). (66) Alt-fv = {a boat, a bicycle, motorcycle…} On the other hand, the interpretation of the sentence Mary bought [a car] F implies a set of alternative propositions. These propositions are obtained by applying the presupposed part of the sentence (which, as we have seen, can be understood as a function) to the elements included in Alt-fv. These propositions also form a set. I call this set Set of alternative propositions, or Alt- p. For the example considered, Alt-p may look like (67). (67) Alt-p = {Mary bought a boat, Mary bought a bicycle, Mary bought a motorcycle…} As it is visible, the content of Alt-p depends on the content of Alt-fv. Importantly, the interpretation of focus implies necessarily the set Alt-p. As we will see in the rest of this work, the interpretation of focus always assumes propositional content. 23 7.5. Informative focus vs. contrastive focus The idea of focus presented so far relates the focused part of a sentence with a congruent question. In this sense, the focused part of the sentence provides new information to the hearer. This is why this kind of focus has been called informative focus. 24 Let’s take, for instance, the question in (63), Who did Mary see? From the point of view of the hearer, the contribution of the answer in (64) Mary saw John to the common ground is the identity of the person such that 23 When the whole sentence is in focus, the content of Alt-fv and the content of Alt-p will formally coincide. Note, however, that the sense would still be different. For Alt-fv, each element is a possible value for the focus variable. For Alt-p, each element is a sentence encoding an alternative proposition. 24 Also called information focus. 36 Mary saw that person, namely John. Although the question in (63) presupposes that some other focus values could have been assigned to the focused unit, nothing is being stated about those other potential values in terms of truth values. Now let’s imagine that the hearer knows that, in the relevant context, Mary could have seen only one person from the restricted set formed by John, Bill, and Tom exclusively. In this case, the answer in (64) is giving a second kind of information in the form of two additional propositions (usually implicit): Mary didn’t see Bill and Mary didn’t see Tom. We can describe this situation by saying that Alt-fv is closed and is part of the common ground. Consequently, Alt-p is also closed. In such a case, the choice of some focus value automatically implies the negation of other potential focus values in Alt-fv. This kind of focus is known as contrastive focus (Krifka 2008a, Neeleman & Vermeulen 2012, among others). In terms of truth values for the propositions in Alt-p, we can say that there are two possible basic values for each proposition: true (T) and undefined (U)—the truth value ‘false’ is derived as ‘no-p is true’ (Vergnaud & Zubizarreta 2001). The main idea is as follows. Both informative and contrastive focus assign the truth value T to the proposition in which the presupposition has been applied to the focus. In the case of informative focus, all other propositions in Alt-p are assigned the truth value U. In the case of contrastive focus, all other propositions in Alt-p are negated, so the truth value T is assigned to all and each ‘no-p’ in the relevant set. 25 In terms of syntax, the distinction between informative focus and contrastive focus is important, since some operations are licensed by contrastive focus only, not by informative focus. I take the following example from Neeleman & Vermeulen (2012: 9). In (68), suppose that speakers A and B know that John has read only one book in the relevant situation. 25 The situation that I call ‘Oscar winner announcement’ illustrates a typical case of contrastive focus. Before the announcement is made, the idea that there’s just one winner is part of the common ground. The set of candidates is closed and explicit, so it is also part of the common ground. Suppose the candidates are A, B, C, D, and E. At some point, the announcement is made that A is the winner. The truth of the proposition ‘A is the winner’ automatically implies the truth of ‘It is not the case that B is the winner’, ‘It is not the case that C is the winner’, and so on. 37 (68) a. A: John read The Extended Phenotype. b. B: (No, you’re wrong.) The Selfish Gene he read. The answer in (68b) shows the DO located at the beginning of the sentence. It is usually assumed that this is the result of a syntactic operation known as focus movement or focus fronting. This is felicitous in English with a contrastive focus interpretation only (in this case, the contrast has a corrective intention). If the focus is simply informative, focus fronting is perceived as infelicitous. To appreciate this, let’s consider the dialogue in (69). Speaker A knows neither the titles nor the number of books that John has read. (69) a. A: What did John read? b. B: # The Selfish Gene he read. The distinction between informative focus and contrastive focus is also relevant for prosody. In Spanish, for instance, subjects can be focus-marked in situ via prosodic prominence, but this is felicitous only if the focus is contrastive, as the examples (70) and (71) show. 26 (70) Context: Just one of them, Juan or Pedro, bought the beer. a. A: Juan compró la cerveza. J. bought the beer b. B: (No, no es cierto.) PEDRO la compró. (No, that’s not true.) PEDRO bought it. (71) Context: Juan, Pedro, or any other person could have bought the beer. a. A: Quién compró la cerveza? Who bought the beer? b. B: #PEDRO la compró. 26 I represent prosodic prominence with capitals. 38 This distinction made, I point out that, unless explicitly mentioned, the kind of focus involved in the reduplicative structures proposed in this work is informative focus, not contrastive focus. Therefore, I will not deal here with the kind of syntactic operations licensed by contrastive focus. In particular, I don’t deal here with focus fronting. 27 8. Summary In this chapter, I present sentence adverbs as lexical units able to modify clauses. In the first section, a fundamental difference is established between sentence adverbs and verbal adverbs. These latter only modify VPs. Some adverbs only appear as sentence adverbs, others only appear as VP-adverbs, and a third group may surface alternatively at both levels. In Spanish, coordination of adverbs ending in –mente ‘-ly’ can be used to make explicit the difference. Negation can also be used as a test: sentence adverbs do not fall under the scope of negation. This is related to the fact that, syntactically, the value of polarity is located at the edge of the CP domain. Among other values, sentence adverbs encode sentential force, point of view, and topic, and this content precisely corresponds to syntactic nodes located in the CP field. In section 2, a sub-class of sentence adverbs is introduced: the class of sentence modal adverbs. This class is the main object of the present study. Sentence modal adverbs express the degree of confidence of the speaker with respect to the truth of the proposition he/she encodes in the sentence. In this sense, they are part of the group of adverbs known as speaker oriented adverbs (SpOAs). It has been proposed that SpOAs are syntactically ordered with respect to subject oriented adverbs and manner adverbs in a higher position. More specifically, sentence modal adverbs have been included in a highly articulated syntactic hierarchy of adverbials. This order is briefly reviewed in section 3. The hierarchy is obtained through successive comparisons of pairs of sentences containing the adverbs to be ordered. Modal adverbs are located below evaluative adverbs and above aspectual adverbs. The 27 On a terminological detour, the labels contrastive and informative to designate these two types of foci may not have been the best choice. In a broad sense, all foci are contrastive (all foci are interpreted on the basis of a set Alt- p), and all foci are informative (all foci assign a value to a variable). If we take into account the idea put forth in Vergnaud & Zubizarreta (2001) according to which contrastive focus assigns a defined truth value to all and each of its propositions in Alt-p, it may be better to refer to contrastive focus with the term exhaustive focus. Informative focus in turn could be simply labelled non-exhaustive focus. This observation made, in what follows I stick to the traditional terms contrastive focus vs. informative focus. 39 final order postulated is simultaneously linear and scopal. The relation obtained is transitive, asymmetric and connected. Originally evaluated for Italian and French, this hierarchy has been ultimately postulated as cross-linguistically—even universally—valid. Since the proposed universal hierarchy is rich and complex (which may be at odds with explanatory adequacy), an alternative view suggests that semantic properties must also be taking into consideration when evaluating precedence relations among sentence adverbs. This alternative is presented in section 4. In this line, SpOAs are viewed as positive polarity items (PPIs). PPIs require an affirmative environment to appear grammatically. In consequence, SpOAs are infelicitous in non-affirmative contexts like, for instance, questions, imperatives, and antecedents of conditionals. Since some sentence adverbs can only provide a non-affirmative context, it can be predicted that other sentence adverbs will not appear under their scope. This fact could be used to explain precedence relations among different classes of sentence adverbs. In section 5, a semantic approach to the interpretation of sentence modal adverbs is briefly considered. Sentence modal adverbs affect the meaning (i.e. the truth conditions) of the proposition encoded by the sentence they c-command. Propositions are understood as functions going from circumstances into truth values, and in turn circumstances are pairs formed by possible worlds and time. Modal expressions can be interpreted in contrast with a modal base containing what the speakers know about the possible worlds taken as standard in their conversational background. Intuitively, a given circumstance is more probable if it comes closer to the shared standards of normality. Since the truth value of the sentence they modify is affected, the conclusion is that sentence modal adverbs are necessarily propositional in scope. In section 6, I present what I call a challenge for the analysis of sentence modal adverbs. In spite of being semantically propositional, it is shown that these adverbs may eventually surface as adjacent to different types of sub-clausal units. This behavior, termed as ‘transportability’ in early formal approaches, can be interpreted as a mismatch between the semantic content of the adverbs and their syntactic distribution. To deal with this mismatch, three possibilities have been succinctly presented. The first assumes that sentence adverbs are not always propositional, but instead they would be underspecified with respect to their semantic 40 type. If so, they may eventually merge with sub-clausal units bearing different semantic types. This option does not seem appropriate, since it overloads the semantic component and is not plausible from the acquisitional point of view. The second line of thought resorts to syntactic movement of the sentence adverb. If it is assumed that the adverb is base-generated in CP, this adverb would have to move downwards, which is not an option for the current derivational approaches to syntax. If it is assumed that the adverb is base-generated in low positions and from there it moves upwards, either overtly or covertly, the problem of type-underspecification would remain unsolved. A third approach is considered as more convenient in this work. It stems on the fact that the sentence adverb is able to surface in the linear string immediately adjacent to the unit in focus. Assuming the basic idea of the Copy Theory of Movement, I propose that focus marking triggers copy of the modified clause. A later process of deletion applied on the formed bi-clausal structure would yield the observed superficial adjacency. Finally, in section 7, I present the definitions of presupposition and focus to which I will stick in this work. The former is defined as what the speakers assume to be the case at the moment any of them utters a sentence in a discourse. The focus, in turn, is the non-presupposed part of that sentence. This divide can be formalized in terms of assertion structure. In this structure, the unit in focus expresses the value assigned to a variable, whereas the presupposition acts as the restrictor for that variable. On the other hand, the divide is related to the congruence conditions of question-answer pairs (Q-A), as proposed from the semantic approach known as Structurated Meaning. To be congruent, Qs and As must share the same background, expressed in terms of λ-functions. This background can be identified with the presupposition, which implies that the presupposed part of a sentence can be represented as a λ-function. Also, the distinction was made between the set of alternative focus values Alt-fv and the set of alternative propositions Alt-p. The interpretation of a sentence that contains a unit in focus implies the set Alt-p, since the interpretation of focus assumes propositional content. The distinction between informative focus and contrastive focus was also made. Informative focus provides new information, in terms of identification of the value of a variable, so one proposition in Alt-p, namely the proposition created by applying the actual focus value to the presupposition, is considered true. Contrastive focus does the same with that proposition, but it also provides additional information about the truth value of all other alternative propositions in Alt-p (they are 41 considered false). In this sense, contrastive focus assumes that Alt-fv and Alt-p are closed sets and are part of the common ground. Importantly, some syntactic and prosodic operations are licensed by contrastive focus only—not by informative focus. In this sense, I indicated that the structural proposal defended in this work concerns to informative focus as the default case. 42 Chapter 2 Theoretical framework In this chapter I present the theoretical framework of this work. Mainly, this framework is based on the notion of movement as Copy+Merge, a proposal known as the Copy Theory of Movement (CTM). The main idea is that movement can (and must) be interpreted as the result of an operation Copy+Merge applied on some unit that has been already merged at a previous stage of the derivation (Chomsky 1993, Wilder 1995, Nunes 2001, 2004, Corver & Nunes 2007, among others). The CTM has been incorporated to the tool-kit of minimalism, 1 since it successfully complies with the Inclusiveness Condition (Chomsky 1995: 228 ff). According to this condition, the objects shipped to LF must have been built exclusively from the features of the lexical items already present in the Numeration. Notably, the Inclusiveness Condition makes problematic the existence of traces, since these units are not present in the Numeration, but instead they are created in the course of the derivation as a consequence of movement—a moved unit must leave a trace in its ‘original’ position, so a trace is created there. The chapter is organized as follows. In section 1, I briefly present the property known as displacement, which is a universal characteristic of human language. Section 2 presents movement from the point of view of the GB model. In section 3, I present a proposal by Kayne (2000) to explain transportability through a series of upward overt movements of constituents known as remnant movement. In section 4, I present the proposal of considering movement as Copy+Merge, along with the idea of deletion. This latter is treated as a byproduct of the demands imposed by Linearization (Hornstein 2001, Grohmann 2003, Nunes 2001, 2004), which in turn satisfies the Linear Correspondence Axiom, or LCA (Kayne 1994). Importantly, deletion can operate in a special fashion, known as scattered deletion, by which pieces of different links of a chain are phonetically realized (Bošković & Nunes 2007: 36 ff). I provide examples of analyses that resort to scattered deletion in different languages. Finally, section 5 summarizes the chapter. 1 Naturally, this doesn’t mean that there exists unanimity among grammarians with respect to Copy+Merge as an alternative to traditional traces. 43 1. Displacement The property of displacement is among the most uncontroversial facts about human language. According to this property, “expressions pronounced in one position are interpreted in another” (Hornstein 2001:4). 2 See, for instance, the wh-word what in (1). (1) What did Mary buy ___ ? In (1), what appears at the left edge of the linear string. However, the word is understood as the DO of the transitive verb buy. That position is indicated by the gap ( “___”) at the end of (1). In early stages of (Chomskyan) Generative Grammar (late 1950s, 1960s), displacement was encoded via transformational rules, whose main role was the re-ordering of the strings created by core base rules (Chomsky 2002, 1965). The main metaphor was that of movement. In words of Pesetsky, “The term movement describes a situation in which a syntactic unit—for example, a phrase—appears to occupy more than one position in syntactic structure.” (Pesetsky 2000: 2). In the following two decades (1970s, 1980s), the approach known as Principles and Parameters (P&P) became dominant in generative grammar. Within the P&P framework, the Government and Binding (GB) theory reduced the transformations to just one rule (“move-α”), although it still contained a specific sub-theory dealing with the licensing of traces (Chomsky 1988). By the middle of the 1990s, when the minimalist program for grammar was firmly proposed, movement was again re-interpreted, this time as Copy and Merge: the ‘moved’ 2 This basic fact is presented as such in most introductory texts of formal syntax. See, for instance, Hornstein et al. (2005: 7) and Eguren & Fernández Soriano (2004: 65 ff). For a totally different sense of the term displacement in General Linguistics (“The ability to speak about things other than the here and now”), see Trask & Stockwell (2007). 44 element was understood as copied from an ‘original’ that remained in situ, while the copy created was re-merged in a higher position in the syntactic tree. In more recent versions of minimalism, Copy+Merge has been further reinterpreted as a case of internal Merge. The basic idea is that the operation Merge is unconstrained. There would be two kinds of Merge. If Merge applies to two separate objects α and β, that is a case of external Merge (the usual kind of Merge since Chomsky 1993). If Merge applies to the units α and β and one of them is already part of the other, then it is case of internal Merge. 3 In words of Chomsky (2004), Merge yields the property of “displacement,” which is ubiquitous in language and must be captured in some manner in any theory. It is hard to think of a simpler approach than allowing internal Merge (a grammatical transformation), an operation that is freely available. Accordingly, displacement is not an “imperfection” of language; its absence would be an imperfection. (Chomsky 2004: 110). The idea of not taking displacement as an ‘imperfection’ anymore is relatively recent. In a note to the quote presented above, Chomsky (2004) declares that this new idea is contrary to his own earlier assumptions about language, [This is c]ontrary to what I have assumed in earlier work. For over forty years, there have been efforts to motivate displacement. That seems to have been a mistake. Recourse to any device to account for the displacement phenomena also is mistaken, unless it is independently motivated (as is internal Merge). If this is correct, then the radically simplified form of transformational grammar that has become familiar (“Move-α” and its variants) is a kind of conceptual necessity, 3 I assume here that the copied element merges in the usual fashion with some XP that contains the ‘original’. The two or more copies—alternatively, the original and its copies—are always assumed to be the same object. As we will see below, this fact is important for linearization purposes. In principle, linearization enforces the deletion of one of the copies (usually, the lower one, although not in all cases; see below). 45 given the undeniable existence of the displacement phenomena. (Chomsky 2004: 125, fn. 29). In spite of their differences, all the aforementioned approaches have in common the underlying metaphor by which syntactic displacement is represented as (some sort of) movement. 4 I will now present the GB view of movement, which will be useful for my purposes in this chapter. 2. The GB view of movement: traces, ECP, remnant movement 2.1. The GB theory Much of the research done in the framework of generative grammar can be seen as an effort to solve what Chomsky calls the Plato problem, namely “how we can know so much given that we have such limited evidence” (Chomsky 1986: xxv). The knowledge referred to by this quote is the knowledge that children have internalized towards the end of their linguistic development process, that is to say when they become competent speakers of any natural language. The limitation referred to is the information available for the child in this process, namely the primary linguistic data (PLD). There are firm evidence in support of the insufficiency of PLD with respect to the scope and high complexity of the linguistic knowledge the child obtains. One way to think about this problem—that generative grammar plainly assumed—is the supposition that children are innately equipped with what is called Universal Grammar (UG). UG is understood as a set of principles, common to every human natural language. These principles contain a (presumably finite) number of unvalued parameters. The role of PLD is to provide the data necessary to assign a value to each of those parameters. The different assignation of values explains the variation among the languages of the world. In this context, the role of grammarians is to model those principles and parameters in order to gain some understanding of the kind of linguistic knowledge that children obtain by the end of their linguistic acquisition processes. On 4 For a critical point of view of the very notion of movement as a representation for displacement, see Koster (2005), who proposes an alternative mechanism of upward percolation of features within restricted local domains. 46 the basis of these main hypotheses, this theoretical approach was appropriately named the Principles and Parameters approach (P&P). By the late 1970s, and within the P&P framework, a model of linguistic knowledge was consolidated. It came to be known as the Government and Binding (GB) theory (Chomsky 1986, 1988). The GB theory has four defining characteristics, 5 which I now will briefly review. First, its model of grammar is modular, in the sense that is composed by a set of sub- theories that monitor the well-formedness of linguistic expressions. Some of these sub-theories or modules are the Government theory, the Binding theory, the Theta theory, the Case theory, the Bounding theory, and the Control theory, among others (see Chomsky 1986: 5). Second, in spite of their differences, the different modules of grammar share a common basic structural relation, namely government. As Hornstein (2001) points out, […] the central grammatical relation in GB is government. This relation is ubiquitous and appears in every module of the grammar. Government lends conceptual unity to otherwise rather diverse components. […]. Theta-roles and structural cases are assigned under government, binding domains are defined in terms of government, the ECP [Empty Category Principle] licenses traces that are in certain government configurations with their antecedents or heads, the subjacency condition on movement is defined in terms of barriers, which are in turn defined in terms of government. (Hornstein 2001: 3). Third, the system of grammar is formed by four subsystems or components: (i) Lexicon, (ii) Syntax, (iii) Phonetic Form, and (iv) Logic Form (Chomsky 1988: 5). In turn, the syntactic component is sub-divided into two levels of representation: DS and SS. 6 The GB theory assumes 5 I use Hornstein (2001), chapter 1 (pp.1-23) as the main source for this characterization. 6 In previous models of grammar, these abbreviations stood for Deep Structure and Surface Structure, respectively. These latter terms were adapted in other fields in the human sciences, like Psychology, Semiotics, and Literary Analysis. Apparently, not all of these adaptations were felicitous. This leaded Chomsky to reduce the terms to the initials only. About DS, Chomsky declared, “Called “deep structures” in earlier works, I have stopped using the term because it led to too much misinterpretation” (Chomsky 1986: 205, n. 8). 47 that the different sub-theories are verified in one or more specific levels or sub-systems. For instance, Theta theory applies at DS, whereas Case theory is checked at SS (and perhaps at PF too); ECP and Subjacency are monitored also at SS, while Binding theory may apply at SS and/or LF. In the heydays of this approach, a main goal of the theory was precisely to find out the exact level or levels of representation where a given module should apply. Importantly, while DS and SS are considered ‘internal levels’ with respect to the grammar, PF and LF are understood as ‘external’ in the sense that these latter interact with extra-grammatical systems: the articulatory/phonetic (A-P) system and the conceptual/intentional (C-I) system, respectively. In this sense, both PF and LF feed interface levels. Although they share a common input (i.e. SS), PF and LF are considered blind to each other, since they manipulate features of different kind. Fourth, as opposed to previous models of grammar that included a large number of construction-specific transformational rules, the GB theory reduced all transformations to one single rule, namely “move-α”. Move-α mediates between DS and SS. In essence, this rule gives permission to any element α to move to any other part of the tree. Of course, such a liberal rule would quickly overgenerate, unless a way of limiting its power is invoked. Indeed, the GB theory put forth a number of constraints, such as Case, Theta criterion, Projection Principle, Empty Category Principle, etc. The architecture of grammar we have just outlined can be summarized in (2), which I adapt from Chomsky (1986: 68ff.). 48 (2) Lexicon DS Syntax “Move-α” SS PF LF -------- -------- A-P system C-I system 2.2. Traces A main principle within the P&P approach is the so-called Projection Principle. This principle demands that all the information encoded at DS must be preserved at SS and at LF. In consequence, no piece of information must be lost because of movement. In particular, for each movement, it would be necessary to somehow indicate the ‘original’ position of the moved element. This is the basic function attributed to traces. They allow the grammar to go “backwards”, if needed—in fact, this is precisely what happens in the operation known as reconstruction, usually understood as a LF operation. The example of trace in (3) is adapted from Chomsky (1988: 30). (3) It is unclear [ CP who i PRO to see t i ] With respect to (3), Chomsky indicates, “The presence of the trace t i is determined by the projection principle, given that see has a direct object as a lexical property” (ibidem). While (3) illustrate a case of wh-movement, the example in (4) shows a case of NP-movement (indeed, DP- 49 movement), which was considered to be driven by assignation of (abstract) Case. The example is taken from Chomsky (1988: 58). (4) John i seems [t i to like ice cream] In the GB theory, the distinction was made between a movement like the one exemplified in (4), whose target [Spec,IP] is an argument position, and a movement like the one illustrated in (3), whose target [Spec,CP] is not an argument position. The first one was referred to as A- movement, while the second one was called Ā or A’ (A bar)-movement. In particular, A bar- movement makes easy to create long distance movements, operating in a cyclic manner. In those cases, traces build a ‘chain’ (Ch) formed by the head (the only phonetically realized element in the chain) and all its traces. See, for instance, (5a) adapted from Boeckx (2012: 32). (5) a. Who i do you think that t i3 Mary said that t i2 the dog bit t i1 ? b. Ch = {Who i , t i3 , t i2 , t i1 } As seen in (5b), the relevant chain Ch has four links, but just one of them (the head) has phonetic content. Indeed, it is assumed by definition that traces lack phonetic realization. However, they are relevant for the semantic interpretation. In GB theory, the distribution of traces is regulated via the Empty Category Principle (ECP), which establishes that, in order to be licensed, traces must be properly governed (see Chomsky 1988: 250 ff. for relevant definitions). This principle can also be seen as a way of restricting potential illicit movements. From a general point of view, traces make justice to the basic characteristic of human language commented in the previous section, namely displacement. Traces indicate the initial and the intermediate steps taken by the head of a chain Ch, so its displacement can be evident. In sum, since displacement is universal, traces are a central and indispensable ingredient of the GB theory. 50 2.3. Remnant movement A complex situation arises when what moves is an XP that already contains a trace. This is the case of the operation known as remnant movement (Müller: 2002: 209). 7 Remnant movement can be seen as a late development about traces within the GB approach. The sentence in (6) illustrates a relatively simple case of this kind of movement, namely fronting of a VP that already contained a trace—example (6) is adapted from Hunter (2012). (6) [ VP Arrested t 1 by the police] 2 , John 1 was t 2 . A structurally similar example, sentence (7), is taken from Müller (2002). (7) [ VP Critiziced t 1 by his boss] 2 , John 1 has never been t 2 . The subscripted numbers indicated a sequence of movements. In (6), for instance, the DP John moves first to the subject position of the main verb, leaving a trace t 1 in the lower VP. Then, this VP (including t 1 ) is fronted to a high position in the tree, which leaves in turn a trace t 2 . The relevant sequence of movements can be seen in (8a,b). (8) a. John 1 was [arrested t 1 by the police] 7 About remnant movement, see also Stabler (1999), Abels (2008), and Hunter (2012), among others. 51 b. [arrested t 1 by the police] 2 John 1 was t 2 Example (9), taken from Müller (2002), involves three traces. A ditransitive-VP moves up after both of its objects have been previously moved. According to this author, before VP- fronting, the IO moves to a position below the position of the moved DO. (9) John [ VP gave t 1 t 2 ] 3 no-books 1 to Mary 2 t 3 Hunter (2012) points out that an early motivation for remnant movement came from the analysis of what was called incomplete category fronting in German grammar. This phenomenon, illustrated in (10), shows a fronted unit that appears to be a non-constituent. The example, cited by Hunter (2012), is taken from Den Besten & Webelhuth (1990). (10) [ t 1 Verkaufen wollen ] 2 wird er [das Pferd] 1 t 2 sell want will he the horse ‘He will want to sell the horse’ In turn, Stabler (1999) exemplifies remnant movement with the analysis of sentence (11) in Nweh, a Bantu language. This analysis was originally proposed in Nkemnji (1995: 181-183). In this case, the verbal head moves out of VP before AspP-fronting occurs. Thus, this AspP already contains a trace when it moves up. The two consecutive movements are represented in (12a,b). 52 (11) a. Njikèm à kε? [ AspP te t 1 akèndòŋ] 2 pfέt 1 t 2 N. Agr P-l Neg plantains eat ‘Njikèm did not eat plantains’ (12) a. AgrP Njikèm Agr’ à TP kε? NegP Neg’ pfέt 1 AspP te Asp’ t 1 VP t 1 akèndòŋ 53 b. AgrP Njikèm Agr’ à TP kε? NegP [ AspP te t 1 akèndòŋ] 2 Neg’ pfέt 1 AspP t 2 As examples (6)-(12) show, remnant movement creates some sort of ‘nested’ dependencies, in virtue of which a moved element—for instance, the verb pfέt ‘to eat’ in (12)— may end up in a position below its trace. This latter is pied-piped to a higher position via upwards movement of the unit that contains it—for instance, AspP in (12). This is a departure from the earlier analyses in the context of the GB theory, since moved traces are not c- commanded by their antecedents. In some cases, a sequence of remnant movements can yield an output whose linear order is exactly the same as that in the input, modulo traces. The following examples of such derivations are taken from Kayne (2000: 228-230). 8 8 In order to represent—somehow redundantly—the sequence of movements, I use numbers for the subscripts instead of the letters used in Kayne’s original examples. 54 (13) a. John reads no-novels → (neg phrase preposing) b. John no-novels 1 reads t 1 → (VP-preposing) c. John [reads t 1 ] 2 no-novels 1 t 2 (14) a. John invited no-strangers in → (particle preposing) b. John in 1 invited no-strangers t 1 → (neg phrase preposing) c. John no-strangers 2 in 1 invited t 2 t 1 → (VP-preposing) d. John [invited t 2 t 1 ] 3 no-strangers 2 in 1 t 3 (15) a. John [considers no-linguist smart] → (predicate raising) b. John smart 1 [considers no-linguist t 1 ] → (neg phrase preposing) c. John no-linguist 2 smart 1 [considers t 2 t 1 ] → (VP-preposing) d. John [considers t 2 t 1 ] 3 no-linguist 2 smart 1 t 3 At first sight, the derivations in (13)-(15) may seem strange. Why should the syntax worry about moving units just to arrive to the same linear order held at the input? This situation may have seemed particularly baroque for earlier transformational approaches. Furthermore, note that except for the subject, all other elements in the outputs (13c), (14d), and (15d) have been moved out of their initial positions. 9 The question becomes particularly relevant in the context of a model grammar whose desideratum is the economy of grammatical resources. This aspect of the P&P approach, which was already present in the GB theory, was strongly highlighted in the context of the minimalist program. Within the minimalist context, movement is not optional. Since it is considered a ‘costly’ operation for the computational system, movement may occur only if strictly necessary (Chomsky 1995: 261 ff.). This means that movement should occur only if there’s no other way to eliminate some non-interpretable features that may lead the derivation to crash at any interface 9 Kayne explicitly declares that, for the sake of clarity, he would put subjects temporarily away. As Hunter (2012) points out (fn. 2), “If external arguments originate VP-internally, then any instance of VP-fronting will involve remnant movement”. Indeed, according to the standard assumptions of the Predicate-Internal Subject Hypothesis (PISH), all subjects originate inside VP/vP, and from there, they may move up. In case they move, all elements in the outputs would end up in a different (higher) position than the one they occupied in the input. 55 level. For instance, DP subjects must move to [Spec,TP] to eliminate their Case feature 10 . In this sense, every step of the derivation involving movement must be justified. 11 For the case of examples (13)-(15), Kayne proposes the following reasons. First, “Negative phrases of the (standard) English sort are subject to overt movement to Spec,Neg in all languages” (Kayne 2000: 249). 12 This means that the negative phrases no-novels, no-strangers, and no-linguist, in (13), (14), and (15) respectively, must undergo raising to [Spec,NegP]. 13 We see those movements in (13a), (14b), and (15b). Second, (14) and (15) have an additional step at the beginning, namely particle preposing and predicate raising, respectively. As for particle preposing, Kayne considers particles (as, for instance, in in (14)) as small-predicates, and thus assimilates them to obligatory predicate raising, a well-established operation in languages like Dutch (Kayne 2000: 229). Therefore, the initial steps (14a) and (15a) are both instances of the same operation, namely (mandatory) predicate raising. In virtue of this operation, the small-predicates in and smart, in (14) and (15) respectively, move to [Spec,PredP]. Since this latter movement precedes negative-DP preposing, it can be deduced that PredP must be located below NegP (Kayne 2000: 230). Finally, as it is visible, the final step in all three derivations is VP-preposing. To justify this movement, Kayne supposes that there must exist in English an abstract high head that he calls “W”, which is a mnemonic for ‘word order’. WP would be located above NegP. A feature in W would trigger movement of VP to its Spec position. Thus VP-preposing would be understood as movement of VP to [Spec,WP] (Kayne 2000: 239). 14 10 Initially, it was considered that the moving DP carried the non-interpretable Case feature. Later, the idea was that the non-interpretable feature was on the head T 0 instead. In any case, movement was justified. 11 This principle was referred to with the term Greed: “Move raises α only if morphological properties of α itself would not otherwise be satisfied in the derivation” (Chomsky 1995: 261). 12 Kayne (2000: 274, fn. 88) explains, “I take ‘of the (standard) English sort’ to mean at least phrases that convey negation without the presence of a negative morpheme on the verb or auxiliary”. 13 Note that NegP is a maximal projection whose head is (abstract) Neg 0 . According to this proposal, negative phrases move to Spec of NegP. In this sense, the use of the label “negative phrase” to refer to negative-DPs (such as no-novels, no-strangers, or no-linguist) is unfortunate, since it may lead to confusion with (the content of) “NegP”. 14 In a somehow similar vein, Müller (2002) explains the motivation for what this author calls secondary remnant movement and secondary object fronting (like the IO-fronting in (9)) with a constraint, attributed to Williams (1999), according to which “Feature checking must not change the linear order of lexical items established in vP”. 56 Derivations like those in (13)-(15) that resort to remnant movement are particularly useful to model some cases of linear variation. An example of this can be seen in the pair (16a,b). (16) a. John invited no-strangers in. b. John invited in no-strangers. In principle, the two sentences in (16) can be considered as interpretively equivalent. The derivation of (16a) was given in (14). The derivation of (16b) is given in (17) (Kayne: 2000: 228). (17) a. John invited no-strangers in → (neg phrase preposing) b. John no-strangers 1 [invited t 1 in] → (VP-preposing) c. John [invited t 1 in] 2 no-strangers 1 t 2 When comparing the two derivations (14) and (17), we can see that the latter lacks the step particle preposing (i.e. movement to [Spec,PredP]). Although this is not stated explicitly by Kayne, it is reasonable to deduce that the difference lies in the fact that there is no Pred 0 head in (17), so the particle just remains inside VP. On the other hand, Kayne uses derivations including remnant movement to explain scope ambiguities like the one in (18) (example attributed to Klima 1964). (18) I will force you to marry no-one According to Kayne, (18) has a narrow-scope reading that corresponds to the movement of no-one to [Spec,NegP], followed by VP-preposing, always within the infinitive phrase, i.e. the embedded clause (Kayne 2000: 231). See the derivation in (19). This constraint, called Shape Conservation, would enforce (i.e. would motivate) secondary movements in order to keep the linear surface order. 57 (19) a. I will force you to marry no-one → (neg phrase preposing) b. I will force you to no-one 1 marry t 1 → (VP-preposing) c. I will force you to [marry t 1 ] 2 no-one 1 t 2 The wide-scope reading of (18) obtains with the same operations, in the same sequence. In this case, however, movement is at “long distance”, since it applies within the matrix clause. The correspondent derivation is shown in (20). (20) a. I will force you to marry no-one → (neg phrase preposing) b. I will no-one 1 force you to marry t 1 → (VP-preposing) c. I will [force you to marry t 1 ] 2 no-one 1 t 2 Indeed, Kayne’s central aim in this work 15 is to show that structural ambiguities as the one (18) must be treated as cases of overt movement, as opposed to the more traditional view that assumes covert movement (in FL only). I have argued that, in a number of cases where covert phrasal movement had been postulated, it is possible and advantageous to dispense with covert movement […] and replace it with a combination of overt movements. The strongest interpretation of this conclusion is that the cases explicitly considered in this essay are typical and that it is not accidental that those cases lend themselves to analysis in terms of overt movement. It is rather that UG leaves no choice: Scope must be expressed hierarchically, there are no covert phrasal movements permitted by UG, and neither can the effect of covert phrasal movement be achieved by feature raising. Scope reflects the interaction of merge and overt movement. (Kayne 2000: 261). Since it serves the purpose of dispensing with covert movement, remnant movement à-la- Kayne is overt by definition. In the same work (i.e. Kayne 2000, chapter 13) this author 15 Kayne (2000), chapter 13, Overt versus covert movement, pp. 223-281. 58 discusses remnant movement to account for cases of negation, heavy-NP shift, and scope ambiguities with two quantifiers. Importantly, it also extends its possibilities to association with focus. This is the topic of the following section. 3. A remnant-movement analysis of association with focus As seen in the previous chapter, only is a focus sensitive unit (FSU), that is to say, it associates with another unit (the focused unit) in its relevant domain, with which some sort of interpretive dependence is established. When discussing the syntax of the FSU only in English, Kayne states that, in sentence (21), (21) John criticized only Bill. […] only cannot be combined with a DP like Bill, despite the fact that [(21)] is acceptable. […] The derivation of [(21)] will not involve merging only with Bill. Instead, Bill is merged into the VP in the standard fashion, after which only is merged with the whole VP (or perhaps some larger constituent). (Kayne 2000: 238). This observation is reflected in (22a), the first step in the derivation proposed by Kayne for (21), where only appears before VP in the input. (22) a. John only criticized Bill → (attraction by only) b. John Bill 1 only criticized t 1 → (raising of only to W) c. John only 2 Bill 1 t 2 criticized t 1 → (VP-preposing to [Spec,WP]) d. John [criticized t 1 ] 3 only 2 Bill 1 t 2 t 3 In (23), a parallel derivation is proposed for John spoke only to Bill, where only does not properly form a constituent with the PP to Bill, although both units are immediately adjacent to each other. (23) a. John only spoke to Bill → (attraction by only) b. John to Bill 1 only spoke t 1 → (raising of only to W) 59 c. John only 2 to Bill 1 t 2 spoke t 1 → (VP-preposing to [Spec,WP]) d. John [spoke t 1 ] 3 only 2 to Bill 1 t 2 t 3 Interestingly, in (22) and (23) the correspondent outputs show an FSU immediately adjacent to the unit in focus (DP-DO and PP, respectively), despite the fact that that is not its ‘original’ position in the input. This phenomenon was presented in previous chapter under the label ‘transportability’. What these examples show is that the set of mechanisms available from remnant movement can be used to yield the immediate linear adjacency of the FSU to the focused unit—precisely the central problem addressed in the present work. As in sentences (13)-(15), the movements invoked in (22)-(23) must also be justified. The third step, VP-preposing to [Spec, WP], has the justification already presented for (13)-(15), namely attraction by an abstract head W. The first step, attraction by only, indicates that this latter FSU bears some feature X that allows it to attract the unit in focus to its Spec. On the other hand, the second step, raising of only to W, suggest a movement by which the lexical FSU adjoins to the head W in order to check some feature on the FSU itself (indeed, Kayne names it [+w] feature). As for the informational ambiguity implicit in a sentence like John only gave Bill a book, where the unit in focus could be alternatively the whole VP, the verb, or any of the two objects, Kayne observes that it is possible that the whole VP can be attracted to [Spec,only] (Kayne 2000: 243). 16 Since it is been claimed that only cannot directly merge with DP, only Bill in (21) is not really a constituent. Thus, the subject of a sentence like (24) requires further explanation. (24) Only John came to the party. The derivation proposed for (24) is as follows (Kayne 2000: 243). 16 Importantly, this observation coincides with the central proposal defended in this work, in which attraction of the whole c-commanded unit is re-interpreted in terms of Copy+Merge. The unit attracted can be either CP or vP, considered as phases by Chomsky (2001). 60 (25) a. Only John came to the party 17 → (attraction by only) b. John 1 only t 1 came to the party → (raising of only to W) c. Only 2 +W John t 2 t 1 came to the party Since John is in [Spec,only], the correct reading is obtained in (25c). Kayne extends the remnant-movement analysis to the FSUs even and too. These units are subject to ‘transportability’, although with different restrictions. The common idea is that, similarly to only, these FSUs also attract a unit in focus to their respective [Spec,FSU]. Even is similar to only in that is unable to affect the whole IP. In a sentence like Even John came to the party, even relates only to John, not to the whole sentence. Finally, Kayne also presents a remnant-movement proposal to account for cases of F- marking in general (although this author doesn’t use this latter term) independently of the presence or absence of FSUs. The core part of the proposal is that there exists an abstract head Foc 0 that acts like an attractor. This head is later adjoined to the head W, in a fashion similar to only, even, and too. Thus the derivation of (26) appears in (27) (Kayne 2000: 249). (26) (What did he look up?) He looked up [a linguistics-term] F (27) a. He Foc 0 looked a linguistics-term up → (attraction to Foc 0 ) b. He a linguistics-term 1 Foc 0 looked t 1 up → (raising of Foc 0 to W) c. He Foc 0 2 +W a linguistics-term 1 t 2 looked t 1 up → (VP-preposing) d. He [looked t 1 up] 3 Foc 0 2 +W a linguistics-term 1 t 2 Interestingly, Kayne locates Foc 0 in the left periphery of VP/vP at the initial step of the derivation. Via the usual set of rolling-up movements, Foc 0 ends up located immediately 17 Note that the FSU only cannot associate with the whole IP, since it has no scope over sentences. Although Kayne is not structurally explicit in this point, I consider that it is fair to deduce that, in the input (25a), only has syntactic scope over vP. Remember that, according to the PISH, the subject is generated as a vP-internal unit. 61 adjacent to the unit in focus, which, in general, is the same position that FSUs occupy in cases of ‘transportability’. In certain sense, the silent Foc 0 could also be seen as the under-specified default case, of which concrete FSUs would be ‘fleshed’ realizations. 18 In any case, this aspect of Kayne’s proposal has the very desirable consequence of treating all cases of F-marking in a systematic way, independently of the presence or absence of a lexical FSU in the clause. If this intuition is on the right track, it would imply that F-marking always implies movement—or, equivalently, that F-marking always implies Copy+Merge. In sum, Kayne’s analysis of association with focus via remnant movement opens interesting ways to the research on the topic. I find particularly insightful, (i) the idea that there exists attraction exerted from a FSU over the focused unit, in virtue of which this latter moves to [Spec,FSU]; (ii) the notion that any XP can undergo focus-raising, including VP/vP and IP/TP; and (iii) the (implicit) suggestion that the immediate adjacency of FSUs to the unit in focus (i.e. ‘transportability’) can be analyzed in terms of upward movement. On the other hand, it is hard not to feel worried about (i) the high complexity of the sequence of nested movements that characterized this approach, especially within the context of Plato’s problem (i.e. the acquisition of L1 by the child); (ii) the resort to arbitrary abstract categories—for instance, WP—whose existence seems to be motivated solely as place-holders of the relevant Spec positions; and (iii) the systematic use of traces, units that are absent in the Numeration and are created in the course of the derivation, which violates the Inclusiveness Condition. 19 4. Movement as Copy+Merge. Scattered complementary deletion 4.1. The minimalist program 20 By the end of the 1980s, the goal of explanatory adequacy within the P&P approach seemed to have reached a point of maturity. Beyond the apparently extreme variation among human 18 Each FSU, of course, would add its specific semantic values and its own distributional restrictions. 19 See (29) below. 20 For an introductory presentation of the minimalist program, see Chomsky et al. (2002), Boeckx (2006a), Horstein et al. (2005), Eguren & Fernández Soriano (2004), and Radford (1997). I use Boeckx (2006a) as the main guide for this brief presentation. For the primary sources, see Chomsky (1993, 1995, 2000, 2001, 2004, 2008). For a contact with the research in minimalist framework, see Epstein & Horstein (1999), Martin et al. (2000), Boeckx (2006b, 2008), Horstein (2009), and Nunes (2009), among many others. 62 languages, the set of modular sub-theories within the GB theory seemed capable of providing a fairly complete account of the main principles of UG. On the other hand, the considerable amount of research done about particular grammars within the P&P approach shed light on the side of parametric variation. From the observed external data, the GB theory had already identified the basic principles that informed the sub-theories, and the related parametric variation. By the beginnings of the 1990’s, it could be said that, in relative terms, this task had been successfully completed. This situation made possible to somehow deepen the scope of the research questions. Having determined a certain number of general principles of UG, it was time to ask questions of a higher level of abstraction. The focus of attention shifted progressively from the content of principles to questions like why those principles hold, what their purpose could be, how they relate to the interfaces, in which sense grammar is optimally designed to fulfill the demands of the interpretive systems, etc. As Boeckx (2006) explains, The minimalist program grew out of the perceived success of the P&P approach. It took the principles-and-parameters shape of the language faculty for granted, and asked how much of this shape could be the direct result of optimal, computationally efficient design. […] linguists began to ask whether the principles they discovered could be understood in terms of higher standards of inquiry. Do linguistic principles display interesting symmetry, uniformity, economy? Why do we have these principles and not others? How much of these linguistic principles follow from the most basic assumptions/axioms everyone has to make when they begin to investigate language (what Chomsky has called ‘virtual conceptual necessity’)? The last question gives meaning to the term ‘minimalist.’ (Boeckx 2006: 3-4). Since its earlier days, generative grammar—as any other scientific enterprise—has been concerned about its methodological aspect. Theories are object of continuous improvements, reviews, and changes. If two grammars are equally successful in generating the set of grammatical/acceptable sentences and only those sentences in a given language, then the simplest, more economical, more ‘elegant’ one would be chosen (Chomsky 1965). This kind of methodological concerns has been constant in generative grammar, and remains important 63 nowadays. The novelty brought about by the Minimalist Program by the 1990s was that questions about optimal design were posed not only about the methods, but also about the object itself. Chomsky et al. (2002) explain this change of perspective with a graphic analogy. For instance, if you try to develop a theory of an automobile that doesn’t work, with terrible design, which breaks down, say the old car you had in Amherst for example: if you wanted to develop a theory of that car, you would still try to make the theory as good as possible. I mean, you may have a terrible object, but still want to make the theory as good as possible. So there are really two separate questions, similar but separate. One is: let’s make our theories as good as we can whatever the object is – a snowflake, your car in Amherst, whatever it may be. And the other question is: is there some sense in which the device is optimal? Is it the best possible solution to some set of conditions that it must satisfy? These are somewhat different questions and there was a shift from the first question, which is always appropriate (let’s construct the best theory), to the second question: does the thing that we are studying have a certain kind of optimal character? (Chomsky et al. 2002: 97). So far, the provisional, methodological, and programmatic answer to the question of whether human language is optimally designed is yes. It is assumed that UG perfectly satisfies the conditions imposed by the interface systems. This idea is called the strong minimalist thesis (Boeckx 2006: 4), and, although it may result too ambitious at the end, it is currently used as a work hypothesis in minimalism. It’s important to highlight the hypothetical nature of this assertion. As Hornstein et al. point out, it is possible that, at the end of the day, minimalist research will find that UG is imperfect, redundant, poorly designed, or “just ugly”, but one can’t know if it is so without trying (Hornstein et al. 2005: 7). As we have seen, the minimalist program plainly assumes the P&P approach and integrate it in a new framework of research. In this sense, minimalism is not a new theory of language. As its name already indicates, the minimalist program refers to a certain mode of doing investigation, which is at the base of very broad questions that may be answered from very 64 different perspectives (Boeckx 2006: 5). In this sense, a minimalist program may also be sought in disciplines other than linguistics. This said, it is important to bear in mind that all theoretical devices that were confidently used in the GB theory were put under hard scrutiny. Notions as basic as government, traces, DS, SS, or X-bar phrase structure were questioned with respect to their conceptual necessity for optimal design. It is true that what Chomsky (2000: 92) calls the tension between descriptive adequacy and explanatory adequacy has always existed in generative grammar. Descriptive adequacy frequently leads to postulate dense systems of rules. Those complex systems, unfortunately, usually go against the demands of the explanatory adequacy, namely the plausibility for those systems to be what the children internalize when acquiring their L1. The evolution from the transformational systems of rules, which were central in the Standard Theory (Chomsky 1965), to the single Move-α rule, plus the simple X-bar structure postulated in the GB theory (Chomsky 1988), exemplifies a shift towards explanatory adequacy, without losing descriptive power. On this issue, minimalism is even more radical. The minimalist project—as the title of Chomsky (2004) already suggests—wants to go beyond explanatory adequacy. Linguistic research is put in the context of a broader bio-linguistic framework. A successful L1 acquisition process was no longer seen as the result of the interaction of PLD and UG exclusively. A third factor was added: the “General properties of organic systems” (Chomsky 2004: 104). The idea is that some principles of efficient computation are indeed general properties of all organic systems, and not specific properties of linguistic models. This supposition is oriented by the question about the reasons behind the observed properties. In words of Chomsky, “In principle, then, we can seek a level of explanation deeper than explanatory adequacy, asking not only what the properties of language are but also why they are that way” (Chomsky 2004: 105). In this context, since linguistic minimalism programmatically assumes the perfection of its object, it has been aligned with the scientific view of the astronomer Galileo, thus it is a Galilean research project. 21 Before leaving this succinct presentation of minimalism in linguistics, I want to mention here four conditions of linguistic computation, proposed within the minimalist framework, which are especially relevant for the purposes of this work. 21 For deeper considerations about a Galilean scientific project in grammar, see Freidin & Vergnaud (2001). 65 (28) Interface condition The information in the expressions generated by the linguistic system (L) must be accessible (legible) to other systems, including the sensorimotor (SM) and conceptual-intentional (C-I) systems. (Chomsky 2004: 106). (29) Inclusiveness condition Any structure (π, λ) must be built exclusively from elements already present in the Numeration 22 ; no new objects can be added in the course of the computation. (Chomsky 1995: 228). 23 (30) Last resort Movement is licensed only if it leads to the elimination of non-interpretable features. (Horstein 2001: 8 ff). 24 (31) Uniformity condition The operations available in the covert syntax must be the same operations available in overt syntax. (Chomsky 1995: 229; Horstein et al. 2005: 368). 22 The notion of Numeration can be seen intuitively as a set containing all the lexical pieces that will be used in a given derivation before it starts. If any of those pieces will be used more than once, the number of its occurrences is indicated in the correspondent subscript. For instance, a sentence like The lady will buy the picture of John, the Numeration will look like the set N, in (i). (i) N = {the 2 , lady 1 , will 1 , buy 1 , picture 1 , of 1 , John 1 } 23 Chomsky (Ibidem) explicitly adds, “in particular, no indices, bar levels in the sense of X-bar theory, etc.”. 24 The original formulation of this condition, from Chomsky (1993), was that of Greed, which means that an element moves to check its own non-interpretable features. The alternative formulation in Hornstein (2001) is more flexible: the non-interpretable feature belongs either to the moved element or to the host of movement. 66 4.2. The Copy Theory of Movement (CTM) 25 The Inclusiveness Condition makes traces no longer available for linguistic computations. Since traces are objects created in the course of the derivation, they are not present in the original Numeration, so, according to (29), they are barred from syntax. In principle, this creates a problem for the computational system, since traces have a valuable function for interpretation. Importantly, traces are used as devices that mark the departure point for each step taken during each cycle of movement. Successive marks create a non-trivial chain, and this is a necessary piece of information for the interpretive interfaces. If that information is lost, the linguistic system would not be properly serving the interpretive conditions imposed on it, in violation of the Interface Condition, in (28). Thus, in order to comply with both the Interface Condition and the Inclusiveness Condition, the grammar must find a different way to fulfill the functions that traces used to perform. Chomsky (1993) proposes that grammar can dispense with traces if movement is reinterpreted as an operation involving copying plus merging of the moved element. An approach that has occasionally been suggested is the “copy theory” of movement: the trace left behind is a copy of the moved element, deleted by a principle of the PF component in the case of overt movement. But at LF the copy remains, providing the materials for “reconstruction”. (Chomsky 1993: 34-35). This suggestion can be seen as the starting point of the renewed interest about the CTM in minimalism. Traces can be advantageously replaced with copies. Accordingly, sentence (32), which had the analysis in (33a) within the GB theory of traces, would have the analysis in (33b) within the CTM (Nunes 2001: 305). (32) John was kissed. 25 About the CTM, see Chomsky (1993, 1995), Nunes (1999, 2001, 2004), and Horstein (2001). I use Nunes (2004) as the main source for this brief presentation of the CTM. For more specific applications, see Wilder (1995), Fanselow & Ćavar (2002), Jo (2013), and the collection of articles reunited in Corver & Nunes (2007); see especially the introductory note by the editors (pp. 1-9). For a logical-semantic approach to copying in grammar, see Kobele (2006). 67 (33) a. [John i [was [kissed t i ]]] b. [John i [was [kissed John i ]]] 26 In principle, the main advantage of (33b) over (33a) is the fact that no new material has been added during the derivation. The Inclusiveness Condition is not affected by the creation of a copy of any given element, provided that it has been selected for the Numeration since before the computation started. An abbreviated derivation for (33b) is presented in (34). (34) N = {Jonh 1 , was 1 , kiss 1 , -ed 1 } a. K = [ TP T [was [ VP kissed [ DP John i ]]]] b. Copy DP → L = [ DP John i ] c. Merge K,L → [ TP [ DP John i ]T [was [ VP kissed [ DP John i ]]]] The Numeration indicates that the item John has been selected only once. In (34b), this unit has been copied, which means that the appearing material (the copy) is not really ‘new’. Then, (34c) only invokes Merge, a ubiquitous, cost-less operation in the context of minimalism. The derivation does not violate the Inclusiveness Condition. Later operations of deletion will yield the expected linear sequence John was kissed, as we will see below. 4.3. Some advantages of the CTM over traces So far we have noted that the CTM satisfies the Inclusiveness Condition and thus allow us to dispense with traces. Nunes (2004: 10-14) presents four additional the advantages of the CTM in the general framework of Minimalism, which I summarize as follows. (i) Elimination of non-interface levels Since minimalism is committed to eliminate all non-essential components of grammar, the existence of the so-called ‘internal’ levels of grammar, namely DS and SS, came into questioning. The CTM may provide support to the idea of considering LF as the appropriate 26 I adopt the convention in Nunes (2001, 2004) about using superscripts to identify all copies of the same unit. 68 level to check some anaphor-related facts that were previously considered as DS/SS phenomena. In this direction, Chomsky (1993: 37) discusses the interpretive ambiguity in (35) marked with subscripts. (35) John i wondered [which picture of himself i/j ] k Bill j saw t k As denoted by the subscripts, sentence (35) is ambiguous. Under the standard GB assumptions, the reading in which Bill and the anaphor himself are co-referential indicate that the binding has occurred before the wh-movement took place, that is to say, at DS level. This would indicate that Principle A must be checked at that level. On the other hand, the reading in which the anaphor is bound by the matrix subject (John) would indicate that Principle A should be verified either at SS or LF. Of course, the only option available to capture the two possibilities would be to consider that Principle A is applicable at either DS or SS/LF, that is to say, at all non-phonetic levels. This is, however, an undesirable result from a minimalist perspective. The alternative suggested by Chomsky starts by interpreting the trace in (35) as a lower copy of the moved element, as shown in (36). (36) John wondered [which picture of himself ] i Bill saw [which picture of himself] i The structure in (36), created by syntax, is shipped as such both to PF and to LF. At PF, an operation of linearization, driven by the Linear Correspondence Axiom (LCA), deletes the lower copy. 27 This is the usual PF-deletion, which would render the PF linear sequence in (37). 28 (37) PF-structure John wondered [which picture of himself ] i Bill saw [which picture of himself] i 27 Although this is the case in (37), the lower copy is not always the deleted one. For details about linearization and the LCA, see below. 28 Note that deletion has been traditionally understood as a phonetic/phonological operation only, and as such, it has been taken as unrelated to LF. 69 As for LF, it has two possibilities—which are precisely at the base of the ambiguity. In case the anaphor is considered bound by the matrix-subject John, LF would not read the lower copy, but only the higher one. On the other hand, in case the embedded-subject Bill is taken as antecedent, LF would chose not to read the higher copy. This ‘not-reading’ choice has also been described as LF-deletion: whichever copy is not read at PF would be considered as LF-deleted. 29 Thus, in the first interpretation mentioned, i.e. binding by the matrix subject, the LF-structure would look like (38a); if the embedded subject is chosen as antecedent, the correspondent LF- structure would look like (38b). (38) a. LF-structure (when John is chosen as antecedent) John wondered [which picture of himself ] i Bill saw [which picture of himself] i b. LF-structure (when Bill is chosen as antecedent) John wondered [which picture of himself ] i Bill saw [which picture of himself] i This analysis places the monitoring of Principle A in LF exclusively. And this is a desirable consequence for minimalism, since neither DS nor SS are longer required for that checking. Before moving to more advantages of the CTM, I would like to highlight two important facts. First, there are two types of deletion: PF-deletion and LF-deletion. This fact conforms to the Uniformity Condition (see (31)). Second, these two types of deletion may not coincide. For instance, sentence (35), repeated here (with the relevant subscripts only) as (39a), can have a representation (39b) at PF, which does not coincide with its LF representation (39c). (39) a. John wondered which picture of himself i Bill i saw. b. PF-structure (PF-deletion targets the lower copy) John wondered [which picture of himself ] i Bill saw [which picture of himself] i c. LF-structure (LF-deletion targets the higher copy) John wondered [which picture of himself ] i Bill saw [which picture of himself] i 29 In this work, I use a single strikethrough to indicate PF-deletion and a double strikethrough to mark LF-deletion. 70 Structures in (39) show a mismatch between PF and LF, which is due to different targets for PF-deletion and LF-deletion, respectively. Such a mismatch is hard to capture under a trace- based representation. (ii) Interpretation of ‘discontinuous’ complex predicates (idioms) Consider the idiomatic expression “(to) take pictures” with the meaning of ‘(to) photograph’. Question (40) keeps the idiomatic meaning, although the idiom has been ‘split’ in two parts. (40) How many pictures of John did you take? This fact has been considered as evidence for the existence of the divide DS/SS. According to this view, we understand the idiomatic meaning in (40) because its idiomatic (unified) interpretation occurs at DS, in spite of the visible split, which supposedly occurs at SS only. The CTM version of this phenomenon considers it instead as a reflex of the divide PF/LF, so we can dispense with the levels DS and SS. The reinterpretation of (40) is shown in (41). (41) [how many pictures of John] i did you take [how many pictures of John] i The split that creates the PF-discontinuity can be seen as the result of PF-deletion in (42a), whereas the idiomatic unified reading can be interpreted as the result of LF-deletion in (42b). If this is on the right track, the phenomenon belongs to the interface levels PF and LF only, and the divide DS/SS is no longer required. (42) a. PF structure (discontinuity as a result of PF-deletion) [how many pictures of John] i did you take [how many pictures of John] i ? b. LF structure (unified interpretation as a result of LF-deletion) [how many pictures of John] i did you take [how many pictures of John] i ? 71 (iii) Elimination of reconstruction An alternative account for the ambiguity in (35) appeals to an LF operation known as reconstruction. This operation consists of putting a moved unit back to its trace position in the covert component. Since this operation is optional, one of the two possible readings may be the result of its application. Specifically, reconstruction makes possible binding of the anaphor by the embedded subject in (35). However, Chomsky (1993: 38 ff) points out that reconstruction is undesirable from a minimalist point of view, since this operation undoes a movement that is necessary for convergence. First, Chomsky claims that the idiomatic value of “(to) take pictures” as ‘(to) photograph’, in (43) is possible only if the anaphor is bound by the embedded subject. 30 (43) John wondered which picture of himself Bill took. If Principle A is applied at SS, as it was traditionally assumed, the matrix subject can bind the anaphor. Now, if reconstruction is applied at LF after this binding occurs, the idiomatic reading of “(to) take pictures” becomes available. According to Chomsky, this would lead to the wrong prediction, since, according to his grammaticality judgment, only the embedded subject is available as an antecedent with the idiomatic reading. The CTM explains straightforwardly these facts without resorting to reconstruction. If LF-deletion targets the higher copy, as in (44a), the idiomatic reading becomes available and, at the same time, the anaphor can be bound by the embedded subject only. On the other hand, if LF-deletion targets the lower copy, as in (44b), the idiomatic reading becomes unavailable and, simultaneously, the anaphor can be bound by the matrix subject only. (44) a. John wondered [which picture of himself ] i Bill took [which picture of himself] i b. John wondered [which picture of himself ] i Bill took [which picture of himself] i 30 However, Hornstein et al. (2005: 273 fn. 14) warn that this judgment is not unanimous among native speakers of English. 72 Note that PF-deletion matches, in this case, only with the LF-deletion shown in (44b). For the interpretation shown in (44a), a PF-/LF-deletion mismatch has to be assumed. (iv) Realization of ‘traces’ and predictability of their phonetic shape In the typical case, only the head of a chain (the higher link) is pronounced, while all traces left behind remain silent. However, in certain cases, more than one link in a chain is pronounced. Nunes (2004: 40) explains that, in these cases, it is always the intermediate wh-copy that is pronounced, and this fact suggests that this intermediate copy has undergone a morphological process of fusion with some intermediate head C 0 (which can be silent) to form one single word. Since the LCA (see next subsection) cannot ‘see’ inside words, the morphological fused copies become ‘invisible' for the axiom, so they are not deleted. The following examples are offered in Nunes (2004: 6-13; 38). (45) German (from McDaniel 1986) Mit wem glaubst du mit wem Hans spricht? With whom think you with whom H. talks ‘With whom do you think Hans is talking?’ (46) Romani (from McDaniel 1986) Kas misline kas o Demìri dikhlà? whom you-think whom D. saw ‘Who do you think Demir saw?’ (47) German (from Fanselow & Mahajan 1995) Wovon glaubst Du wovon sie träumt? What-of think you what-of she dreams ‘What do you believe that she dreams of?’ (48) Brazilian Sign Language (from Nunes & Quadros, in preparation) I LOSE BOOK LOSE ‘I LOST the book’ 73 (49) English child grammar (from Thornton 1990) Who do you think really who is in the can? Interestingly, the realized ‘traces’ have the same phonetic shape of the head of the chain. This is precisely what one would expect if traces are, indeed, copies. 4.4. Linearization of chains and the LCA In the general case, only the higher link of a chain is overtly realized. In the CTM framework, this fact can be interpreted as the fact that PF-deletion targets all copies, except the highest one. Nunes (2001, 2004) relates PF-deletion to the linearization of chains driven by the Linear Correspondence Axiom (LCA), initially proposed by Kayne (1994). The LCA puts forward the idea that precedence, a linear one-dimensional relation, is universally derived from asymmetrical c-command, a two-dimensional relation. In this view, the LCA would ‘prepare’ the hierarchical structure to enter phonological processes. See a formulation of the LCA in (50). 31 (50) Linear Correspondence Axiom (LCA) A lexical item α precedes a lexical item β iff (i) α asymmetrically c-commands β, or (ii) an XP dominating α asymmetrically c-commands β. As an example of the role played the LCA in the linearization of chains, let’s consider again sentence (33b), repeated here as (51). (51) [John i [was [kissed John i ]]] In this sentence, there are two non-distinct copies of the lexical item John. Now, imagine we want to order John with respect to was. The question is which of the items should precede the other, according to the LCA. If we compare the high copy of John with was, we must 31 This formulation of the LCA is taken from Hornstein et al. (2005: 227). Alternatively, Nunes (2004: 15) presents the LCA with the following equivalent wording: “Let X, Y be non-terminals and x, y terminals such that X dominates x and Y dominates y. Then if X asymmetrically c-commands Y, x precedes y”. 74 conclude that John should precede was, since the former asymmetrically c-commands the latter. But if we compare was with the lower copy of John, we must conclude that was should precede John, again for reasons of c-command. Naturally, there’s no way of ordering the two items if these two contradictory conclusions hold, so the derivation will be cancelled. On the other hand, precedence is, by definition, an asymmetrical relation: if α precedes β, then β does not precede α. Also, precedence is irreflexive: if α precedes β, then α ≠ β. Since John is one and the same lexical item in (51), both asymmetry and irreflexivity conditions are violated. In order to overcome these impasses, one of the two copies must undergo PF-deletion. 4.5. Chain Reduction and Formal Feature Elimination Nunes (2004: 25 ff) proposes that this type of PF-deletion is guided by reasons of economy. This claim is illustrated with the following example, involving three copies: high, intermediate, and low. (52) [ TP [ DP the [ NP tall man]] i appears [ TP [ DP the [ NP tall man]] i to have been kissed [ DP the [ NP tall man]] i ]] As we have seen, the structure in (52) cannot be linearized as is, because the LCA would be unable to establish the necessary precedence relations. Interestingly, the restrictions imposed by the LCA would be circumvented if PF-deletion targets repeated parts in different links in such a way that only one copy of each repeated item survives. This kind of deletion has been properly named scattered deletion (as opposed to full deletion). In (53), for instance, the NP tall man has been deleted in the higher copy, the Det the and the N man are deleted in the intermediate copy, and the Det the and the adjective tall are deleted in the lower copy. (53) [ TP [ DP the [ NP tall man]] i appears [ TP [ DP the [ NP tall man]] i to have been kissed [ DP the [ NP tall man]] i ]] The structure in (53) can now be linearized, since there are no violations to the LCA. The resultant sequence, shown in (54), will converge at PF. However, the resultant string is unacceptable. 75 (54) * The appears tall to have been kissed man. At this point, the notion of economy becomes decisive. In minimalism, two convergent derivations created from the same numeration can be compared in terms of the number of steps involved. The derivation with the fewer steps will be chosen as optimal. Consider (55), in which full deletion has been applied to both the intermediate copy and the lower copy of the DP the tall man. (55) [ TP [ DP the [ NP tall man]] i appears [ TP [ DP the [ NP tall man]] i to have been kissed [ DP the [ NP tall man]] i ]] The PF-deletion represented in (55) is also convergent at PF. In this case, the resultant linear sequence is fully grammatical. (56) The tall man appears to have been kissed. When comparing the convergent structure in (53) (created by scattered deletion) with the convergent structure in (55) (created by full deletion), it can be seen that, in the former, PF- deletion has applied five times, whereas, in the latter, PF-deletion has applied only two. Importantly, deletion applies to constituents. 32 Then, the counting considers one application of PF-deletion per deleted constituent. For reasons of economy then, full deletion is preferable over scattered deletion, so (55) is selected as optimal. This idea is elaborated in the PF-operation known as Chain Reduction (Nunes 2004: 27). (57) Chain Reduction (ChR) Delete the minimal number of constituents of a non-trivial chain Ch that suffices for Ch to be mapped into a linear order in accordance with the LCA. 32 Note that a single word can be a constituent. For instance, the DP the man is formed by two immediate constituents: the and man. Thus, if deletion targets the DP, it targets one constituent. If it targets just the, it targets one constituent too. 76 Nunes explains that he keeps (for expository purposes) the formulation of Chain Reduction as it appears in (57), but he is aware of the fact that it is unnecessary to explicitly mention “the minimal” number of deleted constituents, since economy considerations will always select the derivation with fewer steps. This does not mean, however, that scattered deletion is ruled out in all cases. This form of PF-deletion may be selected when it is a way to avoid fatal violations of high ranked constraints, as we will see in the next section. Before moving to that point, however, it is convenient to remember that we have considered that the optimal derivation for (52) (repeated here as (58)) is (55) (repeated as (59a)). We compared this latter derivation with the derivation in (53), which included five applications of (scattered) deletion. Since (59a) only contains two applications of deletion, it is more economical, and thus it is declared optimal. However, derivations (59b,c) also involve only two instances of full deletion, and are also convergent at PF. (58) [ TP [ DP the [ NP tall man]] i appears [ TP [ DP the [ NP tall man]] i to have been kissed [ DP the [ NP tall man]] i ]] (59) a. [ TP [ DP the [ NP tall man]] i appears [ TP [ DP the [ NP tall man]] i to have been kissed [ DP the [ NP tall man]] i ]] b. [ TP [ DP the [ NP tall man]] i appears [ TP [ DP the [ NP tall man]] i to have been kissed [ DP the [ NP tall man]] i ]] c. [ TP [ DP the [ NP tall man]] i appears [ TP [ DP the [ NP tall man]] i to have been kissed [ DP the [ NP tall man]] i ]] In terms of Chain Reduction, the three structures in (59) tie. However, only the first linearization (59a) would produce an acceptable linear string in English. This fact indicates that the grammar may need (at least) one additional criterion of choice. On this issue, Nunes (2004: 30 ff) recalls a proposal made by Chomsky (1995) according to which there must exist an operation that eliminates formal features that may be still visible at PF after morphological processes applies. Since no visible formal features are allowed at PF, the 77 derivation will crash if those features are not eliminated. This operation is called Formal Feature Elimination. (60) Formal Feature Elimination (FFE) Given the sequence of pairs σ = < (F, P) 1 , (F, P) 2 , …(F, P) n > such that σ is the output of Linearize, F is a set of formal features, and P is a set of phonological features, delete the minimal number of features of each set of formal features in order for σ to satisfy Full Interpretation 33 at PF. (Chomsky 1995: 230). The idea is that, in the general case, the head of a chain Ch (i.e. the highest copy) has checked more formal features than its ‘traces’ (the lower copies) while moving up the tree, so if deletion applies on those ‘traces’, fewer applications of FFE (or none) will be required for the derivation to converge. From this reasoning, deletion on ‘traces’ would always be, in principle, more economical than deletion on the head, modulo other violation of any high ranked constraints. Let’s take, for instance, the structures in (59). According to the standard assumptions, A- movement is triggered for Case-checking considerations. The most embedded object moves cyclically to the Spec (subject) position of the two successive TPs. This can be made explicit for the structures in (59) as it appears in (61). Nunes (2004: 32) represents the unchecked Case feature with boldface and the checked deleted feature as subscript. (61) a. [ TP [ DP the [ NP tall man]] i Case appears [ TP [ DP the [ NP tall man]] i -CASE to have been kissed [ DP the [ NP tall man]] i -CASE]] b. [ TP [ DP the [ NP tall man]] i Case appears [ TP [ DP the [ NP tall man]] i -CASE to have been kissed [ DP the [ NP tall man]] i -CASE ]] c. [ TP [ DP the [ NP tall man]] i Case appears [ TP [ DP the [ NP tall man]] i -CASE to have been kissed [ DP the [ NP tall man]] i -CASE ]] 33 Full interpretation is “a principle of representational economy (itself part of substantive economy) that requires that all the features of the pair [(π, λ)] be legible at the relevant interfaces”. (Hornstein et al. 2005: 15). 78 Both derivations in (61b,c) contain an unchecked formal Case feature, so FFE must apply at least once in each of them. On the contrary, (61a) does not contain any unchecked formal feature, so no application of FFE is required. Thus, this latter derivation will be considered as more economical, and as such, it will be declared optimal. 4.6. Scattered deletion It is a partial and complementary deletion that applies inside the links of a given chain. It occurs when some material is deleted in one of the copies, and the complementary material is deleted in one (or more) of the other copies. Although scattered deletion is, in principle, more costly than full deletion, it can be applied whenever full deletion produces a violation of a higher ranked constraint in the language, in the sense of optimality. In the following paragraphs, I present some cases of scattered deletion that have been reported in the literature by different authors, to explain certain constructions in various languages. 4.6.1. Order of clitics in Bulgarian (Bošković 2001) An interesting case in which scattered deletion is chosen as optimal can be found in Bulgarian, a language with a high ranked constraint about clitics. The phenomenon can be better appreciated by contrast with Macedonian (Nunes: 2004: 28). Note that both the linear sequences and their interpretations are the same for both languages (they are genetically close to each other); crucially, what changes is the acceptability. (62) Macedonian (from Rudin et al. 1999) a. Si mu (gi) dal li parite? Are him-DAT them given Q the-money b. * Dal li si mu (gi) parite? Given Q are him-DAT them the-money ‘Have you given him the money?’ 79 (63) Bulgarian (from Rudin et al. 1999) a. * Si mu (gi) dal li parite? Are him-DAT them given Q the-money b. Dal li si mu (gi) parite? Given Q are him-DAT them the-money ‘Have you given him the money?’ Bošković (2001) proposes one and the same starting point for the analysis of (62) and (63). This starting point, shown in (64), involves copying a complex head that is left-adjoined to the interrogative particle li. (64) [si+mu+gi+dal] i + li … [si+mu+gi+dal] i In Macedonian, pronominal clitics are proclitic and the particle li is enclitic. In this case, the most economic option is just one application of full deletion on the lower copy. See (65). (65) [si+mu+gi+dal] i + li … [si+mu+gi+dal] i The structure in (65) corresponds to the fully-grammatical sentence (62a). The sentence in (62b), on the other hand, can be ruled out, since it would be the result of scattered deletion, that is to say, it would involve more than one application of deletion. Now, according to Bošković, Bulgarian grammar considers pronominal clitics and the particle li as enclitics. This morphological constraint would be strong enough to make unacceptable full deletion on the lower copy (63a). 34 On the other hand, this constraint suffices to make optimal the version in (63b) with scattered deletion, that is to say, the version with two applications of PF-deletion, one in each copy, as seen in (66). 34 Roumyana Pancheva (p.c.) specifies that the constraint involved in this case is non-initially, that is li cannot surface in sentence initial position. 80 (66) [si+mu+gi+dal] i + li … [si+mu+gi+dal] i Interestingly, the deleted part on the higher copy is complementary with respect to the deleted part in the lower copy, so at the end, one copy of each item survives. This ‘survival’ of at least one copy (and no more than one copy) of each item is necessary to comply with the Interface Condition (see (28)). 35 4.6.2. English stress assignment and tone-sandhi in Taiwanese (Vergnaud 2014) The complementary character of the surviving parts in copy and (scattered) deletion is explicitly highlighted in Vergnaud (2014), along with the idea that deletion can be an alternative to reconstruction. To summarize, there is a reduplication (copy) transformation which applies to certain domains. Complementary parts of the two identical domains are pronounced. The process of complementary deletion is formally identical to the LF “reconstruction process” assumed in Chomsky 1993. (Vergnaud 2014: 193). The previous quote is based on two cases for which copy and complementary deletion is invoked: the assignation of sentential stress to wh-phrases in English and a tone sandhi phenomenon (TS) related to the complementizer kong in Taiwanese (Vergnaud 2014: 191 ff). As for stress in English, it is noted that, according to Chomsky & Halle (1968) and Cinque (1990), sentential stress in English is normally assigned to a unit in sentence-final position, as shown in (67). 36 (67) a. Helen has written [some BOOKS] b. Helen has written [what BOOKS] 35 Since the scattered deletion in (66) implies two applications of deletion, it would tie, in terms of Chain Reduction, with a derivation in which full deletion is applied to both copies. In this scenario, however, the Interface Condition would suffice to select (66) as optimal. 36 The stressed unit is written with capitals in the original. The superscripts are mine. 81 Now the wh-phrase in (67b) may be raised later to form an indirect question, as in (68). (68) (One knows) [what BOOKS] Helen has written As Bresnan (1971) already noted, it looks like the displaced wh-phrase in (68) were in situ for purposes of sentential stress assignment. Vergnaud (2014: 192) considers that “the simplest and most natural way” to account for this phenomenon is to suppose a copy process targeting the clause in (67b), followed by a complementary (scattered) deletion, as seen in (69). (69) … [Helen has written what BOOKS] i [Helen has written what BOOKS] i As for the TS phenomenon in Taiwanese, Vergnaud (2014) presents some data from Simpson & Wu (2002). In Taiwanese, the complementizer kong appears normally in clause- initial position. However, in some cases, it may appear in final position, as in (70). (70) A-hui siong A-sin m lai kong A-hui think A-sin NEG come KONG ‘A-hui thinks A-sin is not coming’ The presence of kong in final position has a topic/focus interpretation: the content of IP is understood as topic, while the focus falls on the emphatic complementizer itself. The speaker uses this structure to express his/her strong commitment with the content of IP. Simpson & Wu (2002) suggest that, in this use, kong is similar to sentence-final I’m telling you in (71). (71) He’s gone, I’m telling you! Now with respect to TS, some facts taken as unexpected for Taiwanese grammar arise when kong is in sentence-final position. First, the final syllable of the IP (lai) does not undergo TS. Second, the complementizer kong does undergo TS. Interestingly, however, these same two facts are naturally expected when kong appears before its complement, as in (72). 82 (72) A-hui siong kong A-sin m lai A-hui think KONG A-sin NEG come ‘A-hui thought that A-sin was not coming’ According to Vergnaud (2014: 193), also in this case the simplest and most natural explanation for the unexpected facts related to TS in (70) is to suppose that there has been a process of copy targeting the clause kong-IP, so a bi-clausal structure is built. Interpreting this idea, I propose in (73) the following bi-clausal structure, with the correspondent complementary/scattered deletion. (73) (A-hui siong) [ [kong A-sin m lai] i [kong A-sin m lai] i ] If a structure like (73) underlies the sentence in (70), the TS facts that were considered unexpected are no longer unusual. Scattered deletion has applied here to yield kong in final position. As previously indicated, this position puts the focus of the sentence on this particle. 4.6.3. English extrapositions (Wilder 1995) Another application of scattered deletion in the context of CTM is the analysis proposed in Wilder (1995) to account for structures known as extrapositions in English. See an example of this in (74). (74) We talked about the claim yesterday that Mary will hire Peter. In this example, the head of a complex DP gets separated from its complement CP by an intervening adverb, which creates an apparent discontinuity in the aforementioned complex DP (the underlined part in (74)). The traditional analysis of (74) appealed to rightwards movement (i.e. extraposition) of the complement CP (that Mary will hire Peter), as seen in (75). (75) We talked about the claim t i yesterday [ CP that Mary will hire Peter] i 83 The traditional view supposed that the complement CP was rightwards moved to any higher position, usually IP or VP. However, since the anti-symmetrical view of syntax put forward by Kayne (1994) became the new standard, right adjunction was no longer available as a licit operation in syntax. Wilder (1995: 292) proposes an alternative analysis from the point of view of movement as copy (CTM). In this proposal, the PP complement of the main verb is generated to the right of the adverb, and from there it undergoes leftward movement, that is to say, it is copied and merged in a higher position, as shown in (76). 37 (76) We talked [ PP about the claim that Mary will hire Peter] i yesterday [ PP about the claim that Mary will hire Peter] i Now, the structure in (76) is subject to what Wilder calls Chain Internal Selective Deletion, as described in (77). (77) Chain Internal Selective Deletion (CISD) Phonological deletion can remove part of the antecedent and the complementary part of the trace (Wilder 1995: 292). According to (77), CISD is an operation by which PF-deletion is applied in a scattered fashion. Crucially, CISD removes complementary material, which makes possible for the LCA to compute precedence relations. When CISD is applied to (76), we get the structure in (78). (78) We talked [ PP about the claim that Mary will hire Peter] i yesterday [ PP about the claim that Mary will hire Peter] i 38 37 Interestingly, in (76) the temporal adverb is taken as ‘pivot’ for the copy operation, which suggests that the host of the (newly created) high copy is [Spec,AdvP]. Unfortunately, Wilder (1995) does not further elaborate on this intuition. On the other hand, weight of constituents may be at stake. The adverb is considerably lighter than the PP complement. In that case, however, the adverb may have surfaced optimally after the verb and before the PP. 38 The author considers the CISD as composed of two interacting operations: Backward Deletion (BWD) and Forward Deletion (FWD). BWD delete rightmost material in the high copy, and FWD deletes the complementary leftmost material in the lower copy (Wilder 1995: 294). Interestingly, as a result, the topic-like content survives to the left in the higher copy of PP, and the focus-like content survives to the right in the lower copy of PP. 84 Note that in (77) deletion has applied once in the high copy (on the CP complement of the complex DP) and three times in the lower copy (on P, Det, and N), which gives four applications in total. In minimalist terms, this is a costly derivation. When compared with the two convergent structures in (79a,b), which appeal to just one application of deletion each, the competing derivation (78) should be taken as sub-optimal. (79) a. We talked [ PP about the claim that Mary will hire Peter] i yesterday [ PP about the claim that Mary will hire Peter] i b. We talked [ PP about the claim that Mary will hire Peter] i yesterday [ PP about the claim that Mary will hire Peter] i No explanation for this is offered in Wilder’s paper. Nunes (2004: 30) speculates that a possible reason to take (78) as optimal in spite of the more-economical versions in (79) may be related to some prosodic mark on the CP complement. However, the issue remains obscure. Leaving these details aside, Wilder’s proposal can be taken as a valuable early attempt of solving a problem that emerged when a restricted version of grammar (one that assumes Kayne’s anti- symmetry) became dominant. Importantly for this work, Wilder appealed to the CTM, plus selective scattered deletion for that purpose. 4.6.4. XP-split constructions in German and Croatian (Fanselow & Ćavar 2002) A more elaborated proposal about selective scattered deletion is offered in Fanselow & Ćavar (2002) under the name of distributed deletion. In this proposal, a CTM-view is explicitly assumed. The authors focus their attention in what they call XP-split constructions, where XP can be DP or PP only. These XPs surface as discontinuous units. See in (80) some relevant examples (Fanselow & Ćavar 2002: 65). 39 (80) German a. Interessante Bücher hat sie mir keine aus Indien empfohlen Interesting books has she me none from India recommended ‘She has not recommended any interesting books from India to me’ 39 I keep here the authors’ good choice of underlining the split-XPs to make evident their discontinuity. 85 Croatian b. Knijge mi je Marija zanimljive preporučila Books me has Mary interesting recommended ‘Mary has recommended interesting books to me’ German c. Mit was hast du für Frauen gesprochen? With what have you for women spoken ‘With what kind of women did you speak?’ Croatian d. Na kakvo se Ivan stablo penje? On what-kind-of self I. tree climbs ‘On what kind of tree does Ivan climb?’ Sentences (80a,b) illustrate discontinuous DPs, while (80c,d) exemplify discontinuous PPs. 40 For the case of DPs, Fanselow & Ćavar (2002: 66) point out that the standard analysis is the one proposed in van Riemsdijk (1989), according to which “the part of the XP that appears in clause-initial position is moved out of XP, stranding the material left behind”. As for PPs, the standard analysis involves remnant movement of an XP such that some material has been taken out of that XP before it undergoes fronting movement. Nevertheless, these standard-movement analyses have been challenged by what seem to be systematic violations to island constraints, which the authors review in detail in their article, among other shortcomings. In order to circumvent those challenges, Fanselow & Ćavar (2002: 66) propose a CTM-based analysis, exemplified for the Croatian sentence (80b), as follows. 40 The authors observe, “DPs and PPs may simply be split in a considerable number of languages such as German, Croatian, Polish, Russian, Hungarian, Finnish, Latin, Ancient Greek, and Warlpiri” (Fanselow & Ćavar 2002: 66). 86 (81) Croatian a. mi je Marija [zanimljive knijge] preporučila me has M. interesting books recomended b. Copy+Merge DP → [zanimljive knijge] i mi je Marija [zanimljive knijge] i preporučila c. Partial deletion in the upper copy → [zanimljive knijge] i mi je Marija [zanimljive knijge] i preporučila d. Complementary deletion in the lower copy → [zanimljive knijge] i mi je Marija [zanimljive knijge] i preporučila ‘Mary has recommended interesting books to me’ Fanselow & Ćavar (2002: 84) show that the same structure (82b) underlies two kinds of topicalization in German: standard topicalization (82c) and split topicalization (82c’). The result would depend on the kind of deletion: full deletion or distributed (scattered) deletion. (82) German a. Hat er [keine Bücher] gelesen Has he no books read b. Copy+Merge DP → [keine Bücher] i hat er [keine Bücher] i gelesen c. Full deletion of the lower copy (a.k.a. 'overt movement') → [keine Bücher] i hat er [keine Bücher] i gelesen c’. Distributed (scattered) deletion in both copies → [keine Bücher] i hat er [keine Bücher] i gelesen 87 ‘He has read no books’ The authors also differentiate two types of split-XP. If the split-XP has the same linear order than the equivalent non-split XP, it is called pull split, as in examples (83). If the split-XP reverses that order, then it is called inverted split; see (85). (Fanselow & Ćavar 2002: 68). (83) Croatian a. Na kakav je Ivan krov skočio? On what-kind has I. roof jumped? 'On what kind of roof has Ivan jumped?' German b. Wieviel hat er Bücher gelesen? How-many has he books read 'How many books has he read?' (84) Croatian a. Crveni je Ivan auto kupio. Red has I. car bought ‘Ivan has bought a red car’ German b. Autos besitzt er (nur) schnelle. Cars owns he only fast ‘As for cars, he owns only fast ones’ Now, for the purposes of this work, it is useful to review the abstract analysis of split-XPs proposed by Fanselow & Ćavar (2002: 88 ff). In first place, the authors note that split-XPs are usually related to the divide topic/focus. The right part of XP encodes the topic, whereas the left 88 part of XP encodes the focus. 41 Second, the authors consider that features as [+wh], [+focus], [+link-topic], etc. are semantic or pragmatic features. As such, they are considered “operator features”. Now, let’s consider an abstract structure like the one in (85). (85) [ H q … [H p … [ XP a p [b c] q ]]]] Superscripts are understood as features of a head (H) or of a lexical item. In (85), the most external head bears the feature q, and the next (embedded) head bears the feature p. The (soon-to-be-split) XP bears both p and q. In this configuration, both heads may, in principle, attract XP. However, this attraction is subject to the Minimal Link Condition, so initially only p, not q, can be attracted. Suppose that the correspondent head H p attract XP to its Spec. Then a copy of XP merges into [Spec,H p ]. The result is shown in (86). (86) [ H q … [[ XP a p [b c] q ] H p … [ XP a p [b c] q ]]]] Once in [Spec,H p ], the feature p is checked, so a p becomes inactive and ceases to block the attraction exerted by H q in terms of relativized minimality (Rizzi 1990). Thus, this time, H q attracts XP to its Spec position. A copy of XP merges into [Spec,H q ], as shown in (87). (87) [[ XP a p [b c] q ] H q … [[ XP a p [b c] q ] H p … [ XP a p [b c] q ]]]] The structure in (87) contains three non-distinct copies of XP, which cannot be linearized as is. Thus, in order to prepare the string for linearization, deletion must apply. For that deletion to be distributed, Fanselow & Ćavar (2002: 86) propose the following generalization. (88) Suppose [the chain] C = <C 1 ,C 2 > is formed because a strong feature of H has attracted XP and suppose that H checks the operator features f 1 ... f k of XP. Then the categories bearing f 1 ... f k must be spelt out in C 1 . 41 Interestingly, the main proposal of this work contains a similar divide left-right, although in terms of the divide presupposition (to the left) vs. focus (to the right. 89 If we further suppose that the features p and q in (87) are both strong features, given (88), the part of XP bearing the operator feature p should overtly appear next to the head H p , whereas the part of XP bearing the operator feature q should overtly appear next to the head H q . If this is on the right track, the following distributed deletion takes place. (89) [[ XP a p [b c] q ] H q … [[ XP a p [b c] q ] H p … [ XP a p [b c] q ]]]] The abstract structure in (89) corresponds to the inverted splits, exemplified in (84). Since, according to the authors, these are the split structures by default, this is seen as a welcome result. Now, in order to get the pull splits, like the ones illustrated in (83), an additional constraint is invoked. This is called the Parallel Movement Constraint, and it is attributed to Müller (2001). (90) Parallel Movement Constraint (PMC) If A c-commands B at level L, then A c-commands B at level L’ According to this proposal, the PMC would determine that the base-generated c- command relations are to be preserved in the phonetic component. This preservation would generate pull splits (Fanselow & Ćavar 2002: 89). The authors then present what they call the “traffic rules” of the system they have just sketched. For the purposes of this work, however, it suffices to say that the abstract analysis summarized in (85)-(90) provides foundations for a scattered style of deletion in terms of attraction, Copy+Merge, and feature checking. These concepts are independently motivated and are fully legitimate in the context of minimalism. 42 42 In turn, Ott (2009) expands the scope of the split-XP analysis put forward by Fanselow & Ćavar (2002) with what the former author, following Pafel (1995), calls multiple NP splits. These constructions show a discontinuous unit whose parts surface in three positions of the string, as illustrated in (i) for German. (i) Bücher hat man damals interessante in den Osten keine mitnehmen dürfen books has one then interesting in the East no with-take may ‘As for books, one could not take any interesting ones to the East then’ The correspondent analysis proposed by Ott (2009: 77) resorts to distributed deletion applied on complementary parts of the three copies, as seen in (ii). 90 4.6.5. R- and Ha-constructions in Korean (Jo 2013) More recently, Jo (2013) applies a CTM analysis to account for two types of constructions in Korean, namely R-constructions and Ha-constructions, which express the information status known as predicate contrastive topic (CT). To begin with, let’s consider (91), which is a simple declarative sentence (Jo: 2013: 86). (91) Lee kyoswunim-kkeyse nolay-lul pwulu-si-ess-ta L. professor-NOM song-ACC sing-HON-PAST-DECL ‘Professor Lee sang a song’ Let’s consider a dialogue in which speaker A asks, Did professor Lee sing a song? A possible answer for that question is Yes, professor Lee sang a song, as expressed in (91). Now, imagine that speaker B wants to answer the question affirmatively, but at the same time, he or she conveys a contrast with another contrasting thought, usually implicit (for instance, that professor Lee was out of tune). This response may have the form Yes, professor Lee did sing a song (,but…). This second kind of answer contains a predicate contrastive topic (CT). Whereas in English this contrastive/concessive character is expressed via intonation (the so-called “B” fall-rise accent), it may be expressed in Korean through some specific constructions. Two of those constructions are the aforementioned R- and Ha-constructions. According to Jo (2013), they are considered interpretively equivalent. Both constructions contain the particle nun, which encodes the CT value. The R-construction is shown in (92a); the Ha-construction appears in (92b). (ii) [keine interessante Bücher [Top] ] hat man damals [keine interessante Bücher] in den Osten [keine [Foc] interessante Bücher] mitnehmen dürfen Interestingly, Ott’s analysis marks as [+topic] a unit in the leftmost copy, and marks as [+focus] a unit in the rightmost copy. Noteworthy, these marked units are not deleted—and, according to Ott, they must not be. I interpret this latter constraint as the fact that deletion cannot erase data coming from the information structure (IS), because, crucially, IS is independent from LF/PF. 91 (92) R-construction a. Lee kyoswunim-kkeyse nolay-lul pwulu-si-ki-nun pwulu-si-ess-e L. professor-NOM song-ACC sing-HON-KI-CT sing-HON-PAST-TOP Ha-construction b. Lee kyoswunim-kkeyse nolay-lul pwulu-si-ki-nun hay-ss-e L. professor-NOM song-ACC sing-HON-KI-CT do-PAST-TOP ‘Professor Lee did sing a song (,but…)’ As Jo (2013) points out, when compared with the simple sentence in (91), some differences become evident. The R-construction shows a reduplicative part (hence the “R” in its name), while the Ha-construction shows the ‘dummy’ supporting verb ha (hence the “Ha” in its name). 43 In both cases, the sentence-final roots are hosts for the affixes of tense (T) and mood (M). Korean grammars have discussed the relation between these two types of constructions. Traditionally, it has been assumed that the choice between the two types of construction is purely optional. Jo (2013) proposes that, if one assumes a model of grammar that allows morphological (post-syntactic) insertion, in the vein of Distributed Morphology (Halle & Marantz 1993), then, at some level, R-constructions and Ha-constructions are indeed the same construction. This would explain their interpretive equivalence. For the case of (92), the common source is the reduplicative structure shown in (93), whose ‘pivot’ is the CT-encoding particle nun—that is to say, the high copy is attracted to [Spec,nun]. 44 43 In this role, ha is similar to do in English. Precisely, when used as a main verb, ha means ‘to do’. The form hay in (94b) is just an allomorphic variant. Also, -ess and -ss are allomorphs of the same morpheme, PAST. 44 In turn, it is assumed that the particle ki is inserted post-syntactically. 92 (93) [Lee kyoswunim-kkeyse nolay-lul pwulu-si]-ki-nun [Lee kyoswunim-kkeyse nolay-lul pwulu-si]-ess-e Now, if deletion applies partially on the lower copy, as shown in (94), the R-construction (92a) arises. (94) [Lee kyoswunim-kkeyse nolay-lul pwulu-si]-ki-nun [Lee kyoswunim-kkeyse nolay-lul pwulu-si]-ess-e On the other hand, when full-deletion applies on the lower copy, the verbal root (pwulu) is erased, so the morphemes T and M are left stranded. As a rescue strategy, the dummy verb ha is inserted post-syntactically. This yields the structure in (95), which is at the base of the Ha- construction in (92b). (95) [Lee kyoswunim-kkeyse nolay-lul pwulu-si]-ki-nun [Lee kyoswunim-kkeyse nolay-lul pwulu-si]-hay-ss-e The process of copy and deletion summarized here is extended to cover variants of both R- and Ha-constructions, as well as other related structures in Korean. From the point of view of the CTM, this means a positive extension of its explanatory capability. I must mention, however, an aspect of Jo’s analysis that attracts my attention when compared with other proposals previously reviewed here (Bošković 2001, Vergnaud 2014, Wilder 1995, Fanselow & Ćavar 2002, Ott 2009). The fact is that the partial deletion proposed by Jo is not really complementary. Deletion on the lower copy in (95) is total, so as expected, the higher copy is not affected. Deletion in (94) is partial, but there’s no deletion of the complementary material in any other copy in the chain. In the schema (97), related to the tree in (96), none of the partial deletions is complementary (Jo 2013: 95). 45 45 Jo declares that “VP*” is a verbal category that includes an internal subject. Although the label “vP” is not used, it seems that the two labels refer to the same object. 93 (96) FP -ki insertion after syntax VP* F’ X Y V F MP -nun TP M AgrP T VP* Agr X VP Y V (97) Variations in PF due to deletion process in the lower copy a. [X Y V]-ki-nun [X Y V]-Agr-T-M R-construction b. [X Y V]-ki-nun [X Y V]-Agr-T-M R-construction c. [X Y V]-ki-nun [X Y V]-Agr-T-M Ha-construction d. * [X Y V]-ki-nun [X Y V]-Agr-T-M 46 Ha-construction e. * [X Y V]-ki-nun [X Y V]-Agr-T-M Ha-construction The fact that deletion is not complementary poses a challenge for linearization. Let’s consider, for instance, the sequence in (97a). Partial deletion targets the lower copy of VP*, so the constituent X has been deleted. According to the tree in (96), the higher copy of VP* (and all the units contained therein) asymmetrically c-commands the CT-head –nun. At the same time, – 46 Deletions (97d,e) are considered ungrammatical because “deletion targets constituents, and the deletion of the head must accompany all its dependents” (Jo 2013: 95). 94 nun asymmetrically c-commands the lower copy of VP*. If we assume that the linearization is driven by the LCA (see (50)), we would have, on the one hand, the instruction that the copy of Y inside the higher copy of VP*, must precede –nun (i.e. Y > –nun) and, on the other hand, the instruction that –nun must precede the copy of Y contained in the lower copy of VP* (–nun > Y). In this scenario, we would have two contradictory instructions, so the linearization process should be cancelled. On the other hand, it is also possible that a morphological process of fusion has taken place in these cases, so some copies may have became invisible to the LCA, and thus are not deleted. To my view, however, it is not clear in the proposal whether such processes have occurred. In this work, following Nunes (2004), I consider that, in the context of the CTM, scattered deletion is complementary by nature, since I assume that linearization is guided by the LCA. If this supposition is correct, some additional elaboration of Jo’s proposal is required. 5. Summary In this chapter I have presented the theoretical foundations for the main proposal defended in this work. To do so, I have introduced first the idea of displacement as a key characteristic of human language, in virtue of which some linguistic units are interpreted in a position different from the one they are perceived. Then I have shown the way in which generative grammar, in particular from the approach known as Government and Binding (GB) theory, has dealt with displacement. In the generative framework, the main metaphor is that of movement. For the GB theory, a simple but powerful rule, namely Move-α, interacting with a limited set of sub-theories or modules, successfully explains displacement, at least for most of the cases. Importantly, moved elements unavoidably leave traces whenever they are displaced. These traces are a special kind of linguistic unit. They are created by movement and, while considered phonetically empty, they are taken as full in terms of formal content. Any element that moves in a cyclic fashion forms a chain with all its traces. Usually, only the head of the chain is pronounced, while all the traces remain silent. A relatively late development within the GB theory considers that elements containing traces can themselves move. This kind of movement is known as remnant movement. 95 Naturally, remnant movement creates complex series of nested dependencies. Among other applications, remnant movement has been invoked to explain the adjacency between FSUs and the focused units, known as ‘transportability’. While remnant movement accomplishes this purpose quite effectively, its high structural complexity opens questions about its plausibility facing the process of L1 acquisition. Next, I have presented a more recent direction in generative grammar known as minimalism. Just like the GB theory, this approach, assumes a conception of grammar by which there exists a set of innate universal principles, known as Universal Grammar, which is complemented with a set of open rules or parameters, whose values are to be fixed via external experience. This general conception came to be known as the Principles and Parameters approach (P&P). In the context of P&P, minimalism assumes the programmatic assumption that human language is optimally designed in order to satisfy its interface conditions. Stemming from this idea, the minimalist program has questioned the set of analytical tools at use in the GB theory. In particular, minimalism puts forward the condition by which the computational system can operate only with elements already present at the beginning of the computation. This requirement, known as the Inclusiveness Condition, eliminates traces as licit grammatical units, since these units are certainly absent at the starting point. Having dispensed with traces, the alternative is to consider that movement of some unit X consists of two successive steps: Copy X and Merge X. This basic idea gives origin to what is known as the Copy Theory of Movement (CTM). In this framework, movement is reinterpreted as Copy+Merge, and traces are understood as copies. Crucially, this approach assumes the idea that the linear order of units in a string depends on the hierarchical (syntactic) relations among those units. This dependency has been formalized in the so-called Linear Correspondence Axiom (LCA). In rough terms, the LCA invites to consider four units: a non-terminal unit X that dominates a terminal unit x, and a non-terminal unit Y that dominates a terminal unit y. Now, if X asymmetrically c-commands Y, then x will precede y in the linear string. Since copies are considered the same syntactic unit, and since one copy is merged in a c-commanding position with respect to other copies, a challenge for linearization arises. This challenge is usually solved by deleting the lower copies, on the basis of the idea that the higher copy has checked more 96 features, which explains why it is usually the head of the chain the only pronounced copy. However, it is not always the case that the lower copies are fully deleted. Higher ranked constraints in a given language may enforce partial deletion inside the copies of a chain. This means that deletion may target only part of the content of a given copy. This form of deletion is known as scattered deletion. Linearization imposes a constraint to scattered deletion: it has to be complementary. Otherwise, contradictory instructions for linearization will arise and the derivation will collapse. Deletion is complementary when at least one and just one copy of each unit inside the copies survives its application. The notion of deletion inside copies in a chain cannot be captured by any means within the traditional traces approach. This gives a descriptive advantage to the CTM with respect to the theory of traces. In general, I have shown that, so far, CTM ideas have been fruitfully used in the analysis of linguistic phenomena in English, German, Croatian, Bulgarian, Macedonian, Taiwanese, and Korean. On top of that, the CTM complies with the main conditions imposed by the minimalist approach. 97 Chapter 3 Analysis Reduplication and ellipsis in sentence modal adverbs in Spanish This chapter is oriented to present the main proposal of analysis for the apparent mismatch between surface low positions of sentence modal adverbs and their propositional syntactic scope. The high modal adverb surfaces immediately adjacent to the left-edge of the unit in focus. Since the focus-marked unit can be of virtually any category (DP, PP, VP…), this fact poses a challenge to the propositional nature of the high adverb—being propositional, one would expect it to be related to only to TPs, which syntactically encode propositions. I propose that these occurrences, preliminarily attributed to a condition called transportability (Keyser 1968), are related to the F(ocus)-marking of low constituents and to the fact that sentence adverbs are focus sensitive units (FSU). The chapter is organized as follows. In section 1, I present an analysis of some specific constructions in Spanish under the notions of reduplication and ellipsis as an alternative to focus movement (Zubizarreta 2013). In section 2, I present the main proposal of this work, namely the idea that Copy+Merge operations (reduplication) create a bi-clausal structure with the form [CP i [Mod 0 CP i ]], where both CPs are non-distinct copies. 1 This structure is the input for deletion (ellipsis), both at PF and at LF. With regards to the motivation for the proposed Copy+Merge operations, I suggest that they are triggered by the F-marking of some constituent of the sentence. The linear adjacency of the sentence modal to the unit in focus is attributed to its focus sensitivity. If the FSU is not immediately adjacent to focus, reduplication and ellipsis applies. Focus sensitivity is encoded in some feature on the modal adverb, which can be called [+foc]. This feature allows the unit to associate with focus. In accordance with the Inclusiveness Condition, [+foc] is already present in the adverb when it is selected for the Numeration, but it is active only if F-marking occurs in its relevant domain. In section 3, the main proposal is tested for Spanish on different functional configurations for the unit in focus (DO, IO, adjunct, etc), 1 For the metaphor to be more precise (think of a copy machine), one may like to consider the low-XP i as the ‘original’ from which the high-XP i is created as the first ‘copy proper’. As visible, copies of the same object are identified by the same superscript. 98 along with other structures involving topics, internal/external deletions, and prosodic prominence. Finally, section 4 summarizes the chapter. 1. Reduplication and ellipsis in Spanish Zubizarreta (2013) develops an analysis of three specific constructions in Spanish in terms of reduplication and ellipsis. These two notions are akin to the operations of Copy+Merge and deletion, respectively. 2 The specific constructions analyzed are (i) pseudo-clefts in colloquial Peninsular Spanish, (ii) bare-copula constructions in Caribbean Spanish, and (iii) standard Spanish constructions with a marked word order. It is claimed that these three constructions, in this order, represent increasing stages of grammaticalization of the assertion structure (i.e. the divide focus/ presupposition). Informational notions, like contrast, topic, and focus, may be expressed via specific syntactic configurations in natural languages. The family of constructions known as clefts in English is probably the most studied of these configurations. For the case of specificational pseudo-clefts in Spanish, Zubizarreta (2013) follows the main proposal put forth by Schlenker (2003) and Den Dikken et al. (2000), according to which there is a hidden bi-clausal structure in this kind of constructions. The initial motivation for this idea comes from the so-called “connectivity problem”. In sentence (1), where the gap in the first clause corresponds to the wh- word and the copula is understood as an equative operator Eq, the full grammaticality of the anaphoric dependency is hard to explain, since John does not c-command himself, in overt violation of Principle A (example from Schlenker 2003: 158). 2 Reduplication is a well-established notion in (morpho-)phonology. In some sense, Copy+Merge, as I have presented it here, can be seen as its syntactic counterpart. On the other hand, ellipsis is a general notion that refers to the fact that, under certain conditions involving identity, the hearer understands some semantic content from silence. In the following example, adapted from Winkler (2005: 7), the hearer only perceives (i), but interprets (ii). (i) Now you hear Ernie eating a cookie, now you don’t. (ii) Now you hear Ernie eating a cookie, now you don’t hear Ernie eating a cookie. (iii) Now you hear Ernie eating a cookie, now you don’t [hear Ernie eating a cookie]. Facing this phenomenon, the following question may arise: “Is ellipsis the result of a deletion operation or is it a base-generated empty category?” (Winkler 2005: 21). That this question can be validly posed indicates that the notion of ellipsis is more general than the notion of deletion. Moreover, even if we think that the ellipsis in (i) is the result of deletion applied to (ii), as represented in (iii), this latter operation is not complementary. Importantly, the identity that licenses the ellipsis in (i) has not necessarily been created via Copy+Merge, so deletion—in case is indeed involved—does not need to be complementary. 99 (1) [What John i likes ___ ] is himself i Schlenker (2003) claims that the solution to this puzzle lies in what this author calls the “question-in-disguise” approach. According to this view, the pseudo-cleft is indeed a Question- Answer pair a in which part of the answer has undergone ellipsis. Importantly, in terms of information structure (IS), the pre-copular clause contains the presupposition and the post- copular clause conveys the main assertion, as represented in (2). (2) [What John i likes ___ ] is [John i likes himself i ] Since ellipsis here is understood as a PF process only, according to (2) John does c- command the anaphor at LF, as seen in the post-copular clause, so no violation of Principle A is involved. In turn, Zubizarreta (2013) observes that, in Spanish, the pre-copular clause is clearly introduced by a relative pronoun, so the pre-copular clause does not really encode a question. 3 This, however, does not invalidate the bi-clausal approach per se. Sentence (3), for instance, would have the bi-clausal analysis in (4). (3) Con quien Juan i está enojado es consigo-mismo i With whom J. is angry is with.himself 3 Spanish prosody makes easy to distinguish between wh- words marked [+int] (interrogative wh-words) and wh- words marked [-int] (relative wh-words). The former, not the latter, carries primary stress. This fact is reflected even in the conventional orthography. Only the interrogative versions carry an acute accent mark (´). (i) Adivina qué [+int] compró María. Guess what bought M. ‘Guess what M. bought’ (ii) El libro que [-int] compró María The book that bought M. ‘The book that M. bought’ 100 (4) [Con quien Juan está enojado ___ ] es [Juan está enojado consigo-mismo] The bi-clausal proposal is particularly revealing in the analysis of pseudo-clefts in colloquial Peninsular Spanish (CPS). 4 In these constructions, there is an “ellipsis mismatch” between PF and LF. Zubizarreta (2013) illustrates this with sentence (5), among other examples. (5) Lo que no puedes prorrumpir es en sollozos 3SG.N that NEG (you) can burst is in tears ‘What you cannot do is burst into tears’ The interpretively equivalent pseudo-cleft in standard Spanish for (5) is shown in (6). (6) Lo que no puedes es prorrumpir en sollozos 3SG.N that NEG (you) can is burst in tears ‘What you cannot do is burst into tears’ Crucially, the bi-clausal analysis proposed for (5) involves different targets for PF- deletion and LF-deletion. 5 (7) a. PF structure [Lo que [pro no puedes [ vP prorrumpir en sollozos]]] es [pro no puedes [ vP prorrumpir en sollozos]] b. LF structure [Lo k que [pro no puedes [ vPk prorrumpir en sollozos]]] es [pro no puedes [ vP prorrumpir en sollozos]] Interestingly, in both structures in (7), deletion is complementary. In our terms, the PF- structure (7a) provides a valid input for linearization, since there are no violations of the LCA. In turn, Zubizarreta (2003) points out that the LF-structure (7b) provides an ideal input for 4 Zubizarreta (2013) indicates that these constructions have been first reported by Fernández Soriano (2009). 5 A similar mismatch has been presented when discussing the advantages of the CTM with respect to the traditional traces approach. 101 interpretation, since the LF-deleted vP in the pre copular clause can be understood as a variable bound by the determiner lo (as indicated by the subscript k), which would function as a binding operator. The presupposition then would be encoded in the pre-copular clause, and the focus phrase would be interpreted in the post-copular clause, exactly as it is in the standard pseudo- clefts. On this issue, the analysis of the standard pseudo-cleft (6) is shown in (8). (8) a. PF structure [Lo que [pro no puedes [ vP prorrumpir en sollozos]]] es [pro no puedes [ vP prorrumpir en sollozos]] b. LF structure [Lo k que [pro no puedes [ vPk prorrumpir en sollozos]]] es [pro no puedes [ vP prorrumpir en sollozos]] Unlike the colloquial version, the standard case shows total coincidence in targets for PF- deletion and LF-deletion. Again, deletion is complementary, so both structures are optimal for their respective interpretive mechanisms. 6 Now, if we consider that the standard and the dialectal versions have the same interpretation, their LF structures should be the same (this is indeed what we see in (7b) and (8b)). Thus, in both cases the relation operator-variable should have the same origin, namely the variable should have been created via LF deletion. 7 With respect to the bare copula constructions in Caribbean Spanish, 8 it is claimed that they represent a higher degree of grammaticalization of the assertion structure in comparison to the specificational pseudo-clefts. Unlike the latter, the former have no overt operator, so the variable is bound at LF by a general lambda-type operator (Op). Also Zubizarreta (2013) points out that the post-copular unit is a reduced vP (or AspP), such that the copular structure is adjoined to the matrix vP. Although defective, the reduced vP is still able of encoding the 6 The derivations in (7) and (8) start from the same numeration, so in principle they are comparable with respect to economy considerations. In order to explain the colloquial Peninsular Spanish construction, it would be necessary to find an explanation (presumably, in terms of some high ranked constraint in that dialect) for the preference for (7), which involves four applications of PF-deletion, over (8), which only requires three. 7 Zubizarreta (2013) claims instead that “in standard pseudo-clefts, the presuppositional variable is created before spell-out”, that is to say, before LF. 8 As for this kind of constructions, the main data in Zubizarreta (2013) comes from Méndez-Vallejo (2009). See also Bosque (1999) and Camacho (2006), among others. 102 presupposition, since it contains a dependent variable bound by a c-commanding T, whose tense values the dependent T-variable via Agree. An example of this is shown in (9). (9) Clara le trajo a Armando fue galletas C. DAT.Cl=brought to A. COP.3SG cookies ‘It was cookies that Clara brought for Armando’ As (10) below shows, there are no PF-LF mismatches in this kind of construction. At LF, the variable DP k in the pre-copular clause, created via LF-deletion, is bound by a general operator Op k . The bi-clausal structure in (10b) shows the divide between the pre-copular presupposition (“there exists x such that Clara brought x to Armando”) and the post-copular focus value (“x = cookies”). As Zubizarreta (2013: 21) indicates, “The copula, besides functioning as a focusing device, also marks the scope of the assertion, which determines the scope of the focus phrase contained within it”. (10) a. PF structure [Clara j [ T’ le trajo v [[ vP e j v [[ DP galletas] V a Armando]]]]] [ CopP fue [ vP e j tra√ [galletas V a Armando]]] b. LF structure Op k [Clara j [ T’ le trajo v [[ vP e j v [[ DPk galletas] V k a Armando]]]]] [ CopP fue [ vP e j tra√ [[ DPk galletas] V k a Armando]]]] Notably, all deletions in (10) are complementary. Finally, the bi-clausal structure is extended to account for marked word order in Spanish, by which the focused unit appears at the end of the sentence. 9 In that position, it is assigned nuclear stress, as proposed in Zubizarreta (1998). Consider the following examples (Zubizarreta 2013: 23 ff). 10 9 Being SVO the unmarked order for Spanish, the marked orders considered are VSO, VPSS, VOPPS, and SVPPO. 10 Following the original source, the focused unit appears in italics. 103 (11) Order VOS Me regaló el libro María DAT.Cl=gave the book M. ‘It was María who gave me the book’ (12) Order SVPPO El niño escondió debajo de la cama el juguete The boy hid under the bed the toy ‘It was the toy that the boy hid under the bed’ The basic idea is the same that was presented for the Caribbean Spanish construction, namely reduplication of vP, creation of a bi-clausal structure via adjunction, and ulterior deletion, both at PF and LF. There’s no copula though, and that is the main difference with respect to the Caribbean construction. The variable is built via LF-deletion and is bound by a general operator Op. The analysis for (11) and (12) is presented in (13) and (14), respectively. (13) a. PF structure 11 [ TP me regaló v [ vP [ vP María [v el libro]] [ vP María [regal√ el libro]]] b. LF structure Op k [ TP me regaló v [Ev-T [ vP [ vP [ DPk María] [v el libro]] [ vP [ DPk María] [regal√ el libro]]]] (14) a. PF structure [ TP El niño j escondió v [ vP [ vP e j v [ VP el juguete V debajo de la cama]] [ vP pro j escond√ [ VP el juguete V debajo de la cama]]]] b. LF structure [Op k [ TP El niño j [escondió [Ev-T [ vP [ vP e j v [ VP [ DPk el juguete] V debajo de la cama]] [ vP pro j escond√ [ VP [ DPK el juguete] V debajo de la cama]] 11 As in Zubizarreta (1998: 127), the analysis in (13) supposes that the order VOS is derived from the order VSO. 104 In (13b) and (14b), the expression “Ev(ent)-T(ime)” refers to the internal time argument, provided by vP (as opposed to Utt(erance)-T(ime), which refers to the external time argument). This ensures the propositional interpretation for vPs under the scope of the matrix tense. Although the PF-deletions can be considered complementary, that’s not the case for the LF- deletions. Now, in principle, complementary deletion is a PF requisite only, since it is related to linearization. From a minimalist point of view, however, it would be desirable to try to maintain the parallelism PF/LF as an extension of the Uniformity Condition. 12 From this analysis, Zubizarreta (2013) concludes, “focus movement is unnecessary for the cases at hand, since ellipsis accomplishes what movement would otherwise do”. In general, the set formed by reduplication/adjunction plus ellipsis is seen as an advantageous alternative to previous analyses that appealed to movement. Particularly for the case of the marked orders in Spanish, it is observed that, in previous approaches, movement was invoked, either as prosodically motivated movement (p-movement, in Zubizarreta 1998) or as syntactically motivated movement driven by feature-checking (Ordóñez 1998, Belletti 2004). But there again, movement seem to be unmotivated and anti-economical. Why do we have recourse to vP adjunction (or vP reduplication […]) in conjunction with ellipsis to locate the focus at the left edge of the vP phase, instead of having recourse to focus-movement to the (left) edge of the vP phase? A possible answer to that question is the following: there is no syntactic 12 In turn, Ortega-Santos (2006) proposes an alternative analysis of the marked word orders in Spanish explicitly situated in the CTM framework. This approach takes as its starting point that DPs check their structural Case outside their theta domains. In consequence, it supposes that all the arguments of the verb vacate VP and must move out of VP to check their Case in different Agr positions (the verb also moves out of VP, but for independent reasons). Since movement is understood as copying, this movement creates successive copies of each argument- plus-verb over which deletion applies selectively to give rise to the different orders. For instance, from the basic sentence Pedro le dio un libro a María ‘P. gave a book to M.’, deletion applies to generate the marked word orders (i-ii), with focus on the final constituent. In (i), the order is V-DO-IO-S. In (ii), the order is S-V-IO-DO (examples from Ortega-Santos 2006: 57). (i) [ TP Pedro le dio [ AgrS Pedro dio [ AgrDO un libro dio [ AgrIO a María dio [ V1 Pedro dio [ V2 un libro dio a María]]]]]] (ii) [ TP Pedro le dio [ AgrS Pedro dio [ AgrDO un libro dio [ AgrIO a María dio [ V1 Pedro dio [ V2 un libro dio a María]]]]]] Although my own proposal is based on the CTM, it does not understand the main structure of a sentence as formed by a series of Agr positions, so I don’t follow Ortega-Santos’ suggestion on this issue. 105 motivation for such a movement because focus is not a syntactic feature; in other words, there is no focus-feature checking operation. (Zubizarreta 2013: 29). In this work, I plainly assume the idea that focus is not a syntactic feature. Focus is neither a semantic nor a syntactic notion, but instead it belongs to a different structure, namely IS. On the other hand, it is true that focus marking has syntactic, interpretive, and prosodic effects. In that sense, I also assume that that there may be a formal feature encoding sensitivity to focus. This feature may explain how certain lexical items can associate with focus-marked units once they are put together in the same relevant syntactic configuration. More specifically, I propose that sensitivity to focus is lexically encoded, that is to say, certain lexical units (the FSUs) are already loaded in the lexicon with some feature that captures its sensitivity to focus and also explains their immediate adjacency to the the F-marked units in certain cases. I provide a possible explanation on these lines in the next section. For the time being, I consider as particularly insightful for the purposes of this work the idea of ellipsis as a way to locate the focused unit at the edge of the vP-phase, the suggestion that a variable can be created at LF via LF-deletion, and most importantly, the idea of a hidden bi-clausal structure, on which partial deletion operates selectively to finally yield the observed linear string and the perceived semantic interpretation. 2. Main proposal. The [+foc] feature 2.1. An inventory of resources Let’s start with a brief account of the conceptual bases for the main proposal reviewed so far in this chapter. (15) a. Movement as Copy+Merge (CTM) In compliance with the Inclusiveness Condition (see (31)), traces are replaced by non-distinct copies of a given unit. These copies form a chain. Each copy reproduces the formal, semantic, and phonetic features of the ‘original’. b. Linearization of chains Linearization of chains is driven by the LCA. In principle, low copies are full- 106 deleted at PF, unless this yields a violation of some high ranked constraint in a given language. c. Scattered deletion There exists a variety of deletion in which only a part of the material in a copy is deleted. This kind of partial deletion is called scattered deletion. At least at PF, scattered deletion is complementary by nature, since it forms the input for linearization. d. Deletion at two levels In compliance with the Uniformity Condition, deletion, full or scattered, can apply both at PF (PF-deletion) and at LF (LF-deletion). For one and the same chain of copies created by Copy+Merge, the targets of PF-deletion and LF-deletion may not coincide. A PF/LF mismatch is possible. e. Creation of variables At LF, a variable can be created by LF-deletion plus Op. insertion. 2.2. ‘Transportability’ as focus sensitivity The sentence modal adverb probablemente ‘probably’ has propositional scope. It is expected then that this adverb appear in a high position of the sentence. This is indeed the case in (16). (16) Probablemente Manolo comprará una novela de Vargas Llosa. Probably M. will.buy a novel by V.L. ‘Probably M. will buy a novel by V.L.’ However, it may also surface in low sentence positions. This condition of adverbs has been called ‘transportability’ (Keyser 1968). Let’s consider the example in (17). (17) Manolo comprará una novela probablemente de Vargas Llosa. M. will.buy a novel probably by V.L. ‘It is probably V.L. the author of the novel that M. will buy’ 107 As indicated by the English gloss, the focus on sentence (17) is on the author of the novel, so the set of alternative focus-values (Alt-fv) contains only other novelists. According to a generalization presented in chapter 1, transportability can be seen as a result of sensitivity to focus. The sentence adverb probablemente is a focus sensitive unit (FSU). The hypothesis is that the adverb surfaces immediately adjacent to the focused unit in order to delimit its scope by marking its left edge, which otherwise would go unmarked, giving rise to IS ambiguity (the other edge is usually marked prosodically by the usual means that mark the end of an ip). In Velásquez (2011b), the results of an experiment involving reading of sentences containing the adverb probablemente are reported. A prosodic juncture (#), whose value in the Sp_ToBI system is 3, has been consistently found just before the adverb, which would indicate the delimitative function of the aforementioned adjacency. 13 If this hypothesis is on the right track, sentence (17) has the IS in (18). 14 (18) Manolo comprará una novela (#) [probablemente] FSU [de [Vargas Llosa] F ] fcp Association with fcP Since probablemente ‘probably’ is a high sentence adverb, with propositional scope, it is necessary to explain, in terms of syntactic structure, how come it can surface grammatically in a low position, in this case, for instance, adjacent to the sentence final DP in (18). One possibility is to assume that the adverb is some sort of ‘all-rounder’ that can directly merge with any kind of word or phrase, as long as it is focus-marked. In principle, another possibility would be to 13 Among other results, Velásquez (2011b) reports that, in a reading experiment carried out with native speakers of Spanish from Lima, Peru, the juncture (#) before the adverb tends to be realized as a pause (Ps) i.e., as an interruption of the F0 continuity, although an alternative strategy to mark the juncture, namely pitch reset (PR), was also found. Sentence % PR % Ps (i) Manolo (#) probablemente [ VP venderá mermelada de manzana] F 25.0 58.3 M. probably will.sell marmalade (made) of apples (ii) Manolo venderá (#) probablemente [ DP mermelada de manzana] F 18.2 81.8 (iii) Manolo venderá mermelada (#) probablemente [ PP de manzana] F 00.0 91.6 (iv) Manolo le venderá mermelada (#) probablemente [ PP a Ana] F 00.0 100.0 M. Cl.DAT=will.sell marmalade probably to A. 14 The material outside brackets is considered as presupposed. 108 assume that the adverb is indeed generated in a high sentence position and from there it moves downwards to adjoin the unit in focus. Now, since downwards movement is considered odd in most current derivational approaches, a third possibility is that the adverb is generated in the low position and from there it undergoes upwards movement to the CP periphery. But then again this third possibility needs to assume that the adverb can merge from the base with no matter which kind of phrase or category. These alternatives are hard to maintain for a point of view (like the one adopted in this work) that assumes that focus is not a syntactic feature and also considers syntax as a projection of features encoded in the lexical units. My proposal then is as follows. First, probablemente ‘probably’ in (18) is indeed base- generated in a high sentence position. There it selects a clause, so that its propositional scope is verified. Following Cinque (1999), I consider that that high position corresponds to the node Mod(al)P (more specifically, to the node Modal-Epistemic). Unlike Cinque, however, I don’t consider that the adverb is in a Spec position of a functional head, but rather I think it directly occupies the head position, as in the more traditional adjunction approach. Once the CP is built in core syntax, the modal adverb merges directly with CP. Second, some element in the syntactic structure of CP is F-marked. For the case of (18), it is the DP Vargas Llosa. I assume that the adverb contains a feature such that it activates the association with focus or, more precisely, the association with the focus-containing phrase (fcP), as pointed out by Krifka (2006) and highlighted by Zubizarreta (2013). This phrase contains the F-marked unit, and, in principle, can be identified as the correspondent pied-piping domain for wh-movement. For instance, the question in (19) moves the PP-fcP and the result is well- formed, unlike (20), which illicitly moves just the PP-internal focused DP and is unacceptable. (19) De qué autor comprará una novela ____ Manolo? By which author will.buy a novel M.? ‘Which author’s novel will M. buy?’ (20) * Qué autor comprará una novela Manolo de ____? 109 Third, the feature in the Mod 0 attracts the c-commanded CP to its Spec, in an EPP-like fashion. Thus, a copy of CP merges then with Mod’, and occupies [Spec,ModP]. The focus feature [+foc] in Mod 0 encodes sensitivity to focus, and at the same time, triggers syntactic leftwards CP movement—understood here as Copy+Merge. When the copy of CP has been merged, a bi-clausal structure appears, with the general form indicated in (21a). For the example in (18), the bi-clausal structure will look like (21b). (21) a. [ XP CP i [ X’ FSU [+foc] [CP i ]]] b. [ ModP [ CP Manolo comprará una novela [ PP de V.L.] ] i [ Mod’ probablemente [+foc] [ CP Manolo comprará una novela [ PP de V.L.] fcp ] i ]] Fourth, a process of PF-deletion targets the low copy, driven by the Linear Correspondence Axiom (LCA) and the constraint Formal Feature Elimination (FFE). Deletion in the low copy proceeds. Crucially, this deletion stops when it arrives to the left edge of the fcP, which cannot be deleted. This is due to the following. According to the CTM expressed in (15a), the process of copying reproduces the formal, semantic, and phonetic features of the ‘original’ unit. Importantly, focus is neither a formal (syntactic), nor a semantic, nor a phonetic feature, so Copy cannot reproduce F-marking. The high copy of the syntactic unit marked as fcP does not convey focus in the higher clause. This implies that deletion cannot apply on the fcP, because, otherwise, some information relevant for interpretation will be lost, and then a violation of Full Interpretation and the Interface Condition would arise. 15 Actually, this restriction has been presented by Han & Romero (2004) as a constraint militating against deletion of a focus- marked constituent. 15 In this line of reasoning, Ott (2009), when discussing distributed deletion in German, writes, “Notice that in the cases discussed here, the lowest copy always contains a focus-bearing element. For this reason, it cannot be deleted fully (a focus-marked element must be pronounced)” (Ott 2009: 78, fn. 15). 110 (22) Focus Deletion Constraint (FDC) Focus-marked constituents at LF or their phonological locus cannot delete at Spell- Out. (Han & Romero 2004: 199). In order to have just one copy of each unit in the input for linearization, as anticipated in (15c), the (F-unmarked) high copy of the fcP-marked unit must undergo PF-deletion. Therefore, this PF-deletion is, in principle, scattered and complementary. 16 For the case of (18), the result of these processes at PF would look like (23a,b). (23) a. PF-structure [ ModP [ CP Manolo comprará una novela [ PP de V.L.] ] i [ Mod’ probablemente [+foc] [ CP Manolo comprará una novela [ PP de V.L.] fcP ] i ]] b. ModP CP i Mod’ M. comprará una novela de V.L. CP i Mod 0 M. comprará una novela [de V.L.] fcP probablemente [+foc] Now with respect to the LF-structure, there are no reasons for a mismatch, so we can assume that LF-deletion targets the same constituents as PF-deletion. Importantly, by deleting the high copy of the PP, a variable is created, according to (15e). Now we can interpret the 16 Unless the whole CP is in focus, or a mismatched deletion occurs. On these issues, see below. 111 presupposition in the high copy by considering that the variable is bound by a silent lambda-type operator Op. (24) LF-structure OP k [ ModP [ CP Manolo comprará una novela [ PPk de V.L.] ] i [ Mod’ probablemente [+foc] [ CP Manolo comprará una novela [ PP de V.L.] fcP ] i ]] Note that, in (24), the first line expresses to the presupposition, whereas the second line encodes the main assertion. Note that it is the identification of the focus value what is seen under the lens of probability. 2.3. The bi-clausal derivation. Basic analytical remarks Now, (25) shows the proposed derivation for sentence (19). (25) N = {Probablemente 1 , C 0 1 , Manolo 1 , compr- 1 , -ará 1 , una 1 , novela 1 , de 1 , Vargas-Llosa 1 } a. Usual operations in narrow syntax to yield CP 17 → [ CP [ TP Manolo comprará una novela de Vargas-Llosa]] b. Merge probablemente + CP → [ Probablemente [ CP [ TP Manolo comprará una novela de Vargas-Llosa]]] c. F-marking on the sentence-final DP Vargas-Llosa → [Probablemente [ CP [ TP Manolo comprará una novela de [Vargas-Llosa] F ]]] d. fcP-marking of the pied-piping domain, the PP de Vargas-Llosa → [Probablemente [ CP [ TP Manolo comprará una novela [de [Vargas-Llosa] F ]] fcP ]] e. Copy CP, triggered by [+foc] on Mod 0 → [ CP [ TP Manolo comprará una novela de Vargas-Llosa]] f. Merge CP + Mod’ (CP appears on [Spec,ModP]) → [ ModP [ CP [ TP Manolo comprará una novela de Vargas-Llosa]] [ Mod’ probablemente [ CP [ TP Manolo comprará una novela [de [Vargas-Llosa] F ]] fcP ]]] 17 I am considering CP, and not only TP, as a means to recognize the fact that there may be other sentential projections below ModP, as for instance TopP. In (25), C 0 is just an empty position. 112 g. PF-deletion on the low copy, driven by the LCA → [ ModP [ CP [ TP Manolo comprará una novela de Vargas-Llosa]] [ Mod’ probablemente [ CP [ TP Manolo comprará una novela [de [Vargas-Llosa] F ]] fcP ]]] h. Complementary PF-deletion on the high copy → [ ModP [ CP [ TP Manolo comprará una novela de Vargas-Llosa]] [ Mod’ probablemente [ CP [ TP Manolo comprará una novela [de [Vargas-Llosa] F ]] fcP ]]] g’. LF-deletion on the low copy → [ ModP [ CP [ TP Manolo comprará una novela de Vargas-Llosa]] [ Mod’ probablemente [ CP [ TP Manolo comprará una novela [de [Vargas-Llosa] F ]] fcP ]]] h’. Complementary LF-deletion on the high copy (creation of a variable) → [ ModP [ CP [ TP Manolo comprará una novela de Vargas-Llosa]] [ Mod’ probablemente [ CP [ TP Manolo comprará una novela [de [Vargas-Llosa] F ]] fcP ]]] i’. Binding of the variable by an abstract operator Op → Op k [ ModP [ CP [ TP Manolo comprará una novela [ PPk de Vargas-Llosa]]] [ Mod’ probablemente [ CP [ TP Manolo comprará una novela [de [Vargas-Llosa] F ]] fcP ]]] As seen in (25b), the adverb is merged at the left periphery of CP, following standard syntactic procedures. Importantly, this position corresponds directly to its inherent propositional scope. In (25c), some constituent of the sentence gets F-marked; in this case, it is the final DP, which in turn is embedded inside the DP-DO. Notably, F-marking at this point and fcP-marking in the next step imply that IS-motivated steps can be interspersed with syntactically-motivated steps within the same derivation. This possibility, however, is not unconstrained. The main restriction is given by the notion of phase. As it is well-known, by its own nature, focus introduces alternatives. The interpretation considers a set of alternative focus-values (Alt-fv), and also a set of alternative propositions (Alt-p). For this latter set to exist, a plain (not yet F- marked) proposition must previously exist. In other words, F-marking can introduce alternative propositions only if a minimal propositional construction already exists in the derivation. Following Chomsky (2001), I consider here that the minimal syntactic units capable of encoding 113 propositions are vP and CP. Precisely, these units are considered phases. 18 Thus, F-marking can only proceed when a minimal phase has been already built. 19 From the prosodic point of view, being in sentence final position, the final DP in (25) is in the right place to be assigned nuclear stress, according to the NSR. In the case of Spanish, like other Romance languages, the phrasal prominence related to non-contrastive focus is guided by the C-NSR version of the NSR, as indicated in the following formulation from Zubizarreta (1998). 20 (26) Constituent-driven Nuclear Stress Rule (C-NSR) Given two sister categories C i and C j , the one lower in the asymmetric c-command ordering is more prominent. (Zubizarreta 1998:19) Being the F-marked unit in final position, the nuclear stress assigned by the C-NSR to that constituent will not be in conflict with the prosodic salience derived from the Focus Prominence Rule (FPR). (27) Focus Prominence Rule (FPR) Given two sister nodes C i (marked [+F]) and C j (marked [-F]), C i is more prominent than C j . (Zubizarreta 1998: 88) 18 For a similar point of view with respect to phases as the syntactic environment for modal adverbs, see Giorgi (2013) and Giorgi (to appear), where an analysis of modal sentence adverbs used as parentheticals is proposed. 19 At first sight, it may seem that F-marking can independently target a sub-sentential constituent: The lady with the [black] F dress. However, there’s no way to build the set of alternative propositions (Alt-p) from that F-marking, if taken in isolation. Alt-p can be built only if DP is already part of a propositional unit. Say we include the DP in a sentence: John invited the lady with the [black] F dress. Thus Alt-p = {John invited the lady with the black dress; John invited the lady with the red dress; John invited the lady with the purple dress,…}, etc. 20 Germanic languages follow another NSR rule, namely the S(election-driven)-NSR. The S-NSR is defined as follows, “Given two sister categories C i and C j , if C i and C j are selectionally ordered, the one lower in the selectional ordering is more prominent” (Zubizarreta 1998: 19). 114 Having factored out possible conflicts due to prosody, the next step in the derivation, (25d), shows the delimitation of the focus-containing phrase (fcP). 21 I present this notion in the following sub-section. 2.4. Association with focus-containing phrase (fcP) Following a suggestion by Krifka (2006), I assume that association with focus is, indeed, association with focus-containing phrase (fcP). 22 As its name indicates, the fcP includes the F- marked unit. Naturally, the relation of inclusion itself previews that both units may coincide in some cases, but crucially not in all. The F-marked unit may also be properly contained within the fcP. When discussing sensitivity to islands in association with focus, Krifka (2006) showed that the fcP corresponds to the pied-piping unit that undergoes LF-movement in questions. Krifka (2006: 131) attributes the original observation to Nisigauchi (1990), for Japanese. 23 (28) a. John introduced [the author of [which] F novel] to Sue? b. LF: [the author of [which] F novel] i [John introduced t i to Sue] According to Krifka’s semantic approach, known as Structurated Meaning, a question and a congruent answer share the same presupposition (background, in his terms). In (29a,b), we see two possible congruent answers for the question in (127a), whose common LF is represented in (29c). On the other hand, Krifka judges that (29d) is not a good answer. (29) a. John introduced [the author of [Ulysses] F ] to Sue. b. The author of [Ulysses] F . c. LF: [the author of [Ulysses] F ] i [John introduced t i to Sue] 21 Indeed, steps (25c,d) may occur in one single step, but I take them separately here for the sake of clarity. 22 Krifka (2006) uses the term “association with focus phrase”, abbreviated as “FP”. I prefer to use the more explicit term “focus-containing phrase”, abbreviated as “fcP” mainly to avoid potential confusions with the term “focus phrase”, abbreviated “FocP”, used in structural descriptions within the split-CP cartographic tradition, as for instance in Rizzi (1997, 2004), among many others. 23 Krifka (2006) represents the focused unit with accent marks. I use the subscript “F” on the left bracket for the same purpose. 115 d. * [Ulysses] F . Interestingly, Krifka observes that the answer in (29b), which this author calls the short answer, is the result of deletion of the background (since the common background has already appeared in the question) applied on (29a). Crucially, deletion cannot target material inside the fcP, even if it may seem ‘given’ in the sense that it already appeared explicitly mentioned in the question. In Krifka’s own words, “these are short, elliptical answers that are the result of deletion. The material that can maximally be deleted is the material outside of the focus phrase [=fcP], the phrase that has to be LF-moved to create a question meaning” (Krifka 2006: 132). That kind of deletion is seen in (30). (30) a. John introduced [the author of [Ulysses] F ] fcP to Sue b. Deletion of common background → John introduced [the author of [Ulysses] F ] fcP to Sue. c. = The author of [Ulysses] F . (= The short answer, in (29b)) In turn, the inappropriateness of the answer in (29d) can be related to the ungrammaticality of the correspondent parallel LF movement (31b), which violates a syntactic island (31a). 24 (31) a. LF: * [[which] F novel] i [John introduced the author of t i to Sue] b. LF: * [[Ulysses] F ] i [John introduced the author of t i to Sue] Additional support for this view comes from what Krifka calls ellipsis in explicit contrast (Krifka 2006: 122 ff). As illustrated in (32c-e) and (33c-e), deletion under identity in the second clause cannot apply inside the second fcP, which is necessarily parallel to the first fcP in the first clause. 24 Now that the fcP can—at least provisionally—be identified with the correspondent pied-piping domain for questions, a deeper issue arises, namely how come the computational system gets informed about the boundaries of the fcP during the derivation. For the time being, I descriptively assume a fcP-marking process either parallel or posterior to the normal F-marking process. However, this double F-marking seems anti-economical, so it would be desirable to reduce it to just one single process. Although some sort of upwards percolation of [F] comes to mind, I will leave this question open. For more about pied-piping and its relation to focus, see Sportiche (1995), Koopman (1996), Cable (2012), Erlewine & Kotek (to appear), and Erlewine (to appear), among others. 116 (32) Mary didn’t invite [the man in a [black] F suit] fcP to the party, a. … but she invited the man in a [purple] F suit. b. … but the man in a [purple] F suit. c. … * but in a [purple] F suit. d. … * but a [purple] F suit. e. … * but a [purple] F (one). (33) John only invited [the woman in a [black] F dress] fcP to the party, a. … he didn’t invited the woman in a [red] F dress. b. … not the woman in a [red] F dress. c. … * not in a [red] F dress. d. … * not a [red] F dress. e. … * not a [red] F (one). For the case of the association with the sentence modal adverb probablemente ‘probably’ in Spanish, it is important to remember that the association is established between the adverb and a unit marked with informational focus. Informational focus can always be obtained as a response to a grammatical question. See for instance (34), where the adverb appears deeply embedded, adjacent to the focused unit. (34) Juan comprará mermelada de naranja probablemente [agria] F J. will.buy marmalade of orange probably sour ‘Juan will probably buy marmalade made with SOUR oranges’ The sentence in (34) can be felicitously used as a valid answer to the grammatical question posed in (35). (35) De qué (clase de) naranja comprará Juan mermelada? Of which (kind of) orange will.buy J. marmalade ‘Juan will probably buy marmalade made of what kind of oranges?’ 117 Note that both the question in (35) and the answer in (34) share the presupposition that Juan will buy orange marmalade, so they form a congruent Q-A pair, in the sense of Krifka (2008b). Precisely, the answer assigns a value to the variable introduced in the question. In this sense, being the answer to a grammatical question, the focus in (34) is informative, so the adverb can surface immediately adjacent to it. Now, to be grammatical, the wh-phrase in the question (35) needs to pied-pipe the unit naranja ‘oranges’, which is part of the presupposition. Otherwise, the question would be ungrammatical. (36) * Qué comprará Juan mermelada de naranja ___? Which will.buy J. marmalade of oranges The fact that naranja ‘oranges’ must be pied-piped in order to form the grammatical question is at the base of the fact that the short answer for (34) also needs to include this noun, even if it is presupposed in the question. Let’s compare (37a) with (37b). (37) Question (= (35)) De qué (clase de) naranja comprará Juan mermelada? ‘Juan will probably buy marmalade made with what kind of oranges?’ a. Probablemente de naranja [agria] F Probably of orange sour b. # Probablemente [agria] F Probably sour There seems to be an enforced parallelism between the form of the wh-phrase and the form of the unit that survives ellipsis to form the short question. In this case, the wh-phrase is a PP that includes the noun naranja (i.e. De qué (clase de) naranja), and that is exactly what we 118 get in the short answer (De naranja agria). At least provisionally, we can conclude that the adjacent placement of the modal and the formation of the short answer respond to different (although related) conditions. Whereas the placement of the modal adverb in plain declaratives is regulated by the informational focus boundary, the short answer is directly shaped by the grammatical question to which the informational focus responds. This question is derived via movement and is subject to island constraints. Ellipsis, on the other hand, does not obey islands. This explains why probablemente ‘probably’ can surface immediately next to agria ‘sour’ in the normal declarative (34), but not in the short answer (37b). 25 On the other hand, it is also important to take into consideration the effect of definiteness on focus. Let’s consider the sentences in (38), which contain the same F-marked unit. (38) a. Probablemente, Susana comprará un vestido [de seda] F Probably S. will.buy a dress of silk ‘Susana will buy a dress made of silk’ b. Probablemente, Susana comprará el vestido [de seda] F Probably S. will.buy the dress of silk ‘Susana will buy the dress made of silk’ While both F-markings are normal and fine, they are different in nature. The use of a definite determiner in (38b) indicates that the alluded dress is salient in a certain context. For the case of focus marking, the context is given by the closed set Alt-fv (the dress made of silk vs. the dress made of cotton, for instance). This means that el vestido de seda lit. ‘the dress made of silk’ is a definite description. In that sense, (38b) would be felicitous as answer for a disjunctive question as, Which dress will Susana buy, the silk dress or the cotton dress? 26 On the other hand, (38a) will be infelicitous as answer for that question. This is so because for (38a), Alt-fv is an open set. In terms of IS, the focus in (38a) is informative, whereas the focus in (38b) is 25 There is another possibility of modalized short answer: De naranja probablemente agria. To my judgment, this answer is plainly grammatical and felicitous. It shows the adverb immediately adjacent to the focused unit and, obviously, it includes both the preposition and the noun. Undoubtedly, more research is needed on this issue. 26 Obviously more options can be added, as long as they are also definite descriptions of a dress, being the material the only relevant difference. 119 contrastive. Contrastive focus implies that the speakers have in their common ground a closed, explicit, and presumably reduced, set of alternatives. This difference can be better appreciated in (39), where the sentence adverb appears immediately adjacent to the focused unit. (39) a. Susana comprará un vestido probablemente [de seda] F S. will.buy a dress probably of silk b. # Susana comprará el vestido probablemente [de seda] F 27 S. will.buy the dress probably of silk The F-marking in (39b) is contrastive, not informative. Thus, the sentence adverb cannot surface next to it. Its only option is to appear immediately before the definiteness mark: Susana comprará probablemente el vestido de seda. This distinction is important for the structural proposal defended in this work. As anticipated in Chapter 1, this analysis applies to informative focus only. The position immediately adjacent to the contrastive focus in (39b) is not accessible for the adverb. Definiteness blocks association with focus in any internal position. Contrastive focus, on the other hand, can be expressed by other means, as for instance focus-fronting or prosodic prominence. Turning back to Kirfka’s examples (29), (32), and (33), we see in each case that the contrastive focus is contained inside a definite description. It is correctly observed that deletion cannot target material inside the definite description, even if that material can be considered as given in a previous question or statement. From this perspective, what is observed is the fact that contrastive focus can be placed inside a definite description and, in that case, the FSU associates not directly with the unit that is contrastively F-marked, but instead with the minimal unit that 27 It is important to leave aside a parenthetical intonation for the adverb. On the other hand, a similar structure may be felicitous with other adjectives such that can function as secondary predicates. This is the case of cruda ‘raw’ in the sentence Susana comió la carne probablemente cruda lit. ‘S. ate the meat probably raw”. 120 encodes the definite description. I claim that this minimal definite description is, precisely, the focus-containing phrase (fcP). 2.5. Linearization and the Inclusiveness Condition Going back to the derivation, the step (25e) represents the process of copy of the c-commanded CP. In more traditional terms, this step would have been called leftwards movement of CP. To some extent, moving a CP may have been seen as unfamiliar under the GB approach. Usually, the units considered as undergoing movement have been DPs, wh-phrases, or just heads, 28 all of them normally ‘smaller’ than IP or CP. The term heavy pied-piping have been coined to describe movement that targets more ‘masive’ material within the framework of remnant movement (Njkemnji 1995, Koopman 1996). An early proposal involving movement of a whole sentence can be found in Sportiche (1995). This author proposed movement of a whole IP to a higher left position for intonational yes/no questions in French. The example in (40) is analyzed as the derivation in (41), where [+Q] represents an interrogative complementizer. (40) Louis a mangé un œuf? L. has eaten an egg ‘Has Louis eaten an egg?’ (41) a. [ CP [+Q] [ IP Louis a mangé un œuf]] b. IP-movement, triggered by Q → [ CP [ IP Louis a mangé un œuf] i [+Q] t i ] In support of this proposal, Sportiche (1995) notes that NPIs in French are not licensed in intonational yes/no questions. In (42a), the NPI qui que ce soit ‘anybody’ is licensed by negation; in (42b), the NPI is licensed by a syntactic (non-intonational) question, in which IP is c-commanded by [+Q]; (42c) shows that the NPI is not compatible with statements (where no [+Q] or similar head is involved); finally, the ungrammaticality of (42d) indicates that the NPI is not c-commanded by [+Q] in intonational yes/no questions, precisely as proposed in (41). 28 These units roughly correspond to the traditional kinds of movement, namely NP-movement, wh-movement, and head-to-head movement, respectively. 121 (42) a. Il n’a pas vu qui que ce soit. ‘He has not seen anybody’ b. A-t-il vu qui que ce soit? ‘Has he seen anybody?’ c. * Il a vu qui que ce soit. ‘He has seen anybody’ d. * Il a vu qui que ce soit? ‘Has he seen anybody?’ Sportiche’s proposal can be easily transposed to a CTM representation, involving Copy+Merge triggered by an EPP-like feature [+Ft] on Q. Over the newly created bi-sentential structure, PF-deletion applies, targeting the lower copy in full. (43) [ QP [ IP Louis a mangé un œuf] i Q [Ft] [ IP Louis a mangé un œuf]] i Within the minimalist framework, movement is understood as an operation necessary to check some features that would make the derivation crash if unchecked. If this is so, there’s no limit a priori with respect to the dimension or the label of the unit to be moved—or rather, to be copied—as long as the right features on the right units are considered. The derivation in (25) continues now with a Merge operation involving the copy of CP and an intermediate projection of the modal, Mod’, as seen in (25f). 29 As a result, the high copy of CP is located now in [Spec,ModP]. The FSU is taken as the ‘pivot’ for the Copy+Merge operation. Behold the bi-clausal structure. This is the final product of syntax proper, that is to 29 The expression “Copy+Merge” suggests that steps (25e) and (25f) may be collapsed in just one single step. Although this may be indeed the case, I consider them here as separated steps for illustration purposes. 122 say, the final structure that will be shipped both to PF and LF for their respective deletions to take place. Its general formulation was already presented in (21a), and is now repeated in (44). (44) [ XP CP i [ X’ FSU [+foc] [CP] i ]] The following steps (25g,h) occur at PF. In (25g), PF-deletion applies to the low copy, driven by the LCA. This principle can be expressed as follows. Let be a non-terminal unit X that dominates a terminal unit x, and a non-terminal unit Y that dominates a terminal unit y. Now, if X asymmetrically c-commands Y, then x will precede y in the linear string. In this case, X and Y would be the high and the low copy of CP, respectively. Note that the high copy of CP asymmetrically c-commands the low copy of CP. Then the units contained in the high copy will precede the units contained in the low copy. For the case of (25), let’s take for instance the DP Manolo. The high copy of Manolo asymmetrically c-commands the low copy of Manolo. Additionally, the high copy of Manolo asymmetrically c-commands probablemente ‘probably’. This will mean that the high copy of Manolo should precede probablemente. But probablemente, in turn, asymmetrically c-commands the low copy of Manolo. This will mean that probablemente should precede the lower copy of Manolo. Now we need to remember that both copies of Manolo are the same object. At this point, Linearization has received contradictory instructions. On the one hand, Manolo should precede probablemente, but on the other hand probablemente should precede Manolo. Additionally, the higher copy of Manolo precedes the lower copy of Manolo, which amounts to say that Manolo precedes itself. We can predict that a similar situation will arise for the other units in the copied CP. Thus the following corollary to the LCA, the precedence conditions, would be fatally violated. Things being so, the derivation should be cancelled. (45) Precedence conditions (i) Asymmetry: If α precedes β, then β does not precede α. (ii) Irreflexivity: If α precedes β, then β ≠ α. In order to solve this conflict, one of the two copies should be deleted (more generally, just one copy must survive). Normally, the higher copy has checked more features with respect 123 to the lower copy. This is why the lower copies are normally deleted, and only the highest one survives. I assume that this is applicable to the case in (25). Now, for the sake of the presentation, let’s further assume that deletion acts from left to right. In this case, the low copies of the units Manolo, comprará, and una novela are successively deleted in the lower clause, so their higher copies survive. However, this process must stop at the F-marked unit boundary. Since focus marking cannot be copied, deletion of this unit will lead to loss of information. This is why the F-marked unit cannot be deleted, so it remains in situ. Two desirable consequences can be derived from these facts. First, once the non-focused elements in the lower clause have been deleted, the FSU can surface immediately adjacent to the left edge of fcP in the linear string, which explains ‘transportability’. Second, in order to prepare the string for linearization, deletion needs to be complementary, as indicated in (15c). Since the unit that is fcP-marked in the low copy has not been deleted, its reflex in the high CP must undergo PF-deletion, which is precisely what the step (25h) shows. The final PF-arrangement of units forms the right input for linearization. In parallel to the steps (25g,h), the steps (25g’-i’) take place at LF. The modal expresses probability on one specific element of the proposition. In the case of (25), this element is the author (of a certain novel). All other elements in the proposition form the presupposition. 30 To get that interpretation, LF must delete all the presupposed elements in the lower copy—that is to say, the copy that is under the scope of the modal FSU. This occurs in (25g’). As in the case of PF-deletion, LF-deletion cannot apply inside the fcP either, and this is in accordance with the Uniformity Condition. This will force the LF-deletion of the reflex of fcP on the high copy, which is what we see in (25h’). This deletion is able to create a variable, as anticipated in (15f). Finally, this newly created variable is bound by a silent lambda-type operator Op, as seen in 30 We can imagine a conversation between two speakers, A and B. A may ask B about the identity of the author whose novel Manolo will buy. B is not completely sure, but he thinks that that author could be Vargas Llosa, so he responds as in (19). From this perspective, what is seen under the lens of probability is not the whole proposition, but the focused element only. The fact that Manolo will buy a novel is part of the common ground of both speakers at the moment their dialogue takes place. In this sense, the assertion that Manolo will buy a novel written by some novelist x (whose identity is temporarily ignored) equals the shared presupposition. This content is taken not as probable, but as a fact. It is the identity of the novelist what is seen as probable. Precisely, this is the element in focus and corresponds to the main assertion (x PROBABLY equals Vargas Llosa). 124 (25i’). This binding yields the operator-variable relation naturally expected for presuppositions at LF. And this is the final step in the derivation. Note that PF and LF are, in principle, blind to each other. In (25) the two types of deletion match, but this is not necessarily the case at all times, as previewed in (15d). Globally considered, the derivation in (25) poses an important question within minimalism. As we have indicated, the Inclusiveness Condition bans the addition of any feature, primitive, or material not present in the computation as part of the Numeration from the very beginning. If this condition is to be respected, what is the status of F-marking? We have already indicated that F is neither a formal, nor a semantic, nor a phonetic feature. It comes from a different structure, namely the Information Structure (IS). Neither F features nor any F-marked units are present in the Numeration. However, (25) shows that F-marking is able to occur in the middle of an otherwise ‘purely’ syntactic derivation. Moreover, in spite of not being a formal feature, F has syntactic consequences—Copy and Merge—and from there on, its effects are further projected into LF and PF in the form of deletions and prosodic prominence. It looks like F-marking adds to the derivation a transcendental element, namely F, which was certainly absent from the Numeration. From observations like these, it seems fair to conclude that the derivation shown in (25) does violate the Inclusiveness Condition. A second possibility, however, is at hand. As Chomsky (1995) indicates, minimalism does not intend to be able to cover all linguistic phenomena at once, Notice that I am sweeping under the rug questions of considerable significance, notably, questions about what in the earlier Extended Standard Theory (EST) framework were called “surface effects” on interpretation. These are manifold, involving topic-focus and theme-rheme structures, figure-ground properties, effects on adjacency and linearity, and many others. Prima facie, they seem to involve some additional level or levels internal to the phonological component, postmorphology but prephonetic, accessed at the interface along with PF (Phonetic Form) and LF (Logical Form). (Chomsky 1995: 220). 125 On the basis of this acknowledgment of some of the minimalism’s own limits, it is possible to think about modifying the Inclusiveness Condition so that it will be able to include focus-presupposition phenomena like the one presented in (25). Actually, such an extension has already been proposed in Zubizarreta (1998)—whose revised version of Inclusiveness also incorporates prosodic phenomena, (46) Inclusiveness Principle (revised) 31 The interface levels consist of nothing more than arrangements of lexical features and interpretations of the arrangement of categories within the phrase marker in terms of the focus/non-focus distinction and in terms of relative prosodic prominence. (Zubizarreta 1998: 33) Note that (46) does not imply that F, F/fcP-marking, or prosodic prominence should be already encoded from the Numeration. Its perspective is at the interface levels. So to speak, (46) worries about quality of the computation’s output, without taking position about its input at the starting point. 3. Preliminary applications of the bi-clausal reduplicative analysis 3.1. Short answers Sentence (18), whose analysis occupies the previous section, is repeated here as (47). (47) Manolo comprará una novela probablemente de Vargas Llosa. M. will.buy a novel probably by V.L. ‘It is probably V.L. the author of the novel that M. will buy’ This sentence can be felicitously used as an answer for the question in (48), which has also been presented above as (19). 31 It is called Inclusiveness Principle in the source (not Inclusiveness Condition), but, beyond the different wording, the content is the same. 126 (48) De qué autor comprará una novela ____ Manolo? By which author will.buy a novel M.? ‘Which author’s novel will M. buy?’ Following an idea presented in Krifka (2006), the question in (48) can also be felicitously answered by what this author calls a short answer. This kind of answer, already presented in this work in example (30), is attributed to deletion of the common background. I assume that view in (49). (49) Question: De qué autor comprará una novela Manolo? Manolo will buy a novel by which author? a. Manolo comprará una novela probablemente de Vargas Llosa. b. PF-deletion of common background → Manolo comprará una novela probablemente de Vargas Llosa. c. = Probablemente de Vargas Llosa (short answer) Now, in terms of the bi-clausal structure, the short answer in (49c) can be easily accounted for by assuming that a PF-deletion process has applied to the ‘given’ content in the high copy, as represented in (50). 32 32 In Velásquez (2011b: 37) this kind of deletion is presented as omitted presupposition. 127 (50) ModP CP i Mod’ M. comprará una novela de V.L. CP i Mod 0 M. comprará una novela [de V.L.] fcP probablemente [+foc] The syntactic tree in (50) shows that only the modal adverb probablemente and the fcP- DP de Vargas Llosa have survived the five applications of PF-deletion. 33 The bi-clausal structure seems to straightly accommodate the formation of short answers. Note, however, that the deletion on the higher copy shown in (49b) creates an imbalance: this deletion is not complementary. Indeed, it cannot be complementary, since its counterpart in the lower copy has already been deleted. On this imbalance issue, I consider that it is necessary to distinguish between deletion due to the linearization of the bi-clausal structure itself, and deletion due to common background. On the one hand, the first kind of PF-deletion is sentence-internal, since the units over which it operates are two copies of the same unit, belonging to the same sentence (indeed, to the same phase). Internal deletion is driven by the LCA and FFE (in cases of focus marking, it respects the FDC). This is why this kind of deletion is always complementary—deletion is complementary when at-least-one and only-one copy of each copied element survives after it 33 There have been two application of PF-deletion in the low clause: DP Manolo, and VP comprará una novela; plus three applications in the high clause: PP de Vargas Llosa, DP Manolo, and VP comprará una novela. The first three deletions are due to linearization needs of the bi-clausal structure. The short answer arises when the other two deletions are added. 128 applies. On the other hand, deletion due to common background is sentence-external. This kind of deletion is produced by identity with the content of another construction (in this case, the question), such that this construction is different from the phase-unit where deletion applies. It is guided by some principle of pragmatic communicative economy, presumably in the line of Grice’s conversational maxims, and is free of the requisite of being complementary, since it does not apply to copies, but to two units in different phases. The two constructions happen to have elements in common and are uttered in the same communicative act—usually consecutively, usually by different speakers—but the key point is that they are different phase constructions. This situation is expected in a Question-Answer pair, since those two constructions share a common presupposition. From a grammatical point of view, it is significant that sentence- internal deletion is an obligatory process (under penalty of crashing or cancelling the derivation), whereas deletion by common background is normally optional—although certainly very common in real-life communicative situations. From now on then, I will distinguish between two types of deletion, according to the origin of the units affected: (a) internal deletion, which applies to copies belonging to one and the same chain, and (b) external deletion, which is licensed by identity with a construction different from the construction where deletion applies. 34 While Krifka’s short answers are created via ulterior external deletion, the bi-clausal structure presented and analyzed in this work requires internal (complementary) deletion. 3.2. Focus association at a distance One of the main advantages of the CTM with respect to the Trace Theory is the fact that the former can account for PF/LF interpretive mismatches in terms of different targets for their 34 The idea of “different construction” may be understood here as “different phase”. In example (i), adapted from Merchant (2001: 14), there’s just one matrix sentence, so presumably there’s just one initial Numeration (in which, crucially, the verbal root sing√ has been selected twice). However, there are two phases (in this case, CP). VP- deletion in the second phase, as seen in (ii), gets licensed by identity with the first phase. (i) Abby sang because Ben did. (ii) Abby sang because Ben did sing. For more about this kind of deletion (treated as ellipsis) and its relation to focus, see Merchant (2001), in particular chapter 1 (“Identity in ellipsis: focus and isomorphism”, pp.10-38). 129 respective deletion processes. Indeed, this possibility has been listed as the item (15d) in our inventory of resources. In the previous chapter, we already saw an example of mismatch, due to Chomsky (1993), which deals with the interpretation of some ambiguity related to Principle A. In this section, I want to put forward the suggestion that association with focus at a distance can also be explained in PF/LF mismatch terms. Consider sentence (16), repeated here as (51), whose IS is represented in (52). (51) Probablemente Manolo comprará una novela de Vargas Llosa. Probably M. will.buy a novel by V.L. ‘Probably M. will buy a novel by V.L.’ (52) [Probablemente] FSU Manolo comprará una novela [de [Vargas Llosa] F ] fcp Association with fcP at a distance The IS in (52) illustrates a well-known phenomenon, already reported by Jackendoff (1972) among many others, namely association with focus at a distance. 35 In principle, some lexical material of undefined size can be found mediating between the FSU and the fcP in the linear string. In other words, immediate linear adjacency between the FSU and the fcP is not required for them to be understood as associated. I propose that the sentence modal adverb comes from the lexicon already loaded with the feature [+foc]. In virtue of this feature, Copy and Merge of the c-commanded CP are enforced, provided that F-marking has occurred within its sentential domain. If this is on the right track, a bi-clausal structure like (53) will be formed. 36 35 Indeed, this is the reason why it is called association with focus, in first place. Unlike other kind of (syntactic) dependencies requiring fixed structural configurations (head-complement, spec-head, etc.), association with focus seemed to be structurally ‘free’. In early studies, as for instance Jackendoff (1972), an undefined ‘distance’ between FSUs and focused unit was assumed as the norm. Linear adjacency between them was taken as the marked situation. 36 This, of course, is the same bi-clausal structure seen in (25f). 130 (53) [ ModP [ CP [ TP Manolo comprará una novela de Vargas-Llosa]] [ Mod’ probablemente [ CP [ TP Manolo comprará una novela [de [Vargas-Llosa] F ]] fcP ]]] From this point on, PF and LF, each one from its side, will impose deletions in parallel, since PF and LF are, in principle, blind to each other. PF-deletion may proceed. This time, it targets the high copy, as seen in (54). (54) PF-structure [ ModP [ CP [ TP Manolo comprará una novela de Vargas-Llosa]] [ Mod’ probablemente [ CP [ TP Manolo comprará una novela [de [Vargas-Llosa] F ]] fcP ]]] By deleting high copies, this PF-deletion seems to violate the constraint FFE. However, the deletion as such is complementary, so it is linearizable according to the LCA. In that sense, we conclude that the eventual violation of FFE is tolerable. Whatever the case, the structure (54), as such, converges at PF. In a parallel fashion, starting in (53), LF-deletion may proceed. In this case, LF-deletion targets different constituents with respect to PF-deletion (so it is an unmatched deletion). As a consequence, the relation operator-variable can be established, and the semantic scope of the modal on the focused unit (in this case, what is understood as probable is the identity of the author) is obtained via unmatched LF-deletion on the low copy, as seen in (55). (55) LF-structure Op k [ ModP [ CP [ TP Manolo comprará una novela [ PPk de Vargas-Llosa]]] [ Mod’ probablemente [ CP [ TP Manolo comprará una novela [de [Vargas-Llosa] F ]] fcP ]]] Thus, according to this proposal, movement of CP to [Spec,ModP] is always triggered by the feature [+foc] on the modal at LF, provided that some unit within its domain has been F- marked. For the case of focus association at a distance, PF-deletion ‘misbehaves’ at this point, and targets the high copy instead of the lower one, so eventually violating the FFE constraint. 131 The result is phonetically tolerable, however, since it complies with the requisites for linearization demanded by the LCA. In turn, LF-deletion shows ‘good conduct’, and targets the fcP reflex on the higher copy in order to get the right input for interpretation. The FSU then is ‘covertly’ adjacent to the fcP, and an operator-variable relation is created in the high copy, which yields the presuppositional reading required by the interpretation. Ideally, this mismatch is at the base of all cases of association with focus at a distance, not only with sentence modals, but with other FSUs also. This possibility will be explored in the next chapter. 3.3. Sentence modal adverbs in different functional positions A modal sentence adverb may appear adjacent to vPs, DOs, IOs, some adjuncts, etc. In principle, as long as the fcP is located at the end of the sentence, ‘transportability’ can be straightforwardly captured by the bi-clausal proposal involving matched deletions. In (56), the modal is adjacent to the F-marked vP. The correspondent analysis is shown in (57a-c). (56) Manolo probablemente leerá una novela M. probably will.read a novel ‘It is probably reading a novel what M. will do’ (57) a. PF-structure [ ModP [ CP [ TP Manolo leerá una novela]] [ Mod’ probablemente [ CP [ TP Manolo [leerá una novela] fcp ]]]] b. LF-structure Op k [ ModP [ CP [ TP Manolo [ vPk leerá una novela]]] [ Mod’ probablemente [ CP [ TP Manolo [leerá una novela] fcp ]]]] 132 c. ModP CP i Mod’ Manolo leerá una novela CP i Mod 0 Manolo [leerá una novela] fcP probablemente [+foc] In turn, in (58), the modal appears next to the F-marked DO. 37 The bi-clausal analysis is presented in (59a-c). (58) Manolo leerá probablemente una novela M. will.read probably a novel ‘It is probably a novel what M. will read’ (59) a. PF-structure [ ModP [ CP [ TP Manolo leerá una novela]] [ Mod’ probablemente [ CP [ TP Manolo leerá [una novela] fcP ]]]] b. LF-structure Op k [ ModP [ CP [ TP Manolo leerá [ DPk una novela]]] [ Mod’ probablemente [ CP [ TP Manolo leerá [una novela] fcp ]]]] 37 The equivalent structure (SVAdvDO) is not grammatical in English: *Manolo will read probably a novel. On this issue, Zubizarreta (1982: 47-48) indicates, “Adverbs may not appear between the verb and its direct object in English. This is probably due to an independent reason: the adjacency condition on case-assignment”. 133 c. ModP CP i Mod’ Manolo leerá una novela CP i Mod 0 Manolo leerá [una novela] fcP probablemente [+foc] In turn, in sentence (60), the modal is adjacent to the F-marked IO. The bi-clausal analysis is presented in (61a-c). (60) Manolo le regalará una novela probablemente a su sobrino. M. Cl.DAT=will.give a novel probably to his nephew ‘It is probably to his nephew that M. will give a novel’ (61) a. PF-structure [ ModP [ CP [ TP Manolo le regalará una novela a su sobrino]] [ Mod’ probablemente [ CP [ TP Manolo le regalará una novela [a su sobrino] fcp ]]]] b. LF-structure Op k [ ModP [ CP [ TP Manolo le regalará una novela [ PPk a su sobrino]]] [ Mod’ probablemente [ CP [ TP Manolo le regalará una novela [a su sobrino] fcp ]]]] 134 c. ModP CP i Mod’ M. le regalará una novela a su sobrino CP i Mod 0 M. le regalará una novela [a su sobrino] fcP probablemente [+foc] In sentence (62), the modal appears next to the F-marked temporal adjunct. The bi-clausal analysis is presented in (63a-c). (62) Manolo regresará probablemente a las cuatro M. will.come-back probably by the four ‘It is probably by four that M. will come back’ (63) a. PF-structure [ ModP [ CP [ TP Manolo regresará a las cuatro]] [ Mod’ probablemente [ CP [ TP Manolo regresará [a las cuatro] fcp ]]]] b. LF-structure Op k [ ModP [ CP [ TP Manolo regresará [ PPk a las cuatro]]] [ Mod’ probablemente [ CP [ TP Manolo regresará [a las cuatro] fcp ]]]] 135 c. ModP CP i Mod’ Manolo regresará a las cuatro CP i Mod 0 Manolo regresará [a las cuatro] fcP probablemente [+foc] Now, with respect to subjects, there are two basic possibilities. One is that the subject appears in final sentence position. In this case, the prosodic rules C-NSR and FPR would apply directly, so no further prosodic action would be required. This possibility is reflected in (64). The modal appears next to the F-marked subject in final position. The bi-clausal analysis is presented in (65a-c). (64) Compró galletas probablemente Manolo Bought.3SG cookies probably M. ‘It was probably Manolo who bought cookies’ (65) a. PF-structure [ ModP [ CP [ TP Manolo compró galletas]] [ Mod’ probablemente [ CP [ TP [Manolo] fcp [compró galletas]]]]] b. LF-structure Op k [ ModP [ CP [ TP [ DPk Manolo] [compró galletas]]] [ Mod’ probablemente [ CP [ TP [Manolo] fcp [compró galletas]]]]] 136 c. ModP CP i Mod’ Manolo compró galletas CP i Mod 0 [Manolo] fcP [compró galletas] probablemente [+foc] The second basic possibility is that the subject appears F-marked in situ. In principle, this case is different from all the previously presented, since, in Spanish, marking of subjects in situ is felicitous only in contrastive contexts (Zubizarreta 1998, Molnár & Winkler 2010). In general terms, a focus is contrastive when the set of alternative focus values (Alt-fv) is restricted and explicit for the participants in the conversation. 38 This situation implies a clash. The prosodic rules C-NSR and FPR would be in conflict. While the former would assign prominence to the sentence-final constituent, the latter would choose to assign it to the focus marked constituent, which is not in final position. Certainly, the modal adverb would be immediately adjacent to the fcP, so marking the left edge of the fcP would not be a problem in this case. However, this adjacency is the same adjacency that we find in broad/wide focus, where the whole sentence is F-marked, as seen in (66b). What we would need is a way of marking the right edge of the fcP, so (66b), with wide focus on the whole TP, can be perceived as distinct from (66c), with narrow focus on the subject only. The right edge of the fcP indicates the point where this phrase ends (i.e. the right bracket with the subscript). 38 For the case of (165c), imagine that speakers A and B in conversation explicitly share the idea that either Juan or Manolo (and no one else) have bought the cookies. If one of them, say A, wants to state that Manolo (not Juan) is the right value for the focus variable, he or she may use the construction (66c), with focus in situ. 137 When this phrase is located in sentence final position, that right edge coincides with the end of the sentence and the prosodic rules are satisfied without conflict. In the case of narrow focus on the subject in situ, the right edge is located at the end of that subject, after which more lexical material appears. If this edge is not marked, wide focus on TP will be wrongly perceived. (66) a. Probablemente Manolo compró galletas Probably M. bought.3SG cookies b. Probabablemente [Manolo compró galletas] fcP c. Probabablemente [Manolo] fcP compró galletas 3.4. The problem with F-marking of subjects in situ Facing these problems, prosody comes to the rescue. The solution it provides consists of assigning an additional prominence to the focused subject in situ. This prominence is manifested as a higher-than-usual value for F0 on the in situ focus, followed by an abrupt fall in that value. In certain cases, the fall may prolong itself in a lower-than-usual value for F0 on the following material (flat low contour, maybe including deaccenting). Additionally, the peak F0 value of the subject will be aligned with its primary stressed syllable. This also contrasts with the alignment of that peak in the post-tonic syllable, which the usual outcome in other contexts. This kind of solution means that the conflict between C-NSR and FPR will be finally solved in favor of the second rule. In addition, the problem of marking of the right edge of the fcP will be also solved, since the aforementioned drastic fall in F0 will be perceived as an ip boundary. As a way to illustrate this prosodic solution, I present two figures taken from O’Rourke’s analysis of contrastive focus in Peruvian Spanish—which happens to be my own native dialect. 39 39 For an accessible explanation to the basic elements of prosodic annotation in Spanish, plus some practical tools on the issue, see Aguilar et al. (2009). For an expansion on this, see Estebas-Vilaplana & Prieto (2009). For more specific details about prosody of contrastive focus in Spanish, see Face (2002). 138 Fig.1 (from O’Rourke 2012: 499) The F0 contour in Fig.1a shows a succession of three peaks in a descending pattern known as downstep, typical of Spanish declaratives. The first peak corresponds to the subject su familia ‘his family’. It can be noticed that this peak is misaligned with respect to the stressed syllable mi; indeed, this peak occurs in the post-tonic syllable. Precisely, this kind of pitch accent is what one would expect in these cases. 40 According to Aguilar et al. (2009), this pitch accent is “the predominant choice for pre-nuclear accents in broad-focus statements”. 41 On the other hand, in Fig.1b, the first peak is neatly aligned with the stressed syllable. Then, an abrupt fall in F0 occurs in the post-tonic syllable, and, from there on, a flat contour for F0 is observed. This contour clearly contrasts with the downstep pattern seen in Fig.1a. Also, the value in Hz of the first peak in Fig.1b is higher than the value in Hz of the equivalent first peak in Fig.1a. Beyond this quantitative difference, what seems to count more is the inequality between F0’s highs and lows. It seems as if the speaker in Fig.1b would like to maximize the contrast between the focused unit (highest F0 possible) and the non-focused material (lowest F0 possible). 40 This pitch accent is represented as L+>H*. Aguilar et al. (2009) describe it as follows, “This pitch accent is phonetically realized as a rising pitch movement. Typically, the L tone is aligned with the onset of the accented syllable, and the H tone is aligned with the postaccentual syllable”. 41 Similar results were obtained in Velásquez (2011b), which also studied data from Lima native speakers. 139 Fig.2 (from O’Rourke 2012: 500) As for Fig.2a, a misalignment between the first stressed syllable and the first F0 peak is observed, which is similar to the one seen in Fig.1a. In comparison, the equivalent part in Fig.2b shows clear alignment. The succession of the second and third peaks is similar in both figures (the downstep is more subtle than in Fig.1), but, in this case, the main the difference between Fig.2a and Fig.2b lies on the F0 values of their respective first peaks. This difference makes the fall of F0 in the post-tonic syllable more dramatic, even if their ‘landing-site’ value is similar. In sum, although their strategies may vary in details, both figures show that, in Peruvian Spanish, an additional prosodic prominence occurs whenever the subject is contrastively focused in situ. 3.5. F-marked subjects with prosodic prominence: the structural view One of the possible reasons behind ‘transportability’ is the functional advantage of marking the left edge of the fcP. The right edge is normally marked by its alignment to the sentence-final ip. The unmarkedness of left edge, on the other hand, creates ambiguity with respect to the limits of the fcP whenever the modal adverb associates with it at a distance. This ambiguity disappears once the modal is ‘transported’—I use this improper term on purpose here—to a position where it is adjacent to the fcP. Now, since both edges of the focused subject in (66c) have been marked via prosodic prominence, it would be licit to ask whether the bi-clausal structure is (or should be) built anyway. 140 My answer to this question is yes. The reason for this is that the modal adverb contains the [+foc] feature from the lexicon, and this feature will trigger ‘movement’ (that is to say, Copy+Merge) whenever some unit within its relevant domain is F-marked. This is indeed the case in (66c). The correspondent bi-clausal analysis then would be as follows. (67) Probablemente MANOLO compró galletas Probably M. bought.3SG cookies ‘It was probably Manolo who bought cookies’ (68) a. PF-structure [ ModP [ CP [ TP Manolo compró galletas]] [ Mod’ probablemente [ CP [ TP [MANOLO] fcp [compró galletas]]]]] b. LF-structure Op k [ ModP [ CP [ TP [ DPk Manolo] [compró galletas]]] [ Mod’ probablemente [ CP [ TP [Manolo] fcp [compró galletas]]]]] 141 c. ModP CP i Mod’ Manolo compró galletas CP i Mod 0 [MANOLO] fcP [compró galletas] probablemente [+foc] The analysis in (68) indicates that also here we find a PF/LF mismatch. While PF- deletion targets the high copy as a whole, LF-deletion targets the (copy of) the subject in the high copy, plus the VP in the low copy. This deletion is complementary, which means that the structure will finally be linearizable. On the other hand, LF-deletion creates the right variable on the high copy, and binds it with a silent operator, which opportunely yields the intended interpretation for the presupposition. 3.6. The reduplicative structure, HT and CLLD constructions Before leaving this section, I want to briefly consider a structure involving topicalization known as Clitic Left Disclocation or CLLD (Cinque 1990, Rizzi 1997, Casielles-Suárez 1997, 2001, 2003, 2004) in Spanish. This structure connects syntax and information structure by assigning a fixed syntactic position to the topic at the high left periphery. In this sense, CLLD presents topic, which is an IS notion, as directly encoded in a specific syntactic node. In words of Reinhart (1981: 63), “[…] certain syntactic structures have fixed, or structurally marked topic position, which means that NPs in these positions are obligatorily topics. This holds, to a certain extent, to all fronted NPs, but the clearest case is the structure known as Left Dislocation”. For Spanish, those structures are exemplified in (69) and (70). 142 (69) La novela la compró Manolo. The novel Cl.ACC=bought.3SG M. ‘The novel, MANOLO bought it’ (70) La comida Manolo la preparó . The meal M. Cl.ACC=prepared.3SG ‘The meal, Manolo PREPARED it’ In (69), the first DP la novela ‘the novel’ is located in a high sentence position. This is also the case for the DP la comida ‘the meal’ in (70). Importantly, in (69), the DP subject Manolo appears in sentence final position, where it conveys new information. In that sense, (69) may be used as felicitous answer to the question Quién compró la novela? ‘Who bought the novel?’ From this, we can assume that Manolo is F-marked. A similar reasoning can be invoked for the complex clitic-verb la preparó ‘prepared it’, which I take as the fcP in (70). 42 An important question about CLLD in Spanish is whether the DP-topic has been moved from a sentence-internal position. Casielles-Suárez (2001) argues that the units appearing in high topic position may have been directly generated there. According to this author, who cites Contreras (1991) and Vallduví (1988) as her sources, the examples (71) and (72) show that CLLDs are not sensitive to islands constraints (Casielles-Suárez 2001: 70). (71) La sidra no conozco a nadie que la odie The cider NEG know.1SG A nobody who Cl.acc=hate ‘Cider, I don’t know anybody that hates it’ (72) La sidra no sabemos quién la trajo 42 However, the situation is less clear here, since (70) can also be felicitously used with focus on the subject, provided that prosodic prominence has been added: La comida MANOLO la preparó. With neutral prosody, I tend to think that this sentence is more appropriately used in a contrastive context, where the closed set of focus values is formed by a restricted number of possibilities, for instance, when the set Alt-fv limited to three values: Alt-fv = {prepare it, bought it, asked for it by delivery}. 143 The cider NEG know.1PL who Cl.acc=brought ‘The cider, we don’t know who brought it’ A different position can be found in Zubizarreta (1998, 1999). Quoting Cinque (1990) for Italian, and Dolci (1986) for Spanish, this author indicates that CLLD is sensitive to strong islands. The following relevant examples are offered (Zubizarreta (1998: 186, n. 23). (73) Relative clause * Estoy segura de que a Pedro, conocemos a la mujer que lo traicionó. I.am sure of that A P. we.know A the woman who Cl.ACC.3SG=betrayed (74) Adjunct * Me parece mejor que a Pedro, terminemos la tarea antes de llamarlo To.me (it) seems better that A P. we.finish the task before call=Cl.ACC.3SG (75) Sentential subject * Estoy segura de que a Pedro, que María lo haya invitado sorprendió I.am sure of that A P. that M. Cl.acc.3sg=had invited surprised a todo el mundo. A everyone. Zubizarreta (1998, 1999) distinguishes between two different topical constructions: CLLD and Hanging Topic constructions (HT). HTs are the expressions used to introduce a new topic in the discourse, as for instance En cuanto a X ‘as for X’ or Con respecto a X ‘with respect to X’. In oral informal discourse, those expressions are frequently omitted. 43 Now, whereas 43 “Como me hace notar I. Bosque, las construcciones con ‘tema vinculante’ [=HT] son propias de la lengua oral, pero no de la lengua escrita. En efecto, la construcción con tema vinculante resultaría chocante en un ensayo científico. Por ejemplo, llamaría la atención encontrar en un libro de Física una secuencia (por lo demás, plenamente gramatical) como La radioactividad, existe un acuerdo general en que se mide por el número de desintegraciones que se producen cada segundo”. (Zubizarreta 1999: 4220, fn. 6). 144 CLLD is sensitive to islands, HT is not. 44 Compare the CLLD construction in (75) vs. the HT construction in (76). (76) Hanging Topic (HT) (En cuanto a) Pedro, que María lo haya invitado sorprendió (As for) P. that M. Cl.acc.3sg=had invited surprised a todo el mundo. A everyone. Zubizarreta (1998: 187) Given their insensitivity to islands, the examples in (71) and (72) may be better described as cases of HTs. By the same token, their topics—and all HTs in general—may have been indeed base generated in high sentence position. If at least some kind topics are indeed base generated in their correspondent high positions, the simplified structures for (69) and (70) would look like (77) and (78), respectively. Note the fcP-marking in the string-final constituents. 45 44 See Zubizarreta (1999: 4224) for a detailed comparison CLLD vs. HT. 45 F-marking on the final constituent seems to be at stake also in example (i), from Casielles-Suárez (2003: 6). (i) El periódico lo compré yo. The newspaper Cl.ACC=bought1sg I ‘It is me/I who bought the newspaper’ Being a pro-drop language, Spanish may optionally replace a full pronoun with a phonetically null pro in many contexts, particularly when a 1SG-pronoun (yo) is in subject position: {Yo/pro} compré el periódico ‘I bought the newspaper’. This option is not available in (i), however. *El periódico lo compré pro. When it is F-marked (in this case, with contrastive focus), the subject must be fully pronounced (otherwise, the linear string produced will be understood with focus on the DO complex lo-compré). Intuitively, this seems related to the ban on deleting the fcP. 145 (77) TopP La novela TP T vP la compr-ó [Manolo] fcP VP la compr- (78) TopP La comida TP Manolo T’ T vP [la prepar-ó] fcP Manolo VP la prepar- 146 Since the structures just described include an fcP, they are compatible with a sentence modal adverb immediately adjacent to that focused unit. (79) La novela la compró probablemente Manolo The novel Cl.ACC=bought.3SG probably M. ‘The novel, probably MANOLO bought it’ (80) La comida Manolo probablemente la preparó . The meal M. probably Cl.ACC=prepared.3SG ‘The meal, Manolo probably PREPARED it’ A central claim in this work is that the modal adverb, being an FSU, contains a feature, namely [+foc], which triggers movement—understood as Copy and Merge—of the c- commanded CP to [Spec,FSU] whenever a unit is F-marked within its relevant domain. This is indeed the case for (79) and (80). The crucial assumption here is that the modal adverb merges with an already formed CLLD or HTC structure. If this idea is on the right track, the bi-clausal proposal would provide the following analyses for (79) and (80), presented in (81) and (82), respectively. (81) a. PF-structure [ ModP [ CP [ TopP La novela la compró Manolo]] [ Mod’ probablemente [ CP [ TopP la novela la compró [Manolo] fcp ]]]] b. LF-structure Op k [ ModP [ CP [ TopP La novela la compró [ DPk Manolo]]] [ Mod’ probablemente [ CP [ TopP la novela la compró [Manolo] fcp ]]]] 147 c. ModP CP i Mod’ La novela la compró Manolo CP i Mod 0 la novela la compró [Manolo] fcP probablemente [+foc] (82) a. PF-structure [ ModP [ CP [ TopP La comida Manolo la preparó]] [ Mod’ probablemente [ CP [ TopP la comida Manolo [la preparó] fcp ]]]] b. LF-structure Op k [ ModP [ CP [ TopP La comida Manolo [ VPk la preparó]]] [ Mod’ probablemente [ CP [ TopP la comida Manolo [la preparó] fcp ]]]] c. ModP CP i Mod’ La comida Manolo la preparó CP i Mod 0 la comida Manolo [la preparó] fcP probablemente [+foc] 148 3.7. Sentence modal adverbs in sentence-final position When considering parenthetical uses of modal adverbs in Italian, Giorgi (2013) notes a contrast in acceptability due to prosody when these adverbs are located in sentence-final position. (83) a. * Gianni ha mangiato la torta probabilmente b. Gianni ha mangiato la torta, probabilmente G. has eaten the cake probably The prosodic boundary orthographically represented with a comma in (83b) saves the unacceptability of (83a), in which such a boundary is not present. This seems to be true also for English. It is definitely true in the case of Spanish. (84) a. * Juan ha comido la torta probablemente b. Juan ha comido la torta, probablemente J. has eaten the cake probably If we assume—as I assume here—that probablemente bears the [+foc] feature from the lexicon, the unacceptability of (84a) can be explained as follows. The feature will trigger movement when a unit in its domain gets F-marked. Let’s assume, for the sake of the example, that the DP-DO la torta ‘the cake’ is F-marked. Let’s additionally assume that the modal adverb is merged with CP, as usual. The bi-clausal structure will arise. (85) [ ModP [ CP [ TP Juan ha comido la torta]] [ Mod’ probablemente [ CP [ TP Juan ha comido [la torta] fcp ]]]] Now in order to obtain the linear order seen in (84a), PF-deletion will need to apply on the lower copy, as a whole. 149 (86) PF-structure [ ModP [ CP [ TP Juan ha comido la torta]] [ Mod’ probablemente [ CP [ TP Juan ha comido [la torta] fcP ]]]] From the point of view of linearization, (86) would have no problems, since the deletion is complementary. However, a fatal violation is involved: the deletion has applied inside the fcP. We have seen that this is not possible. Indeed, the bi-clausal analysis is heavily based on this assumption. PF-deletion cannot touch the fcP, so it must stop once it arrives to the fcP boundaries. This is not what happens in (86), thus the derivation crashes. The prohibition is based in the fact that F cannot be copied, since it is neither a syntactic, nor a semantic, nor a phonetic feature. If F cannot be copied, the F-marked unit must remain in situ once all deletions have applied. Note that the same problem would arise no matter what unit is chosen to be F- marked. Be it the subject, the VP, even the whole TP, a fatal violation will always occur. Now, with respect to the acceptability of (84b), it is clear that a PF-structure including full deletion of the low copy, like (86), cannot be invoked. The fact that an intonational boundary saves the grammaticality of the sentence suggests that we must look for the right difference in the syntax-prosody interface. This is precisely what Giorgi (2013) does. In first place, this author considers a proposal by Selkirk (2005) according to which there exists a [+comma] feature. This feature is manifested as a comma intonation in various structures, as parentheticals, non-restrictive relatives, nominal appositives, etc. Structurally, this can be expressed as comma phrases, which are further mapped into Intonational Phrases (IP). This is minimally illustrated in sentence (87), taken from Giorgi (to appear: 6). (87) John, as everybody knows, likes to go to parties. The phrase as everybody knows is a comma phrase. This phrase is demarcated by two pauses. Prosodically, it is mapped into its own IP. In order to represent this fact from a syntactic point of view, Giorgi (to appear) proposes then that the pause or the comma is indeed a head, which this author calls K. 150 I propose that the pause, or the Comma, is a head, K, and projects its own constituent. Since there are two intonational breaks, there are two heads K, the higher one taking the parenthetical as its complement and the lower one taking the host sentence as its. The notion of complement here is to be understood purely in structural terms, and not in lexical ones. Therefore, a certain degree of syntactic ‘permeability’–as variously pointed by many scholars–between the supplement and the host can be allowed, but there is no subordination relation, due to the nature of the head K, which is not a complementizer (Giorgi, to appear: 6). According to Giorgi, being a non-lexical head, K mediates between syntax and prosody. In the case of (84b), the KP would be present from the point where the modal merges with the CP. Since a comma phrase in the sense of Selkirk (2005) is typically delimited by two pauses, the structure attributed to KP considers two K heads, as seen in (88). (88) [ KP K [ KP Probablemente [K [ CP Juan ha comido la torta]]]] Now, on the basis of this structure, Giorgi (2013) proposes upward movement of the whole sentence under (the lower) head K. (89) [ KP [ CP Juan ha comido la torta] i K [ KP probablemente [ K t i ]]] In terms of movement as Copy and Merge, (89) should look like (90). (90) [ KP [ CP Juan ha comido la torta] i K [ KP probablemente K [ CP Juan ha comido la torta] i ]]] At this point, it is necessary to establish the following precedence relation: KP- movements must occur prior to F-marking. Otherwise, deletion of the lower copy would be banned, exactly as in the ungrammatical structure (86). Being Ks non-lexical heads, related 151 exclusively to the relative order between pauses and comma phrases (as Giorgi proposes), this precedence relation seems reasonable. With this warning in mind, PF-deletion on the lower copy can proceed, as presented in (91). (91) [ KP [ CP Juan ha comido la torta] i K [ KP probablemente K [ CP Juan ha comido la torta] i ]]] The structure in (91) is the base of the grammatical sentence (84b), repeated here as (92). (92) Juan ha comido la torta, probablemente. Now, in terms of F-marking, sentence (92) is hopelessly ambiguous. As is, (92) is compatible with at least three different non-contrastive F-markings. (93) a. [Juan ha comido la torta] fcP , probablemente. b. Juan [ha comido la torta] fcP , probablemente. c. Juan ha comido [la torta] fcP , probablemente. This time, not even prosody can come to the rescue—unless contrastive focus is involved, as in (94). Even in this case, however, it is not clear if the adverb really establishes a focus-association relation with the fcP. (94) Juan ha comido [LA TORTA] fcP , probablemente. Taken globally, these observations show that sentence-final position is not a ‘transportability’ position. The formal mechanisms that gives origin to this configuration are qualitative different from the bi-clausal structure responsible of all previous structures. In this vein, I follow Giorgi (2013) in considering that the modal adverb in sentence-final position is really a parenthetical. 46 This means that there’s no F-marking under its syntactic scope. If F- 46 A break index 4 would be expected then for the comma in (94). Intuitively, this expectation seems to be correct. 152 marking takes place, as in (93), it happens out of the reach of the adverb. Consequently, its [+foc] feature remains inactive. 47 4. Summary In this chapter I have presented the main proposal of this work. First, I presented an analysis of certain phenomena in Spanish in terms of ellipsis and reduplication, two notions akin to deletion and copy, respectively. The phenomena studied are pseudo-clefts in colloquial Peninsular Spanish, bare copular constructions in Caribbean Spanish, and marked word orders in standard Spanish. Putting aside details, these analyses showed that ellipsis and reduplication can advantageously account for the observed linear strings and their correspondent interpretation in the three cases considered. Importantly, the structures resultant from ellipsis may show a mismatch between what is deleted at PF and what is deleted at LF. It is further argued that, since ellipsis and reduplication give the same results as focus movement, it is recommendable to dispense with focus movement. On the basis of all the previous points, I presented the main proposal of the work. To start with, the sentence modal adverb always merges with CP, occupying a high position at the left periphery. This assumption is in accordance with the propositional scope of the modal and also with its syntactic selection properties. Now, I consider that the modal is loaded from the lexicon with a feature, called [+foc] (= focus feature), which is sensitive to focus. Whenever a unit in its domain is F-marked, that feature is activated. This activation triggers a copy process, targeting the whole c-commanded CP, which is then copied. This copy of CP is merged at [Spec,ModP]. This creates a bi-clausal structure with the form [ ModP CP i [Mod 0 [CP] i ]]. Crucially, F can be found in the low copy only. Since F is neither a phonetic, nor a semantic, nor a syntactic feature, but a mark stemmed from IS, it cannot be copied. Now, the bi-clausal structure must be linearized in accordance with the LCA. The low clause as a whole would be targeted for deletion, unless a high ranked constraint enforces a different outcome. I assume that such a constraint exists, namely a constraint forbidding the deletion of the F-marked unit (FDC). The F-marked unit is necessarily located in the low copy. At this point, scattered complementary 47 On the other hand, its semantic features are not cancelled, so the adverb can contribute with them to the global meaning of the sentence. 153 deletion applies. As a result, the modal adverb ends up by being immediately adjacent to the focused phrase in the linear string. On the interpretive side, the complementary deletion of the (F-unmarked) high copy of the focused phrase produces a variable, which is bound by an abstract operator Op. As a consequence, the presuppositional interpretation of the pre-adverb part of the sentence is felicitously obtained. Finally, the proposed bi-clausal structure has been put to the test in four environments. First, I showed that short answers, in which only the modal plus the focused phrase are pronounced, are the product of a different kind of deletion, namely external deletion. Deletion is external when is licensed by identity with some material located in a phase different from the phase where that deletion applies. In this sense, the scattered deletion inherent to the bi-clausal proposal is not external, but internal (it targets copies belonging to the same chain). Second, association with focus at a distance has been interpreted as the result of a mismatch between PF- and LF-deletions that take as their base one and the same bi-clausal structure. Third, it was shown that the bi-clausal analysis efficiently accounts for ‘transportability’ in Spanish whenever the unit in focus is located in sentence final position. This includes vP, DO, IO, some adjuncts, and certain fixed structures conveying topics, namely CLLD and HT. In case the focused unit is not in final position, prosodic means should intervene. A special prosodic prominence is added to the unit in focus that is not in sentence-final position. Fourth, I claimed that sentence-final is not a ‘transportability’ position for the modal, but instead is a parenthetical position. In consequence, its [+foc] feature is not activated there. 154 Chapter 4 Possible extensions of the main proposal I: Reduplication with other focus sensitive units This chapter is oriented to explore some possible extensions of the main proposal presented in the previous chapter. In section 1, I start by presenting a generalized Reduplicative Structure (RS), which, in principle, applies with informative focus in-situ. 1 Section 2 shows an analysis in terms of the RS for the FSUs Sp. sólo ‘only’, también ‘also’, and incluso ‘even’, along with some of their possible combinations with the modal probablemente ‘probably’. Next, in section 3, I explore the possibility of extending the RS analysis to other –mente adverbs in Spanish and also to sub-sentential negation. In section 4, I apply the RS to derive some marked word orders in this language. Section 5 summarizes the chapter. 1. The reduplicative structure (RS) As presented in the previous chapter, the central goal of this work is to explain how come a sentence adverb like probablemente ‘probably’, with clear propositional scope, can at times surface immediately adjacent to units located in low-sentence positions, like VP, DP, AdjP, PP, etc. As an initial observation, it was noted that the modal adverb surfaces immediately adjacent to the unit marked as informative focus. In structural terms, my proposal can be summarized as follows. (1) a. Probablemente ‘probably’ is base generated in high sentence position (where it gets propositional scope) in all cases. b. Probablemente has a feature [+focus] (=[+foc]), sensitive to F-marking. c. If F-marking occurs, [+foc] triggers Copy of the c-commanded clause. d. The copy is attracted to [Spec,ModP], and this creates a bi-clausal structure. The two copies form a non-trivial chain. This syntactic structure is shipped both to PF and LF. 1 In this sense, the RS analysis presented here does not apply for cases of what is known as focus movement, which, in Spanish, involves fronting of contrastive focus. 155 e. At PF level, PF-deletion applies for linearization purposes. This deletion is driven by the Linear Correspondence Axiom (LCA). According to the LCA (see (2)), since the two clauses are identical copies, the structure cannot be linearized as is. Thus, one of the copies undergoes PF-deletion, via the operation Chain Reduction (ChR) (see (3)). In principle, the operation Formal Feature Elimination (FFE) (see (4)) would choose deletion of the lower copy as a whole; however, the high ranked constraint Focus Deletion Constraint (FDC) (see (5)) enforces the conservation of the F-marked unit in the lower clause and the complementary PF-deletion of its F-unmarked clone in the higher clause. 2 f. At LF level, LF-deletion applies for interpretive purposes. LF-deletion in the lower clause is also driven by the FDC, so the F-marked unit is also preserved in the lower clause at LF. At the same time, LF-deletion of the non-focused material between the modal and the focused unit produces interpretive adjacency between these two units. Because of the preservation of the F-marked unit in the lower clause, complementary LF-deletion on the F-unmarked copy in the higher clause is enforced. This latter deletion creates a variable, which is then bounded by an abstract lambda-type operator. This binding creates the presuppositional reading in the higher clause. (2) Linear Correspondence Axiom (LCA) A lexical item α precedes a lexical item β iff (i) α asymmetrically c-commands β, or (ii) an XP dominating α asymmetrically c-commands β. (Kayne 1994; Hornstein et al. 2005: 227 ). (3) Chain Reduction (ChR) Delete the minimal number of constituents of a non-trivial chain Ch that suffices for Ch to be mapped into a linear order in accordance with the LCA. (Nunes 2004: 27). 2 Operations and constraints in (2)-(5) have been introduced in the previous chapter. 156 (4) Formal Feature Elimination (FFE) Given the sequence of pairs σ = < (F, P) 1 , (F, P) 2 , …(F, P) n > such that σ is the output of Linearize, F is a set of formal features, and P is a set of phonological features, delete the minimal number of features of each set of formal features in order for σ to satisfy Full Interpretation at PF. (Chomsky 1995: 230). (5) Focus Deletion Constraint (FDC) Focus-marked constituents at LF or their phonological locus cannot delete at Spell- Out. (Han & Romero 2004: 199). On the basis of the analysis summarized in (1), I want to explore the possibility of extending its scope to other cases involving informative focus. First, I present in (6) the general proposal of the RS, and I will justify its applicability in the following sub-sections. (6) Reduplicative Structure (RS) XP Sentence Advs. Other FSUs X = NEG YP i X’ Identity COP Ø (empty head) X 0 YP i [… [α] …] CP YP = vP [+foc] TP [… [α] fcP …] In (6), X is a head that had been identified in the previous chapter with the sentence modal adverb probablemente ‘probably’. Now, if my insight is on the right track, X might also 157 be identified with other focus sensitive adverbs (FSUs), certain sentence adverbs, sub-sentential negation, some forms of the identity copula SER (‘to be’), and even a phonetically empty head Ø. What all these heads have in common is the feature [+foc]. The syntactic scope of X is YP, identifiable with either CP or vP, precisely the syntactic units considered as phases in Chomsky (2001). It seems that YP can also be TP. Crucially, these units may encode a propositional content. 3 Whenever a unit within the YP domain gets F-marked, the feature [+foc] triggers Copy of YP. The newly created copy of YP is then merged in [Spec,XP]. PF-deletion and LF-deletion apply to this structure. In all cases, the F-marked unit α in the low copy remains (that is to say, it is not deleted) both at PF and LF. On the other hand, the high copy of this unit is always complementarily deleted both at PF and at LF, and this latter deletion creates a variable. This latter is bounded by an abstract lambda-type operator. In terms of Information Structure (IS), the RS in (6) encodes the presupposition to the left of the head, whereas the main assertion, which contains the informative focus, is located to its right. This kind of divide is constant at the interpretive interface LF. At PF, however, there may be either a perfect match or a mismatch with LF in with respect to the deletion targets within the limits imposed by the FDC. In case where LF-/PF-deletion mismatch occurs, the FSU remains in situ at PF (that is to say, to the left edge of the phase), and then focus association at a distance obtains. 2. The RS and other focus sensitive units 2.1. Sp. sólo ‘only’ The first difference that sólo ‘only’ shows with respect to the sentence modal adverb reviewed in the previous chapter is that it cannot modify a sentence. Note that, in principle, (7) in Spanish 3 A proposition is anchored on time, so at first sight it may look like vP, being tenseless, would be unable to encode propositions. We can assume, however, that the propositional reading at LF can obtain from the tense value of the immediately c-commanding TP. Since focus is interpretively propositional, Alt-p requires this kind of reading in all cases. 158 and (8) in English may be intended as potential answers to the question Qué pasó? ‘What happened?’, respectively. (7) * Sólo [María compró un carro nuevo] F (8) * Only [Mary bought a new car] F We saw in the previous chapter that the remnant movement analysis of only proposed in Kayne (2000), reproduced here in (9), assumed that this FSU is base generated as a modifier of VP. (9) a. John only criticized Bill → (attraction by only) b. John Bill 1 only criticized t 1 → (raising of only to W) c. John only 2 Bill 1 t 2 criticized t 1 → (VP-preposing to [Spec,WP]) d. John [criticized t 1 ] 3 only 2 Bill 1 t 2 t 3 (Kayne 2000: 239) For the reasons presented in the previous chapter, I don’t follow the remnant movement analysis of focus in this work. However, there’s a similarity in terms of immediate adjacency between only and the focused unit. Note also that the unit to the left of the FSU is presupposed. As for the structural merging point for only, I assume that this unit (and its equivalente in Sp., sólo) is base generated in polarity position. This choice is related to the fact that only introduces negative assertions. Suppose that John was at the greengrocer and he bought apples, and nothing else. In this context, the truth of the proposition John bought only apples implies the truth of the set of propositions John didn’t buy oranges, John didn’t buy bananas, John didn’t buy grapes, and so on. In this sense, only associates with contrastive focus. The node PolP, however, is located above TP, and, as the example (8) illustrates, only has no sentential scope. Tentatively, I assume that TPs cannot be F-marked with contrastive focus. This fact invokes reduplication and ellipsis, so that the FSU can appear adjacent to a sub-sentential unit that is F- marked with contrastive focus. 159 Let’s take for instance the sentence in (10), taken from den Dikken (2006: 28). (10) (What does Imogen eat?) Imogen eats only [biscuits] F/fcP In (10), the DP biscuits is the unit in focus—as shown by the question in parentheses— with which only associates. This latter adverb surfaces in a position immediately adjacent to the F-marked DP, in a manner similar to what was previously observed for the modal probablemente ‘probably’. If these hints can be trusted, the correspondent RS analysis would be as follows. (11) a. Creation of the TP in syntax proper → [ TP Imogen eats biscuits] b. Merge of only + TP → only [ TP Imogen eats biscuits] c. F/fcP-marking on the DP-DO → only [ TP Imogen eats [biscuits] F/fcP ] d. Copy TP and merge it in [Spec,PolP] → [ PolP [ TP Imogen eats biscuits] i [ Pol’ only [ TP Imogen eats [biscuits] F/fcP ] i ]] e. At PF level, complementary PF-deletion → [ PolP [ TP Imogen eats biscuits] i [ Pol’ only [ TP Imogen eats [biscuits] F/fcP ] i ]] e’. At LF level, deletion & creation of variable via binding by Op.→ Op k [ PolP [ TP Imogen eats [ DPk biscuits] i ] [ Pol’ only [ TP Imogen eats [biscuits] F/fcP ] i ]] As it can be seen in the parallel steps (11e) and (11e’), in this case, both PF- and LF- deletion have the same targets, so deletion is matched. The correspondent RS simplified phrase marker is presented in (12). 160 (12) PolP TP i Pol’ only [+foc] TP i [Imogen eats [cookies]] [Imogen eats [cookies] F/fcP ] The sentence in Spanish equivalent to (10) is shown in (13). (13) María com-e sólo [galletas] F/fcP M. eat-3SG.PRS only cookies The correspondent equivalent derivation is presented in (14). (14) a. Creation of the TP in syntax proper → [ TP María come galletas] b. Merge of sólo + TP → sólo [ TP María come galletas] c. F/fcP-marking on the DP-DO → sólo [ TP María come [galletas] F/fcP ] d. Copy of TP and merge of it in [Spec,PolP] → [ PolP [ TP María come galletas] i [ Pol’ sólo [ TP María come [galletas] F/fcP ] i ]] e. At PF level, PF-deletion → [ PolP [ TP María come galletas] i [ Pol’ sólo [ TP María come [galletas] F/fcP ] i ]] e’. At LF level, deletion & creation of variable via binding by Op.→ Op k [ PolP [ TP María come [ DPk galletas] i ] [ Pol’ sólo [ TP María come [galletas] F/fcP ] i ]] 161 The syntactic tree for (14) is shown in (15). (15) PolP TP i Pol’ sólo [+foc] TP i [María come [galletas]] [María come [galletas] F/fcP ] As anticipated, the association with focus can also take place at a distance. I illustrate this case for Spanish in (16). (16) María sólo le entreg-ará las llaves [a [Susana] F ] fcP M. only Cl.DAT.3SG=give-3SG.FUT the keys to S. ‘María will give the keys only to Susana’ In this case, the F-marked unit Susana is inside the PP-IO, which in turn is fcP-marked. The derivation proposed for (16), which contains a PF-/LF-deletion mismatch, is shown in (17). (17) a. Creation of the TP in syntax proper → [ TP María le entregará las llaves a Susana] b. Merge of sólo + TP → sólo [ TP María le entregará las llaves a Susana] c. F/fcP-marking on the DP-IO → sólo [ TP María le entregará las llaves [a [Susana] F ] fcP ] d. Copy TP and merge it in [Spec,PolP] → 162 [ PolP [ TP María le entregará las llaves [ PP a Susana]] i [ Pol’ sólo [ TP María las llaves le entregará [a [Susana] F ] fcP ] i ]] e. At PF, vP deletion on the high copy & complementary deletion on the low copy→ [ PolP [ vP María le entregará las llaves [ PP a Susana]] i [ Pol’ sólo [ vP María le entregará las llaves [a [Susana] F ] fcP ] i ]] e’. At LF level, deletion & creation of variable via binding by Op.→ Op k [ PolP [ TP María le entregará las llaves [ PPk a Susana]] i [ Pol’ sólo [ TP María las llaves le entregará [a [Susana] F ] fcP ] i ]] The phrase-marker in (18a) summarizes the PF-structure proposed for the sentence in (16). The interpretation that associates the FSU with the focus marked unit even if they are not adjacent derives from the unmatched LF-structure in (18b). (18) a. PolP TP i Pol’ sólo [+foc] TP i María le entregará las llaves a Susana María le entregará las llaves [a[Susana] F ] fcP 163 (18) b. PolP TP i Pol’ sólo [+foc] TP i Op k María le entregará las llaves [ PPk a Susana ] María le entregará las llaves [a[Susana] F ] fcP It seems then that the RS analysis works relatively fine for cases as (13), where the DO is in focus, for (16), where the IO is focused, and also for (19)-(20), where a temporal adjunct is fcP-marked in sentence-final position. (19) Juan beb-e vodka sólo [los sábados] F/fcP J. drink-3SG.PRS vodka only the Saturdays ‘Juan drinks vodka only on Saturdays’ (20) PolP TP i Pol’ sólo [+foc] vP i [Juan bebe vodka [los sábados]] [Juan bebe vodka [los sábados] F ] 164 A special situation appears when the subject is in focus. For Spanish, there exists the possibility of pronouncing the focused subject in final-sentence position, as in (21). (21) Com-erá galletas sólo [Amanda] F/fcP Eat-3SG.FUT cookies only A. ‘Only Amanda will eat cookies’ The correspondent RS analysis would account at the same time both for the adjacency between the FSU and the focused unit, and for the final-sentence position of the subject. (22) PolP TP i Pol’ sólo [+foc] TP i [Amanda] comerá galletas [Amanda] F/fcP comerá galletas Example (23) illustrates the case when the subject is in sentence-initial position. (23) Sólo [Amanda] F/fcP com-erá galletas Only A. eat-3SG.FUT cookies ‘Only Amanda will eat cookies’ In this case, deletions are unmatched. The PF-structure of (23) is shown in (24a), whereas its LF-strucutre is shown in (24b). 165 (24) a. PolP TP i Pol’ sólo [+foc] TP i [Amanda] comerá galletas [Amanda] F/fcP comerá galletas b. PolP TP i Pol’ sólo [+foc] TP i Op k [ DPk Amanda] comerá galletas [Amanda] F/fcP comerá galletas 2.2. Sp. también ‘also’ and incluso ‘even’ Other frequent FSUs in Spanish are también ‘also’ and incluso ‘even’. In principle, these units share some contexts with sólo ‘only’. For instance, compare (13), repeated here as (25), with (26) and (27). These sentences share the characteristic that any of them can be a felicitous answer to the context-question Sp. Qué come María? ‘What does María eat?’ (25) María com-e sólo [galletas] F/fcP M. eat-3SG.PRS only cookies 166 (26) María com-e también [galletas] F/fcP M. eat-3SG.PRS also cookies (27) María com-e incluso [galletas] F/fcP M. eat-3SG.PRS even cookies I propose that, in the contexts of (26) and (27), the RS analysis is similar to the one previously proposed for (25) (=(13)). See (28) and (29), respectively. (28) FSUP TP i FSU’ también [+foc] TP i [María come [galletas]] [María come [galletas] F/fcP ] (29) FSUP TP i FSU’ incluso [+foc] TP i [María come [galletas]] [María come [galletas] F/fcP ] 167 Since the RS is similar, I assume that, at least for these three simple cases, the interpretive difference comes from the semantic content of each of the units. This difference can be visualized through the kind of representation named Assertion Structure (AS) (see Zubizarreta 1998, especially ch.1). In (30)-(32), the AS for sections (a) appears in sections (b). (30) a. María com-e sólo [galletas] F/fcP M. eat-3SG.PRS only cookies b. A 1 : There is an x, x ∈ Alt-fv, such that María eats x. (Existential presupposition) A 2 : There is no y, y ∈ Alt-fv, y ≠ x, such that María eats y. A 3 : The x, such that María eats x = [cookies] (Main assertion) (31) a. María com-e también [galletas] F/fcP M. eat-3SG.PRS also cookies b. A 1 : There is an x, x ∈ Alt-fv, such that María eats x. (Existential presupposition) A 2 : There is at least one y, y ∈ Alt-fv, y ≠ x, such that María eats y. A 3 : The x, such that María eats x = [cookies] (Main assertion) (32) a. María com-e incluso [galletas] F/fcP M. eat-3SG.PRS even cookies b. A 1 : There is an x, x ∈ Alt-fv, such that María eats x. (Existential presupposition) A 2 : There is at least one y, y ∈ Alt-fv, y ≠ x, such that María eats y. A 3 : The x, x ∈ Alt-fv, such that María eats x, is an extreme point in an ordered scale (assumed to be part of the conversational background) of the elements of Alt-fv. A 4 : The x, such that María eats x = [cookies] (Main assertion) 168 In (30)-(32), the assertion A 1 , which is common for all three FSUs, represents the existential presupposition. This presupposition includes necessarily a focus variable x, existentially bound. 4 On the other hand, the main assertion, which include the unit in focus that gives a specific value to the variable, is also the same in all three cases, since the unit in focus, namely galletas ‘cookies’, is the same too. These two parts of the AS (presupposition/ main assertion) are precisely the parts that are grammaticalized through the RS. The other ‘intermediate’ presuppositions come from the specific content of each FSU. 2.3. Sp. sólo, también, incluso: Some similarities and differences Sólo ‘only’ introduces a negative existential presupposition: there’s no other element y from the relevant set of alternative values Alt-fv such that y can occupy the focus-variable position. In the case of (30), this means that there is no other food y, apart from cookies, such that María eats y. On the other hand, también ‘also’ introduces a positive existential assertion according to which there exists at least one element y from Alt-fv such that y can occupy the focus variable position. For the example in (31), this indicates that, besides cookies, there exists at least one more thing that Mary eats—say, for instance, bread. In turn, incluso ‘even’ adds two assertions. One is existential, and is similar to the one presented for también ‘also’; the other one is scalar, and affirms that x (the focus value) occupies an extreme point in a scale assumed to be part of the speakers’ common ground. In the example (32), the scale may refer, for instance, to the quality of food, so the cookies may be considered either as the most valuable or as the less valuable food from a relevant list. Note that different FSUs may share some specific assertions. In this case, both también ‘also’ and incluso ‘even’ share an existential presupposition, but the former lacks the scalar presupposition that characterizes the latter. 4 On the basis of examples as [Nobody] F likes Bill, Herburger (2000) questions the analytical relevance of existential presuppositions like those presented in the assertions A 1 in (30)-(32). Commenting some ideas by Jackendoff (1972), Herbuger states that the aforementioned example “cannot be seen to presuppose that someone likes Bill, then going on to assert that it was nobody. If it did, [it] would be contradictory” (Herburger 2000: 14). Moreover, quoting the conclusions of Boer & Lycan (1976), Herburger criticizes the presuppositional analysis of focus itself, stating that presupposition “is mainly useful as a cover term, since presuppositions are not a unified phenomenon” (Herburger 2000: 17). Thus, this author proposes an alternative analysis called Structured Davidsonian Decomposition, which relates focus to quantification on events. In turn, Zubizarreta (1998), commenting a similar example, [Nobody] F lied to me, notes that, in this case, the focus is contrastive, since this kind of focus “may negate the assertion introduced by its context statement” (Zubizarreta 1998: 7). For the purposes of this work (which is not centered on the semantics of focus), the intuitive difference between presupposition (what the speakers assume to be the case at the point their conversation takes place) and main assertion (the assertion identifying the relevant focus value) will suffice. For a deeper characterization of the Davidsonian view of focus, I refer the interested reader to Herburger’s book (2000), and the sources cited therein. 169 On top of the differences about specific assertions, there may be some syntactic contexts that are not shared by the three FSUs presented here. Let’s illustrate this with sentences (33)- (35), which, in principle, are ambiguous in IS terms. In all three cases, the focused unit can be alternatively the whole vP, the DO, or just the tensed verb, provided that a special prosodic emphasis is added to this latter. These would be cases of ‘forward’ association with focus. Now an interesting difference arises with respect to the subject. While (33) is clearly ungrammatical, (34)-(35), with apparent ‘backwards’ association, are not. The context question for the three sentences is Quién come galletas? ‘Who eats cookies?’ (33) * María sólo com-e galletas M. only eat-3SG.PRS cookies (Ungrammatical with the reading ‘Only María eats cookies’) (34) María también com-e galletas 5 M. also eat-3SG.PRS cookies ‘Also María eats cookies’ (35) ? María incluso com-e galletas M. even eat-3SG.PRS cookies ‘Even María eats cookies’ These sentences show an intriguing structural contrast. We must take into account that, as presented here, the RS grammaticalizes the divide presupposition/main assertion by locating the former to the left and the latter to the right of the FSU head. In this sense, the linear order in (34) and (35) is unexpected. In these cases the divide is reversed—the focused unit appears to the left and the presupposition appears to the right. It is true that we can suppose that the reduplication has indeed applied below T, and a later raising of the FSU explains its surface position to the left of the tensed verb. Moreover, we can further assume that María has raised to some higher [Spec,TopP], so this move would explain its sentence-initial occurrence. This 5 There is a prosodic juncture after también. This is related to its condition of proform (see below). 170 explanation, however, would now have the burden on the side of (33): the problem would be why this mechanism does not work for sólo ‘only’. At least for the cases of también ‘also’, I think another possibility can be entertained. The FSU también can be used as a proform, that is to say, as a unit able to referentially stand for some elided content in certain contexts—in this case, vP/VPs, or even entire sentences. Let’s suppose two speakers A and B involved in a conversation in which they are considering who can play the guitar in a certain musical activity they are planning. (36) a. A: Susana toca la guitarra. Y María? S. plays the guitar And M. ‘Susana plays guitar. What about María?’ b. B 1 : María también toca la guitarra. M. also plays the guitar ‘Also María plays guitar’ B 2 : María también. M. also ‘Also María plays guitar’ B 3 : También. Also ‘Also María plays guitar’ To my understanding, the three alternative interventions of speaker B (B 1 , B 2 , and B 3 ) are perfectly synonymous. In the intervention B 1 the material to the right side of the FSU presents presupposed content, so presumably it would (tend to) be pronounced with a flat F0 contour, maybe even deaccented. In turn, the intervention B 2 —a ‘text-book’ example of VP ellipsis 171 under identity—seems to result from ellipsis on the presupposed part of the sentence, which keeps audible only the unit (we suppose is) in focus, namely the subject María. The comparison between B 1 and B 2 seems to confirm the view of ellipsis as an extreme form of deaccenting, and idea suggested by Tancredi (1992), Chomsky & Lasnik (1995), Vanden Wyngaerd & Zwart (1999), among others. 6 On this issue, with the proviso that their correspondent domains of applicability are not identical, Winkler (2005) articulates the similarities between deaccenting and ellipsis through the following two related hypotheses. (37) a. Null Suprasegmental Feature Hypothesis Deaccentuation results from marking the suprasegmental level as null. b. Null Segmental Feature Hypothesis Deaccentuation results from marking the suprasegmental and the segmental level as null. (Winkler 2005: 68). In the case of intervention B 2 in (36), ellipsis has targeted the presupposed part of the sentence, as expected. Interestingly, in the intervention B 3 , it looks like a more extreme case of ellipsis has occurred: the (supposed) focused unit María seems to have undergone ellipsis, violating the FDC, while the (supposed) FSU is the only surviving element. For the idea of deletion we have presented so far in this work, the intervention B 3 is surprising, so it deserves some special consideration. My interpretation of B 3 is as follows. Sp. también ‘also’ is a proform that refers to the proposition encoded by the sentence María toca la guitarra ‘María plays the guitar’, and simultaneously states its truth value as true. Note that all the elements in this sentence, except también, are given in the intervention of speaker A. In this sense, the whole content of that 6 For VP-ellipsis in particular, Tancredi (1992) claims, “VP-ellipsis is no more but an extreme case of deaccenting where a VP ceases to be audible altogether” (Tancredi 1992: 120). Now if ellipsis is only an extreme form of deaccenting, the environments of applicability for these two operations should be the same. According to Winkler (2005) however, this is not the case. See also Merchant (2001) for a critical view of the idea of deletion as an extreme case of deaccenting. 172 sentence can be assumed as presupposed. 7 Crucially, the question What about María? at the end of intervention of speaker A clearly indicates that this unit is not the focus, but instead it is introduced as the topic. I consider that proform itself is in focus—and this is why it is not deleted, so ultimately the FDC is not violated. The same proform is present in B 1 and B 2 , but in these cases, the presupposition (or a part of it) is pronounced, which is usually an option available in cases where ellipsis is realized via external deletion. To put this analysis to the test, let’s introduce a slight change in (36), by switching the topic from the instrument player to the instrument being played. (38) a. A: María toca la guitarra. Y el piano? M. plays the guitar And the piano ‘María plays guitar. What about piano?’ b. B 1 : María también toca el piano. M. also plays the piano ‘María also plays piano’ B 2 : El piano también. The piano also ‘María also plays piano’ B 3 : También. Also ‘María also plays piano’ 7 I am not implying with this that ‘given’ and ‘presupposed’ are one and the same concept. My only claim is that, in the context of the conversation in (36), the content of the sentence is already explicit in the intervention of speaker A, so it can be assumed as known (in this sense, presupposed) in the alternative interventions of speaker B. In fact, the analysis I entertain here militates in favor of the difference. For relevant discussion about the distinction between the pair presupposition/focus and the pair given/new (information), see Zubizarreta (1998: 160, n.4). 173 Starting this time with B 1 , we see that the proform states the truth value true (or “1”) for the proposition María toca el piano ‘María plays piano’. Along with the proform también ‘also’, the rest of the sentence expresses information already given in the intervention of speaker A, crucially including the topic, namely el piano. In turn, in B 2 only the topic and the proform survive ellipsis, as expected. Finally, in B 3 , only the proform is pronounced, which amounts to say that only the focused unit, i.e. the proform itself, has survived ellipsis under identity. Although no word belonging to the sentence that encodes the proposition has been actually pronounced, its true value is stated as 1. If this analysis is on the right track, we may conclude —at least provisionally—that there’s no ‘backwards’ association with focus with the Sp. adverb también ‘also’, as it appeared to be the case in (34). Also, there may be two kinds of también. One is an FSU; the other one is a proform. Now, sólo ‘only’ does not function as a proform, so is predictable that, on top of never surfacing with apparent ‘backwards’ association, it will be unable to participate in configurations similar to the ones illustrated in (36) and (38). This prediction is borne out (consider the same situational context for speakers A and B in these examples). (39) a. A: (Creo que) Susana toca la guitarra. Y María? I.think that S. plays the guitar And M. ‘(I think) Susana plays guitar. What about María?’ b. B 1 : Sólo María toca la guitarra. Only M. plays the guitar ‘Only María plays guitar’ B 2 : ? Sólo María. Only M. ‘Only María plays guitar’ 174 B 3 : * Sólo. Only ‘Only María plays guitar’ In this case, each alternative intervention of the speaker B corrects a belief of speaker A, so it is reasonable to consider that the focus on the subject is contrastive. Now, the answer B 3 is ungrammatical, since it supposes deletion of the focus marked unit. In turn, the status of B 2 is unclear. While it is not clearly ungrammatical, it seems somehow marginal. The reason for this is probably related to the fact that the sentence that would license the VP ellipsis is not immediately adjacent to B 2 in the discourse. We can arrive to similar conclusions from (40), where the question in A refers to the instrument played. (40) a. A: (Creo que) María toca la guitarra. Y el piano? I.think that M. plays the guitar And the piano ‘(I think) María plays guitar. What about piano?’ b. B 1 : (María/ella) toca sólo el piano. M./she plays only the piano ‘María only plays piano’ B 2 : ? Sólo el piano. Only the piano. ‘María only plays piano’ B 3 : * Sólo. Only ‘María only plays piano’ In sum, unlike también ‘also’, the adverb sólo ‘only’ is always an FSU, it associates with focus to its right only, and cannot function as a proform. 175 As for the adverb incluso ‘even’, while its status with respect to ‘backwards’ association is unclear, I consider it should be grouped with sólo, since—at least to my judgment—it cannot appear grammatically in the kind of answer B 3 . This time, let’s assume that the speaker B considers that the speaker A is likely to consider María (who belongs to a family of famous writers) too young to have already published a novel. (41) a. A: Varios miembros de esa familia han publicado novelas. Y María? Several members of that family have published novels. What about María b. B 1 : ? María incluso ha publicado una novela. 8 M. even has published a novel ‘Even María has published a novel’ B 2 : * María incluso. M. even ‘Even María has published a novel’ B 3 : * Incluso. Also ‘Even María has published a novel’ The partial conclusion at this point is that, while the RS analysis can be effectively applied to the three Sp. FSUs sólo ‘only’, también ‘too’, e incluso ‘even’ when they associate with an F-marked unit in the relevant domain, some particularities in their respective distributions can be explained in terms different from F-marking or association with focus. 2.4. Sp. probablemente ‘probably’ and sólo ‘only’ in the same sentence In this work, I have argued in favor of the idea that the sentence modal adverb probablemente ‘probably’ takes scope over CP, whereas the FSU sólo ‘only’ is base generated at PolP. Thus the 8 The acceptability of B 1 somehow improves if a prosodic ip boundary (intermediate phrase, value 3 in the Sp_ToBI notation) is inserted after the adverb: (María incluso)ip # (ha publicado una novela)ip. This may be interpreted as a sign that the adverb itself is in focus, whereas the second ip encodes the presupposition. 176 RS predicts that (i) the two adverbs can co-exist in the same sentence associated with the same focused unit, and (ii) in these cases, the linear order probablemente > sólo (i.e., the sentence modal precedes sólo) is fixed—that is to say, the reverse order sólo > probablemente will be unacceptable. In the following lines, I show that these predictions are borne out. As a way of illustration, let’s consider (42) and (43), where the DO galletas ‘cookies’ is interpreted as bearing informational focus. (42) María comerá probablemente sólo galletas. M. will.eat probably only cookies ‘Probably, María will eat only cookies’ (43) * María comerá sólo probablemente galletas. 9 (Ungrammatical with the reading ‘Probably, María will eat only cookies’) Both adverbs are sensitive to fcP-marking, so both of them will associate with the same focused unit, galletas ‘cookies’. This is shown in (42). On the other hand, the sentence in (43) shows that, assuming association with the same unit, the order sólo > probablemente is not grammatical. The derivation of a sentence containing the two adverbs proceeds in two cycles. In the first cycle, the structure María comerá solo galletas ‘M. will eat only cookies’ is created via reduplication and ellipsis. In the second cycle, the adverb probablemente merges with the sentence created in the first cycle. This triggers a second process of reduplication and ellipsis. This is summarized in (44) and (45). 9 Sentence (42) may have an acceptable reading if galletas ‘cookies’ is presupposed, and the modal adverb itself is taken as contrastively focus marked, via prosodic prominence. Interestingly, the set of alternatives Alt-fv would contain other modal adverbials expressing certainty: Alt-fv = {indudablemente ‘undoubtedly’, ciertamente ‘certainly’, con seguridad/de seguro ‘for sure’,…}. Naturally, this is not the intended reading for (42). 177 (44) Cycle I → María comerá sólo galletas. PolP TP i Pol’ sólo [+foc] TP i María comerá galletas María comerá [galletas] F/fcP (45) Cycle II → María comerá probablemente sólo galletas. ModP CP i Mod’ probablemente [+foc] CP i María comerá sólo galletas [María comerá sólo [+foc] [galletas] F/fcP ]] As indicated by the single strikethrough, (45) represents the PF-structure of (42). Now, since the deletions applied in both phases are matched, we would expect a parallel LF-structure. This one is shown in (46). 178 (46) ModP CP i Mod’ probablemente [+foc] CP i Op k [María comerá sólo [ DPk galletas]] [María comerá [ FSUP sólo [+foc] [galletas] F/fcP ]] My proposal now goes on to say that the LF-structure in (46) also applies for sentence (47), where probablemente associates at a distance, and for sentence (48), where both probablemente and sólo associate with focus at a distance. (47) Probablemente, María comerá sólo galletas. Probably M. will.eat only cookies ‘Probably, María will eat only cookies’ (48) Probablemente, María sólo comerá galletas. Probably M. only will.eat cookies ‘Probably, María will eat only cookies’ To derivate the PF-structure for (47), the computation in cycle I would be the same as the one presented in (44), which generates María comerá sólo galletas. This sentence is used as input for cycle II, where it merges with probablemente. As seen in the previous chapter, in this case PF-deletion would not match the correspondent LF-deletion. The high copy of CP gets PF- deleted, so probablemente ends up associated with the focused unit at a distance. The resulting (simplified) PF-structure for (47) is shown in (49). 179 (49) ModP CP i Mod’ Probablemente [+foc] CP i [María comerá sólo [galletas]] [María comerá [ FSUP sólo [+foc] [galletas] F/fcP ]] In the case of (48), since there are two associations at a distance, I assume two unmatched deletions. The first one occurs in cycle I. At this point, the high copy of vP is deleted, so the resultant PF-structure would be María sólo comerá galletas. This structure is taken as input for cycle II, where this sentence merges with probablemente. The second unmatched deletion occurs here. The high CP gets PF-deleted, so the PF-structure (50), which corresponds to (48), is formed. (50) ModP CP i Mod’ Probablemente [+foc] CP i [María sólo comerá [galletas]] [María sólo comerá [ FSUP [galletas] F/fcP ]] At this point, the following question may arise. We have assumed that both probablemente and sólo associate with the same unit within the sentence. Could they associate 180 with different units? In order to test this, let’s imagine for (51) a situational context in which Juan is a ‘treasure hunter’, so he usually looks for valuable objects in different pawn shops and thrifty stores. (51) a. Juan encontró probablemente un Picasso en el Goodwill de Figueroa J. found probably a (painting by) P. in the Goodwill of F. ‘It is probably a painting by Picasso what Juan found at the Goodwill at Figueroa St.’ b. Juan encontró un Picasso sólo en el Goodwill de Figueroa J. found a (painting by) P. only in the Goodwill of F. ‘It is only at the Goodwill at Figueroa St. where Juan found a painting by Picasso’ c. ? Juan encontró probablemente un Picasso sólo en el Goodwill de Figueroa J. found probably a (painting by) P. only in the Goodwill of F. The sentence in (51a) expresses that some painting found by Juan at the Goodwill store located at Figueroa Street. is probably an authentic painting by Picasso. In turn, (51b) expresses that the only place where Juan has found a painting by Picasso is the Goodwill at Figueroa Street. Now, (51c) attempts to combine both contents in just one sentence. If this attempt could be successful, it would amount to say that indeed two different FSUs can associate with two different units within the same sentence. To my view, however, this attempt fails. The limited degree of acceptability of (51c)— which justifies the <?> mark—is due to the interpretation of probablemente un Picasso as part of the presupposition. To a certain degree, (51c) could be acceptable if it is taken as an answer to the question in (52)—which, in turn, is not fully grammatical either, but for independent reasons. 10 10 As presented in Chapter 1, the modal probablemente is a positive polarity item (PPI), so it cannot appear as fully grammatical in questions (like (52)), if-clauses, imperatives, and other environments that lack declarative/affirmative force. This fact also explains why probablemente or its equivalent in English probably cannot appear under negation: *No students probably eat the wheaties (from Bellert 1977). On PPIs in general, see Progovac (1994) and Giannakidou (2011). For relevant discussion about high modals understood as PPIs, see Nilsen (2003). 181 (52) ? Dónde encontró Juan probablemente un Picasso? Where found J. probably a (painting by) P. ‘Where did Juan find a probable painting by Picasso?’ Note that what is being considered under the lens of probability is the authorship of the painting, both in (51c) and in (52). As the gloss shows, this probability is expressed in English via the adjective probable. The pair Adv. probably/Adj. probable is reproduced in Spanish as the pair Adv. probablemente/Adj. probable. 11 In this sense, (53), which is a ‘fixed’ version of (51c), is fully grammatical. In (53), we see just one FSU and just one association with focus, namely that of sólo with the PP en el Goodwill de Figueroa, which is F-marked. (53) Juan encontró un probable Picasso sólo en el Goodwill de Figueroa J. found a probable (painting by) P. only in the Goodwill of F. Indeed, the bottom line of these examples is the question of whether more than one unit can be F-marked with informational focus within the same sentence. If two (or more) syntactic units can be F-marked within the same sentence, it would be possible for two different FSUs to associate with different F-marked units. Following Zubizarreta (1998), my view on this issue is that just one unit can be F-marked within the same sentence. Another property worth mentioning is that there is one and only one focus per sentence. This is suggested by the fact that multiple wh-questions and their correspondent answers obligatorily have a linked interpretation; see [(54)] and its AS in [(55)]. (54) [Mary] F bought [the newspaper] F , [Peter] F bought [the book] F , … (Who bought what?) 11 About the syntactic relation between adverbs in –mente in Spanish and the adjectives from which they originate, see Fábregas (2003). 182 (55) a. There is an (x,y) such that x bought y b. The (x,y) such that x bought y = <Mary, the newspaper>, <Peter, the book>, … (Zubizarreta 1998: 6) Before leaving this sub-section, let me consider a case, involving enumerative coordination, where the linear order sólo > probablemente seems legitimate, being both adverbs clearly associated with focus. This case is illustrated in (56). For interpretive purposes, let’s assume a context in which the list of guests for María’s party is under discussion. (56) María ha invitado sólo a Susana, a Pedro y probablemente a Luis. M. has invited only A S., A P. and probably A L. 12 ‘María has invited only Susana, Pedro, and probably Luis’ In (56), the FSU sólo associates with the DO, in this case the list of guests, which is in focus. In turn, the modal probablemente has associated with one single guest, namely Luis. This seems to go against two conclusions we have just reached in the previous pages, namely (i) that the linear order probablemente > sólo is fixed (thus the order solo > probablemente is disallowed) when the two units associates with focus, and (ii) there can be just one focused unit per sentence. Moreover, in this case, it seems as if one focused unit (Luis) is nested inside the other focused unit (the DO). I maintain that the counterexample to our conclusions is only apparent. Example (56) conveys an illusion due to ellipsis. My guess is that in (56) there are structurally two clauses. The coordinative copula & = y ‘and’ located before the modal adverb marks the relevant boundary between them, as seen in (57). (57) a. [ &P [María ha invitado sólo [ &P a Susana & a Pedro] F ] & [probablemente María ha invitado [a Luis] F ]] 12 The occurrences of a in (56) correspond to the so-called personal-a, which appears before [+human] accusatives and is not considered syntactically as a real preposition (Zagona 2002). 183 b. [ &P [María ha invitado sólo [ &P a Susana & a Pedro] F ] & [probablemente María ha invitado [a Luis] F ]] The structure in (57a) shows the two coordinated clauses before ellipsis. To start with, I interpret the comma between a Susana and a Pedro as form of realization (possibly due to deletion) of the coordinative copula, 13 so indeed there are two coordinative phrases (&Ps) at different levels in (56). In the first clause, the FSU associates with the F-marked &P [a Susana & a Pedro]. In the second clause, the modal has associated with the F-marked DP [a Luis]. In turn, (57b) shows the relevant external PF-deletion that gives rise to the linear order seen in (56). At first sight, the content of the first clause may seem contradictory with respect to the content of the second clause. We have postulated that sólo ‘only’ introduces a negative existential assertion. Thus in the first clause, the use of sólo would indicate that, apart from Susana and Pedro, there is no other element from the set of potential guests who has been invited to the party. In spite of this, the second clause introduces the probability of including one more guest to the relevant list. This might be perceived as contradictory. A closer view would reveal, however, that there’s no such a contradiction. The second clause does not affirm positively that Luis has been invited to the party. It only states a probability. In that sense, there is no contradiction, but expansion of the common conversational background. A certain state of facts, expressed in the first clause, is complemented with a probable state of facts, expressed in the second clause. 14 I propose that a similar analysis can be applied to the expression in (58), where the linear order sólo > probablemente is found in subject position. Again I claim that it should be taken as a concealed bi-clausal structure reduced via ellipsis (external PF-deletion), as seen in (59). The 13 In enumerations, commas can be replaced with a lexical coordinative copula in the rhetorical scheme known as polysyndeton: María ha invitado a Juan y a Susana y a Pedro y a Luis. In the reverse rhetorical scheme called asyndeton, all coordinative copulas in the enumeration are realized as commas: María ha invitado a Juan, a Susana, a Pedro, a Luis. This would indicate that both a comma and a lexical coordinative copula are, in principle, alternative ways of realization for the syntactic head “&”. 14 Note that this expansion can be expressed with two clauses explicitly realized as such: Mary has invited only Sue and Peter. And she has probably invited John (too). 184 context for (58) is a conversation about the national teams participating in the upcoming soccer world cup. (58) Sólo Brasil, Alemania y probablemente España pueden ganar la copa. Only Brazil Germany and probably Spain can win the cup (59) a. [ &P Sólo [ &P Brasil & Alemania] F [pueden ganar la copa]] & [probablemente [España] F [puede ganar la copa]] b. [ &P Sólo [ &P Brasil & Alemania] F [pueden ganar la copa]] & [probablemente [España] F pueden ganar la copa] Interestingly, as shown in (59b), the final agreement is plural, which may indicate a post- syntactic reanalysis at the morphophonemic level. Alternatively, it may be the case that the second clause appears as a parenthetical between the subject and the predicate of the first clause. 15 In that case, the comma-analysis of parentheticals proposed in Giorgi (to appear), where K represents the ‘comma-head’, would look like (60). (60) a. [ KP [ &P Sólo [ &P Brasil & Alemania] F K y [probablemente [España] F [puede ganar la copa]] [ KP K sólo [ &P Brasil & Alemania] F [pueden ganar la copa]]] In any case, the analyses proposed show that the conclusions to which we have arrived in this sub-section, namely (i) the order probablemente > sólo is fixed when the two FSUs associate with the same unit, and (ii) there is just one F-marked unit per sentence, can still be considered valid. 3. Other –mente ‘-ly’ sentence adverbs, and negation in Spanish In this sub-section, I explore the possibility of extending the RS analysis to other sentence adverbs in Spanish. Just as a way to keep a common standard through the different kinds of 15 The value of the prosodic break index would be key to decide between these two alternatives. Parentheticals demand a neat, clearly perceptible IP boundary, whose break index value in the Sp_ToBI system is 4, the highest possible (see Aguilar et al. 2009). 185 adverbs in this language—some sort of Ariadne’s thread—I adopt here a morphological criterion, namely, the –mente ‘-ly’ ending. Thus, in principle, I will focus my attention on these adverbs and its adverbial equivalents when convenient for illustration purposes. Additionally, I will consider some cases of sub-sentential negation in Spanish. 3.1. Other –mente ‘-ly’ sentence adverbs Given their complexity and heterogeneity, there are several ways to classify adverbs. In general, the classifications proposed can be divided in two big groups: semantic classifications and syntactic classifications (see Delfitto 2006: 89 ff. for a summary). Being this work mostly aligned on the syntactic side, I assume a syntactic classification roughly based on a simplified version of Cinque’s universal hierarchy of adverbials (Cinque 1999). In this classification, adverbs and their equivalent adverbials are classified according to their mutual structural precedence relations. Initially proposed for Italian and French, and later found also for several genetically non-related languages of the world, Cinque puts forward the general order presented in (61). (61) “Higher” (sentence) AdvPs > “Lower” AdvPs > (DPsubj) (V) complements > Place, time, manner, etc. adverbials > (focused) “Lower” AdvPs > de-accented material (Cinque 1999: 16) For Italian and French, Cinque (1999) states that lower AdvPs are those that can occupy the clausal “space” between the auxiliary verb and the past participle, as is the cases of adverbs in italics in examples (62a,b) (from Cinque 1999: 4-5). (62) a. It. Alle due, Gianni non ha solitamente mica mangiato, ancora. b. Fr. A deux heures, Gianni n’a généralement pas mange, encore, ‘At two, G. has usually not eaten yet.’ As for higher sentence adverbs, Cinque (1999) remits to Jackendoff’s intuitions about adverbs taking scope over sentences, which also show an “orientation” towards either the speaker or the subject. The English adverb frankly in the sentence Frankly, John lied to Bill is 186 presented as an example of speaker oriented adverb (SpOA), whereas the adverb carefully in the sentence Carefully, John spilled the beans is presented as an example of subject oriented adverb (SbOA) (Jackendoff 1972: 56ff). Interestingly, this author notes that, when coexisting within the same sentence, SpOAs always precede SbOAs, so the order *[SbOA > SpOA] is banned. This claim is illustrated in (63). (63) Evidently, Max carefully was trying to climb the walls of the garden. * Carefully, Max evidently was trying to climb the walls of the garden. (Jackendoff 1972: 89) As seen so far in this work, the adverb probably and its counterpart in Spanish probablemente belong to the class of (higher) sentence adverbs, a fact that is easy to perceive intuitively because of the propositional scope of the modal. Now following (non-exhaustively) Cinque’s structural proposal, I will consider the following basic types of sentence adverbs, from higher to lower (Cinque 1999: 106). 16 (64) a. Speech-Act Advs.: Francamente ‘frankly’, honestamente ‘honestly’, sinceramente ‘sincerely’… b. Evaluative Advs.: Afortunadamente ‘fortunately’, felizmente ‘happily’, desafortunadamente/desgraciadamente ‘unfortunately’… c. Evidential Advs.: Evidentemente ‘evidently’, obviamente ‘obviously’, visiblemente ‘visibly’… d. Modal Advs.: Probablemente ‘probably’, necesariamente ‘necessarily’, posiblemente ‘posibly’ 17 … 16 Note that I am taking Cinque’s Universal Hierarchy as an ordering device only. I am not necessarily assuming all aspects of that proposal. On this issue, see chapter 1 on this work. See also Velásquez (2009) (which in turn is based on Nilsen 2003) for a general overview of Cinque’s hierarchy for Spanish. See also Zyman (2012), Smit (2013), and Bhatia (n.d.); these works put Cinque’s hierarchy to the test for English, Afrikaans, and Hindi, respectively. 17 In Cinque’s hierarchy, each of these three adverbs belong to a different functional projection: Mod-epistemic, Mod-necessity, and Mod-possibility, respectively. It is assumed that this order is also their correspondent hierarchical order. Since this difference is not relevant for my analysis at this point, I put them all together in one single group here, attending to their common modal semantics. 187 e. Aspectual Advs.: Usualmente ‘usually’, nuevamente ‘again’, frecuentemente ‘frequently’ 18 … f. Volitional Advs.: Intencionalmente ‘intentionally’, deliberadamente ‘deliberately’, sin intención/sin querer ‘unintentionally’… Groups (64a-d) can be considered as subdivisions of the SpOAs distinguished in Jackendoff (1972). In turn, volitional adverbs in (64f) are considered as SbOAs in Cinque’s proposal. At this point, I would like to consider in (65a) some other SbOAs that, following Toner Castells (2005), I call Agent evaluative adverbs. These are obviously related to volitional adverbs, but since they are not limited to volitional aspects, I include them in a separate group. Additionally, I will add domain adverbs in (65b) to the classification. In a way analogous to topics, these adverbs introduce the field or point of view from which something is asserted. 19 (65) a. Agent evaluative Advs: Inteligentemente ‘cleverly’, estúpidamente ‘stupidly’, cuidadosamente ‘carefully’, tontamente ‘foolishsly’… b. Domain Advs.: Económicamente ‘economically’, socialmente ‘socially’, políticamente ‘from a political point of view’… Before moving to the correspondent examples, two previous warnings are in order. In first place, the difference between sentence adverbs and VP/vP adverbs seems to be, at least for some cases, only positional. By this I mean that one and ‘the same’ adverb may appear as sentential or as VP-internal. 20 18 Again, in Cinque’s proposal, each of these adverbs correspond to its own functional projection: Asp-habitual, Asp-repetitive (I), and Asp-frequentative (I), respectively. It is assumed that there are also a lower repetitive adverb and a lower frequentative adverb in the hierarchy (hence, the “(I)” for the higher projections). In this case, I put the three adverbs together attending to their shared aspectual value. 19 For deeper considerations about domain adverbs in Spanish, see Kovacci (1999) and Rodríguez Ramalle (2009). 20 The question about ‘sameness’ is pertinent in this case. It is obvious that the set of phonetic features is exactly the same for both uses. It is reasonable to consider that the set of semantic features is also the same. Now, if we assume that the potential syntactic scope is already encoded in the lexicon, via subcategorization, the set of formal/syntactic features may be considered different. In that case, while the adverb in (66a) would be encoded as a sentence scope- taker, the one in (66b) would be listed as a VP-scope taker. 188 (66) a. Sincerely, the movie was boring. b. The witness spoke sincerely. In (66a), the adverb sincerely is a SpOA such that it qualifies the speaker’s attitude towards his/her own statement. It could be paraphrased as, I am being sincere when I say that the movie was boring. On the other hand, in (66b) the same adverb is used as a manner adverb related to the verb. It could be paraphrased as, The witness spoke in a sincere manner. This difference is important, since I will be concerned here only with sentence adverbs that appear located in low-sentence positions, presumably due to focus marking, not with VP- adverbs themselves, whose lower location is unmarked, that is to say, they are base generated in low sentence positions. 21 It has been proposed that negation can be used as a test to differentiate between these two potential uses. I illustrate this for Spanish. The example in (67) is taken from Torner Castells (2005). 22 (67) a. El asunto no ha terminado desgraciadamente. The issue NEG has ended unfortunately ‘The issue didn’t end in an unfortunate manner’ b. El asunto no ha terminado, desgraciadamente. The issue NEG has ended unfortunately ‘Unfortunately, the issue has not ended’ (Torner Castells 2005: 16) 21 This warning is especially applicable to the low agent evaluative adverbs. When used in lower positions, the evaluative adverb itself tends to be in focus: John answered the questions [stupidly] F . 22 This author, in turn, attributes the original example to Hernanz & Brucart (1987). 189 In (67a), as shown by the English gloss, the adverb desgraciadamente ‘unfortunately’ is a VP-manner adverb that falls under the scope of negation. In (67b), however, the same adverb is a sentence adverb, and, as such, it is not affected by negation. Importantly for the present work, sentence modal adverbs cannot appear as VP-adverbs. The following example for Spanish is taken from Rodríguez Ramalle (2001). (68) a. Probablemente, María analiza el problema. Probably, M. analyzes the problem b. * De manera probable, María analiza el problema. In manner probable, M. analyzes the problem Rodríguez Ramalle (2001: 118). This author points out that the Sp. adjective probable ‘probable’ can only be applied to propositions, so it cannot be used as modifier of a noun as manera ‘manner’. 23 Since this author claims that the –mente adverb inherits the semantic features of the adjective from which it originates, this implies that probablemente ‘probably’ cannot be used as VP-internal adverb, as seen in the following contrast. (69) a. * María analiza el problema probablemente. M. analyzes the problem probably b. María analiza el problema, probablemente. 23 “La agramaticalidad de la oración [(68b)] se debe a que el adjetivo probable sólo se predica de las proposiciones, con lo que nunca podrá predicarse del nombre manera, al no cumplir los requisitos de su selección. […] A este respecto, adviértase que un adverbio como probablemente nunca podrá interpretarse como un modificador del verbo, ya que la base adjetiva no denota propiedades relativas a la manera de realizar las acciones” (Rodríguez Ramalle 2001: 118). 190 The second warning concerns to the fact that, unlike sentence modal adverbs and domain adverbs in (70), all other sentence adverbs considered in this section entail the truth of the sentence they modify; see (71). (70) a. Probably, John will buy a novel. Does not entail: John will buy a novel. b. Economically, your proposal is unquestionable. Does not entail: Your proposal is unquestionable. (71) a. Frankly, your movie is boring. Entails: Your movie is boring. b. Fortunately, John bought the tickets in January. Entails: John bought the tickets in January. c. Evidently, the current administration favors the richest. Entails: The current administration favors the richest. d. Usually, Peter buys presents for his nephews. Entails: Peter buys presents for his nephews. e. Intentionally, Tom gave the documents to the wrong person. Entails: Tom gave the documents to the wrong person. f. Intelligently, Sue booked her tour to Japan for the low season. Entails: Sue booked her tour to Japan for the low season. This difference is important in semantic grounds. For the adverbs in (70), the truth-value of S is not implied by the truth-value of [SAdv+S]. This lack of entailment can be seen as some sort of ‘guarantee’ of their propositional scope. In the case of the adverbs in (71), the entailment 191 does not provide a similar guarantee, so their propositional scope has to be checked via independent tests (for instance, via the negation-test mentioned in (67)). The sentences in (71) may be paraphrased as a coordination of sentences [ &P S & S’], where S’ translates the content contributed by the adverb to the global meaning. (72) a. Frankly, your movie is boring. ≈ Your movie is boring, and I am being frank when I say it. b. Fortunately, John bought the tickets in January. ≈ John bought the tickets in January, and this fact is fortunate (for me, the speaker). 24 c. Evidently, the current administration favors the richest. ≈ The current administration favors the richest, and this fact is evident (for me, the speaker). d. Usually, Peter buys presents for his nephews. ≈ Peter buys presents for his nephews, and this occurrence is usual. e. Intentionally, Tom gave the documents to the wrong person. ≈ Tom gave the documents to the wrong person, and he did that on purpose. f. Intelligently, Sue booked her trip to Japan for the low season. ≈ Sue booked her trip to Japan for the low season, and that was intelligent from her. In the case of the adverbs in (70), the truth-value of S is unknown/undefined, precisely in virtue of the modification by the sentence adverb, so a similar equivalence [ &P S & S’] fails as a paraphrase. 24 In this case, the orientation towards the speaker is assumed by default. An alternative orientation is possible, but then it must be explicitly stated: Fortunately for you guys, John bought the tickets in January. 192 (73) a. Probably, John will buy a novel. * John will buy a novel, and this fact is probable. b. Economically, your proposal is unquestionable. * Your proposal is unquestionable, and that fact is stated from the point of view of economy. This being said, let’s evaluate the RS with respect to these sentence adverbs. I exclude sentence modals, since they have been object of RS analysis in the previous chapter. The following examples are the Spanish equivalent of sentences (71a-f) and (70b), with an explicit F- marking presented via bracketing. As (74) shows, these adverbs can associate at a distance with an F/fcP-marked unit within their correspondent c-command domain. To simplify the analysis, I only consider sentences with the F-market unit located in sentence-final position. (74) a. Francamente, tu película es [aburrida] fcP Frankly, your movie is boring b. Afortunadamente, Juan compró los boletos [en enero] fcP Fortunately, J. bought the tickets in January c. Evidentemente, el actual gobierno favorece [a los más ricos] fcP Evidently the current government favors A the most rich ‘Evidently, the current administration favors the richest’ d. Usualmente, Pedro compra regalos [para sus sobrinos] fcP Usually P. buys presents for his nephews e. Intencionalmente, Tomás le dio los documentos [a la persona equivocada] fcP Intentionally, T. Cl.=gave the documents to the person wrong ‘Intentionally, Tomás gave the documents to the wrong person’ 193 f. Inteligentemente, Susana reservó su viaje a Japón [para la temporada baja] fcP Intelligently, S. booked her trip to Japan for the season low ‘Intelligently, Susana booked her trip to Japan for the low season’ g. Económicamente, tu propuesta es [incuestionable] fcP Economically, your proposal is unquestionable Now, let’s assume that the focus association between the high sentence adverbs in (75) and the F/fcP-marked units is analogue to the one presented in the previous chapter for sentence modal adverbs. If this is so, the other sentence adverbs should also be able to surface (non- parenthetically) immediately adjacent to the fcP-marked unit. This prediction is borne out. (75) a. Tu película es francamente [aburrida] fcP Your movie is frankly boring b. Juan compró los boletos afortunadamente [en enero] fcP J. bought the tickets fortunately in January c. El actual gobierno favorece evidentemente [a los más ricos] fcP The current government favors evidently A the most rich ‘Evidently, the current administration favors the richest’ d. Pedro compra regalos usualmente [para sus sobrinos] fcP P. buys presents usually for his nephews e. Tomás le dio los documentos intencionalmente [a la persona equivocada] fcP T. Cl.=gave the documents intentionally to the person wrong ‘Intentionally, Tomás gave the documents to the wrong person’ 194 f. Susana reservó su viaje a Japón inteligentemente [para la temporada baja] fcP S. booked her trip to Japan intelligently for the season low ‘Intelligently, Susana booked her trip to Japan for the low season’ g. Tu propuesta es económicamente [incuestionable] fcP Your proposal is economically unquestionable I take the sentences in (75) as evidence in favor of an RS analysis for (at least these) non- modal sentence adverbs. This means that their sensitivity to focus is also lexically encoded through a [+focus] feature. This feature is able to trigger Copy+Merge of the c-commanded clause. A high copy of the clause is then merged into [Spec,FSU]. Complementary deletion, guided by the LCA and the constraint FDC, which proscribes deletion of the focus marked unit, produces the linear adjacency (a.k.a. ‘transportability’) between the adverb and the F/fcP-marked unit at PF level. In turn, at LF level, deletion of the (F-unmarked) high copy of the F-marked unit creates a variable that is then bounded by an abstract lambda-type operator. Let’s illustrate this structural claim for (75b). The correspondent (simplified) PF- structure is shown in (76), whereas the (matched) LF-structure appears in (77). (76) FSUP CP i FSU’ afortunadamente [+foc] CP i [Juan compró los boletos [en enero]] [Juan compró los boletos [en [enero] F ] fcP ] 195 (77) FSUP CP i FSU’ afortunadamente [+foc] CP i Op k [Juan compró los boletos [ PPk en enero]] [Juan compró los boletos [en [enero] F ] fcP ] A similar RS analysis can be postulated for all other cases in (75), with the obvious (but irrelevant) differences. The sentences in (74), which show focus association at a distance, have a LF-structure similar to the one in (77), with an operator binding a variable created via deletion. In turn, their PF-structures result from PF-deletion applied to the high copy of the clause. As suggested in the previous chapter, focus association at a distance can be seen, also for these adverbs, as the result of unmatched deletions applied at PF- and LF-levels. I illustrate this idea with the unmatched structures for (74a). (78) FSUP CP i FSU’ Francamente [+foc] CP i [Tu película es [aburrida]] [tu pelicula es [aburrida] fcP ] 196 (79) FSUP CP i FSU’ francamente [+foc] CP i Op k [Tu película es [ AdjPk aburrida]] [Tu pelicula es [aburrida] fcP ] Interestingly, two (or more) sentence adverbs may coexist within one and the same sentence, associated with the same unit in focus—as predicted, indeed, in Cinque’s hierarchy. 25 (80) Afortunadamente, Juan regresará a casa probablemente en diciembre Fortunately, J. will.come.back to house probably in December ‘Fortunately, Juan will come back home probably in December’ The sentence in (80) assumes a context where Juan is, say, a soldier currently stationed in war zone. It can be felicitously used as response to the question Cuándo regresará Juan a casa? ‘When will Juan come back home?’. This indicates that the unit in focus is the PP en diciembre ‘in December’. As it is visible, the evaluative associates at a distance, whereas the modal associates via immediate adjacency. 26 25 There may be some pragmatic/psycholinguistic constraint militating against more than two high-FSUs co- occurring within the same sentence. For instance, although comprehensible, the following sentence seem hard to process: ?Frankly, probably, economically, your proposal is inadequate. However, with only two high-FSUs, one located in high position and the other located immediately adjacent to the F-marked unit, the acceptability seems to increase: Frankly, your proposal is probably inadequate; Frankly, your proposal is economically inadequate; Probably, your proposal is economically inadequate. I leave this issue open. 26 In principle, it is grammatically possible to have the two high FSU adverbs either both at a distance or both immediately adjacent to the F-marked unit. However, Spanish grammar bans two –mente adverbs surfacing consecutively (presumably for phonological reasons): *Afortunadamente, probablemente, Juan regresará en diciembre; *Juan regresará afortunadamente probablemente en diciembre. In these cases, the use of equivalent adverbials can help: ?Por fortuna, probablemente Juan regresará en diciembre; ?Juan regresará por fortuna 197 If both adverbs are FSUs, then we must conclude that the derivation for (80) would include two reduplications, one per FSU. To reach that result, I assume that the derivation is carried out in two cycles, correspondent to two CP-phases, in a way somehow similar to what has been proposed in sub-section 2.3 for the co-occurrence of a CP-adverb (probablemente ‘probably’) and a TP/vP-adverb (sólo ‘only’). For (80), in the first cycle, a RS is created via Copy+Merge of the initial clause, triggered by the modal probablemente. In the second cycle, the result of cycle I, shipped to Spell-Out, is used as input for a second RS, triggered this time by the [+foc] feature on the evaluative afortunadamente. The reduplication in this second cycle is complemented by an unmatched deletion, which yields focus association at a distance for this latter adverb. Before moving into focus and negation, I would like to suggest that those –mente adverbs that are base generated in low positions don’t have a [+foc] feature, thus are unable to trigger reduplication. They do show some sort of ‘transportability’ though, so this fact is in need of some explanation. Let’s illustrate this with the Sp. Adverb completamente ‘completely’—the lowest possible adverb in Cinque’s hierarchy. (81) a. Juan limpió el garaje completamente J. cleaned the garage completely b. Juan limpió completamente el garaje J. cleaned completely the garage c. Juan completamente limpió el garaje J. completely cleaned the garage probablemente en diciembre. In any case, a different location for the two adverbials gives even better results (at least to my judgment): Por fortuna, Juan regresará probablemente en diciembre. 198 d. # Completamente, Juan limpió el garaje Completely J. cleaned the garage I tentatively assume that (81a) represents the base position for this adverb. Moreover, I assume that this is the base position for all manner adverbs. In that position, it can be F-marked. This happens when (81a) is used as felicitous answer to the question Cómo limpió el garage Juan? ‘How did Juan clean the garage?’. Sentence (81b), in turn, shows the case in which the adverb is part of the presupposition, while the DO el garaje ‘the garage’ is F-marked; this sentence can be used as a good answer for the question Qué limpió Juan completamente? ‘What did Juan completely cleaned? In turn, (81c) implies that the adverb is part of the presupposition and the focus marked unit is the VP; the correspondent question for this case would be Qué hizo Juan completamente? ‘What did Juan completely do?’. Finally, (81d), which in principle could be interpreted as the result of topicalizing the adverb, is nevertheless degraded, presumably because, being a manner adverb, completamente cannot take scope over the whole sentence. If this explanation makes sense, ‘transportability’ in low adverbs would also be derived from the divide focus/presupposition, but unlike sentence adverbs, the source of this divide would not be a feature located in the adverb itself. On the contrary, the adverb can be F-marked in its base position. I propose that cases like this can be better treated as part of the more general phenomenon of marked word orders in Spanish, which I deal with below. 3.2. Focus and negation Negation (NEG) in Spanish seems sensitive to focus. This is exemplified in (82). (82) a. Juan no le regaló una novela a Susana por su cumpleaños (, sino por Navidad) J. NEG Cl.=gave a novel to S. for her birthday (, but for Christmas) ‘Juan didn’t give a novel to Susana for her birthday (, but for Christmas)’ 199 b. Juan no le regaló una novela a SUSANA por su cumpleaños (, sino a Javier) ‘Juan didn’t give a novel to SUSANA for her birthday (, but to Javier)’ c. Juan no le regaló una NOVELA a Susana por su cumpleaños (, sino un CD) ‘Juan didn’t give a NOVEL to Susana for her birthday (, but a CD)’ d. Juan no le REGALÓ una novela a Susana por su cumpleaños (, sino que se la prestó) J. NEG Cl.=gave a novel to S. for her BD (, but that Cl.DAT=Cl.ACC=lend) ‘Juan didn’t GIVE a novel to Susana for her birthday (, but he lend it to her)’ In (82a), the PP por su cumpleaños ‘for her birthday’ is focus marked in sentence-final position. In all other cases (82b-d), there is F-marking in situ, so prosodic prominence is expected on the focused unit, as indicated by the capitals. On the other hand, the unit in parentheses introduces an alternative value for the unit in focus, which corrects the value expressed in the previous part—in this sense, these would be cases of contrastive focus. The content in parentheses also makes explicit the fact that, for each sentence in (82a-d), there is a different unit falling under the scope of negation. If this is on the right track, NEG in Spanish can be considered as an FSU, capable of establishing association with focus at a distance. Interestingly, NEG in Spanish can also surface immediately adjacent to the focused unit. 27 This is shown in (83). 28 27 Contra Zagona (2002), who states that Sp. no (=NEG) “must be adjacent to the verb […] Only clitics can intervene between no and the verb” (Zagona 2002: 195). For this author, a violation of this constraint would be the source of the ungrammaticality of (i). (i) * Juan no ayer leyó el diario J. not yesterday read the newspaper ‘Juan didn’t read the newspaper yesterday’ (Ibidem) As we will immediately see, the linear sequence no ayer is perfectly grammatical: Juan leyó el diario no ayer, sino el domingo pasado ‘J. read the newspaper not yesterday, but last Sunday’. I propose that the ill-formedness of (i) is related to IS: the focus-value (ayer) has been negated, so an interpretive expectation is created, but not satisfied, since no alternative (positive) focus-value is offered. See below my comments about the AS of (86). 28 In order to simplify the analysis (again), in (83) the focused unit appears in sentence-final position. 200 (83) a. Juan le regaló una novela a Susana no por su cumpleaños *(, sino por Navidad) J. Cl.=gave a novel to S. NEG for her birthday *(, but for Christmas) ‘Juan didn’t give a novel to Susana for her birthday , but for Christmas’ b. Juan le regaló una novela no a Susana *(, sino a Isabel) J. Cl.=gave a novel NEG to S. *(, but to Isabel) ‘Juan didn’t give a novel to Susana , but to Javier’ c. Juan le regaló a Susana no una novela *(, sino un CD) J. Cl.=gave to S. NEG a novel *(, but a CD) ‘Juan didn’t give Susana a novel, but a CD’ Since NEG in Spanish can associate with focus at a distance, and since it can additionally appear immediately adjacent to the F-marked unit, an RS analysis seems plausible for that unit. To start implementing it, I consider that NEG corresponds to the node PolP. From the lexicon, NEG is loaded the [+foc] feature. This feature explains its sensitivity to focus, and triggers Copy+Merge of TP when some unit in its domain gets F-marked. Deletion applies then in the already known fashion: it is complementary, it is carried out constrained by the LCA and the FDC, it can be either matched or unmatched, and it always creates a variable at LF. Taking the sentence in (83b) as example, I propose the PF-structure in (84) and the matched LF-structure in (85). 201 (84) PolP TP i Pol’ no [+foc] TP i [Juan le regaló una novela [a Susana]] [Juan le regaló una novela [a Susana] fcP ] (85) PolP TP i Pol’ no [+foc] TP i Op k [Juan le regaló una novela [ PPk a Susana]] [Juan le regaló una novela [a Susana] fcP ] As indicated by the asterisks, in (83) the units in parentheses are no longer optional: the unit that starts with the negative coordinative sino ‘but’ must be pronounced. This is due to the fact that it contains a unit in focus—precisely, the one that ‘corrects’ the previous focused unit (considered wrong, thus negated). 202 To syntactically represent this, I propose a bi-clausal coordination in which external PF- deletion has applied. The first clause is the CP that includes the RS represented in (85). The structure suggested for (83b) is shown in (86). (86) &P CP &’ & CP Juan le regaló una novela no [a Susana] fcP sino Juan le regaló una novela [a Isabel] fcP Importantly, I am neither assuming that sino ‘but’ 29 contains a [+foc] feature, nor that the two CPs are copies with respect to each other (note that there are no superscripts). 30 The structure in (86) is not an RS itself. It contains an RS at the vP level, namely the one represented in (84)-(85), in the high clause, but there is no reduplication triggered by sino. On the other hand, I do claim that the PF-deletion applied in the low CP is external—and since external 29 The coordinative unit sino ‘but’ is a NPI (unlike pero, which is also usually translated as ‘but’, but is not an NPI). This means that sino requires a negative environment to surface, as in (i) (= (86)). If this negative environment is absent, as in (ii), sino is not licensed. Interestingly, sino itself cannot license another NPI, as for instance tampoco ‘either’, in its complement (see (iii)). In this sense, sino appears like a ‘pivot’ for polarity contrast: negative environment in its Spec vs. positive environment in its Compl. This makes it ideal to express contrastive focus. (i) Juan no le regaló una novela a Susana, sino a Isabel. (ii) Juan le regaló una novela a Susana {*sino/ y} a Isabel. (iii) Juan no le regaló una novela a Susana, sino (*tampoco) a Isabel. For the structure in (86), I propose that the highest label of the high CP can be taken as a polarity phrase (PolP) marked with a negative value, presumably via a high operator encoding sentential negation. Since this high clause is located in [Spec,&P], the licensing of the NPI would be checked in a Spec-Head configuration. 30 Although the two clauses are not copies of each other as a whole, it is possible to entertain the idea that the high clause, which is derivationally created after the low clause, has been built using copies of the relevant units (Juan, le regaló, una novela) taken from the low clause, instead of taken directly from the lexicon. I will return to this idea when discussing the derivation of pseudo-clefts. 203 deletion is not the result of reduplication, it doesn’t need to be complementary. 31 I am further saying that there are two foci, one for each clause, and, crucially, the units in focus are different in the two clauses, since the focus-value in the low clause provides the focus-value that is absent in the high clause precisely because of negation. Note, on this issue, that the assertion structure (AS) for the first clause is unable to pick a positive value for the focus variable. The lack of a positive assertion leads to an unfulfilled expectation on the interpretive side. Thus, the coordination with a second clause containing the same presupposition with a positive focus-value is enforced—and this is why the pronunciation of the content in parentheses in (83) is not optional. (87a) corresponds to the high clause, whereas (87b) shows its correspondent AS. In turn, (88a) corresponds to the low clause, and (88b) shows its AS. Note that their respective existential presuppositions are identical. (87) a. Juan le regal-ó una novela no [a Susana] F/fcP (…) (High clause) J. Cl.DAT.3SG=gave a novel NEG to S. b. A 1 : There is an x, x ∈ Alt-fv, such that Juan gave a novel to x. (Existential presupposition) A 2 : The x, x ∈ Alt-fv, such that Juan gave a novel to x ≠ [Susana] (Negative main assertion) (88) a. (…sino que) Juan le regal-ó una novela [a Isabel] F/fcP (Low clause) (…but) J. Cl.DAT.3SG=gave a novel to I. b. A 1 : There is an x, x ∈ Alt-fv, such that Juan gave a novel to x. (Existential presupposition) 31 While external deletion (up to the focus) seems obligatory on the lower clause, I think it is also possible to flesh it out in some cases. In one possible scenario, there is an overt complementizer que—which, by the way, confirms its clausal nature—, the subject is realized as pro, and the DO is expressed pronominally via the accusative clitic (la), which in turn enforces a morphological readjustment on the dative clitic (le → se): Juan le regaló una novela no a Susana, sino que se la regaló a Isabel. 204 A 2 : The x, x ∈ Alt-fv, such that Juan gave a novel to x = [Isabel] (Positive main assertion) If these ideas are on the right track, at least for some cases, NEG could be analyzed as an FSU loaded with a [+foc] feature, which is merged with TP and creates an RS, provided that some unit has been F-marked in its relevant domain. 32 NEG’s association with focus can be realized either via adjacency, in the cases of matched deletion, or at a distance, in the cases of unmatched deletion. Additionally, I consider that these ideas may shed some light on the analogous structure not only X, but also Y, as used in the sentence Mary bought not only books, but also CDs. I want to suggest that this structure, in which two overt FSUs are involved, also comprises F/fcP- marking, reduplication, a concealed bi-clausal structure, contrastive/corrective focus, and association with focus via immediate linear adjacency. 33 4. Marked word order in Spanish. The RS approach Although frequently considered as an SVO language, Spanish may be better characterized as a relatively flexible language with respect to order of constituents. 34 In principle, the three orders 32 Lorenzo (1998: 105 ff) suggests an analysis of sub-sentential negation in Spanish in which NEG is base generated as directly adjacent to the unit in focus, as seen in (ii). From that position, it moves upwards to adjoin the complex [V+T] by the left, which yields the SS shown in (i). (i) SS: El niño no com-ió [las galletas] F The boy NEG eat -PST.3SG the cookies ‘The boy didn’t eat the cookies’ (ii) DS: El niño -ió com-√ no [las galletas] F Although the RS analysis may adopt the landing site (= [V+T]) for the rasing of NEG (which, by the way, is not obligatory, as (83) shows), the departing point would be different. In the RS perspective, NEG is generated at PolP. The linear adjacency to the unit in focus is not base generated, but instead is the result of F-marking, which triggers Copy+Merge, which, in turn, is followed by complementary deletion. From a minimalist perspective, Lorenzo’s proposal would mean that NEG merges with whatever the unit in focus may be before a phase structure (vP/CP) has been reached. This is hard to sustain, since focus is propositional by nature. A detailed comparison between the two approaches, however, is beyond the scope of the present work. 33 I leave the analysis of this structure for future research. 34 About word order from the point of view of typology, see Comrie (1989)—especially ch. 4 (p. 86 ff). For word order in Spanish, see Zubizarreta (1998, 1999, 2013), Ordóñez (1997, 2000), and Zagona (2002), among others. 205 SVO, VOS, and VSO are available for declaratives, although the former two may be perceived as more natural than the latter (Zagona 2002). 35 (89) a. Mi hermana escribió una carta (SVO) My sister wrote a letter b. Escribió una carta mi hermana (VOS) c. Escribió mi hermana una carta (VSO) (Adapted from Zagona 2002: 27) It has also been observed that there is a relation between the aforementioned flexibility with respect to word order and the position of informative focus in Spanish (Zubizarreta 1998, 1999). 36 Unlike contrastive focus, which is marked via emphatic prominence, non-contrastive focus is signaled through nuclear stress (NS). In Spanish, NS is assigned following the C-NSR. When focus is involved, the C-NSR must interact with the FPR. When NS is assigned to the rightmost constituent within the intonational phrase (IP), both rules are satisfactorily fulfilled. For the sentences in (89a) and (89b), this means that the unit in focus will be aligned to the rightmost edge of IP. Now, while the sentence in (89a) is ambiguous in IS terms, the sentence in (89b) is not; in this latter, only the subject mi hermana ‘my sister’ can be assumed as 35 This author further observes that the grammaticality (acceptability) of the VSO order varies among speakers and seems to depend on the lexical content of the sentence. Thus, (ii), where the subject has an experiencer θ-role, is usually perceived as more natural than (i), whose subject bears an agent θ-role. (i) ? Pintó el artista retratos terribles Painted the artist portraits terrible ‘The artist painted terrible portraits’ (ii) Sufrió el paciente dolores terribles Suffered the patient pains terrible ‘The patient suffered terrible pains’ (Zagona 2002: 27) 36 See also Zubizarreta & Vergnaud (2006) for a discussion about the relation between syntax, focus, and nuclear stress, for English and some other languages. 206 bearer of informative focus. This difference in terms of IS is shown in (90) and (91), respectively (the questions in parentheses provide the relevant context). (90) a. (Qué escribió tu hermana? ‘What did your sister write?’) Mi hermana escribió [una carta] F My sister wrote a letter b. (Qué hizo tu hermana? ‘What did your sister do?’) Mi hermana [escribió una carta] F c. (Qué pasó? ‘What happened?’) [Mi hermana escribió una carta] F (91) (Quién escribió una carta? ‘Who wrote a letter?’) Escribió una carta [mi hermana] F The sentence in (91) is unambiguous in IS terms, since only the subject, being in IP-final position, can be felicitous as focused unit. Sentence (91) is also in full compliance with both the C-NSR and the FPR, without resorting to emphatic prominence—the NS has been assigned to the rightmost unit of IP, which is F-marked. These facts may lead to the consideration of the order VOS as a derived (i.e. marked) word order. As Zubizarreta (1998) points out, […] the VOS order has two salient properties: the subject bears unambiguously nuclear stress and the subject is focused. On the basis of this observation, it is reasonable to hypothesize that the underlying structure of VOS is either SVO or VSO, and the motivation for reordering is to put the focused subject in a position to receive NS via the C-NSR […], an output compatible with the FPR. (Zubizarreta 1998: 126). 207 In this same source, it is suggested that VOS is derived from VSO. Additionally, other derived orders for Spanish are presented, namely VPPS, V[OPP]S, and VPPO. In these three cases, focus falls unambiguously on the rightmost constituent of IP. See examples of these orders in (92). (92) a. (Quién habló ayer contra el gobierno? ‘Who spoke yesterday against the government?’) Ayer habló contra el gobierno [el sindicato] F (VPPS) Yesterday spoke against the government the union ‘It was the union that spoke yesterday against the government’ b. (Quién puso la valija sobre la cama? ‘Who put the suitcase on the bed?’) Puso la valija sobre la cama [la camarera del hotel] F (VOPPS) Put the suitcase on the bed the attendant of.the hotel ‘It was the hotel’s attendant who put the suitcase on the bed’ c. (Qué escondió Ana debajo de la cama? ‘What did Ana hide under the bed?’) Ana escondió debajo de la cama [la muñeca] F (VPPO) A. hid under of the bed the doll ‘It was the doll what Ana hid under the bed’ (Zubizarreta 1998: 127 ff) According Zubizarreta (1998), the mechanism that derives the marked orders is a type of movement motivated by prosody. It is called p-movement. I have uncovered a type of scrambling in the grammar of MS [Modern Spanish] with particular properties: it gives rise to structures that are prosodically unambiguous and that are associated with a narrow focus interpretation. On the basis of these observations, I have suggested that this case of scrambling is prosodically motivated. A defocalized constituent is moved immediately above 208 the focused constituent in order to ensure that the focused constituent is in a position to be assigned main prominence by the C-NSR; the output structure is thus compatible with the FPR. More precisely, I am suggesting that p-movement in Spanish plays exactly the same role as metrical invisibility in other languages. […] In Spanish all phonologically specified material is metrically visible for the NSR. Therefore, in cases in which the NSR would assign main prominence to a [- F] constituent (generating an output that contradicts the FPR), this language has recourse to an alternative strategy, namely, [p-]movement. (Zubizarreta 1998: 133). Importantly, feature checking is not the source of p-movement, as it is assumed for other types of syntactic movement. Its motivation stems on prosody. As the quote already indicates, the aim of p-movement is to solve a conflict originated whenever the rules C-NSR and FPR select different units—which by definition are in a sisterhood relation—to assign prominence. Let α and β be those sisters. Let’s also use the initials <ph> as an abbreviation for phonological content, and the diacritic <*> to indicate relative strength on [ph]. Then, a prosodically contradictory structure would be represented as in (93). (93) … [ δ [ α ph*]… [ β ph*] ] , where α and β are metrical sisters. (Zubizarreta 1998: 139) Facing the conflict represented in (93), p-movement is invoked to solve the contradiction via re-ordering of constituents. In its more general formulation, p-movement would be expressed as in (94). (94) Affect the nodes α and β iff these nodes have contradictory prosodic properties, where the notion of prosodic contradiction is to be understood as in [(93)]. (Zubizarreta 1998: 140). Now, what should we understand by the term affect in (94)? For the general purposes of this sub-section, the answer offered is crucial. 209 The term affect can be interpreted in the most general syntactic terms—that is, as changing the c-command relation between α and β. This can be achieved via copying, deletion, or both. Clearly, copying one of the nodes without deleting the phonological content of the initial copy will give rise to a non-convergent derivation that will crash at PF. Deletion without copying is also excluded by the condition on recoverability of deletion. Therefore, both copy and deletion must apply, and copying plus deletion equals movement. (Ibidem). 37 As a consequence of p-movement, the troublesome structure in (93) would be transformed in the non-offending structure (95), where β has undergone upwards movement to adjoin δ. It must be noted that cannot move higher than δ, so this kind of movement is strictly local (by definition, it affects only two sisters). This prevents any risk of overgeneration. (95) … [ δ [ β ph*] [ δ [ α ph*]… [ β ] ] ] (Ibidem) More recently, in Zubizarreta (2013), an alternative approach to marked word order in Spanish is suggested. In this case, an analysis involving reduplication plus ellipsis is explicitly assumed—contra the focus-movement account. Consider the sentence in (96), with the linear sequence VOS, and focus on the subject María. (96) Me regal-ó el libro María Cl.DAT1SG=give -PST3SG the book M. ‘It was María who gave me the book’ (Zubizarreta 2013: 23) The analysis of this sentence proposes a bifurcation at the vP level within the same matrix sentence. The bifurcation is created by adjoining a reduplicative copy of vP to the matrix 37 Final italics are mine. As the reader may have already anticipated, I fully assume the final equation, but I would like to present it—slightly enriched—in the reverse order, namely (p-)movement equals Copy+Merge plus deletion. This latter formula summarizes the CTM approach. 210 vP. The correspondent PF- and LF-structure 38 for (96) are presented as in (97) and (98), respectively. (97) [ TP Me regaló v [ vP [ vP María [v el libro ]] [[ vP María [regal√ el libro]]] (98) [Op k [ TP Me regaló v [Ev-T [ vP [ vP [ DPk e] [v el libro]] [ vP [ DPk María] regal√ el libro]]]] (Zubizarreta 2013: 24) As visible in (97), PF-deletion is complementary. Also, note in (98) that the presuppositional variable created at LF gets bound by a high abstract operator Op. Ev-T stands for internal event time, a time argument provided by vP, which in turn orders the utterance time (Utt-T), the external time argument selected by tense in root clauses. The structures in (97) and (98) are reminiscent of the RSs proposed in sub-section 2 for sólo ‘only’ and other FSUs in Spanish (see, in particular, (13)-(15)). In this analysis, a reduplicative TP/vP followed by deletion at PF and LF is also put forward. The adverb itself has been used as a ‘pivot’ for reduplication, having the low copy of vP located in its Compl and the high copy of vP located in its Spec. Naturally, being this adverb (or any visible equivalent) absent from (97) and (98), that structural choice seems no longer available. An option can be the one adopted in Zubizarreta (2013), namely copying vP, put it on hold, and then merge the newly created copy as an adjunct of the ‘original’ matrix vP. On this issue, an idea presented in Kayne (2000)—briefly commented in the previous chapter— may also be entertained. When presenting the remnant movement analysis of focus, Kayne suggests that there may exist a phonetically empty head, which this author calls “Foc 0 ”. This abstract head is base generated at the left periphery of VP. It functions as an attractor for the unit in focus, which ends up raised to its Spec. Then, Foc 0 moves upwards to adjoin the head W—which is a high empty head postulated in Kayne’s analysis that, in turn, attracts FSUs to its 38 Indeed, (98) shows what Zubizarreta (2013) represents as LF Ass , that is to say, the part of LF in which the AS is encoded. It is assumed that PF is visible to LF Ass since, at LF, the focus phrase “is identified by virtue of its relation to PF, namely the part of the assertion structure that is undeleted at PF” (Zubizarreta 2013: 4). 211 Spec. Finally, VP is attracted by the complex [Foc 0 +W], so that, finally, this complex ends up located immediately adjacent to the unit in focus, exactly the final position of FSUs in cases of Copy+Merge followed by matched deletion. For instance, the remnant movement analysis proposed for (99) has the derivation shown in (100). 39 (99) (What did he look up?) He looked up [a linguistics-term] F (100) a. He Foc 0 looked a linguistics-term up → (attraction to Foc 0 ) b. He a linguistics-term 1 Foc 0 looked t 1 up → (raising of Foc 0 to W) c. He Foc 0 2 +W a linguistics-term 1 t 2 looked t 1 up → (VP-preposing) d. He [looked t 1 up] 3 Foc 0 2 +W a linguistics-term 1 t 2 (Kayne 2000: 249). While I am not assuming the remnant movement analysis of focus in this work, I am willing to consider the idea of a phonetically empty head as a possible realization of a FSU. If this is indeed a possibility, this unit would carry the feature [+foc] and, in principle, it would merge at the edge of a phase (vP or CP, maybe TP), just like other FSUs. 40 Provisionally, I would refer to it with the symbol <Ø>. 41 In the case of example (97), Ø would be merged at the vP level. Once a unit within this domain gets F-marked, Ø would trigger reduplication. Then, complementary deletion would derive the final linear order at PF, and the creation and binding of the focus variable at LF. 42 39 This example has been also quoted in the previous chapter. 40 The question may arise about how to determine the size of the complement of FSU. While I have no definitive answer for that question, I consider that the task can be seen as empirical: we have to build sentences, and then evaluate its acceptability. 41 Here again, I want to differentiate between a phonetically empty FSU from the functional category “Foc” as in “FocP”, a position used in the cartographic tradition like the one put forward by (Rizzi 1997, 2004), among many others. 42 In turn, Ø could be seen as a silent focus operator. 212 (101) FSUP vP i FSU’ Ø [+foc] vP i [[María] me regal-√ el libro] [[María] fcP me regal-√ el libro] (102) FSUP vP i FSU’ Ø [+foc] vP i Op k [[ DPk María] me regal-√ el libro] [[María] fcP me regal-√ el libro] In (101) and (102), the bifurcation at vP level assumed in Zubizarreta (2013) is represented in a way analogous to what has proposed here for only, even, also and their Spanish counterparts, all of them sensitive to focus. And sensitivity to focus is precisely what underlies (at least some cases of) reordering of constituents in Spanish. From my point of view, this is a desirable consequence, since this analysis—namely the RS analysis—structurally unifies apparently diverse focus-related phenomena. 213 Before putting to the test this proposal for other marked word orders, I would like to verify the potential scope of Ø. Recall that we had established that neither only nor its Sp. equivalent sólo can modify F-marked sentences (semantically, they cannot modify propositions), so they should be considered as a vP level FSUs only (see examples (7) and (8) above, repeated here as (103) and (104), respectively). (103) * Sólo [ CP María compró un carro nuevo] F (104) * Only [ CP Mary bought a new car] F In principle, there seem to be no obstacle for Ø to modify an entire F-marked sentence, so it would have a scope similar to that of the modal probablemente ‘probably’ and other high – mente adverbs reviewed in this chapter. If this is on the right track, Ø can, in principle, merge with CP, that is to say, it is a CP-level FSU. 43 (105) a. Probablemente [ CP María compró un carro] F Probably M. bought a car b. Ø [+foc] [ CP María compró un carro] F Now, it is interesting to consider the potential interpretation of (105b). While it is intuitively clear that, in (105a), F-marking on CP indicates that the proposition encoded by the sentence will be semantically modified by the modal (it will be modalized), an equivalent intuition seems harder to get for (105b). Let’s observe the F-marking, and recall that focus is to be understood in opposition to presupposition. In principle, the divide focus/presupposition is less obvious when the whole sentence appears F-marked in comparison to cases when F-marking affects a sub-constituent only. Take, for instance, the sentence in (105) and let’s suppose, for the sake of the example, that the DO un carro ‘a car’ is F-marked. Then, it seems clear that the 43 At the current point of development of the proposal, it is not clear to me whether reduplicative Ø can have just one scope or it can have two; if this second option is correct, Ø might alternatively have either CP-scope or vP-scope. In any case, it seems that vP level is the restricted case, so I will temporarily assume that, in absence of such a restriction, Ø merges at CP level. 214 existential presupposition will have a form similar to There exists x such that María bought x. This presupposition is built using elements taken from the original sentence (María, bought), and normally these elements will survive deletion on the high copy of the RS. In the case of (105), however, the presupposition has the form There exists x such that x happened. No element of this presupposition appears in the original sentence—in any case, it may have appeared in the contextual question. F-marking on the entire sentence means that the sentence as a whole, and not just one of its sub-parts, is providing content such that it is not part of the common ground yet. In this sense, the FSU located in high sentence position will necessarily discharge its sensitivity to focus, and will trigger reduplication, even if the copy of CP created and merged in [Spec,FSUP] will be later deleted. In a similar vein, if reduplicative Ø appears in a high sentence position, according to the proposal defended here, it will trigger reduplication, even if deletion will later target the high copy as a whole. Indeed, a similar situation arises with sentence modal adverbs, as in (105a). The CP has been copied, merged, and then deleted—in more traditional terms, it has been moved. 44 If all this seems unnecessary at first sight, recall that this mechanism is responsible of creating the presuppositional variable at LF-level, and this in turn will ensure the alternative interpretation that is typical of focus. Let’s illustrate this with the LF-structure attributable to (105a). (106) ModP CP i Mod’ Probablemente [+foc] CP i Op k [ CPk María compró un carro] [María compró un carro] fcP 44 In traditional terms, it has been moved covertly. 215 The interpretation one can get from (106) is that, from the relevant set of alternative events that might have been the case, a specific one has been picked, namely María compró un carro ‘M. bought a car’, and this occurrence is seen under the lens of probability. In this case, the respective extensions of Alt-fv (the set of alternative focus-values) and Alt-p (the set of alternative propositions) would coincide. These facts considered, let’s turn back to reduplicative Ø and its eventual role in marked word order cases. Thus, considering Ø as a CP-level FSU, the structures in (107)-(109) would correspond to the sentences (92a)-(92c), respectively. (107) Order VPPS (see (92a)). a. FSUP CP i FSU’ Ø [+foc] CP i [Ayer [el sindicato] habló contra el gob.] [ayer [el sindicato] fcP habló contra el gob.ayer] 45 45 I assume that the temporal adverb ayer ‘yesterday’, originally merged at the end of the low clause, undergoes topicalization, so it moves to the front of the low clause before this clause is copied. 216 b. FSUP CP i FSU’ Ø [+foc] CP i Op k [Ayer [ DPk el sindicato] habló contra el gob.] [ayer [el sindicato] fcP habló contra el gob.] (108) Order VOPPS (see (92b)). a. FSUP CP i FSU’ Ø [+foc] CP i [[La camarera del hotel] puso la valija sobre la cama]] [[la camarera del hotel] fcP puso la valija sobre la cama]] 217 b. FSUP CP i FSU’ Ø [+foc] CP i Op k [[ DPk La camarera del hotel] puso la valija sobre la cama]] [[la camarera del hotel] fcP puso la valija sobre la cama]] (109) Order VPPO (see (92c)). a. FSUP CP i FSU’ Ø [+foc] CP i [Ana escondió [la muñeca] debajo de la cama]] [Ana escondió [la muñeca] fcP debajo de la cama]] 218 b. FSUP CP i FSU’ Ø [+foc] CP i Op k [Ana escondió [ DPk la muñeca] debajo de la cama]] [Ana escondió [la muñeca] fcP debajo de la cama]] Before leaving this sub-section, I will consider the case of low –mente ‘-ly’ manner adverbs in Spanish. I have suggested above that these adverbs originate in sentence-final position, where they are frequently put in focus themselves. In some cases, however, these adverbs can appear in other positions (except for sentence initial), and this fact can also be seen under the lens of ‘transportability’. 46 See the examples for Spanish in (110) with the adverb parcialmente ‘partially’. 47 (110) a. Juan limpió la casa parcialmente J. cleaned the house partially b. Juan limpió parcialmente la casa c. Juan parcialmente limpió la casa d. ? Parcialmente, Juan limpió la casa 46 Indeed, Kayser (1968) coined the term ‘transportability’ when considering in which positions the manner adverb immediately can or cannot appear in a sentence. 47 This adverb is in the same semantic field of completamente ‘completely’, which occupies the lowest position in Cinque’s universal hierarchy of adverbials (Cinque 1999). 219 Considering (110a-c), the diverse positions of parcialmente can be attributed to applications of F-marking to different sub-constituents, as the bracketing indicates in (111). (111) a. Juan limpió la casa [ AdvP parcialmente] F b. Juan limpió parcialmente [ DP la casa] F c. Juan parcialmente [ vP limpió la casa] F Interestingly, unlike high (sentence) adverbs, the different positions for the manner adverb are neither due to focus sensitivity, nor to any feature on the adverb itself. I claim that it is due to reordering of constituents triggered by F-marking, of which the manner adverb itself participates. This reordering of constituents can be seen now under the light of the RS analysis, taking reduplicative Ø as the relevant axis for Copy+Merge. The RSs in (112)-(114) correspond to the sentences in (111a-c). (112) a. FSUP CP i FSU’ Ø [+foc] CP i [Juan limpió la casa [ AdvP parcialmente]] [Juan limpió la casa [ AdvP parcialmente] fcP ] 220 b. FSUP CP i FSU’ Ø [+foc] CP i Op k [Juan limpió la casa [ AdvPk parcialmente]] [Juan limpió la casa [ AdvP parcialmente] fcP ] (113) a. FSUP CP i FSU’ Ø [+foc] CP i [Juan limpió [ DP la casa] parcialmente] [Juan limpió [ DP la casa] fcP parcialmente] 221 b. FSUP CP i FSU’ Ø [+foc] CP i Op k [Juan limpió [ DPk la casa] parcialmente] [Juan limpió [ DP la casa] fcP parcialmente] (114) a. FSUP CP i FSU’ Ø [+foc] CP i [Juan [ vP limpió la casa] parcialmente] [Juan [ vP limpió la casa] fcP parcialmente] 222 a. FSUP CP i FSU’ Ø [+foc] CP i Op k [Juan [ vPk limpió la casa] parcialmente] [Juan [ vP limpió la casa] fcP parcialmente] To illustrate the complementary distribution between reduplicative Ø and other phonetically non- empty FSU, let’s consider (115), whose RSs are represented in (116). (115) Juan limpió la casa sólo parcialmente 48 J. cleaned the house only partially ‘It was only partially that Juan cleaned the house’ 48 In Spanish, sólo ‘only’ is the shortened version of a –mente adverb, namely solamente. Whereas both forms have identical meaning and can be used indistinctly in most contexts, only the shortened version is acceptable before another –mente adverb. 223 (116) a. PolP TP i Pol’ sólo [+foc] TP i [Juan limpió la casa [ AdvP parcialmente]] [Juan limpió la casa [ AdvP parcialmente] fcP ] b. PolP TP i Pol’ sólo [+foc] TP i Op k [Juan limpió la casa [ AdvPk parcialmente]] [Juan limpió la casa [ AdvP parcialmente] fcP ] In (116), the axis for reduplication is sólo, which occupies the FSU position. There is neither syntactic place available for Ø [+foc] , nor need for it. 224 As for the sentence modal probablemente, it is interesting to note that it cannot be placed immediately adjacent to the unit in focus (see (117b)), since this latter is also a –mente adverb, and Spanish bans an immediate sequence of two of these adverbs. The high adverb can still associate with the focused unit at a distance, as in (117a). Another possibility is to use an equivalent adverbial, as in (117c) (these are the cases when adverbials come at hand). (117) a. Probablemente, Juan limpió la casa [parcialmente] fcP Probably J. cleaned the house partially b. # Juan limpió la casa probablemente [parcialmente] fcP J. cleaned the house probably partially c. Juan limpió la casa probablemente [en parte] fcP J. cleaned the house probably in part ‘It was probably partially that Juan cleaned the house’ If we want to also include sólo ‘only’ in the sentence, the same versions are available. Although in (118b) probablemente and parcialmente are not contiguous, they are still close enough to make that option dispreferred. Interestingly, in (118), three –mente adverbs coexist harmonically within the same sentence: (i) probablemente, a CP-level FSU; (ii) sólamente, a vP- level FSU, which is shortened to sólo; and (iii) parcialmente, a low manner adverb that is F- marked in sentence-final position. (118) a. Probablemente, Juan limpió la casa sólo [parcialmente] fcP Probably J. cleaned the house only partially b. # Juan limpió la casa probablemente sólo [parcialmente] fcP J. cleaned the house probably only partially 225 c. Juan limpió la casa probablemente sólo [en parte] fcP J. cleaned the house probably only in part ‘It was probably only partially that Juan cleaned the house’ 5. Summary In order to explain cases of sensitivity to focus for the sentence modal adverb probablemente ‘probably’ in Spanish, a reduplicative structure (RS) has been postulated in the previous chapter. In this chapter, it has been shown that an analysis in terms of RS can also be considered as a way to describe other focus-related phenomena in this language. The RS grammaticalizes the divide presupposition/focus by placing the former in [Spec,FSU] and the latter in [Compl,FSU]. The FSU can act either at the CP-level (as is the case of probablemente ‘probably’) or at TP/vP-level (as is the case of sólo ‘only’). When an element in its domain gets F-marked, the [+foc] feature on the FSU triggers reduplication of the c-commanded unit. Deletion applies then, in the form of chain reduction (ChR), guided by the LCA and constrained by the FDC. As a consequence, the FSU may end up immediately adjacent to the unit in focus—more precisely, to the focus- containing unit. I propose that this adjacency occurs at all times at LF, but mismatches with respect to PF are tolerable. In these latter cases, the FSU associates with focus at a distance. The first unit tested for the RS analysis has been sólo ‘only’. This unit is generated at the node polarity. It triggers reduplication plus ellipsis, that is to say Copy+Merge plus deletion, when an element is F-marked in its domain. Next, the FSUs también ‘also’ and incluso ‘even’ were tested. With differences derived from their specific semantic contents, they seem to behave syntactically like sólo, i.e. as ‘pivot’ points for reduplication, with respect to the RS. Some intriguing facts, like an apparent ‘backwards’ focus association with these FSUs, have been analyzed as alternative uses of the adverbs as proforms. As for the combination of the sentence modal probablemente ‘probably’ with sólo ‘only’ in the same sentence, the RS analysis predicts a fixed order (probablemente > sólo) by which the sentence adverb will always precede the vP- adverb when those adverbs associate with the same F-marked unit. I show that this prediction is borne out. The derivation proceeds in two cycles, one for each FSU, which means that two RSs have been created. Association at a distance is explained as cases of PF/LF deletion mismatch. 226 In the course of these structural considerations, it has been provisionally confirmed that there is just one F-marked unit per sentence. Also, some apparent exceptions to the fixed order probablemente > sólo are explained in terms of a concealed dual structure. Next, I tested the RS for high sentence adverbials from the following classes in Spanish: speech-act adverbs, evaluative adverbs, evidential adverbs, aspectual adverbs, volitional adverbs, agent evaluative adverbs, and domain adverbs. I used a morphological criterion as a way to limit the sample of adverbs, namely the –mente ‘-ly’ ending. Like sentence modals, the adverbs considered here have sentential scope and are supposed to be generated as sisters of CP. From that position, they can associate with the F-marked unit at a distance. Again, like sentence modals, these adverbs can surface immediately adjacent to the unit in focus. It has been shown that this possibility can be satisfactorily explained via the RS analysis, so these adverbs are also assumed to carry the [+foc] feature from the lexicon. Since each class corresponds to a different hierarchical level, it is predictable that they can coexist within the same sentence. It has been shown that this is indeed the case, with the limitations derived from the ban Spanish grammar imposes over a sequence of two or more –mente adverbs. On the other hand, some other –mente adverbs in Spanish are generated as low manner vP-adverbs. Although they can also surface in different positions inside the sentence, it has been suggested that, unlike high adverbs, this is not due to a [+foc] feature in the adverb itself, but it is related to the re-ordering of constituents. In the next section, I extended the RS proposal to some uses of the negative particle no ‘NEG’ in Spanish. Whereas, in its typical use, this particle surfaces immediately preceding the inflected verb, I have shown that it can also surface immediately adjacent to the focused unit. In these cases, an RS analysis is neatly applicable. Importantly, in this latter configuration, the focus value is explicitly negated, so for this clause to appear grammatically, it must be inserted in a broader structure whose second clause provides a valid positive focus value to ‘replace’ the negated one. In most cases, this second clause undergoes ellipsis under identity. Interestingly, the existential presupposition for both clauses is necessarily the same. Finally, the RS approach has been extended to cover cases of marked word order in Spanish. In order to implement such an analysis, I assume the existence of a phonetically empty 227 unit <Ø> that is loaded with the [+foc] feature and is capable of triggering reduplication when a unit in its domain gets F-marked, in the same fashion as the full-fledged FSUs previously reviewed in this work. I have shown that the RS analysis provides accounts for the orders VPPS, VOPPS, and VPPO in Spanish. Additionally, it provides an explanation for the different positions that some low –mente manner adverbs can occupy within a sentence. 228 Chapter 5 Possible extensions of the main proposal II: Copy+Merge in certain focal constructions In this chapter, I seek to extend some of the main insights that informed the previous chapters to the explanation of certain constructions related to the grammatical expression of the divide focus/presupposition. The term construction is understood here, in a broad sense, as certain recurrent patterns of organization for lexical items. From a minimalist conception of language, I assume that these patterns are not primitives of grammar, but instead they exist for extra- grammatical purposes. More specifically, I further assume that the constructions reviewed in this chapter are oriented to the satisfaction of informational demands originated in the common ground management. These constructions accommodate the lexical items in such a way that the information is pronounced in a clear, unambiguous way. In this view, it comes as natural that these constructions have fixed focal positions. While the constructions are grounded in discourse, they are internalized as part of the general linguistic knowledge of an adult speaker. In general terms, the novelty of the proposal entertained in this chapter is related to the operation Copy. I suggest that, in the course of the derivation, this operation can apply in one cycle in such a way that the copy generated can be put on hold and merged in a different cycle. The main inspiration for this idea lies in the proposal known as sideward movement (Nunes 2001, 2004). In section 1, I review a construction known as Focalizing SER ‘to be’, at use in Caribbean Spanish. In section 2, I put forth some considerations related to copular structures and their relation with focus marking. Section 3 deals with the constructions known as pseudo-clefts. I suggest a general derivation for them involving Copy+Merge of non-focalized content. Section 4 is concerned with the constructions known as It-clefts in English. Following the spirit of early analyses, I assume that It-clefts are structurally derived from pseudo-clefts. In section 5, I revisit pseudo-clefts in colloquial Peninsular Spanish. I propose that their particularities are derived from the lack of proforms in that dialect. Finally, section 6 summarizes the chapter. 229 1. Focalizing-SER (‘to be’) in Caribbean Spanish. The RS approach 1.1. The phenomenon In this section, I want to investigate the possibility of expanding the RS view to a dialectal construction of Caribbean Spanish known as Focalizing-SER ‘to be’ (FS). 1 This construction has been studied in Bosque (1999), Camacho (2006), Méndez Vallejo (2009), and Zubizarreta (2013), among others. The empirical observation that makes FS constructions potential candidates for the RS analysis is that, in these constructions, there exists some unit, namely a tensed copula SER ‘to be’, that surfaces immediately adjacent to the constituent in focus. This constituent may be of different categorical types (DP, PP, AdjP, AdvP, etc.) as long as it appears post-verbally (Méndez Vallejo 2009: 3) and is located in final-sentence position, where it is assigned nuclear stress (Camacho 2006: 14). In (1) some examples of FS are shown. Importantly, in these sentences the copular verb (translated as ‘to be’) is not the matrix verb. (1) a. Compró los libros fue [Pedro] fcP Bought the books be.PST.3SG P. ‘It was Pedro who bought the books’ b. El gobierno les cedió terrenos fue [a las multinacionales] fcP The government Cl.granted land be.PST.3SG to the transnationals ‘It was transnational corporations that the government granted land to’ c. Los bomberos rescataron fue [a la abuela] fcP The firefighters rescued be.PST.3SG A the grandmother ‘It was the grandmother who the firefighters rescued’ d. Mi hermano est-aba 2 era [triste] fcP 1 Also know as bare copula construction (Zubizarreta 2013). 2 The conjugated form estaba corresponds to the verb ESTAR, which is also translated into English as ‘to be’. The difference SER/ESTAR in Spanish has been traditionally regarded as the lexicalization of the difference between a copular unit introducing individual-level predicates (SER) and a copular unit introducing stage-level predicates (ESTAR). According to this approach, individual-level predicates are considered relatively permanent or essential (María es rubia ‘M. is-SER blonde’), as opposed to stage-level predicates that are taken as temporary or accidental 230 My brother be-PST.IPFV.3SG be.PST.IPFV.3SG sad ‘It was sad that my brother was’ e. Llegamos fue [ayer] fcP We.arrived be.PST.3SG yesterday ‘It was yesterday that we arrived’ f. El presidente de esa época tuvo una fue [parálisis] fcP The president of that time had a be.PST.3SG paralysis ‘It was a paralysis what the president at that time had’ 3 (Camacho 2006: 14) The role of FS as focus markers only (as opposed to main verbs) can be better appreciated by comparing the sentences in (1) with their equivalent sentences in general Spanish, shown in (2). (2) a. Compró los libros [Pedro] fcP Bought the books P. ‘It was Pedro who bought the books’ b. El gobierno les cedió terrenos [a las multinacionales] fcP The government Cl.granted land to the transnationals ‘It was transnational corporations that the government granted land to’ (María está cansada ‘M. is-ESTAR tired’). While this difference is not relevant for our purposes, it must be stated that FS (as its name indicates) is realized with SER only, not with ESTAR—at least to the best of my knowledge. For a critical view about conceiving the difference SER/ESTAR in terms of individual level/stage level predicates, see Maienborn (2005), from which I have borrowed the examples in this fn. 3 Note that in (1f), FS surfaces between D and NP inside the DP-DO, which is a position inaccessible for all other FSUs reviewed so far in this work (El presidente tuvo una (*sólo) parálisis, * El presidente tuvo una (*probablemente) parálisis). Unfortunately, I have found no equivalent examples other sources about FS that I had reviewed so far. I have no explanation for this striking behavior of FS, so—this disclaimer being made—I just will leave this case aside in what follows. 231 c. Los bomberos rescataron [a la abuela] fcP The firefighters rescued A the grandmother ‘It was the grandmother who the firefighters rescued’ d. Mi hermano estaba [triste] fcP My brother was sad ‘It was sad that my brother was’ e. Llegamos [ayer] fcP We.arrived yesterday ‘It was yesterday that we arrived’ f. El presidente de esa época tuvo [una parálisis] fcP The president of that time had a paralysis ‘It was a paralysis what the president at that time had’ According to Bosque (1999: 208), expressions with FS are appropriate in contrastive contexts, so for instance (1a) would be infelicitous as the initial intervention in a conversation; that sentence would be fine though as a correction of a previous wrong statement like Juan compró los libros ‘J. bought the books’. 4 This means that FS introduces contrastive focus. In turn, Méndez Vallejo (2009) recognizes the contrastive use of FS as frequent, but indicates that it can also be used to introduce non-contrastive focus, as in the example (3) where the speaker B expands the common ground, instead of correcting any previous statement of speaker A. (3) a. Speaker A: Y ahora, tienes mascotas? And now have.PRS.2SG pets ‘And now, do you have pets? 4 “[…] el uso más natural de [Llegué fue ayer] no sería el comienzo de una conversación sino más bien una situación en la que el hablante replicara a alguna afirmación previa que situara su llegada en un momento distinto del día anterior, por ejemplo, Cuando llegaste esta mañana…” (Bosque 1999: 208). 232 b. Speaker B: Pues, cuando llegué acá, decidí adoptar un conejo. Well when I.arrived here I.decided to.adopt a rabbit ‘Well, when I arrived here, I decided to adopt a rabbit’ El pobre se murió y ahora tengo es [dos gatos] F The poor Cl.=died and now I.have be.PRS.3SG two cats ‘The poor one died and now it is two cats what I have’ (Méndez Vallejo 2009: 19) On the other hand, FS cannot modify preverbal subjects: compare (1a) with (4). Moreover, FS cannot modify an entire focused sentence either, as seen in (5). As Camacho (2006: 14) puts it, “In general, [FS] cannot modify any category higher than vP”. 5 (4) * Fue [Pedro] fcP compró los libros be.PST.3SG P. bought the books ‘It was Pedro who bought the books’ (5) (Qué pasó? ‘What happened?’) * Fue [hubo un incendio] fcP be.PST.3SG there was a fire ‘What happened was there was a fire’ (Camacho, p.c.) 1.2. The RS analysis of FS From the RS analysis, these observations can be interpreted as follows: the FS copula is a tensed FSU that corresponds to the vP-phase level. As such, it is equipped from the lexicon with [+foc]. This feature will trigger reduplication of vP, provided that a unit within its domain is F- 5 This is related to the fact that there is no tense in the post-copular part. 233 marked (which will always be the case, given the intrinsically focal nature of these constructions). In this sense, the main novelty of FS with respect to other FSUs reviewed so far is its tensed nature. With respect to vP reduplication in FS constructions, a bi-clausal structure plus ellipsis is explicitly proposed in Zubizarreta (2013). Structurally, it is suggested there that the constituent formed by [copula+vP] is adjoined to the matrix vP. With respect to tense, it is considered that the unit [copula+vP] is a reduced clause, so the tense in the copula is defective (underspecified). As a consequence, the embedded tense gets valued by the matrix T, via Agree. 6 Here it is argued that [the copular part of the structure] is not a full-fledged CP; it is a reduced clause and it is this reduction that triggers its demise from having the status of a main (semantically autonomous) predicate to an adjoined one. As we shall see later, the temporal morphology on the copula is not purely morphological, it is semantically active. Yet, its semantic specification is crucially determined by a minimally c-commanding T. (Zubizarreta 2013: 9). On this line of analysis, which further assumes that there are no PF/LF mismatches with respect to deletion, the FS sentence in (6) would have the PF-/LF-structures shown in (7a,b), respectively. (6) Clara le traj -o a Armando fue [galletas] fcP C. Cl=bring-PRT.3SG to A. be.PST.3SG cookies ‘It was cookies what Clara brought for Armando’ (Méndez Vallejo 2009: 119) 6 As for general agreement on the FS copula, Méndez Vallejo writes, “[…] my observations regarding the morphology of FS ser allow me to suggest that it establishes Tense and Aspect agreement with the verb in T (e.g. a matrix verb or an auxiliary verb), on the one hand, and Person and/or Number agreement with the focused constituent, on the other.” (Méndez Vallejo 2009: 3-4). 234 (7) a. PF-structure [ TP Clara j [ T’ le trajo v [[ vP e j v [galletas V a Armando]]]]] [ CopP fue [ vP e j traj√ [galletas V a Armando]]] b. LF-structure Op k [ TP Clara j [ T’ le trajo v [[ vP e j v [[ DPk galletas] V k a Armando]]]]] [ CopP fue [ vP e j traj√ [galletas V k a Armando]]] (Zubizarreta 2013: 9-10) The LF-structure in (7b) shows that the presuppositional variable, which is bound by a high operator Op, is created via LF-deletion of the DP-DO on the high vP. In turn, the DP-DO galletas ‘cookies’ in the low vP specifies the correspondent focus value. These two assumptions are fully compatible with the RS analysis. As for the copula, although it is tensed, it can be considered neither as a full verb nor as an auxiliary. Its role is that of a focus marker. To some extent, this should not come as surprise, considering cross-linguistic evidence pointing towards the grammaticalization of copulas as focus markers (and vice-versa). In a study about copulas in the Niger-Congo language Wolof, Martinović (2013) points out, It is not unusual for copular sentences to contain particles that otherwise act as focus markers. In fact, the grammaticalization of a copula into a focus marker is widely attested in the grammaticalization literature […], and is especially common in African languages. Grammaticalization in the opposite direction, from focus marker to copula, has also been argued for, in some Chadic languages […], and Swahili […]. Therefore, there seems to be an inherent connection between the meanings conveyed with copular sentences in general and focusing. (Martinović 2013: 12). 235 In turn, Green (2004), quoting Pustet (2003), mentions that while in languages like English (and—I add—also in Spanish) the copula is realized as a verb, in other languages (indeed, in the majority of Pustet’s survey of over 150 languages), that function is carried out either by pronominals or by particles (adpositions). Green’s own study on Hausa, a Chadic language, concludes that, in this language, the copula nē/cē is indeed a focus marker (Green 2004: 16 ff). Grammaticalization is a universal linguistic tendency. On these grounds, it doesn’t seem particularly wrongheaded to speculate that the copula in Caribbean Spanish, in spite of being still tensed, is going through the direction pointed out by Green and by Martinović for African languages, namely from a full copula to a mere focus marker. 7 Méndez Vallejo summarizes this change of status locating the FS copula in the context of its current role as an IS marker. […] given that FS SER has lost its original characteristics as a copula or an auxiliary verb, and that it is used to introduce the focused element and to concatenate it with previous information, I argue that FS ser should be viewed as a link between focus and its background. (Méndez Vallejo 2009: 4). 8 This latter is precisely the role that corresponds to the copula in a RS analysis of FS in Caribbean Spanish. Let’s illustrate the proposal with (6), whose PF-/LF-structures are in (8a,b). 9 7 In fact, an important percentage of Caribbean America’s current population is made up by descendants of African immigrants that traveled to America—sadly, in terrible conditions—for centuries, by the thousands. Could this dialectal feature (=FS) be an effect of linguistic substrate? I have no competence to answer that question, so I pass it on to the experts in language contact and/or historical linguistics for their kindly consideration. 8 Italics are mine. As used in this quote, background can be understood as presupposition. 9 Having assumed that the embedded T is valued by the matrix T via Agree, I abstract away from that issue from now on. For more details about tense agreement, see Zubizarreta (2013) and Méndez Vallejo (2009). 236 (8) a. CopP vP i cop’ fue [+foc] vP i [Clara le traj-√ [ DP galletas] a Armando] [Clara le traj-√ [ DP galletas] fcP a Armando] b. CopP vP i cop’ fue [+foc] vP i Op k [Clara le traj-√ [ DPk galletas] a Armando] [Clara le traj-√ [ DP galletas] fcP a Armando] Next, I consider (1a), a sentence combining two phenomena reviewed in this work: FS bare copula and marked word order (in this case, the order is VOS). The correspondent structures appear in (9a,b). 237 (9) a. CopP vP i cop’ fue [+foc] vP i [[ DP Pedro] compr-√ los libros] [[ DP Pedro] fcP compr-√ los libros] b. CopP vP i cop’ fue [+foc] vP i Op k [[ DPk Pedro] compr-√ los libros] [[ DP Pedro] fcP compr-√ los libros] Finally, I present in (10) the RS analysis of (1d), whose matrix verb is a conjugated form of ESTAR realized with imperfective aspect and whose focused unit is an AdjP. 10 10 Naturally, all other remaining examples of FS in (119) will have an RS analysis similar to the ones already seen for other FSUs in this work. 238 (10) a. CopP vP i cop’ era [+foc] vP i [Mi hermano est-√ [ AdjP triste]] [mi hermano est-√ [ AdjP triste] fcP ] b. CopP vP i cop’ era [+foc] vP i Op k [Mi hermano est-√ [ AdjPk triste]] [mi hermano est-√ [ AdjP triste] fcP ] Remember that (8)-(10) presents the RS for vP. From these structures, the subject and the surviving verbal root will move upwards, so finally the matrix verb for the whole sentence is not COP, but the complex [V+T] (trajo ‘brought’, compró ‘bought’, estaba ‘was’, respectively). 239 2. Considerations on copular structures in Spanish. The ‘Hausa-Spanish connection’ In this section, I offer some general considerations about the copular verb SER ‘to be’ in Spanish. To start with, in English, just like in Spanish, there exists a copular verb that connects two DPs, as in the sequence [DP 1 COP DP 2 ]. This copula can be used in at least three different ways, illustrated in (11). (11) a. [ DP1 Clark Kent] is [ DP2 an idiot] (Predicational) b. [ DP1 The journalist] is [ DP2 Clark Kent] (Specificational) c. [ DP1 Clark Kent] is [ DP2 Superman] (Equative) (Adapted from Martinović 2013: 1). In (11a), the copular sentence is said to be predicational. Basically, in this structure, DP 1 denotes an individual via a proper name, in this case Clark Kent, and DP 2 predicates a property of DP 1 , in this case, being an idiot. The copula does not indicate equality, but belonging: the individual denoted by DP 1 belongs to the set of individuals that satisfy the property denoted by DP 2 (in this case, Clark Kent belongs to the set of idiots). In (11b), the specificational sentence, DP 1 denotes an individual via a definite description whose interpretation assumes (presupposes) that there is a unique contextually salient journalist; in turn, DP 2 denotes an individual via a proper name, and by doing so it specifies the identity of the individual referred by DP 1 via the definite description. The copula indicates equality, since two expressions (a definite description and a proper name) are compared to establish that they denote the same individual. Finally, in (11c), the two DPs are obviously referential, and it is established that two proper names, refer to the same individual, hence the name equative sentence. The copula clearly indicates equality, and can be understood as similar to the mathematical operator “=”. Martinović (2013), quoting Mikkelsen (2005), compares predicational and specificational copular sentences in terms of IS. The conclusion is that, unlike predicational sentences, 240 specificational sentences have a fixed informational structure: topic for the pre-copular DP, and focus for the post-copular DP. This is illustrated as follows. (12) Who is the winner? a. [ DP1 The winner] is [ DP2 John] (Specificational) b. [ DP1 JOHN] is [ DP2 the winner] (Predicational) What is John? c. # [ DP1 the WINNER] is [ DP2 John] (Specificational) d. [ DP1 John] is [ DP2 the winner] (Predicational) (Martinović 2013: 2). In this view, the ill-formedness of (12c) would be due to a violation of the fixed structure of specificational copular sentences in English. As for Spanish, I consider that a similar claim can be made for specificational sentences, namely that specificational copular sentences have a fixed focal structure, where the presupposed part is necessarily to the left of the copula and the unit in focus appears to its right. The equivalent to (12) is in (13). (13) Quién es el ganador? ‘Who is the winner?’ a. [ DP1 El ganador] is [ DP2 Juan] (Specificational) b. [ DP1 JUAN] es [ DP2 el ganador] (Predicational) Qué es Juan? ‘What is John?’ c. # [ DP1 EL GANADOR] es [ DP2 Juan] (Specificational) d. [ DP1 Juan] es [ DP2 el ganador] (Predicational) The general structure for specificational copular sentences, common for English and Spanish, can be summarized as follows. 241 (14) [ DP1 X ] [ COP [ DP2 Y ] F ] (Specificational) Now, with respect to equatives in Spanish, I claim that there is also a fixed focal structure for informative focus, namely this focus is always on the post-copular DP. Let’s consider (11c), translated into Spanish as (15). (15) [ DP1 Clark Kent] es [ DP2 Supermán] F Apparently, (15) simply establishes a relation of equality between two proper names: there is one individual x out there that can be alternatively called by two proper names. As a consequence, all the features associated with one of the names that were part of the common ground are now attributed to the individual denotated with the other name. In this sense, (15) would be somehow similar to linguistics expressions that flesh out mathematical equations (hence the term equatives). See an example of this mathematical use in (16). (16) a. 3 * 7 = 21 b. [ DP1 Three by seven] {equals/, / is} [ DP2 twenty one] Crucially, the equation in (16) can be reversed with no change with respect to its truth conditions. (17) a. 21 = 3 * 7 b. [ DP1 Twenty one] {equals/, / is} [ DP2 three by seven] By the same token, the truth conditions for (15) will be not affected if the equative is reversed. (18) [ DP1 Supermán] es [ DP2 Clark Kent] F 242 Now, although the truth conditions may not have changed from (15) to (18), its internal IS have changed (as indicated by the position of the subscript F). 11 This can be seen when considering the relevant contexts. Let’s start with (18), Supermán es Clark Kent. This sentence can be felicitous in a context where the relation between the proper name Supermán and the individual x is already part of the common ground. What (18) adds to the common ground is the fact that x can also be named with a different proper name, with the consequent transfer of associated attributes (informally, ‘the individual we use to know under the name ‘Superman’ from now on can also be called ‘Clark Kent’, because both names refer to the same individual’). Imagine that the director of the Diario El Planeta ‘Daily Planet’ has finally found the truth that (18) expresses. At least in Spanish, there is just one way of reveling the news with an equative— informative focus, literally—, namely (18) Supermán es Clark Kent. 12 On the other hand, despite the prosodic prominence applied on the proper name Clark Kent, sentence (19) is not felicitous, since that kind of stress corresponds in Spanish to contrastive (corrective or emphatic) focus only. (19) # [ DP1 CLARK KENT] F es [ DP2 Supermán] To be felicitous, (19) has to be used contrastively. Suppose that speaker A has heard about the breaking news, but for some reason he/she got the wrong version. Then, speaker B intervenes. (20) a. Speaker A: ¿Oíste la noticia? ¡Supermán es Bruno Díaz! ‘Did you hear the news? Superman is Bruce Wayne!’ 11 In fact, although both expressions state the same mathematical identity, the reversing from (16) to (17) supposes a change in focus, which is related to the contexts in which each linguistic expression may be used. Sentence (16) may be used in multiplication, as for instance in multiplication tables, whereas (17) is expected in factorization. To illustrate this latter operation, let’s suppose someone is trying to solve the algebraic equation 3x = 21. The first step in the solution may be to remember that twenty-one equals three-by-seven (= sentence (17b)), so this identity may be applied to the second side of the equation, which then will be re-written as 3x = 3 * 7 (crucially, we ‘started’ with 21 and ‘ended up’ with 3 * 7). The next and final step may be to eliminate 3 as a factor in both sides of the equation, so that the solution, x = 7, is found. 12 Presumably, since Clark Kent is not a public figure in Metrópolis, his proper name is not informative enough, so something else will need to be added: Supermán es Clark Kent, un periodista de este diario ‘Superman is Clark Kent, a journalist from this newspaper’, etc. 243 b. Speaker B: No, CLARK KENT es Supermán. ‘No, Clark Kent is Superman’ This distinction is formulated in Zubizarreta (1999) for Spanish. Sentences with main prominence on the preverbal subject can only have a contrastive (or emphatic) interpretation. In other words, they are appropriate in a situation in which the presupposition is explicitly negated (Zubizarreta 1999: 76). On the other hand, the context for (15), Clark Kent es Supermán, without prosodic prominence on the subject, is different. This sentence assumes that the relation between the proper name Clark Kent and the salient (known) individual x out there is already part of the common ground. For instance, this sentence would be felicitously uttered among the staff at the editorial office of Diario El Planeta, where Clark Kent routinely works. The focus expresses that a new relation is added to the common ground, namely the individual x, that used to respond to the name Clark Kent, from now on can also be named with the proper name Supermán, with all the associated consequences for the correspondent denotations. 13 On the basis of this, the general structure for equative copular sentences in Spanish would be presented in (21). (21) [ DP1 X ] [ COP [ DP2 Y ] F ] (Equative) Note that (21) is the same general structure postulated in (14) for specificational copular sentences. In fact, the only difference between specificational sentences and equative sentences would be the content of their respective DPs: for specificationals, DP1is a definite description and DP2 is a proper name; for equatives, either both DPs are proper names or both DPs are 13 In this case, being Supermán a public figure, thus already part of the common ground, it may not be necessary to add more information to the proper name. 244 definite descriptions. This being so, predicational sentences could be detached from specificational and equatives in terms of IS. At least for Spanish, predicational sentences have no fixed focus structure, while specificational and equatives have a fixed focal structure, in which the focus is necessarily post-copular, and the post-copular material is necessarily in focus. This divide is related to the semantic content of the copula. Whereas in predicational sentences the copula is semantically empty 14 , in specificational and equative sentences it is not. In terms of semantic types, the copula in predicational sentences and the copula in specificational-equative sentences are different objects, since they correspond to different types. Let’s illustrate this claim for the sentences in (11a-c), translated into Spanish in (22a-c). The correspondent analysis in terms of types is presented in (23)-(25), respectively. (22) a. [ DP1 Clark Kent] es [ DP2 un idiota] (Predicational) b. [ DP1 El periodista] es [ DP2 Clark Kent] (Specificational) c. [ DP1 Clark Kent] es [ DP2 Supermán] (Equative) (23) S <t> DP1 <e> es’ es DP2 <e, t> Clark Kent <e> un idiota <e, t> 14 Abstracting away from TAM (tense, aspect, mood) features, which are not part the semantic contribution of the copula itself. 245 In (23), corresponding to (22a), the DP2 un idiota ‘an idiot’ is a predicate of type <e, t>, which means that it takes an individual of type <e> (provided by DP1) to yield a truth value <t>. This value appears directly in the top node S. As it is evident, the copula has no participation in the predication. (24) S <t> DP1 <e> cop’ <e, t> El periodista es DP2 <e> <<e, t>, e> <e, t> <e, <e, t>> Clark Kent <e> In (24), correspondent to (22b), both DP1 El periodista ‘the journalist’ and DP2 Clark Kent are of type <e>. If the copula were empty, it would be impossible to arrive to type <t> in S. This indicates that the copula has to be of type <e, <e, t>>, which means that it takes an individual of type <e> (provided by DP2) and yields a predicate of type <e, t>. In turn, this predicate (in cop’) takes an individual of type <e> (provided by DP1) to yield a truth value <t>, which now can appear in S. 246 (25) S <t> DP1 <e> cop’ <e, t> es DP2 <e> Clark Kent <e, <e, t>> <e> Supermán <e> Again in (25), both DPs are of type <e>. We can assume that the copula is of type <e, <e, t>>, which allows it to take an individual of type <e> (provided by DP2) and yield a predicate <e, t> in cop’. This predicate in turn takes an individual of type <e> (provided by DP1) to yield a truth-value <t>, which now appears in S. Noteworthy, the analysis on terms of types also groups together specificational and equative copular sentences (a copula of type <e, <e, t>> for both) as opposed to predicational sentences (an empty copula). I would like to suggest that the copula in the first group always introduces an ‘identity revelation’, so provisionally I will call this copula the identity/identificational copula (id-COP), as opposed to the empty copula (e-COP) used in predicatives. The pre-copular DP, be it either a definite description or a proper name, is necessarily part of the common ground. The post-copular DP reveals a previously unknown identity, something that was not part of the common ground, so it will necessarily be in focus. If this idea is on the right track, we can conclude, at least provisionally, that, in these kinds of copular sentences, the divide focus/presupposition has been grammaticalized. Copular constructions with a copula of type <e, <e, t>> (= id-COP) will always introduce a focused 247 element (they will always ‘reveal an identity’). Thus the general structure for identificational copular construction is shown in (26). (26) [ DP1 X ] [ id-COP [ DP2 Y ] F ] (Identificational) An interesting corollary of the fixed-focus structure is that neither specificational sentences nor equative sentences can be felicitous if used as wide focus (i.e. the whole sentence cannot be in focus). (27) Qué ocurre? ‘What happens?’ a. # [Supermán es Clark Kent] F S. is C.K. b. # [El periodista es Clark Kent] F The journalist is C.K. c. ? [La recepcionista del hotel es la hija del dueño] F The receptionist of.the hotel is the daughter of.the owner ‘The receptionist at the hotel is the owner’s daughter’ Sentences (27a,b) cannot be a felicitous answer to the question posed, unless something on the line of Someone has just said that S or It has been revealed that S is implicitly understood—in which case the event of the revelation itself is in focus, not the content of the revelation. Sentence (27c) is interesting in the sense that there is a definite description at both sides of the copula. To my perception, this fact ameliorates its appropriateness for the question posed, but even so it is not completely adequate. 15 Compare the sentences in (27) with the sentences in (28). In this latter, the ‘right questions’ for the correspondent answers have been formulated. 15 I leave this issue open for later research. 248 (28) a. Quién es Supermán? ‘Who is Superman?’ Supermán es [Clark Kent] F S. is C.K. b. Quién es el periodista? ‘Who is the journalist?’ El periodista es [Clark Kent] F The journalist is C.K. c. Quién es la recepcionista del hotel? ‘Who is the receptionist at the hotel?’ La recepcionista del hotel es [la hija del dueño] F The receptionist of.the hotel is the daughter of.the owner ‘The receptionist at the hotel is the owner’s daughter’ These observations confirm the strong link between focus structure and copular structures highlighted by Green (2004) and Martinović (2013) for African languages, as we saw in the previous sub-section. Indeed, if specificational and equative copular sentences in Spanish have a fixed focal structure, then the copula in those constructions would not only be a verb, but also, simultaneously, a true focus marker, just like the copula in Hausa reported by Green (2004). 16 I find this ‘Hausa-Spanish connection’ particularly interesting for at least two reasons. On the one hand, this might be situated at the origins of the (now degraded) copula in Caribbean Spanish. On the other hand, it shows that general Spanish has at least one kind of copula (id- COP) that acts like a focus marker, and this unit is precisely at the heart of two constructions (specificationals, equatives) that grammaticalize the divide focus/presupposition. 16 Indeed, specificational copulas are considered by definition as focus markers (or focus operators). 249 3. Revisiting pseudo-clefts 17 In this section, I present a structural proposal for a specific kind of constructions in English that have come to be known as pseudo-clefts (PsCls). The analysis proposed here takes inspiration from some ideas formulated within the context of the CTM (Nunes 2001, 2004) namely sideward movement. I project the analysis for English to the equivalent of It-clefts in Spanish. 3.1. PsCls: a CTM-inspired proposal Schlenker (2003: 158) invites the reader to consider the case of the PsCl (29) in terms of Principle A. (29) [What John i likes ___ ] is [himself i ] As the indexation shows, a referential dependency between the anaphor himself and the R-expression John is plainly grammatical. Now, according to the basic structure in (29), John does not c-command the anaphor, as demanded by Principle A, so this grammaticality is unexpected. In turn, den Dikken et al. (2000: 42) note that the pronoun his can be bound by nobody in (30), and in (31), the pre-copular (did)n’t licenses the NPI any, in both cases in spite of the apparent lack of c-command. 17 The literature about clefts and pseudoclefts is overwhelmingly abundant, and always growing. As it will become immediately evident, Schlenker (2003), Den Dikken et al. (2000), and Zubizarreta (2013) provide the main starting points for this section. As for other sources, I just give a—hopefully—representative sample here. Reeve (2012a) is a comprehensive study about clefts in general, with emphasis in English and Slavonic languages. Den Dikken (2005) deepens into the relation of clefts and topics, while Frascarelli (2010) suggests the idea that focused units in clefts are predicates of small clauses. In turn, Reeve (2012b) criticizes the idea of focus as a functional head (FP) in clefts. Patten (2012) offers an analysis of It-clefts in English from the point of view of constructional grammar, along with a diachronic review of these constructions. In turn, Percus (1997) proposes an analysis of clefts based on the extraposition of a relative and the spellout as it of a DP containing a definite Det and a CP trace. For cross- linguistic surveys, see Iatridou & Varlokosta (1998), Lambrecht (2001), and Miller (1996). As for language-specific studies, see Reeve (2011) for English; Moreno Cabrera (1999) and Guitart (2013) for Spanish; Boxus (2006) for French, Portuguese, and English; Paul (2001) and Potsdam (2006) for Malagasy, among many others. Collins (2006) deals with the prosody and pragmatics of clefts, while Frascarelli & Ramaglia (2009) work on their syntax, semantics, and also prosody. Hupet & Tilman (1986) investigate the comprehension of clefts from the psycholinguistic perspective, while Weinert & Miller (1996) analyze spontaneous uses of clefts in conversational English. See also Shell (1972) and, especially, Akmajian (1970) for early—but insightful—approaches to the structural relations between clefts and pseudo-clefts. 250 (30) [What nobody i bought ___ ] was [a picture of his i house] (31) [What Mary didn’t buy ___ ] was [any wine] The PsCls in (29)-(31) illustrate what has been called the connectivity problem, which can be stated as the question of how come these PsCls seem to behave as their connected counterparts, represented in (32a-c) respectively, in spite of their lack of visible c-command. (32) a. John i likes himself i b. Nobody i bought a picture of his i house c. Mary didn’t buy any wine In order to solve the connectivity problem, both den Dikken et al. (2000) and Schlenker (2003), propose the general idea that there is a concealed bi-clausal structure in PsCls. The post- copular expression is indeed part of a clause that has been phonetically reduced via ellipsis under identity. These bi-clausal structures guarantee that there’s no c-command violations involved. The structures proposed for (29)-(31) are shown in (33a-c), respectively. (33) a. [What John i likes ___ ] is [John i likes himself i ] b. [What nobody i bought ___ ] was [nobody i bought a picture of his i house] c. [What Mary didn’t buy ___ ] was [Mary didn’t buy any wine] Schlenker (2003: 159) considers that the first clause is a “question in disguise”, for which the second clause is an appropriate answer. In turn, den Dikken et al. (2000) assume a topical structure. The following phrase marker is proposed in den Dikken et al. (2000: 62) for (33c). 251 (34) TopP Spec Top’ Top 0 IP What Mary didn’t buy is/was she didn’t buy any wine Furthermore, Den Dikken et al. (2000: 43) notice that, in some PsCls, the post-copular IP is fully pronounced. See the example in (35). 18 (35) [What John did ___ ] was [ IP he bought some wine] The PsCl in (35) illustrate the fact that the post-copular unit can be a full sentence. Precisely, the basic idea of the bi-clausal proposal is that the post-copular unit is always clausal. The sub-clausal appearance of this unit in other cases is due to ellipsis. Ellipsis applies on the post-copular unit whenever identity is met. In the case of (35), this condition is not met (indeed, it cannot be met), thus ellipsis cannot proceed. In (36), by contrast, ellipsis is licensed. 19 Thus the relevant comparison is (35) vs. (36). (36) [What John bought ___ ] was [ IP he bought some wine] 18 The authors indicate that these constructions are more frequent in oral/informal discourse. 19 Den Dikken et al. (2000) explain that ellipsis is licensed by “sameness of content” and further state that this notion is “far from trivial” (p. 47, fn. 7). However, the authors declare that they abstract away from that issue, along with the switch from the R-expression John to the pronoun he in (35) and (36). 252 In the case of Spanish, PsCls like (35), with a full post-copular sentence, also exist. In (37) I present an example taken from a forum online. 20 (37) Lo que hice fue agarré un tubo de aluminio 21 Det.N.3SG that I.did was I.took a tube of aluminum ‘What I did was I took an aluminum tube’ In cases similar to (37), involving the verb HACER ‘to do’ in the pre-copula, the post- copular unit can eventually be a full-fledged CP. This fact is revealing, since the verb HACER in Spanish, with the meaning ‘to do’, does not select neither a TP nor a CP as its complement. 22 The following examples are also taken from Internet forums. (38) Lo que hice fue que hablé con mi novio 23 Det.N.3SG that I.did was that I.spoke with my boyfriend ‘What I did was (that) I talked to my boyfriend’ (39) Lo que hago es que le llamo a las 11:30 o 12 de la noche 24 Det.N.3SG that I.do is that Cl=I.call at the 11:30 or 12 of the night ‘What I do is (that) I call him at night by 11:30 or 12’ (40) Lo que hicimos fue que cerramos Broadway al tráfico 25 Det.N.3SG that we.did was that we.closed B. to.the traffic ‘What we did was (that) we closed Broadway St. to the traffic’ 20 Also in Spanish, this kind of sentences seems to be more frequent in oral/informal contexts. Participants in forums online usually write quite informally, not too far from oral discourse. 21 http://pistolerosdelmar.com/archive/index.php/t-1891.html 22 See Zubizarreta (2013) for a comment on this fact, presented as evidence that, “in the standard pseudo-cleft, it is the semantic type (not the syntactic type) of the presuppositional variable that must match that of the focus phrase” (Zubizarreta 2013: 4). 23 https://es.answers.yahoo.com/question/index?qid=20120522152357AAJlI3H 24 https://ar.answers.yahoo.com/question/index?qid=20100317164948AAztswN 25 http://comunicaciones.univalle.edu.co/InformesPrensa/2013/octubre/OC-378-2013.html 253 On the other hand, a full CP is regularly expected in Spanish in post-copular position whenever the pre-copular clause refers to an event or occurrence as a whole. These forms are frequently used in a shortened version, starting with the id-COP, as explanations or answers to the question Why X? (41) (Lo que pasa ) es que no tengo dinero 26 (Det.N.3SG that happens) is that NEG I.have money ‘What happens is (that) I have no money’ (42) Lo que sucede es que tenemos dificultades 27 Det.N.3SG that happens is that we.have difficulties ‘What happens is (that) we have problems’ (43) Lo que ocurrió fue que estábamos en un proceso 28 Det.N.3SG that happened was that we.were in a process ‘What happened was (that) we were in a process’ Importantly, the identity that licenses ellipsis is presented as simply given in both Schlenker (2003) and den Dikken et al. (2000) accounts. No computational story is offered to derive PsCls. The motivation for ellipsis is not clear either—identity/sameness is presented as a condition for ellipsis, not as a reason for it. In this work, I start by plainly assuming the bi-clausal proposal for PsCls. Additionally, I will offer a derivational story, and propose some possible motivation for the assumed ellipsis/deletion in minimalist terms. 26 The shortened version Es que no tengo dinero could be used in colloquial Spanish, for instance, as an answer to the question Por qué no compras vino? ‘Why don’t you buy wine?’ In this case, id-COP indentifies the reason, and at the same time, the construction conveys a justification/apology tone, which is absent in the direct answer Porque no tengo dinero ‘Because I have no money’. For a study of these constructions in Spanish from the point of view of pragmatics, see Reig Alamillo (2011). 27 http://www.radionacional.com.ar/?p=9371 28 http://es.wikipedia.org/wiki/Genocidio_de_Ruanda 254 To start with, let’s consider the sentence (36), translated (somehow freely) into Spanish as (44). (44) Lo que Juan compró fue un vino Det.N.3SG that J. bought was a (bottle of) wine ‘What Juan bought was a bottle of wine’ According to the bi-clausal proposal I assume here, the structure for (44) is presented in (45). (45) [ DP Lo [ CP que i C 0 [ TP Juan compró que i ]]] fue [ CP C 0 [ TP Juan compró un vino]] Some observations are in order. With respect to the pre-copular clause, first, the wh- word que ‘that’ is generated as the DO of the verb compró ‘bought’, and then it moves to [Spec,CP]. This wh-movement is represented as Copy+Merge (note the superscripts) plus deletion on the lower copy. Second, the wh-word is [–interrogative], thus the CP que Juan compró ‘that J. bought’ has an adjectival distribution, as in the DP El libro que Juan compró ‘The book that J. bought’. Third, this CP is merged with the neuter Det Lo, so finally the pre- copular DP Lo que Juan compró ‘What J. bought’ is a definite description. This definite description is interpreted as There is a contextually salient x such that Juan bought x. As visible, this interpretation corresponds to the presuppositional variable. Now, since it is felicitous to say, Lo que Juan compró, entre otras cosas, fue un vino ‘What J. bought, among other things, was a bottle of wine’, I don’t assume uniqueness (i.e. There is only one contextually salient x) for this variable. To get that result, it would be necessary to add an explicit restriction, as in Lo único que Juan compró fue un vino ‘The only thing J. bought was a bottle of wine’. Note the ill- formedness of *Lo único que Juan compró, entre otras cosas, fue un vino ‘The only thing J. bought, among other things, was a bottle of wine’. On the other hand, the existential presupposition is enforced. A sentence like *Lo que Juan compró fue nada ‘What Juan bought 255 was nothing’ is ungrammatical if taken literally, 29 although it could be acceptable with some sort of ‘[+irony]’ pragmatic feature (for instance, when J. bought much fewer things than expected). 30 With regards to the post-copular clause, I assume a CP structure with a phonetically empty head C 0 , so it has the same structure of the examples in (38)-(43). In the same vein, C 0 is a complementizer such that the resulting CP has nominal distribution, as the CP (with a lexical C 0 ) that appears in the sentence Sé que Juan compró un vino ‘I know (that) J. bought a bottle of wine’. Now, about the copula, I crucially claim it is an id-COP. This copula links two units: the pre-copular unit, which is a definite description, and the post-copular unit, which is a nominal CP. As suggested in the previous section, copular sentences with an id-COP have a fixed focal structure. According to this structure, the post-copular unit is necessarily in focus. This is indeed the case. As it is visible, the post-copular unit provides the value for the variable introduced in the pre-copular unit. It can be interpreted as the main assertion of the PsCl, namely, The x such that Juan bought x is a (bottle of) wine. 31 The (simplified) phrase marker for (44) is shown in (46). (46) CopP DP Cop’ Cop 0 CP Lo que i Juan compró que i fue Juan compró un vino 29 In this respect, PsCls are clearly different from simple sentences like Juan no compró nada ‘J. bought nothing’, or Nada compró Juan, with the order OSV resulting from object fronting. Obviously, the existential presupposition is not enforced in these sentences. For a relevant discussion and useful references on this issue, see Reeve (2012a: 19 ff). 30 For a study of irony in Spanish from the perspective of pragmatics, see Kočman (2011). 31 Significantly, the metalinguistic phrasing of the main assertion contains an id-COP. 256 At first sight, the structure in (46) seems rather similar to the one in (34), which is the structural proposal for PsCls offered in den Dikken et al. (2000). Although both structural descriptions are akin in spirit, some differences should be observed. First, their proposal resorts to the divide topic/comment, whereas mine builds on the divide focus/presupposition via a specific copular construction with a fixed focal structure. Second, den Dikken et al. consider the copula as a Top 0 head, while I take it as a semantically non-empty head id-COP. 32 Third, while den Dikken et al. label the post-copular unit as IP (=TP), I label this unit as CP, which seems more systematic regarding clear cases of full CPs appearing in the same position in Spanish. On the other hand, (46) also presents some similarities with respect to the RSs seen in this work. Perhaps the most salient is the fact that the pre-copular unit encodes the presupposition, with the important difference that, in this case, the operator is overt and the presuppositional variable binding is obtained via a wh-phrase. Additionally, just like all previously reviewed RSs, the unit in focus is located in the post-copular side. Also, ellipsis (PF- deletion) applies on this latter side up to the element in focus. Finally, whereas the pre-copular unit is not a full copy of the post-copular one, some elements seem to be ‘the same’: the subject Juan and the tensed verb compró. It doesn’t seem attributable to mere coincidence that these are precisely the units that survive matched deletion on the high copy in the RSs reviewed so far in this work. Facing these similarities and differences between PsCls and RSs, and being the case that the copy in the former seem only partial, I adopt here the null hypothesis that reduplication is not involved in PsCls, so the copula does not bear a [+foc] feature. While these assumptions would account for the differences, the similarities would be in need of further explanation. To my understanding, the best way to account for PsCls is to take into consideration its nature as a specific type of construction. From a strictly minimalist point of view, constructions are just epiphenomena. They are not considered as analytical primitives, but instead they are 32 As opposed to the semantically empty head e-COP used in predicational copular sentences, reviewed in the previous section. 257 viewed as the result of a reduced number of basic operations, like Merge or Copy, applied in a certain way, following a certain sequence. Chomsky reflects on this issue as follows, […] the right answer is that there aren’t any constructions anyway, no passive, no raising: there is just the option of dislocating something somewhere else under certain conditions, and in certain cases it gives you what is traditionally called the passive and in other cases it gives you a question and so on, but the grammatical constructions are left as artifacts. In a sense they are real; it is not that there are no relative clauses, but they are a kind of taxonomic artifact. They are like “terrestrial mammal” or something like that. “Terrestrial mammal” is a category, but it is not a biological category. It’s the interaction of several things and that seems to be what the traditional constructions are like, VPs, relative clauses, and so on. (Chomsky et al. 2002: 94-95). In this spirit, PsCls are not primitives of the grammar, but the result of certain operations, carried out in a certain manner, presumably with extra-grammatical purposes, as for instance, common ground management. In this sense, the ‘instructions’, so to speak, to build some specific construction may be part of the general linguistic knowledge (beyond grammar) of the speaker. Now in order to give an account about the grammatical properties of PsCls, it will be necessary to present their derivational story. This story will reveal its structure and may explain at least some of their properties. Ideally, all the steps of this story will be built on elementary operations only. I assume here that those operations are Copy, Merge, and Delete. To start telling the story, the previous observation is that PsCls are necessarily formed by two clauses, so two phases (in this case, two CPs) would be involved. As we have seen in previous sections of this work, the structures in each phase are built in different cycles, until they are put together in a final cycle. For PsCls, I consider that the low clause is built first and put on hold; then high clause is built and also put on hold; finally in the last cycle the two clauses are 258 structured together. 33 The preliminary derivation proposed in (47)-(49) corresponds to the PsCl in (44). (47) Cycle I: low-CP phase Sub-Numeration I = {C 0 1 , -ó 1 , Juan 1 , compr-√ 1 , un 1 , vino 1 } a. Merge un + vino → [ DP un vino] b. Merge compr-√ + DP → [ v’ compr-√ [ DP un vino]] c. Merge Juan + v’ → [ vP Juan [ v’ compr-√ [ DP un vino]]] d. Merge -ó + vP → [ T’ -ó [ vP Juan [ v’ compr-√ [ DP un vino]]]] e. Move compr-√ to adjoin -ó (head-to-head movement) → [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino]]]] f. Move Juan to [Spec,TP] → [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino]]]]] g. Merge C 0 + TP → [ CP C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino]]]]]] In (47), following the CTM, I am using Move as an abbreviation for Copy+Merge plus deletion, 34 so indeed all the steps presented involve basic operations only. All the units in the sub-numeration, and only those units, have been used. 33 The idea is that the construction resulting of a cycle is somehow ‘stored’ (= put on hold) in a syntactic workplace until another construction starts or finishes its derivation. On this issue, Hornstein et al. (2005) point out that, “In fact, it won’t be uncommon that in building a sentence we may have several “treelets” around prior to their combining into a single big tree” (Hornstein et al. 2005: 63). In turn, when considering the case of Caribbean Spanish (CSp), Zubizarreta (2013) writes, “As suggested by Jairo Nunes (p.c.), if we adopt the dynamic structure- building theory of the Minimalist framework (Chomsky 1995, 2001), then the following proposal can be put forth for CSp: the pre-copular vP is a syntactic copy of the post-copular vP. At the point that the vP phase is built, the VP contained within it is still visible. Therefore the vP can be copied, put on hold, and then remerged.” (Zubizarreta 2013: 34 fn. 8). 34 I apply the same convention in all that follows. 259 (48) Cycle II: high-CP phase Sub-Numeration II = {Lo 1 , C 0 1 , -ó 1 , Juan 1 , compr-√ 1 , que 1 } a. Merge compr-√ + que → [ v’ compr-√ [ WhP que]] b. Merge Juan + v’ → [ vP Juan [ v’ compr-√ [ WhP que]]] c. Merge -ó + vP → [ T’ -ó [ vP Juan [ v’ compr-√ [ WhP que]]]] d. Move compr-√ to adjoin -ó (head-to-head movement) → [ T’ compró [ vP Juan [ v’ compr-√ [ WhP que]]]] e. Move Juan to [Spec,TP] → [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ WhP que]]]]] f. Merge C 0 + TP → [ CP C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ WhP que]]]]]] g. Move que to [Spec,CP] → [ CP que C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ WhP que]]]]]] h. Merge Lo + CP → [ DP Lo [ CP que C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ WhP que]]]]]]] Again in (48), all the units in the sub-numeration and only those units have been used. Let’s present in (49) the final cycle. (49) Cycle III: Copular sentence Sub-Numeration III = {fue 1 } a. Merge fue + CP in (48g) → [ Cop’ fue [ CP C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino]]]]]]] b. Merge DP in (48h) + Cop’ → [ CopP [ DP Lo [ CP que C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ WhP que]]]]]]] [ Cop’ fue [ CP C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino]]]]]]]] 260 c. Ellipsis under identity on the low-CP → [ CopP [ DP Lo [ CP que C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ WhP que]]]]]]] [ Cop’ fue [ CP C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino]]]]]]]] At sight of the preliminary derivation proposed, I would like to put forth some comments. (i) In both sub-numerations I and II, the unit Juan is included once. This means that the unit has been pulled out from the lexicon twice. However, those two occurrences of Juan refer to the same object; not only that, but also the two occurrences are in an identity relation, since this is what legitimates ellipsis in (49c). (ii) The same claim in (i) can be made for the verbal root compr-√ ‘buy’. (iii) Both Juan and compr-√ have the same function in the high and the low clause, namely subject and matrix verb, respectively. (iv) The step (49c) does not involve a basic operation. We can refer it to deletion, but we will still be in need for a justification. As previously commented, identity is a condition for ellipsis, not a motivation. At this point, I would like to consider a somehow different possibility. Let’s imagine that we can apply the basic operation Copy to a unit pulled out from the lexicon in the first cycle, and then put that copy on hold until the next cycle. Then, in the second cycle, the stored copy can be used to build whatever the structure intended in that cycle. If something like this is possible, this would mean, on the one hand, that the aforementioned unit has been selected from the lexicon only once, and, on the other hand, that the identity between the two units (the ‘original’ and its copy) is guaranteed, since they would be, by the very definition of Copy, the same unit. In case we could apply this intuition to the previous derivation for the units Juan and compr-√ ‘buy’, then (i) and (ii) would be explained: the units in the high clause would be copies of the same units in the low clause. Also ellipsis in (iv) could be instantiated as deletion driven by the LCA (see (2)) for linearization purposes. In turn, (iii) would be explained as a particularity of PsCls: the high clause expresses the presupposition conveyed in the low clause before deletion applies. 261 From the point of view of the CTM, where movement is interpreted as Copy+Merge, the possibility suggested in the previous paragraphs can be interpreted as the possibility of moving a unit from one cycle to another. This has indeed been suggested in the framework of the CTM under the label of sideward movement (Nunes 2001, 2004). In principle, this kind of movement is restricted to links within a chain. This restriction is imposed to prevent overgeneration (Nunes, p.c.), and it would also justify later deletion via the operation Chain Reduction (see (3)). The basic intuition of sideward movement, as presented in Nunes (2004: 90ff), is that the computational system can copy an element α from a syntactic object K and merge it with a different syntactic object L, which has been independently assembled and is not (yet) connected with K. This idea is represented in the sequence of steps (50a,b). (50) a. [K … α i …] α i [L… ] (← Copy α) b. [K … α i …] [ M α i [L… ]] (← Merge α + L, to yield M) In this scenario, Nunes supposes that “Further computations may then form a unique syntactic object HP containing the two copies” (Nunes 2004: 94). This will gives a result represented in (51). 35 (51) HP [ K … α i …] H’ H [ M α i [ L … ] 35 Although it is not stated as such in Nunes (2004), I see no problem in assuming that HP could have also been formed with M as specifier and K as complement, as in the following structure: [ HP [ M α i [ L … ] [H [ K … α i …]]]. 262 In principle, the two copies of α in (51) do not form a chain because there’s no c- command relation between the two copies. Nunes’ idea is that, in the course of the derivation, another head Y may be introduced, such that Y may require a copy of α in [Spec,YP]. See (52). (52) YP α i Y’ Y … HP [ K … α i …] H’ H [ M α i [ L … ] At this point, the high copy of α participates in two chains. The chain Ch 1 relates the high copy of α with the low copy of α in K. The chain Ch 2 relates the high copy of α with the low copy of α in M. Chain Reduction then applies. Importantly, Nunes considers that Chain Reduction applies to both chains independently (i.e. Ch 1 and Ch 2 don’t form a third chain). Now, since the high copy has normally checked more relations than the lower copies, the former survives deletion. Assuming that this is the case, if Chain Reduction applies on the chains in (52), we get (53). 263 (53) YP α i Y’ Y … HP [ K … α i …] H’ H [ M α i [ L … ] At this point, Nunes (2004) notes, In the output associated with [(53)], α appears to have moved from two different launching sites. Thus, any construction that appears to involve movement of an element from more than one position “at once” is a good candidate for sideward movement analysis. (Nunes: 2004: 95). An interesting application of this analysis is offered for parasitic gaps (PG). 36 Consider the question in (54a), whose traditional analysis is presented in (54b). (54) a. Which paper did you file without reading? b. [ CP [which paper] i did+Q [ TP you [ vP [ vP file t i ] [ PP without reading PG i ]]]] 36 Another phenomenon that Nunes (2004) describes using sideward movement is across-the-board (ATB) extraction. The interested reader may directly consult the aforementioned source. 264 If the sideward analysis applies to (54a), the (simplified) structure would look like (55). For details about the derivation, see Nunes (2004: 98 ff). (55) CP [which paper] i C’ C TP did+Q you T’ T vP’ [ vP file [which paper] i ] [ PP without PRO reading [which paper] i ] From my perspective, an important advantage of the sideward movement analysis in (55) is that it is consistent with the interpretation of which paper in the three positions indicated. It is indeed the same unit in three positions. The traditional GB idea of movement and traces would have considerable difficulties to capture this intuition—if it captures it at all. 3.2. Copies without chains? Coming back to PsCls, the big difference with respect to the sideward movement analysis of PGs is that the copies in PsCls would not form a chain, due to the lack of c-command, so Chain Reduction cannot apply. 37 However, following the LCA, we can be sure that one copy of each 37 In this sense, it would not be fair to say that the computational sequence I suggest here for PsCls strictly follows the sideward movement proposal as presented in Nunes (2001, 2004). Although it is clearly inspired by those ideas, my proposal relies solely on linearization to prevent overgeneration via deletion of low copies. Nunes (p.c.) indicates that, if a safe way to prevent overgeneration is found, the idea of ‘moving’ copies from one tree to another may be interpretively adequate for PsCls. See also Hornstein (2001) for some potential cases of sideward movement without chains. 265 pair of identical copies will be necessarily deleted, and that the unmarked target for deletion is the lower copy (i.e. the ‘original’), according to the constraint FFE (see (4)). If this is somehow on the right track, the reason for ellipsis in PsCls, realized as PF-deletion, would be linearization: since the two copies are the same object, one of them (usually the lower) will have to be deleted to prevent contradictory instructions that could lead the computation to be cancelled at PF. 38 At this point, I would like to review the derivation previously proposed for sentence (44) taking into account the considerations about sideward movement commented above, along with the idea of F-marking acting as a restrictor for Copy (see (56e) below). (56) Cycle I: low-CP phase Sub-Numeration I = {C 0 1 , -ó 1 , Juan 1 , compr-√ 1 , un 1 , vino 1 } a. Merge un + vino → [ DP un vino] b. Merge compr-√ + DP → [ v’ compr-√ [ DP un vino]] c. Merge Juan + v’ → [ vP Juan [ v’ compr-√ [ DP un vino]]] d. F/fcP-marking on the DP-DO → [ vP Juan [ v’ compr-√ [ DP un vino] fcP ]] e. Copy the non-focused units of vP → {Juan c , compr-√ c } 39 (These copies are put on hold) f. Merge -ó + vP → [ T’ -ó [ vP Juan [ v’ compr-√ [ DP un vino] fcP ]]] g. Move compr-√ to adjoin -ó (head-to-head movement) → [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino] fcP ]]] h. Move Juan to [Spec,TP] → 38 Naturally, this doesn’t mean that all cases of ellipsis can be explained by linearization—or, for that matter, by any other single reason. See Saab (2008) for a comprehensive analysis of the kind of identity that licenses ellipsis. 39 I use the subscript c to distinguish the copies created after F-marking with illustration purposes only. 266 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino] fcP ]]]] i. Merge C 0 + TP → [ CP C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino] fcP ]]]]] Note that F-marking on the DO occurs in (56d) after the formation of vP. Recall that vP can also be taken as presuppositional. In (56e), the proper name and the verbal root are copied. The motivation for copying these units in particular comes from IS: these are the non-focused units. 40 It may occur, however, that F-marking takes place only after TP has been formed, in which case T 0 will also be copied as part of the presupposition (i.e., the inflected verb will be copied and then deleted in the lower clause). Whereas this alternative would work for (56), certain agreement patterns in Spanish suggest that, at least for some PsCls, only the verbal root is copied (I omit deletion lines in the presupposed part of the second clause for the sake of clarity). (57) [ CP Quien ____ compr-ó el vino ] fuiste [ CP [ tú ] fcP compr-aste el vino] 41 Who buy-PST.3SG the wine were you buy-PST.2SG the wine ‘It was you who bought the wine’ Let’s consider now the revised version of cycle II. Note, in particular, that neither Juan nor the verbal root compr-√ are included in the sub-numeration, that is to say, those units have not been selected again from the lexicon. Instead, their correspondent copies are directly taken from the syntactic workplace where they have been put on storage in the previous cycle (see (56e)). 42 40 This would be coherent with the idea that presupposed material is copied and ‘moved’ up to [Spec,FSU] in RSs. It will also be akin to the proposal by Den Dikken et al. (2000) about locating the topical content to the pre-copular clause. The significant difference would be that, in the RS, the focused unit is also copied (although without the F- marking) and gets deleted later at LF in order to create a presuppositional variable. In this case however, the unit in focus is not copied, and this can be taken as related to the fact that, in PsCls, the high copy has its own bound wh- variable, located in the parallel slot. 41 In cases like this, the question may arise about the (lack of) identity with respect to T 0 in both clauses. If the inflection is different (thus, there is no identity), how come T 0 gets deleted in the second clause? My provisional suggestion would be in the phonological side. Since an inflected verb in Spanish is a single phonological word ([V+T]), PF-deletion of the verbal root would pied-pipe deletion of the associated inflectional morphemes. 42 For illustration purposes, I write a subscript “c” (for example, Juan c ) to identify the copies that have been put on hold from one cycle to another. 267 (58) Cycle II: high-CP phase Sub-Numeration II = {Lo 1 , C 0 1 , -ó 1 , que 1 } a. Merge compr-√ c (taken from storage space) + que → [ v’ compr-√ c [ WhP que]] b. Merge Juan c (taken from storage space) + v’ → [ vP Juan c [ v’ compr-√ c [ WhP que]]] c. Merge -ó + vP → [ T’ -ó [ vP Juan c [ v’ compr-√ c [ WhP que]]]] d. Move compr-√ c to adjoin -ó (head-to-head movement) → [ T’ compr c -ó [ vP Juan c [ v’ compr-√ c [ WhP que]]]] e. Move Juan to [Spec,TP] → [ TP Juan c [ T’ compr c -ó [ vP Juan c [ v’ compr-√ c [ WhP que]]]]] f. Merge C 0 + TP → [ CP C 0 [ TP Juan c [ T’ compr c -ó [ vP Juan c [ v’ compr-√ c [ WhP que]]]]]] g. Move que to [Spec,CP] → [ CP que C 0 [ TP Juan c [ T’ compr c -ó [ vP Juan [ v’ compr-√ c [ WhP que]]]]]] h. Merge Lo + CP → [ DP Lo [ CP que C 0 [ TP Juan c [ T’ compr c -ó [ vP Juan c [ v’ compr-√ c [ WhP que]]]]]]] Finally, (59) presents the final step in the derivation of the PsCl. Importantly, ellipsis gets instantiated as deletion (a basic operation in this framework) on the lower clause, driven by the LCA in (59c). (59) Cycle III: Copular sentence Sub-Numeration III = {fue 1 } a. Merge fue + CP in (56i) → [ Cop’ fue [ CP C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino] fcP ]]]]]] b. Merge DP in (58h) + Cop’ → [ CopP [ DP Lo [ CP que C 0 [ TP Juan c [ T’ compr c -ó [ vP Juan c [ v’ compr-√ c [ WhP que]]]]]]] 268 [ Cop’ fue [ CP C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino] fcP ]]]]]]] c. PF-deletion on the low-CP driven by the LCA → [ CopP [ DP Lo [ CP que C 0 [ TP Juan c [ T’ compr c -ó [ vP Juan c [ v’ compr-√ c [ WhP que]]]]]]] [ Cop’ fue [ CP C 0 [ TP Juan [ T’ compró [ vP Juan [ v’ compr-√ [ DP un vino] fcP ]]]]]]] = c’. Simplified version → [ DP Lo [ CP que [ TP Juan c compró c [ WhP que ]]] fue [ CP C 0 [ TP Juan compró [ DP un vino] fcP ]]] The simplified version in (59c’) allows us to check at a glance the general idea of my structural proposal for (at least a certain sub-set of) PsCls in Spanish. Although there is no total reduplication of the lower clause, the higher clause has been built using some copies taken from the non-focused part of the lower clause, and those copies are placed in the same structural places (in this case, Juan is the subject and compró is the matrix verb) in the two clauses. This fact guarantees the interpretive identity for each pair of copies, and later justifies the PF-deletion of the lower copies following the LCA. Importantly, this deletion creates immediate adjacency between the copula (which is of type id-COP) and the unit in focus. On the other hand, an overtly bounded wh-phrase, located in the high clause in the slot parallel to the unit in focus, creates the presuppositional reading typical of the high clauses in PsCls. As for LF, not having reasons to assume a PF/LF mismatch so far, I will assume, for the time being, that is matched with respect to PF. The simplified LF-structure parallel to (59c’) is shown in (60). (60) [ DP Lo k [ CP que [ TP Juan c compró c [ WhPk que ]]] fue [ CP C 0 [ TP Juan compró [ DP un vino] fcP ]]] In (61), I present a PsCl in Spanish with focus on the subject of the low clause. (61) a. Quienes preparamos las galletas de soda fuimos nosotros Who-PL prepare-PRT.1PL the cookies of soda were we 269 ‘It was us who prepared the soda crackers’ b. [ CP Quienes [ TP [ WhP quienes ] preparamos c [las galletas de soda] c ]] fuimos [ CP C 0 [ TP [ DP nosotros] fcP preparamos [las galletas de soda]]] c. [ CP Quienes k [ TP [ WhPk quienes ] preparamos c [las galletas de soda] c ]] fuimos [ CP C 0 [ TP [ DP nosotros] fcP preparamos [las galletas de soda]]] About (61), note that the idea that the presupposed units in the lower clause are copied, put on storage, and later used to build the higher clause simplifies the job on this clause (hence, it is more economical), since las galletas de soda ‘the soda crackers’ would be copied as a full DP (i.e., one single unit), as opposed to the alternative possibility of selecting four units, las, galletas, de, and soda, separately from the lexicon. The PsCl in (62) illustrate the case of focus on the IO. (62) a. A quienes les dieron la copa fue a los rusos To whom-PL (they) Cl.DAT=give-PRT.3PL the cup was to the Russians ‘It was to the Russians that they gave the cup’ b. [ CP A quienes [ TP pro les-dieron c [la copa] c [ WhP a quienes ]]] fue [ CP C 0 [ TP pro les-dieron [la copa] [ PP a los rusos ] fcP ]] c. [ CP A-quienes k [ TP pro les-dieron c [la copa] c [ WhPk a quienes ]]] fue [ CP C 0 [ TP pro les-dieron [la copa] [ PP a los rusos ] fcP ]] Finally, (63) exemplifies focus on a PP-adjunct. (63) a. Con quien ella habló fue con el director With whom she spoke was with the director ‘It was the director whom she talked to’ b. [ CP Con quien [ TP ella c habló c [ WhP con quien ]]] fue [ CP C 0 [ TP ella habló [ PP con el director] fcP ]] c. [ CP Con-quien k [ TP ella c habló c [ WhPk con quien ]]] fue [ CP C 0 [ TP ella habló [ PP con el director] fcP ]] 270 Similarly to English, the pre-copular clause in Spanish has the form of a free relative, involving a wh-expression. The wh-word quien/quienes ‘who.sg/who.pl’ is frequently used, either alone as in (61), or preceded by preposition, as in (62) and (63). 43 Again as in English, the set of wh-expressions allowed in PsCls in Spanish is restricted with respect to the set of of wh-expressions allowed in direct questions. In general, cual ‘which’, cuyo ‘whose’, and cuanto+NP ‘how much+NP’ are banned. Additionally, PsCls in Spanish cannot be built with (P)+{qué/cuál}+NP ‘(P)+which+NP’, although these wh expressions are perfectly fine in direct questions. 44 (64) a. Contra qué equipo jugamos hoy? (Direct question) Against which team we.play today ‘Against which team do we play today?’ b. * Contra qué equipo jugamos hoy es contra el Galaxy. (* PsCl) Against which team we.play today is against the G. More generally, PsCls in Spanish only admit unstressed wh-words. This indicates that those words are marked as [-interrogative]. For the case of the word que ‘what’, generated in DO position, as in (59c’), its unstressed condition converts the clause in an adjectival clause when it moves to [Spec,CP]. Since, in principle, the id-COP requires a nominal clause (i.e. a definite description) in its Spec, this adjectival-CP must further merge with the neuter determiner lo. As Reeve (2012a: 25 ff) opportunely advises, it is important to differentiate PsCls from restrictive relatives. For instance, the PsCl in (63) should not be confused with (the copular sentence with) a restrictive relative shown in (65a). By definition, a restrictive relative is not a 43 It also seems that not all prepositions are allowed. In principle, it looks like the preposition must be part of the thematic grid of the verb, not part of a ‘real’ free adjunct. In (63) the thematic grid of the verb HABLAR (understood as ‘to talk to sb.’) includes the preposition con ‘with’ to mark an obligatory argument of the verb. Some non- generative grammars have labeled such arguments as objeto preposicional ‘prepositional object’ (see, for instance, Alarcos 1999: 320) to indicate its obligatory condition, in spite of being neither DOs nor IOs. 44 See Den Dikken et al. (2000: 48 fn.6) for details about the correspondent restrictions in English. I leave the question about the reasons behind these restrictions for future research. 271 free relative, and PsCls only admit free relatives. Although both constructions seem interpretively very similar, and both include an identificational copular verb, they are structurally different. For instance, note the lack of preposition in the post-copular clause (which is obligatory for the PsCl) in the case of the restrictive. (65) a. El hombre con quien ella habló es el director (Restrictive relative) The man with whom she spoke is the director ‘It is the director the man who she talked to’ b. El hombre con quien ella habló es (*con) el director (Restrictive relative) (Cfr. Con quien ella habló fue *(con) el director) (PsCl = (63a)) Note that a restrictive relative is perfectly admissible as an answer for the question posed in (64a), Contra qué equipo jugamos hoy? ‘Against which team do we play today?’, while, as we’ve just seen in (64b), a PsCl is not only inappropriate as an answer, but—even more radically—it is also impossible to build grammatically. (66) El equipo contra el que jugamos hoy es el Galaxy. (Restrictive relative) The team against the that we.play today is the G. ‘The team against which we play today is the Galaxy’ (Cfr. * Contra qué equipo jugamos hoy es contra el Galaxy) 45 (*PsCl = (64b)) These disclaimers made, I would like to propose now a general structure for PsCls in Spanish. 45 Of course, this sentence would be perfect with a continuative high-pitch accent and a long tone break before the copula, but this would indicate that the first part of the answer is an echo-question, so the construction is not a PsCl. 272 (67) General Structure for PsCls in Spanish (conservative version) id-CopP (cycle III) High clause (cycle II) id-Cop’ id-Cop 0 Low clause (cycle I) [[WhP] i [(Y c *) [WhP] i (Z c *)]] SER (‘to be’) [(Y*) [X] fcP (Z*)] The structure in (67) contains the following elements. In cycle I, which builds the low clause, there is an element X that is focus marked. X may be, in principle of any category, including vP or the whole TP/CP. Eventually, there are some material Y before the focused element X and some other material Z after X. The asterisk indicates that there may be more than one of such units. Being non-focused, these units are copied, and the copies are put in storage in some separate syntactic workspace. Importantly, no deletion has applied in this cycle yet. In cycle II, there is a wh-phrase (WhP) that merges with the relevant copies Y c or Z c in such a way that the aforementioned WhP occupies the same structural position that X occupies in the low clause. Then WhP undergoes wh-movement, so the high clause is created. In case the element in focus is a verbal phrase (X = vP/VP), no copy from the low clause is used in the high clause; instead the verbal pro-form HACER ‘to do’, or any equivalent, is used. If X=TP/CP (i.e. the whole sentence is in focus), the pro-form used is OCURRIR ‘to occur’, or any other synonym expression. 273 Finally, in cycle III the identity-copula SER merges with the low clause. This operation creates id-Cop’. Then, the high clause merges with this unit by the left to form id-CopP. In case any copies Y c or Z c had been used to build the high clause, deletion applies, driven by the LCA, so the relevant low copies Y or Z get finally deleted. In consequence, the identificational copula surfaces immediately adjacent to focused unit the in the low clause, while the high clause properly encodes the presupposition. The resultant structure is what we call a pseudo-cleft in Spanish. 3.3. Another possibility: a total copy of CP In a somehow conservative fashion, the structural proposal for PsCls just presented considers that the wh-unit originates inside the wh-clause, and from there it moves to [Spec, CP]. However, it is conceivable that the wh-unit has been directly generated in that position. If this is assumed, then another structural possibility is available. Instead of considering that in cycle I only the non-focused units are copied and put on hold, we can assume that the whole clause, including the unit in focus, is copied. Then, in cycle II, the copy of the focused unit will be deleted. This deletion creates a variable in the wh-clause. The wh-unit then binds the variable, so the presuppositional reading is created. On the basis of these observations, a revised general structure for PsCls is presented in (68). 274 (68) General Structure for PsCls in Spanish (non-conservative version) id-CopP (cycle III) High clause (cycle II) id-Cop’ id-Cop 0 Low clause (cycle I) [[WhP] [(Y*) [X] (Z*)]] i SER (‘to be’) [(Y*) [X] fcP (Z*)] i In (68), the entire clause [(Y*) [X] (Z*)] has been copied and merged with WhP in cycle II. After the two clauses become connected in cycle III, deletion applies on the lower clause, so only the focused unit survives. Complementary deletion applies on the higher clause, so this deletion creates a variable. The Wh-unit then binds the variable, creating the presuppositional reading. Let’s illustrate this possibility with sentence (69). (69) Quien trajo la cerveza fue Juan Who brought the beer was J. ‘It was Juan who brought the beer’ The structure assignable to this sentence from this perspective is represented in (70). 275 (70) id-CopP (cycle III) High clause (cycle II) id-Cop’ id-Cop 0 Low clause (cycle I) Quien [[Juan] trajo la cerveza] i fue [[Juan] F trajo la cerveza] i One advantage of this version is that the low clause is copied as a whole. Also this structure would be similar to other cases reviewed in this work involving copy of clauses. From a general point of view, the particularity of PsCls would be that in these constructions the operator that binds the variable is overt. In what follows, this revised version for PsCls is assumed. 4. It-clefts: just a tentative proposal When discussing the focalizing uses of adverbials, Cinque (1999: 31) proposes the following two examples of clefts, in Italian (71), and in French (72). (71) E’ probabilmente per questa ragione che lo hanno licenziato ‘It is probably for this reason that they fired him’ 276 (72) C’est heureusement Paul qui ha vendu sa voiture ‘It is luckily Paul who sold his car’ Cinque considers these sentences as evidence supporting his claim that the adverb and the focused unit form one single constituent (i.e., probabilmente per questa ragione, and heureusement Paul, respectively). The main idea is that the adverb and the unit in focus have been fronted together—as a single unit—from some position inside the wh-clause. As the English glosses show, this would also be true for English. Thus, for this view, probably for this reason and luckily Paul would form a single constituent in each case. Let’s illustrate the idea for the English sentence (73). (73) It is [probably a BMW] i [what Mary bought t i ] At first sight, the idea illustrated by the arrow seems plausible. The interpretation in (73) indicates that the DP-object of some purchase is a BMW, and this fact is seen under the lens of probability. The probable object has (supposedly) been fronted in a special syntactic configuration such that the unit [Adv+DP], probably a BMW, appears in post-copular position of a copular sentence whose subject is the pronoun it. The copular sentence appears before the wh- clause and presumably carries some sort of relevance, both in interpretive and prosodic grounds. In turns out that, at least since Jespersen (1927), syntactic configurations like these have been studied under the name of It-clefts. Now, coming back to Cinque’s claim, if by ‘forming a constituent’ one means that two units (in this case, a modal Adv and a DP) undergo Merge to form a third unit (in this case [Adv+DP], whatever its label may be), this would go against the main proposal defended in this work. Certainly, if a modal adverb can simply merge with an object DP (or, for that matter, with any other category, with any other focused constituent) to form a third unit, there is no need for reduplicative structures of any kind to explain the relative ubiquity of sentence adverbs. 277 Sentence adverbs would simply merge with other units, regardless of their category or position in the syntactic tree—end of the story. In this work, I present arguments against this view—which I consider attractive at first sight, but somehow naïve. I sustain the idea that the sentence modal adverb is always propositional in scope, despite the fact that (centrally in Spanish) it may sometimes surface immediately adjacent to the unit in focus in sentence-internal linear positions. I also endorse the idea that the modal adverb merges only at the TP/CP level, so it doesn’t merge with any other sentence-internal constituents. I claim that the immediate adjacency between the adverb and the focused unit does not mean that just the two of them have merged together to form a third constituent. That is not the whole story. There occurs reduplication and ellipsis; that is to say, Copy, Merge, and Deletion are involved. From these grounds, it seems necessary to respond to Cinque’s claim. In order to do so, I will propose a structural account of It-clefts from the point of view of the CTM. To start with, I would like to focus my attention on the contrastive use of It-clefts in English. This has been the typical use attributed to this kind of clefts, although certainly not the only one. 46 In what follows, I will refer only to the contrastive use of It-clefts. This can be better perceived in context. In (74), the It-cleft is used by speaker C to correct a previous statement in the dialogue considered wrong. 46 Declerck (1984), as quoted by Huber (2006: 552), proposes that, besides the contrastive clefts, there are in English at least two more pragmatic uses for It-clefts, illustrated in (i) and (ii). (i) It was also during these centuries that a vast internal migration […] from the south northwards took place, a process no less momentous than the Amhara expansion during the last part of the 19 th century. (ii) It was just about 50 years ago that Henry Ford gave us the weekend. According to Declerk, in the contrastive use of It-clefts, the copular sentence conveys the focus (usually identified with new information), while the wh-clause expresses the background (typically, identified with given information). On the other hand, in (i), which Declerck labels as unstressed-anaphoric-focus cleft, the copular sentence contains given information and the wh-clause conveys new information, so the contrastive reading is cancelled. In turn, in the example (ii), labeled as discontinuous cleft, all information is new, since this kind of sentences is typically used text-initially. See Huber (2006) for an even richer classification of clefts (in English, German, and Swedish) from the point of view of their pragmatic functions. See also Reeve (2012b) for the argument that these distinctions are problematic for the point of view that takes foci as a functional heads (FP/FocP). 278 (74) a. Speaker A: Which car did Mary buy? b. Speaker B: I think she bought a Ford… c. Speaker C: No, no. It was probably a BMW what Mary bought. c’. Speaker C: No, no. It was probably a BMW. Note that the speaker C has two choices. While in (74c) speaker C pronounces a full It- cleft, in (74c’) he uses what seems a shortened version of that cleft. This shortened version has been called truncated cleft (Mikkelsen 2005: 118 ff). Thus, a full It-cleft will be linearly formed by a truncated cleft, plus a wh-clause (which is pronounced without prosodic prominence, presumably deaccented). On the other hand, this wh-clause looks rather similar to the wh-clauses used in PsCls. Indeed, a PsCl may have been a third option for the speaker C: No, no. What Mary bought was probably a BMW. The resemblance between It-clefts and PsCls has been noticed since the first formal approaches to the issue. Akmajian (1970), working in a rule-based framework, proposes that It- cleft sentences are derived from PsCls via an extraposition rule. This author discusses facts related to verbal agreement and pronoun interpretation in English that are plainly consistent with the proposed relation. Among those observations, the parallelism with respect to the restrictions applicable to wh-phrases—which wh-phrases are (or are not) allowed both in PsCls and in It- clefts—stands out. Although Akmajian’s framework is no longer at use nowadays, his fundamental insight is still valid, at least to my view. In fact, Percus (1997), already working in a minimalist framework, also indicates that there exists a structural relation between PsCls and It-clefts, which would be linked via some version of extraposition. Although I don’t assume the idea of extraposition, in this work, I also propose here that PsCls are at the derivational base for It-clefts. Let’s first consider the PsCl in (75). 279 (75) What Mary bought was a BMW. = [ CopP [ CP What Mary bought a BMW] [ Cop’ was [ CP Mary bought [a BMW] fcP ]]] I have proposed that this sentence has been built following a derivation in three cycles. Now I propose that, in order to build the It-cleft, the aforementioned three cycles also occur, exactly in the same way, but with two important differences: (i) in cycle I (which creates the low clause Mary bought a BMW) a copy of the unit in focus is also put on storage in the syntactic workspace, that is to say, a copy of the DP [a BMW] c is put on hold; and (ii) in cycle III (which creates the PsCl), a copy of the id-COP is made and also put on hold, that is to say, [was] c is added to the stock of available copies. After cycle III has been completed, so when the PsCl is created, I propose, tentatively, that the following derivation takes place. (76) Cycle IV Sub-numeration IV = {It 1 } a. Merge was c + [a BMW] c (both units pulled out from the storage space) → [ Cop’ was c [a BMW] c ] b. Merge It + Cop’→ [ CopP It [ Cop’ was c [a BMW] c ]] c. F-marking on [a BMW] c → [ CopP It [ Cop’ was c [[a BMW] c ] fcP ]] As the reader may have noticed, cycle IV in (76) builds the truncated cleft. In this derivation, the copula is a copy of the identificational copula (id-COP) used in the cycle III. This unit is merged with a copy of the focused unit that was kept on storage in cycle I. Note that the copy is not F-marked as such, since the F-marking cannot be copied. However, since the copula is of the id-COP type, it will necessarily F-mark the unit in its Compl position—so ultimately this copy gets F-marked anyway. 280 At this point, we have two separate ‘treelets’ in our syntactic workbench: the PsCl and the truncated cleft. The next process is the adjunction by the left of the truncated cleft to the PsCl. This creates the It-cleft. (77) Cycle V Sub-numeration V = { } (empty set) a. Merge (adjoin) CopP (= truncated cleft in (76c)) + PsCl → [ CopP [ CopP It [ Cop’ was c [[a BMW] c ] fcP ]] [ CopP [ CP what Mary bought a BMW] [ Cop’ was [ CP Mary bought [a BMW] fcP ]]]] b. Delete low copies, according to the LCA → [ CopP [ CopP It [ Cop’ was c [[a BMW] c ] fcP ]] [ CopP [ CP what Mary bought w BMW] [ Cop’ was [ CP Mary bought [a BMW] fcP ]]]] Simplified version → = [[ It was [[a BMW] c ] fcP ] [ what Mary bought [was [ Mary bought a BMW]]] Linear version → = It was a BMW what Mary bought Some observations are in order. First, there is deletion of the low focus marked unit in (77b). In principle, we had established that an F-marked unit cannot be deleted, since this deletion would lead to loss of information. In this case, however, an exact copy of the low focused unit gets independently F-marked in a higher position (cycle IV), so no information is lost by deletion. Second, in the final step, all the content of the low clause created in cycle I is deleted. One may wonder why the grammar should worry about creating a clause just to have it completely deleted at the end of the computation. To respond this question, note that the low clause is the source of the relevant copies used to build the wh-clause (Mary, bought) and the 281 truncated cleft (a BMW). Additionally, the copula in the PsCl is also copied (was), and that copy is also used later to build the truncated cleft. The fact that these units are copies and not just elements pulled out from the lexicon explains the intuition—expressed, for instance, in Cinque’s claim—that the unit in focus in the It-cleft has been ‘moved’ from some position internal to the wh-clause. Indeed, it has been ‘moved’ (copied) from inside the lower clause, created in cycle I. On the basis of these structural suggestions, let’s consider now the modalized It-cleft in (73), repeated here (without the arrow) as (78). (78) It was probably a BMW what Mary bought. According to (the spirit of) Cinque’s claim, probably a BMW would be a single constituent formed by merging together the adverb and the DP. Allegedly, this constituent has been later extracted from the wh-clause and fronted as a solid unit. But I reply that this is not the case. The relevant difference lies in cycle IV. In (79), I present the relevant change with respect to the non-modalized version (compare with (76)). (79) Cycle IV Sub-numeration IV = {It 1 , probably 1 } a. Merge was c + [a BMW] c (both units pulled out from the storage space) → [ Cop’ was c [a BMW] c ] b. Merge It + Cop’→ [ CopP It [ Cop’ was c [a BMW] c ]] c. F-marking on [a BMW] c → [ CopP It [ Cop’ was c [[a BMW] c ] fcP ]] d. Merge probablemente + CopP → Probablemente [ CopP It [ Cop’ was c [[a BMW] c ] fcP ]] e. Reduplicate (Copy+Merge) CopP → [ CopP It [ Cop’ was c [a BMW] c ] probably [ CopP it [ Cop’ was c [[a BMW] c ] fcP ]] 282 f. Delete the high copy of the F-marked unit → [ CopP It [ Cop’ was c [a BMW] c ] probably [ CopP it [ Cop’ was c [[a BMW] c ] fcP ]] g. Delete low copies, according to the LCA → [ CopP It [ Cop’ was c [a BMW] c ] probably [ CopP it [ Cop’ was c [[a BMW] c ] fcP ]] = [It was probably [[a BMW] c ] fcP ] The modal adverb appears in the sub-numeration IV, so it is merged with the truncated cleft ‘externally’ (see (79d)). Since the truncated cleft contains a unit in focus, the feature [+foc] in the modal triggers reduplication. After reduplication occurs, the usual deletions apply, so the modal ends up adjacent to the focused unit. The modalized-truncated cleft merges later with the PsCl in cycle V, so the modalized It- cleft is finally created. The simplified phrase marker in (80) corresponds to the modalized It-cleft in (78), namely It was probably a BMW what Mary bought. 283 (80) CopP (cycle V) CopP (cycle IV) CopP (cycle III) CP (cycle II) Cop’ It was c a BMW probably it was c [[a BMW] c ] fcP was CP (cycle I) what Mary bought a BMW Mary bought [a BMW] fcP In this syntactic tree, cycle III corresponds to the pseudo-cleft, cycle IV corresponds to the truncated cleft, and cycle V corresponds to the It-cleft. Note that the unit created in cycle II, namely the wh-clause, conveyed the presupposition of the PsCl, and it still conveys the presupposition in cycle V, when the It-cleft is created. This is why this clause is pronounced deaccented or, eventually, is omitted. 47 Additionally, the structure clearly shows that probably a BMW is not a constituent by itself. It also shows that there has been no fronting of such a unit from a position internal to the wh-clause. On the other hand, the phrase marker appropriately represents the intuition according to which the post-copular unit is ‘at the same time’ the DO of the verb in the wh-clause, but it does so by remitting it to the ‘original’ verb in the first cycle, where the clause containing the simple sentence has been initially built. 47 A similar reason may be at the base of the fact that the wh-word may surface as that: It was probably a BMW that Mary bought. This could be seen as a post-syntactic morphophonemic readjustment. I leave this issue open. 284 As for the Spanish version of this It-cleft, it is almost entirely parallel to the English version, with the noticeable exception of the high pronoun it. 48 No overt equivalent of that pronoun is at sight in Spanish. I consider, however, that it is possible to think of a phonetically empty category occupying that syntactic slot, namely the Spec position of the truncated cleft. I tentatively represent that category as PRO. 49 The translation into Spanish of the modalized sentence (78) appears in (81). (81) PRO fue probablemente un BMW lo que María compró PRO was probably a BMW what M. bought ‘It was probably a BMW what María bought’ The derivation of (81) is entirely parallel to that of (78). In cycle I, F-marking on the DP- DO un BMW takes place. Then, copies of the non-focused units María and compró are made. The focused DP is also copied. All copies are put on hold. In cycle II, the copies of the non- focused units María and compró are merged with a wh-phrase, (unstressed) que, which occupies the same syntactic slot of the focused unit. This wh-phrase is then fronted, and the result is merged with the neuter determiner lo, so the wh-clause is created. In cycle III, an identity copula fue merges with the low clause, so the segment cop’ is created. Then the wh-clause merges with cop’, and this creates CopP, that is to say, the pseudo cleft. A copy of the copula is made, and the copy is also put on hold. Next, cycle IV starts by merging the copied copula fue with the copied DP-DO un BMW. PRO then is merged in the [Spec,CopP] position. Since the copula is of the idenitity type, the copy of the DP-DO gets focus marked. The truncated cleft is created. Then, the sentence adverb probablemente ‘probably’ merges with the truncated cleft. This 48 As Reeve (2012a) summarizes, there has been a debate about the nature of this pronoun. Roughly speaking, there are two sides on this debate. On the one hand, some authors consider it as a mere expletive, which makes no semantic contribution to the general meaning of the It-cleft (Chomsky 1977, Rochemont 1986, Kayne 1994, É. Kiss 1998, among many others). On the other hand, some other authors (among them, Percus 1997, Hedberg 2000, and Reeve himself) consider that the pronoun is not expletive, but referential. See Reeve (2012a: 10 ff) for extensive presentation of evidences in favor of the referential nature of it. 49 At this point of the development of my proposal, the issue of the nature of this pronoun for Spanish can be considered as work in progress. To my initial perception, however, this empty pronoun is referential, and it should be interpretively co-indexed with the wh-clause. Interestingly, for French (another Romance language), Reeve (2012a: 14 fn. 10) highlights an observation attributed to Kayne (1983): in this language, clefts are built with the pronoun ce, which is referential (originally, a demonstrative), not with the pronoun il, which is used in atmospheric sentences as a real expletive (Il pleut ‘it rains’). Compare C’est á Marie que Jean pense vs. *Il est á Marie que Jean pense ‘It is about Marie that Jean thinks’. 285 adverb is focus sensitive, so its [+foc] feature triggers reduplication of the truncated cleft. After reduplication, deletion applies following the LCA and the FDC, so the modal adverb ends up adjacent to the unit in focus. Finally, cycle V adjoins the truncated cleft to the pseudo-cleft, and deletes the low copies, according the LCA. The simplified syntactic tree is presented in (82). (82) CopP (cycle V) CopP (cycle IV) CopP (cycle III) CP (cycle II) Cop’ PRO fue c un BMW probablemente PRO fue c [[un BMW] c ] fcP fue CP (cycle I) lo que María compró un BMW María compró [un BMW] fcP In the same line of comparable observations presented in Akmajian (1970) for English, certain patterns about agreement in Spanish confirm the idea that the copula used in building the truncated cleft is indeed a copy of the copula used in building the pseudo-cleft. Let’s consider the PsCls in (83)-(85). (83) Quien compró el vino fui yo Who.SG buy.PRT.3SG the wine be.PRT.1SG I(1SG) ‘It was me who bought the wine’ 286 (84) Quien compró el vino fuiste tú Who.SG buy.PRT.3SG the wine be.PRT.2SG you(2SG) ‘It was you who bought the wine’ (85) Quien compró el vino fue él Who.SG buy.PRT.3SG the wine be.PRT.3SG he(3SG) ‘It was him who bought the wine’ As it is visible, the copula agrees in person with the focused subject of the low clause. On the other hand, the main verb in the wh-clause agrees in number (singular), but is invariably inflected in third person. The interpretively equivalent It-clefts 50 in Spanish are shown in (86)-(88), respectively. A parallel pattern of agreement arises. (86) PRO fui yo quien compró el vino 51 PRO be.PRT.1SG I(1SG) who.SG buy.PRT.3SG the wine ‘It was me who bought the wine’ (87) PRO fuiste tú quien compró el vino PRO be.PRT.2SG you(2SG) who.SG buy.PRT.3SG the wine ‘It was you who bought the wine’ (88) PRO fue él quien compró el vino PRO be.PRT.3SG he(3SG) who.SG buy.PRT.3SG the wine ‘It was him who bought the wine’ 50 I keep the label for the Spanish constructions to indicate their structural affinity with their counterparts in English. 51 It seems that, for some dialects of Spanish (not mine), the verb in the wh-clause can also agree with the copula and pronoun: Fui yo(1SG) quien compré(1SG) el vino; Fuiste tú(2SG) quien compraste(2SG) el vino. In PsCls, however, that variation does not occur, at least to my knowledge: *Quien compré(1SG) el vino fui yo(1SG). In any case, (86) and (87) correspond to general Spanish. 287 Now, when the focused subject is plural, the PsCls strictly demand person agreement in the three positions, namely the main verb in the wh-clause, 52 the copula, and the pronoun. (89) Quienes compramos el vino fuimos nosotros Who.PL buy.PRT.1PL the wine be.PRT.1PL we(1PL) ‘It was us that bought the wine’ (90) Quienes comprasteis el vino fuisteis vosotros 53 Who.PL buy.PRT.2PL the wine be.PRT.2PL you(2PL) ‘It was you(PL) that bought the wine’ (91) Quienes compraron el vino fueron ellos Who.PL buy.PRT.3PL the wine be.PRT.3PL they(3PL) ‘It was them that bought the wine’ The same strict demand of triple agreement is seen for the correspondent It-clefts. (92) PRO fuimos nosotros quienes compramos el vino PRO be.PRT.1PL we(1PL) who.PL buy.PRT.1PL the wine ‘It was us that bought the wine’ (93) PRO fuisteis vosotros quienes comprasteis el vino PRO be.PRT.2PL you(2PL) who.PL buy.PRT.2PL the wine ‘It was you(PL) that bought the wine’ (94) PRO fueron ellos quienes compraron el vino PRO be.PRT.3PL they(3PL) who.PL buy.PRT.3PL the wine ‘It was them that bought the wine’ 52 According to its function as subject of the wh-clause, the wh-word also agrees in number. 53 The pronoun vosotros and its associated inflexional forms are currently used in Peninsular Spanish only, which is not my native dialect. I thank Sergio Robles-Puente for his judgments about sentences (90) and (93). 288 Addtionally, in Spanish, the pronouns in the latter sentences, or their equivalent full DPs, can be placed in sentence-initial position. Those structures are known as inverted-PsCls. Consider, for instance, the sentences in (95), where the noun Pedro takes the place of the pronoun (to simplify the interpretation). (95) a. Quien compró el vino fue Pedro (PsCl) Who.SG buy.PRT.3SG the wine be.PRT.3SG P. ‘It was Pedro who bought the wine’ b. Pedro fue quien compró el vino (Inverted-PsCl) P. be.PRT.1SG who.SG buy.PRT.3SG the wine ‘Pedro was who bought the wine’ c. PRO fue Pedro quien compró el vino (It-cleft) PRO be.PRT.3SG P. who.SG buy.PRT.3SG the wine ‘It was Pedro who bought the wine’ Tentatively, I assume that inverted-PsCls are created in only two cycles. In cycle I, the wh-clause quien compró el vino is created. In cycle II, a copula is merged with the wh-clause, and then the pronoun is merged in [Spec,CopP]. This creates the inverted-PsCl. Interestingly, given the appropriate situational contexts, either the pre-copular or the post-copular units can be contrastively put in focus (in the case of the pre-copular noun, via prosodic prominence). This is impossible in the other two cases, whose focal structure is fixed: only the post-copular unit can be in focus in PsCls and It-clefts. This difference suggests that the copula in inverted-PsCls is not identificational, but is the empty e-COP. 54 In any case, what is important at this point is to distinguish inverted-PsCls from It-clefts. 55 54 In this sense, the inverted-PsCl would be somehow interpreted in the line of Pedro fue el comprador de vino ‘Pedro was the wine-buyer’, a predicational copular sentence. I leave this issue pending for future research. 55 Another difference can be seen in PsCls with the whole low clause in focus. In this case, the inverted-PsCl seems natural, whereas the It-cleft version seems odd. I have no explanation for this fact, for the time being. (i) Lo que pasó fue que el motor se rompió. ‘What happened was (that) the engine broke’ (PsCl) (ii) Que el motor se rompió fue lo que pasó. (Inverted-PsCl) (iii) ? Fue que el motor se rompió lo que pasó. (It-cleft) 289 In sum, PsCls and their correspondent It-clefts in Spanish share the same agreement patterns. This is consistent with the basic idea that Akmajian already put forth by the early 1970s, namely that the two types of construction are structurally related. While Akmajian (1970) postulated an extraposition rule to explain this relation, I propose instead the generation of a parallel construction with a fixed focal structure, namely the truncated cleft, which is adjoined to the PsCl by the left in order to create the It-cleft. The relevant interpretive readings are obtained by copying the relevant units in one given early cycle, and re-merging them in another later cycle. Deletion is responsible of the final linear order. The general structure for It-clefts in Spanish is presented in (96). (96) General Structure for It-clefts in Spanish CopP (cycle V) CopP (cycle IV) CopP (cycle III) CP (cycle II) Cop’ PRO SER c [X c ] fcP Cop 0 CP (cycle I) [[WhP] i [(Y*) [X] (Z*)] i ] SER [(Y*) [X] fcP (Z*)] i 290 Up to the node cycle III, this structure is parallel to the one proposed for PsCls in (67), with the important difference that, during the computation previous to cycle IV, a copy of the identity copula and also a copy of the focused unit are created and put on hold. Precisely, cycle IV takes back these copies from storage and manipulates them to generate the truncated cleft. The final cycle, cycle V, just adjoins the truncated cleft to the pseudo-cleft, and then proceeds to delete the low copies according to the LCA. Before moving to the case of PsCls in colloquial Peninsular Spanish, I would like to put forth a final disclaimer. The eventual success of the general structures in (67) for pseudo-clefts, and in (96) for It-clefts rely on the possibility of copying a certain syntactic unit and using it later, presumably in a different cycle, to build a certain sub-structure which is different from the ‘original’ one. As stated before, this idea is explicitly inspired by the proposal of sideward movement put forth by Nunes (2001, 2004). However, Nunes’ proposal cautiously restricts the applicability of sideward movement to the formation of chains, so to avoid overgeneration (Nunes, p.c.). That prudent attitude seems largely justified. Indeed, if a copy of any unit can be made, put on hold, and merged later in some higher position, islands would not exist, since whatever unit one would like to extract could be liberally copied, put on storage, and merged later in the desired position. However, as seen in the example (97c), this is not the case. Even if one can imagine that a copy what c can be made and put on hold during the cycle that creates the wh-clause, the eventual copy produced cannot be re-merged later in the high [Spec,CP]. It is an undeniable fact that islands exist, solidly, stably and cross-linguistically (Boeckx 2012). (97) a. John loves the woman who wears a black dress (Context) b. John loves the woman who wears what? (In situ wh-question) c. * What c does John loves the woman who wears what? (*wh-movement) These considerations lead us to the relevant question of what licenses (or, at least, seems to license) the sequence copy-storage-remerge in certain cases. This is, of course, a complex 291 question, which demands far deeper study. 56 I don’t claim I have a clear answer for it. However, for the time being, I would like to suggest that one possible source of solutions could be found in the information structure (IS). Note that the general derivations proposed are permeated by IS considerations. In some cases, the unit to be copied is the non-focused unit. In other cases, it is the unit in focus. Their landsites are also fixed by IS stipulations. In PsCls, copies of non- focused material can only be used in building the wh-clause that expresses the presupposition. In It-clefts, the copy of the focused unit can be used only to build the truncated cleft, and only in the [Compl,CopP] position. In sum, copies cannot be produced ad libitum. They must satisfy certain requirements established by IS. They cannot be placed anywhere either. Again, IS fixes beforehand one single potential landsite for each copied piece. If this seem too stipulative to the reader, I would not be surprised, since it is stipulative in a certain sense. Both pseudo-clefts and It-clefts are constructions with fixed focal structure. As Chomsky puts it 57 , constructions are artifacts of grammar, the result of a certain set of basic steps, carried out in certain fixed sequences. Under this lens, PsCls and It-clefts can be seen as objects made by following some sort of ‘instructions’ or ‘recipes’ that, in turn, exist to deliver structures that will result in clear and unambiguous expressions in IS terms. These ‘recipes’ must be part of the general linguistic knowledge of the speaker, which certainly includes grammar, but is not limited to it. This said, it is crucial to recall that what is being modeled here is not the external linguistic production, but the internal grammatical knowledge behind that production—indeed, one among many other factors influencing the externalized production. Thus these ‘instructions’ are not to be understood as what the speaker does when he or she utters a piece of discourse, but instead as what kind of knowledge he or she has internalized in order to produce a certain kind of sentences. 58 56 On this issue, I find very relevant the work by Saab (2008), a extensive study oriented to analyze the conditions that license ellipsis under identity (I thank J. Nunes for putting me in contact with Saab’s work). 57 See Chomsky’s quote on section 3.1. 58 To illustrate the difference, let’s take for instance the sentence in (97a), John loves the woman who wears a black dress. The external production (pronunciation) of this sentence starts with John and continues from left to right, until it reaches the final sound [s] of the word dress. In contrast, the syntactic derivation of the sentence, which represents the grammatical knowledge involved, will start by merging black and dress, and from there on it will 292 As for the basic operation Copy, it is as if UG provides us with a 3D-copy machine, which allows us to reduplicate the ‘Lego pieces’ we need to build certain fixed structures described in our ‘manual of instructions’. We are allowed use the 3D-copy machine, but not at our whim. Only certain marked pieces can be copied, and the copies produced can be placed only in certain pre-determined places. 5. Pseudo-clefts in colloquial Peninsular Spanish (CPS) Fernández Soriano (2009) reports the existence of a special kind of PsCls in colloquial Peninsular Spanish (CPS). In these PsCls, only a part of the sentence appears clefted in the wh- clause; however “the whole VP or even the whole sentence is interpreted as focus” (Fernández Soriano 2009: 90). A key to detect this discrepancy between what is clefted and what is interpreted as focus (besides the context) is the fact that the initial wh-phrase seems to have been reduced with respect to what one would have expected if the unit in focus were indeed the verbal complement only. Let’s start with standard Spanish. Suppose we have the PsCl in (98), where the whole low clause is in focus. (98) Lo que ocurre es que no tienes ganas de ir (Standard Sp. PsCl) Det.N.SG that happens is that NEG you.have desires of to.go ‘What happens is that you don’t feel like going’ In (98), the whole low CP que no tienes ganas de ir ‘that you don’t feel like going’ is in focus. This unit appears after the identificational copula es ‘is’, as expected. Additionally, as we have seen when presenting the general structure for PsCls in Spanish, the wh-clause contains a proform, namely the verb OCURRIR ‘to happen’, which appears there whenever the focused unit is the whole TP/CP. continue ‘upwards’ (from right to left) until John is finally merged in [Spec,TP]. In a certain sense, the ‘order’ of operations seems to have been reversed. In fact, the alluded ‘operations’ are different in nature. 293 Now, the particularity of the CPS-PsCl with respect to the standard PsCl is that, keeping the interpretation of (98), the corrrespondent CPS-PsCl is built with part of the focused CP in pre-copular position, without the complementizer and, crucially, without the proform OCURRIR in the wh-clause. Observe that the English gloss for (99) is the same gloss for (98). (99) Lo que no tienes ganas es de ir (CPS-PsCl) Det.N.SG that NEG you.have desire is of to.go ‘What happens is that you don’t feel like going’ (Fernández Soriano 2009: 90) The aforementioned ‘reduction’ with respect to the initial wh-phrase (de lo que → lo que) can be seen in (100), which shows the standard PsCl with focus indeed restricted to the post- verbal complement de ir. (100) De lo que no tienes ganas es de ir (Standard Sp. PsCl) Of Det.N.SG that NEG you.have desires is of to.go ‘What you don’t want to do is to go’ A similar case occurs with CPS-PsCls with focus on vP. As we have seen in the previous section, in standard Spanish whenever the unit in focus is vP, the wh-clause takes as main verb the proform HACER. In a fashion parallel to (99), CPS-PsCls with focus on vP also dispense with the proform, and also places part of the unit in focus in pre-copular position. (101) Lo que no puedes prorrumpir es en sollozos (CPS-PsCl) Det.N.SG that NEG you.can to.burst is into tears ‘What you cannot do is burst into tears’ (Fernández Soriano 2009: 90) Note that the interpretively equivalent PsCl in standard Spanish contains the proform HACER in the wh-clause. Also the vP prorrumpir en sollozos ‘burst into tears’, which is the unit in focus, appears after the copula. 294 (102) Lo que no puedes hacer es prorrumpir en sollozos (standard Sp. PsCl) Det.N.SG that NEG you.can do is to.burst into tears ‘What you cannot do is burst into tears’ (Fernández Soriano 2009: 90) Naturally, CPS-PsCls pose a challenge for the general structure postulated for PsCl in standard Spanish in the previous section. In the following lines, I will attempt to provide some insights about the issue from the point of view of the RSs and the CTM. In first place, I take the fact that the CPS-PsCl and the ‘equivalent’ standard PsCl have the same interpretation as an indication that the two structures share the same LF-structure. The relevant difference would be at PF-level only. I assume this as the null hypothesis. Indeed a LF/PF mismatch is also at the base of the analyses of this phenomenon proposed both in Fernández Soriano (2009) and Zubizarreta (2013) despite their different analytical perspectives. As before, I consider that the best way of accounting for the structure of a given construction is by looking at its derivational story. Let’s review first the derivation of the standard Sp. PsCl in (98), Lo que ocurre es que no tienes ganas es de ir ‘What happens is that you don’t feel like going’. (103) Cycle I: low-CP phase Sub-Numeration I = {que 1 , pro 1 , no 1 , tien-√ 1 , -es 1 , ganas 1 , de 1 , ir 1 } a. Usual derivations in syntax → [ CP que [ TP pro no tienes ganas de ir]] b. F-marking on CP → [ CP que [ TP pro no tienes ganas de ir]] fcP Note that the F-marking on CP in cycle I has a direct effect in the selection of items for the sub-numeration II, namely a proform OCURRIR ‘to occur’ (or any synonym) must be selected. 295 (104) Cycle II: wh-clause phase Sub-Numeration I = { C 0 1 , lo 1 , que 1 , ocurre 1 } 59 a. Merge que + ocurre → [ TP que ocurre] b. Merge C 0 + TP → [ CP C 0 [ TP que ocurre]] c. Move que to [Spec,CP] → [ CP que C 0 [ TP que ocurre]] d. Merge lo + CP → [ DP lo [ CP que C 0 [ TP que ocurre]]] (105) Cycle III: the PsCl Sub-Numeration III = {es 1 } a. Merge es + low-CP (103b) → [ Cop’ es [ CP que [ TP pro no tienes ganas de ir]] fcP ] b. Merge wh-clause (104d) + Cop’→ [ CopP [ DP lo [ CP que C 0 [ TP que ocurre]]] [ Cop’ es [ CP que [ TP pro no tienes ganas de ir]] fcP ]] Simplified version → = [ CopP [ DP lo [ CP que ocurre]] [ Cop’ es [ CP que pro no tienes ganas de ir] fcP ]] Linear version → = lo que ocurre es que no tienes ganas de ir Not having any reason to assume a mismatch between PF and LF, I assume that both PF and LF structures are represented in (105b). 59 To simplify, I present the proform OCURRIR ‘to occur’ in the sub-numeration as already inflected. Indeed, this verbal proform gets its TAM values from the low verb via agreement. 296 Let’s consider now the derivation of the CPS-PsCl in (99) Lo que no tienes ganas es de ir, with the same interpretation of (98), ‘What happens is that you don’t feel like going’. (106) Cycle I: low-CP phase Sub-Numeration I = {que 1 , pro 1 , no 1 , tien-√ 1 , -es 1 , ganas 1 , de 1 , ir 1 } a. Usual derivations in syntax → [ CP que [ TP pro no tienes ganas de ir]] b. F-marking on CP → [ CP que [ TP pro no tienes ganas de ir]] fcP Up to this point, both derivations for the standard PsCl and the CPS-PsCl are equivalent. Now this is the moment when the difference arises. The F-marking on CP will demand the selection of a proform in the sub-numeration II. For some reason, which is currently beyond my comprehension, the proform OCURRIR (or any similar) is not selected in this dialect. Facing this deficit, a different strategy applies. Instead of the verbal proform, the pre-verbal material plus the verb of the low clause is copied. In this case, the verb is an idiomatic expression: tener ganas, which functions as a single unit. These copies are used to build the wh-clause. For the case of (106), the set of copies is as follows. (107) Copies = {que c , pro c , no c , tienes-ganas c } The next cycle builds the wh-clause of the CPS-PsCl. The simplified derivation is as follows. (108) Cycle II: wh-clause phase Sub-Numeration I = {C 0 1 , lo 1 , que 1 } a. Merge tienes-ganas c + que → [ vP tienes-ganas c que] b. Merge no c + vP → 297 [ vP no c tienes-ganas c que] c. Merge pro c + vP → [ TP pro c [ vP no c tienes-ganas c que]] d. Merge C 0 + TP → [ CP C 0 [ TP pro c [ vP no c tienes-ganas c que]] e. Move que to [Spec,CP] → [ CP que C 0 [ TP pro c [ vP no c tienes-ganas c que]] f. Merge lo + CP → [ DP lo [ CP que C 0 [ TP pro c [ vP no c tienes-ganas c que]]] (109) Cycle III: the PsCl Sub-Numeration III = {es 1 } a. Merge es + low-CP (106b) → [ Cop’ es [ CP que [ TP pro no tienes ganas de ir]] fcP ] b. Merge wh-clause (108f) + Cop’→ [ DP lo [ CP que C 0 [ TP pro c [ vP no c tienes-ganas c que]]] [ Cop’ es [ CP que [ TP pro no tienes ganas de ir]] fcP ]] c. Deletion driven by the LCA → [ DP lo [ CP que C 0 [ TP pro c [ vP no c tienes-ganas c que]]] [ Cop’ es [ CP que [ TP pro no tienes ganas de ir]] fcP ]] 60 Simplified version → = [ CopP [ DP lo [ CP que no tienes ganas]] [ Cop’ es [ CP que pro no tienes ganas de ir] fcP ]] Linear version → = lo que no tienes ganas es de ir The derivation (106)-(109) delivers the PF-structure of (99). As for the LF-structure, starting in (109b), LF-deletion targets the high copies, instead of the lower ones. 60 Interestingly, the complementizer que in the lower clause gets deleted by phonetic identity with the accusative wh- word que in the higher wh-clause. 298 (110) Unmatched LF-deletion on the high wh-clause → [ DP lo k [ CPk que C 0 [ TP pro c [ vP no c tienes-ganas c que]]] [ Cop’ es [ CP que [ TP pro no tienes ganas de ir]] fcP ]] As in previous cases seen in this work, in (110) the variable is created via unmatched LF- deletion. Since the proform OCURRIR is not present in (110), the null hypothesis is not confirmed. In (105b), namely the LF-structure of the standard Sp. PsCl, the structure includes OCURRIR, whereas this is not the case in (110), the LF-structure of the equivalent CPS-PsCl. A similar comparison can be made for the case of examples (101) and (102). Let’s start with this latter, the standard Sp. PsCl Lo que no puedes hacer es prorrumpir en sollozos ‘What you cannot do is burst into tears’. (111) Cycle I: low-CP phase Sub-Numeration I = {C 0, , pro 1 , no 1 , pued-√ 1 , -es 1 , prorrumpir 1 , en 1 , sollozos 1 } a. Usual derivations in syntax → [ CP C 0 [ TP pro no puedes prorrumpir en sollozos]] b. F-marking on vP → [ CP C 0 [ TP pro no puedes [ vP prorrumpir en sollozos]]] c. Copy non-focused material → Copies = {pro c , no c , puedes c } (112) Cycle II: wh-clause phase Sub-Numeration I = {C 0 1 , lo 1 , que 1 , hacer 1 } a. Merge hacer + que → [ vP hacer que] b. Merge puedes c + vP → 299 [ vP puedes [ vP hacer que]] c. Merge no c + vP → [ vP no [ vP puedes [ vP hacer que]]] d. Merge pro c + vP → [ TP pro c [ vP no [ vP puedes [ vP hacer que]]]] e. Merge C 0 + TP → [ CP C 0 [ TP pro c [ vP no [ vP puedes [ vP hacer que]]]]] f. Move que to [Spec,CP] → [ CP que C 0 [ TP pro c [ vP no [ vP puedes [ vP hacer que]]]]] g. Merge lo + CP → [ DP lo [ CP que C 0 [ TP pro c [ vP no [ vP puedes [ vP hacer que]]]]]] (113) Cycle III: the PsCl Sub-Numeration III = {es 1 } a. Merge es + low-CP (111b) → [ Cop’ es [ CP C 0 [ TP pro no puedes [ vP prorrumpir en sollozos]]]] b. Merge wh-clause (112g) + Cop’→ [ CopP [ DP lo [ CP que C 0 [ TP pro c [ vP no [ vP puedes [ vP hacer que]]]]]] [ Cop’ es [ CP C 0 [ TP pro no puedes [ vP prorrumpir en sollozos]]]] c. Deletion driven by the LCA → [ CopP [ DP lo [ CP que C 0 [ TP pro c [ vP no [ vP puedes [ vP hacer que]]]]]] [ Cop’ es [ CP C 0 [ TP pro no puedes [ vP prorrumpir en sollozos]]]] Simplified version → = [ CopP [ DP lo [ CP que no puedes hacer]] [ Cop’ es [ CP pro [ vP prorrumpir en sollozos] fcP ]]] Linear version → = Lo que no puedes hacer es prorrumpir en sollozos 300 Now, the derivation for the equivalent CPS-PsCl would be as follows. (114) Cycle I: low-CP phase Sub-Numeration I = {C 0, , pro 1 , no 1 , pued-√ 1 , -es 1 , prorrumpir 1 , en 1 , sollozos 1 } a. Usual derivations in syntax → [ CP C 0 [ TP pro no puedes prorrumpir en sollozos]] b. F-marking on vP → [ CP C 0 [ TP pro no puedes [ vP prorrumpir en sollozos] fcP ]] c. Copy non-focused material → Copies = {pro c , no c , puedes c } The F-marking on vP in (114b) indicates that a proform HACER must be retrieved from the lexicon in the following cycle. But then again, this dialect doesn’t select for verbal proforms. The alternative strategy is to copy the verb from the low clause, even if it is part of the unit in focus. A copy of prorrumpir c is added to the set of available copies. (115) Copies’ = {pro c , no c , puedes c , prorrumpir c } (116) Cycle II: wh-clause phase Sub-Numeration I = {C 0 1 , lo 1 , que 1 } a. Merge prorrumpir c + que → [ vP prorrumpir c que] b. Merge puedes c + vP → [ vP puedes c [ vP prorrumpir c que]] c. Merge no c + vP → [ vP no c [ vP puedes c [ vP prorrumpir c que]]] d. Merge pro c + vP → [ TP pro c [ vP no c [ vP puedes c [ vP prorrumpir c que]]]] e. Merge C 0 + TP → 301 [ CP C 0 [ TP pro c [ vP no c [ vP puedes c [ vP prorrumpir c que]]]]] f. Move que to [Spec,CP] → [ CP que C 0 [ TP pro c [ vP no c [ vP puedes c [ vP prorrumpir c que]]]]] g. Merge lo + CP → [ DP lo [ CP que C 0 [ TP pro c [ vP no c [ vP puedes c [ vP prorrumpir c que]]]]]] Finally, cycle III creates the CPS-PsCl. (117) Cycle III: the CPS-PsCl Sub-Numeration III = {es 1 } a. Merge es + low-CP (114b) → [ Cop’ es [ CP C 0 [ TP pro no puedes [ vP prorrumpir en sollozos] fcP ]]] b. Merge wh-clause (116g) + Cop’→ [ DP lo [ CP que C 0 [ TP pro c [ vP no c [ vP puedes c [ vP prorrumpir c que]]]]]] [ Cop’ es [ CP C 0 [ TP pro no puedes [ vP prorrumpir en sollozos] fcP ]]] c. Deletion driven by the LCA → [ DP lo [ CP que C 0 [ TP pro c [ vP no c [ vP puedes c [ vP prorrumpir c que]]]]]] [ Cop’ es [ CP C 0 [ TP pro no puedes [ vP prorrumpir en sollozos] fcP ]]] Simplified version → = [ CopP [ DP lo [ CP que no puedes prorrumpir]] [ Cop’ es [ CP pro no puedes prorrumpir en sollozos] fcP ]] Linear version → = Lo que no puedes prorrumpir es en sollozos Note that deletion in (117c) contains a violation of the FDC, the constraint militating against deletion of the focused material. The copy of prorrumpir that gets deleted is the lower one. Being the FDC a high ranked constraint in standard Spanish, one must suppose that, in this dialect, there is an even higher ranked constraint in CPS authorizing this evident violation of the FDC. Indeed, the reason seems to be rather simple: if the high copy of prorrumpir were deleted, there would be no verb in the high clause, so no clause could be formed. 302 As for the LF-structure, deletion is unmatched, as expected. (118) Unmatched LF-deletion → [ DP lo k [ CP que C 0 [ TP pro c [ vP no c [ vP puedes c [ vPk prorrumpir c que]]]]]] [ Cop’ es [ CP C 0 [ TP pro no puedes [ vP prorrumpir en sollozos] fcP ]]] For both examples of CPS-PsCl in (99) and (101), taken from Fernández Soriano (2009), Zubizarreta (2013) already offers comparable analyses that crucially involve a bi-clausal structure and unmatched PF/LF deletion. In any case, the derivations shown here indicate that a key particularity of the CPS is its choice of not retrieving the correspondent verbal proforms represented by OCURRIR and HACER whenever the unit in focus is CP and vP, respectively. Facing this, CPS opts for copying the low verb. This copy is added to the ‘regular’ (standard) set of copies, and it is used later in cycle II as the main verb of the wh-clause. Interestingly, some data reported by Fernández Soriano (2009) indicates that the copy process stops at the verb, so no lexical post-verbal material gets copied. Consider the following contrast. (119) a. Lo que tienes que dar es el regalo a tu hermana Det.N.3SG that you.have to give is the present to your sister ‘What you have to do is give the present to your sister’ b. * Lo que le tienes que dar el regalo es a tu hermana 61 Det.N.3SG that Cl=you.have to give the present is to your sister ‘It is to your sister that you have to give the present’ (Fernández Soriano 2009: 98) 61 An interesting contrast arises with respect to the copula FS in Caribbean Spanish, where a sentence like Le tienes que dar el regalo es a tu hermana, with the interpretation shown in (119b), is expected to be grammatical. 303 From the point of view of the analysis presented here, the ungrammaticality of (119b) can be attributed to the fact that no copies of post-verbal material are allowed in CPS, so the DO el regalo ‘the present’ cannot be copied in the high wh-clause, and thus it cannot surface in pre- copular position. This ban is consistent with the idea that the copy of the low verb is an alternative strategy used in CPS to cover the lack of verbal proforms OCURRIR and HACER at use in standard Spanish. Once the verb is copied, no more copies are needed, nor tolerated. 62 Going back to the metaphor of a ‘manual of instructions’, it seems that CPS and standard Spanish have slightly different ‘editions’ of that book. In the latter, the user builds the PsCl using proforms whenever the F-marked unit is vP or CP (i.e. phases). In the former, there are no proforms, so the user needs to copy the non-focused material and also the verb, even if this latter is included within the focus marked unit. Another difference arises at the step where deletion applies. While standard Spanish strictly verifies the FDC, CPS deletes focus marked material on the low clause—otherwise, there would be no verb in the higher clause, so rigorously speaking, there would be no clause. Our assumption so far has been that the results of those two different editions are somehow equivalent. Naturally, this does not mean they are identical. Consider, for instance, the CPS-PsCl in (119a), Lo que tienes que dar es el regalo a tu hermana. As is, without a situational context, the sentence could be interpreted (by speakers of standard Sp.) as taking only the DO el regalo ‘the present’ as the focused unit. 63 But we know this is not the intended reading in CPS. Abstracting away from prosody, there is some sort of ambiguity in (119a), although the alternative interpretations correspond to different speakers: (i) focus on vP for CPS speakers, vs. (ii) focus on DO for non-CPS speakers. As far as I can tell, this sort ‘interdialectal ambiguity’ is not the case for the standard Sp. PsCl. Standard Sp. PsCls unambiguously mark the post-copular unit as the unit in focus. 64 62 Naturally, the idea of ‘lack’ of proforms does not imply any valorative judgment whatsoever with respect to the CPS as a dialectal variety. All I say is that this dialect uses a different strategy with respect to the one used in standard Spanish. On the other hand, I am not saying that OCURRIR and HACER don’t exist in CPS. What I say is that, for some reason—still obscure to me—they are not used as proforms to build PsCls. 63 In this case, since it is not in final position, the DO el regalo ‘the present’ may carry some form of prosodic prominence. Complementarily, the IO a tu hermana ‘to your sister’ is likely to be pronounced either deaccented or with a low-flat F0 contour. 64 That’s why this conventional variety is considered standard, in the first place. Supposedly, a standard variety aspires to be intelligible for all natural dialects of a given language—or, at least, as much dialects as possible. 304 This said, it comes as no surprise that CPS-PsCls are frequently used with idiomatic expressions, as reported by Fernández Soriano (2009). The fact that those idioms exist beyond the PsCls as part of the speakers’ lexical knowledge ensures that that those expressions will be understood in its integrity, unambiguously, as a unit in focus, in spite of appearing divided by the copula. Consider the following verbal idioms in Spanish. (120) Idiom in Spanish Interpretation Equivalent in English a. Meter la pata ‘Make a mistake or a mess’ Put one’s foot in it b. Costar un riñón ‘Cost too much’ Cost a fortune c. Dar la tabarra ‘Annoy’ Be a nuisance Now let’s consider the following examples offered in Fernández Soriano (2009: 95). I underline the parts corresponding to the idioms and write the copula in boldaface (both in Spanish and in the English equivalent), for illustrative purposes. (121) Lo que no puedes meter es la pata hasta el punto de que te echen. What NEG you.can put is the foot to the point of that Cl.ACC.2SG=they.fire ‘What you cannot do is put your foot in it until you get fired’ (122) No es que no me guste; lo que cuesta es un riñón. NEG is that NEG Cl=like; what costs is a kidney ‘It is not that I don’t like it; what happens is that it costs a fortune’ (123) Lo que tienes que seguir dando es la tabarra. What you.have to keep.on giving is the pain ‘What you have to do is keep on being a nuisance’ The underlining in these examples shows that the idiom appears divided by the copula in CPS. Importantly, only the verb appears as pre-copular. The English gloss presents the equivalent idiom in its integrity in post-copular position. This indicates that the idiom as a whole is in focus, and this interpretation is the same both in English and Spanish. The fact that the low 305 verb appears in pre-copular position in each case is due to the fact that CPS doesn’t resort to proforms, so these verbs must be copied from the low clause, merged in the wh-clause, and then deleted in the low clause. Thus the idiom appears divided at PF, but is understood as a chunk at LF. While Fernández Soriano (2009) proposes an explanation in terms of reconstruction, my proposal links this phenomenon to a PF/LF mismatch due to the lack of proforms in CPS. This lack is not seen as a deficit of CPS with respect to standard Spanish, but only as a different strategy encoded in the internalized knowledge of CPS speakers. In terms of Chomsky (1986, 1988), this would be a case of parametric variation. No matter the dialect, it seems that all Spanish speakers who have reached the steady state of linguistic knowledge are equipped with a set of ‘DIY-instructions’ about how to build certain constructions (PsCls, It-clefts, and the like), oriented to satisfy certain demands from the interface systems. 6. Summary In the first section of this chapter, I have proposed an RS analysis for Focalizing SER (FS) in Caribbean Spanish. From this point of view, the bare copula is seen simultaneously as a defective tensed focus marker, and as a vP-level focus-sensitive unit (FSU). As for its FSU nature, the copula is loaded with a [+foc] feature such that it triggers reduplication when a unit in its c-command domain gets F-marked. As for its defective tense character, the copula takes its values from the matrix verb via Agree. The usual operations of deletion apply on the bi-vP structure in such a way that, at PF, the copula surfaces immediately adjacent to the focused unit. In turn, LF-deletion applies on the high copy, and this creates a variable that is bound by a silent operator, so the precopular material conveys the presupposed content. If this view is on the right track, it would provide a unified analysis for this dialectal phenomenon (FS) and for other vP- level FSUs in standard Spanish, like sólo ‘only’, también ‘also’, or incluso ‘even’. In section 2, I propose to differentiate predicational copular sentences, on the one hand, from specificational and equative copular sentences, on the other hand. I proposed that, while predicational sentences are built with a semantically empty copula (e-COP), specificational and equative sentences are built with a copula of a different semantic type (namely <e, <e, t>>). I 306 call this latter identificational copula (id-COP). Unlike predicational sentences, which allow for different possibilities of focus-marking, in Spanish sentences built with id-COP have a fixed focal structure, where the post-copular unit is necessarily in focus. In this sense, identificational sentences (specificational and equatives) in Spanish grammaticalize the divide between focus and presupposition. The copula (id-COP) in these constructions can be interpreted both as a verb and as a focus marker, in a way somehow similar to what has been reported for copulas in Hausa and other African languages. In section 3, I deal with the structure of pseudo-clefts (PsCls). From the beginning, I assume the bi-clausal proposals that interpret the post-copular part of the PsCl as a clause, usually concealed via ellipsis. From this starting point, I focus my attention on the identity between what is interpreted as deleted in the low clause and what surfaces as lexical content in the high (wh-)clause. While it is not impossible that the aforementioned identity is the result of double selecting of (the same) items directly from the lexicon, I propose that what is at stake is a process of Copy of non-focused elements in the low clause, followed by Merge of those copies in the process that builds the high clause. This idea is adapted from the proposal of sideward movement, by which an element used in a sub-tree can be copied and re-merged in a higher sub- tree, in such a way that the ‘original’ and the copy end up by forming a chain. Later, the operation Chain Reduction applies to the chain to deliver the perceived linear order. Although in the case of PsCls the copies and their ‘originals’ do not form a chain, I claim that deletion of copies is still guaranteed via the LCA. On the other hand, the risk of overgeneration may be moderated (or even cancelled) by considering that IS poses strict requisites with respect to what can be copied and where the copies can merge. With this in mind, a derivation of PsCls involving three cycles is proposed. In the first cycle, the basic clause is created and F-marking takes place. The relevant copies of non-focused material are put on storage in some side of the syntactic workplace. Then, in cycle II, the relevant copies taken in cycle I are merged with a wh- phrase that corresponds to the postion occupied by the unit in focus in the lower clause. In case the units focused in the low clause are vP or the whole CP (i.e. phases), no copy is involved, but rather a proform (HACER ‘to do’ or OCURRIR ‘to occur’) must be pulled out from the lexicon and used as the main verb of the high clause. Later, wh-movement occurs, and this creates the high clause. Finally, in cycle III, an identificational copula links the two clauses, and then deletion 307 procedes according to the LCA. This provides the reasons behind the ellipsis process already postulated by the bi-clausal approaches to PsCls. Now if we are allowed to base generate a wh- unit directly in [Spec,CP], an alternative becomes available. The lower clause created in cycle I can be copied as a whole and merged in cycle II with the wh-unit. Deletion in the higher clause creates a variable. The wh-unit binds the variable, and so the presuppositional reading is obtained. In section 4, I provide an account of It-clefts. I start by assuming that these constructions are structurally derived from PsCls. Beyond the similarity with respect to their interpretation, some common formal characteristics seem to support this claim. For the case of Spanish, a similar pattern of agreement has been shown. In this vein, I assume that the It-cleft is built in a derivation with five cycles. The first three cycles are the same that have been postulated for PsCl, with the important difference that a copy of the unit in focus and a copy of the copula are made, and put on storage. Precisely, in cycle IV, the copy of the copula and the copy of the unit in focus are merged. In turn, this segment is merged with It for the case of English, or PRO for the case of Spanish. The syntactic object created in cycle IV has been labeled as truncated cleft. Cycle V, finally, adjoins the truncated cleft to the pseudo-cleft produced in the first three cycles. Finally, deletion applies following the LCA. As a result, the It-cleft is formed. Importantly, if this proposal is on the right track, it provides a strong structural argument against the common (but naïve) idea that sentence modal adverbs can simply merge with any sub-clausal constituent, of any kind, and that combination can later be fronted to the focus position of the It-cleft. In the last section of this chapter, I offer an analysis of pseudo-clefts in colloquial Peninsular Spanish (CPS). I have shown that these constructions follow a different pattern with respect to PsCls in standard Spanish. In standard Spanish, PsCls with focus on vP or CP in the low clause demand the use of a proform HACER ‘to do’ or OCURRIR ‘to occur’, respectively, as the main verb of the wh-clause. In CPS, no such use of proforms is observed. Instead, the alternative strategy consists of copying the low verb and using that copy to build the high clause. Against the FDC, this pre-copular copy survives deletion for structural reasons. As a result, part of the unit in focus surfaces linearly before the copula, which does not occur in standard Spanish. Interestingly, the dialectal structure is frequently used with idiomatic expressions containing a 308 verb. Since these expressions are part of the lexical knowledge of the speakers, they are easily interpreted as a unit, despite the fact that they appear separated by the copula—the verb as pre- copular and its complements as post-copular. 309 Chapter 6 Conclusions and some projections for future research As for the conclusions of this work, I have confirmed that an analysis in which sentence modal adverbs in Spanish are always propositional in scope is possible. This kind of analysis presents some advantages with respect to alternative sub-clausal approaches, mainly from semantic grounds. Syntactically, the immediate adjacency between the sentence modal adverb and the unit in focus can be explained via a process of reduplication triggered by sensitivity to focus. This process copies the clause locaetd in [Compl,FSUP] and merges this copy in [Spec,FSUP], which in turn yields the reduplicative structure (RS). This bi-clausal structure is the final product of syntax, and from that point, it is shipped from there to the interpretive components PF and LF. A later process of complementary ellipsis, driven by linearization purposes and constrained by the focus marking on the lower copy occurs at PF, whereas at LF a variable is created on the higher copy. I have suggested that, in this case, ellipsis can be understood as deletion, so that a mismatch may occur between PF-deletion and LF-deletion in certain cases (as, for instance, association with focus at a distance). The RS grammaticalizes the divide between focus and background by locating the main assertion—which contains the unit in focus—in [Compl, FSUP] and the presupposition—which corresponds to the background—in [Spec, FSUP]. In the RS, the presupposition always contains a variable bound by an operator, either overt or empty. The actual focus assigns a value to that variable. Beyond sentence modal adverbs, the syntax of other units that have been considered as focus sensitive units can also be represented by the RS. I have shown that this is the case for the adverbs sólo ‘only’, también ‘also’, and incluso ‘even’ in Spanish, and also for some other high sentence adverbs ending in –mente ‘-ly’. The RS seem to be useful to explain negation and certain cases of marked word order in Spanish. 310 The generalized RS may also give insights to represent the syntax of some focus-related constructions in Spanish, such as copular focal constructions and pseudo-clefts. This idea has been applied to some dialectal constructions in this language, namely Focalizing-SER in Caribbean Spanish and pseudo-clefts in Peninsular Colloquial Spanish. As for the projections for future research, on the one hand, a comprehensive prosodic approach to the RS remains pending. On the other hand, it would be necessary to evaluate whether focus fronting, a kind of focus movement licensed by contrastive focus, can also be represented via some sort of reduplicative model, which resorts to the basic operations Copy, Merge, and Delete only. 311 References Abels, Klaus. 2008. Towards a restrictive theory of (remnant) movement. In van Craenenbroeck, Jeroen (ed.). Linguistic variation yearbook 2007. Amsterdam/Philadelphia: John Benjamins, pp. 53-120. Aguilar, Lourdes et al. (coords.). 2009. Sp_ToBI Training materials. Online: http://prosodia.upf.edu/sp_tobi/en/ Akmajian, Adrian. 1970. On deriving cleft sentences from pseudo-cleft sentences. Linguistic Inquiry 1, No. 2, pp. 149-168. Alarcos, Emilio. 1999. Gramática de la lengua española. Madrid: Espasa. Alexiadou, Artemis. 1997. Adverb Placement. A Case Study in Antisymmetric Syntax. Amsterdam/Philadelphia: John Benjamins. Bellert, Irena. 1977. On semantic and distributional properties of sentential adverbs. Linguistic Inquiry 8, pp. 377-351. Belletti, Adriana. 2004. Aspects of the low IP area. In Rizzi, Luigi (ed.). The Structure of CP and IP. Oxford/ New York: Oxford University Press, pp. 16–51. Bhatia, Archna. n.d. Testing Cinque's hierarchy: Adverb Placement in Hindi. Ms. Urbana- Champaign: University of Illinois at Urbana- Champaign. Online: http://vanhise.lss.wisc.edu/ling/files/ling_old_web/lso/wpl/6/bhatia.pdf Boeckx, Cedric. 2006a. Linguistic minimalism. Origins, concepts, methods, and aims. Oxford: Oxford University Press. Boeckx, Cedric (ed.). 2006b. Minimalist essays. Amsterdam/Philadelphia: John Benjamins. Boeckx, Cedric. 2008. Understanding minimalist syntax. Lessons from locality in long distance dependencies. Malden, Mass.: Blackwell. Boeckx, Cedric. 2012. Syntactic islands. Cambridge: Cambridge University Press. Boer, Steven & William Lycan. 1976. The myth of semantic presupposition. Bloomington: Indiana University Linguistics Club. Bošković, Željko. 2001. On the nature of the syntax-phonology interface. Clitizacion and related phenomena. Amsterdam: Elsevier Science. 312 Bošković, Željko & Jairo Nunes. 2007. The copy theory of movement. A view from PF. In Corver, Norbert & Jairo Nunes (eds.). The copy theory of movement. Amsterdam/Philadelphia: John Benjamins, pp. 13-74. Bosque, Ignacio. 1999. Sobre la estructura sintáctica de una construcción focalizadora. BFUCh (Boletín de Filología de la Universidad de Chile) XXXVII - Homenaje al profesor Ambrosio Rabanales, pp. 207-231. Boxus, Marie-Agnes. 2006. Considerations sur les constructions pseudo-clivées en français, portugais et anglais. Revista da Facultade de Letras - Linguas e literatures II Série, Vol. XXIII, pp. 433-460. Bresnan, Joan. 1971. Sentence stress and syntactic transformations. Language 47, pp. 257-281. Cable, Seth. 2012. Pied-piping. Introducing two recent approaches. Language and linguistic compass, Vol. 6, Issue 12, pp. 816-832. Camacho, José. 2006. In Situ focus in Caribbean Spanish: Towards a unified account of focus. In Sagarra, Nuria & Almeida Jacqueline Toribio (eds.). Proceedings of the 9th Hispanic Linguistics Symposium, 13-23, Somerville, Ma: Cascadilla Proceedings Project. Casielles-Suárez, Eugenia. 1997. Topic, focus, and bare nominals in Spanish. PhD dissertation. Amherst: University of Massachusetts. Online: http://scholarworks.umass.edu/dissertations/AAI9737510/ Casielles-Suárez, Eugenia. 2001. The syntax and semantics of preverbal topical phrases in Spanish. In Gutiérrez-Rexach, Javier & Luis Silva-Villar (eds.). Current Issues in Spanish syntax and semantics. Berlin/New York: Mouton de Gruyter, pp. 65-82. Casielles-Suárez, Eugenia. 2003. On the interaction between syntactic and information structures in Spanish. Bulletin of Hispanic Studies 80, pp. 1-20. Casielles-Suárez, Eugenia. 2004. The syntax-information structure interface. Evidence from Spanish and English. New York/London: Routledge. Chierchia, Gennaro & Sally McConnell-Ginet. 2000. Meaning and grammar. An introduction to semantics. 2nd ed. Cambridge, Mass.: The MIT Press. Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1977. On wh-movement. In Culicover, Peter et al. (eds.). Formal syntax. New York: Academic Press, pp. 71-132. Chomsky, Noam. 1986. Knowledge of language. Its nature, origin, and use. New York: Praeger. 313 Chomsky, Noam. 1988 [1981]. Lectures on government and binding. The Pisa lectures. 5 th ed. Dordrecht: Foris. Chomsky, Noam. 1993. A minimalist program for linguistic theory. In Hale, Kenneth & Samuel J. Keyser (eds.). The view from Building 20. Essays in Linguistics in honor of Sylvain Bromberg. Cambridge, Mass.: The MIT Press, pp. 1-52. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: The MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: the framework. In Martin, Roger et al. (eds.). Step by Step. Essays on minimalist syntax in honor of Howard Lasnik. Cambridge, Mass.: The MIT Press, pp. 89-155. Chomsky, Noam. 2001. Derivation by phase. In Kenstowicz, Michael (ed.). Ken Hale. A life in language. Cambridge, Mass.: The MIT Press, pp.1-52. Chomsky, Noam. 2002 [1957]. Syntactic structures. 2 nd ed. Berlin: Mouton de Gruyter. Chomsky, Noam. 2004. Beyond explanatory adequacy. In Belletti, Adriana (ed.). Structures and beyond. The cartography of syntactic structures. Vol III. Oxford: Oxford University Press, pp. 104-131. Chomsky, Noam. 2008. On phases. In Freidin, Robert et al. (eds.). Foundational issues in Linguistic Theory. Essays in honor of Jean-Roger Vergnaud. Cambridge, Mass.: The MIT Press, pp. 133-166. Chomsky, Noam, et al. 2002. On nature and language. Cambridge, UK.: Cambridge University Press. Chomsky, Noam & Howard Lasnik. 1995. The theory of Principles and Parameters. In Chomsky, Noam. The Minimalist Program. Cambridge, Mass.: The MIT Press, pp. 13-127. Chomsky, Noam & Morris Halle. 1968. The sound pattern of English. New York: Harper and Row. Chuaqui Farrú, Claudia. 2006. Dos problemas de sintaxis adverbial. Onomázein 13, pp. 127-134. Cinque, Guglielmo. 1990. Types of A-bar dependencies. Cambridge, Mass.: The MIT Press. Cinque, Guglielmo. 1999. Adverbs and functional heads. A cross-linguistic perspective. New York/Oxford: Oxford University Press. Collins, Peter. 2006. It-clefts and Wh-clefts: Prosody and pragmatics. Journal of Pragmatics 38, pp. 1706-1720. 314 Comrie, Bernard. 1989. Language universals and linguistic typology. Syntax and morphology. 2 nd . ed. Chicago: The University of Chicago Press. Contreras, Heles. 1991. On the position of subjects. Syntax and semantics 25, pp. 63-79. Corver, Norbert & Jairo Nunes (eds.). 2007. The copy theory of movement. Amsterdam/ Philadelphia: John Benjamins. Declerck, Renaat. 1984. The pragmatics of it-clefts and wh-clefts. Lingua 64, pp. 251-289. Dehé, Nicole & Yordanka Kavalova. 2007. Parentheticals. Amsterdam/Philadelphia: John Benjamins. Delfitto, Denis. 2006. Adverb classes and adverb placement. In Everaert, Martin & Henk van Riemsdijk (eds.). The Blackwell companion to syntax. Malden, MA: Blackwell, pp. 83- 120. den Besten, Hans & Gert Webelhuth. 1990. Stranding. In Grewendorf, Günther & Wolfgang Sternefeld (eds.). Scrambling and barriers. Amsterdam/Philadelphia: John Benjamins, pp. 77-92. den Dikken, Marcel. 2005. A comment on the topic of topic-comment. Lingua 115, pp. 691-710. den Dikken, Marcel. 2006. Relators and linkers. The syntax of predication, predicate inversion, and copulas. Cambridge, Mass.: The MIT Press. den Dikken, Marcel, et al. 2000. Pseudoclefts and ellipsis. Studia Linguistica 54, pp. 41-89. Dolci, Roberto. 1986. Algunas construcciones con anteposición de constituyentes oracionales en español. Su determinación y análisis sintáctico. Degree thesis. Venice: Ca' Foscari University of Venice. É. Kiss, Katalin. 1998. Identificational focus vs. information focus. Language 74 (2), pp. 245- 273. Eguren, Luis & Olga Fernández Soriano. 2004. Introducción a una sintaxis minimista. Madrid: Gredos. Epstein, Samuel & Norbert Horstein (eds.). 1999. Working minimalism. Cambridge, Mass.: The MIT Press. Erlewine, Michael. To appear. Explaining leftward focus association with even but not only. (To appear in Sinn und Bedeutung 18). Ms. Cambridge, Mass.: MIT. Erlewine, Michael & Hadas Kotek. To appear. Intervention in focus pied-piping. (To appear in Proceedings of NELS 43). Ms. Cambridge, Mass.: MIT. 315 Estebas Vilaplana, Eva & Pilar Prieto. 2009. La notación prosódica del español. Una revisión del Sp_ToBI. Estudios de Fonética Experimental XVIII, pp. 263-283. Fábregas, Antonio. 2003. Adverbios en –mente y la estructura del adjetivo en español. ELUA (Estudios de Lingüística de la Universidad de Alicante) 17, pp. 1-26. Face, Timothy. 2002. Local intonational marking of Spanish contrastive focus. Probus 14, pp. 71-92. Fanselow, Gisbert & Anoop Mahajan. 1995. Partial movement and successive cyclicity. In Lutz, Uli & Gereon Müller (eds.). Papers on wh-scope marking. Arbeitspapiere des Sonderforschungsbereichs 350, Sprachtheoretische Grundlagen für die Computer Linguistik, Bericht Nr. 76. University of Stuttgart/University of Tübingen, pp. 131-161. Fanselow, Gisbert & Damir Ćavar. 2002. Distributed deletion. In Alexiadou, Artemis (ed.). Theoretical approaches to universals. Amsterdam/Philadelphia: John Benjamins, pp. 65- 107. Fernández Soriano, Olga. 2009. On the nature of covert operations: The case of focus in Spanish pseudo-clefts. In Aboh, E.O. et al. (eds.). Romance Languages and Linguistic Theory. Selected Papers from Going Romance 87-108. Amsterdam/ Philadelphia: John Benjamins. Frascarelli, Mara. 2010. Narrow focus, clefting and predicate inversion. Lingua 120, pp. 2121- 2147. Frascarelli, Mara & Francesca Ramaglia. 2009. (Pseudo)cleft constructions at the interfaces. Ms. Roma: University of Roma Tre. Online: http://ling.auf.net/lingbuzz/000841/current.pdf Freidin, Robert & Jean-Roger Vergnaud. 2001. Exquisite connections: some remarks on the evolution of linguistic theory. Lingua 111, pp. 639-666. Giannakidou, Anastasia. 2011. Negative and positive polarity items. In Maeinborn, Claudia et al. (eds.). Semantics. An international handbook of natural language meaning. Vol. 2. Berlin/Boston: De Gruyter Mouton, pp. 1660-1712. Giorgi, Alessandra. 2013. Epistemic adverbs, the theory of phases and the prosody-syntax interface. Linguistic Symposium on Romance Languages (LSRL) at CUNY, April 2013. Handout. Abstract online: http://lsrl43.commons.gc.cuny.edu/files/2012/09/giorgi.pdf 316 Giorgi, Alessandra. To appear. Prosodic signals as syntactic formatives in the left periphery. Online: http://ling.auf.net/lingbuzz/001625) Green, Melanie. 2004. Predication, equation and information structure: Evidence from Hausa copular sentences. Ms. Brighton: University of Sussex. Online: http://www.sussex.ac.uk/Users/melanieg/Cop.pdf Grohmann, Kleanthes K. 2003. Prolific domains. On the anti-locality of movement dependencies. Amsterdam/Philadelphia: John Benjamins. Guitart, Jorge M. 2013. Del uso de las oraciones hendidas en el español actual. Revista Internacional d’Humanitats 27, pp. 89-104. Halle, Morris & Alec Marantz. 1993. Distributed morphology and the pieces of inflection. In Hale, Kenneth & Samuel J. Keyser (eds.). The view from Building 20. Essays in Linguistics in honor of Sylvain Bromberg. Cambridge, Mass.: The MIT Press, pp. 111- 176. Han, Chung-hye & Maribel Romero. 2004. Disjunction, focus, and scope. Linguistic Inquiry, Vol. 35 No. 2, pp. 179-217. Hedberg, Nancy. 2000. On the referential status of clefts. Language 76 (4), pp. 891-920. Heim, Irene & Angelika Kratzer. 1998. Semantics in generative grammar. Malden, Mass./ Oxford: Blackwell. Herburger, Elena. 2000. What counts. Focus and quantification. Cambridge, Mass.: The MIT Press. Hernanz, Ma. Lluïsa & José Ma. Brucart. 1987. La sintaxis. Barcelona: Crítica. Higashimori, Isao. 1978. On syntactic, semantic and pragmatic properties of possibly as a sentence adverb. Osaka Literary Review 17. 12-27. Online: http://hdl.handle.net/11094/25664 Hornstein, Norbert. 2001. Move! A minimalist theory of construal. Malden, Mass.: Blackwell. Hornstein, Norbert. 2009. A theory of syntax. Minimal operations and Universal Grammar. Cambridge, UK: Cambridge University Press. Hornstein, Norbert et al. 2005. Understanding minimalism. Cambridge, UK: Cambridge University Press. 317 Huber, Stefan. 2006. The complex functions of it-clefts. In Molnár, Valéria & Susanne Winkler (eds.). 2006. The architecture of focus. Berlin/New York: Mouton de Gruyter, pp. 549- 578. Hunter, Tim. 2012. A constraint on remnant movement. In Di Sciullo, Anna Maria (ed.). Towards a biolinguistic understanding of grammar. Essays on interfaces. Amsterdam/Philadelphia: John Benjamins, pp.31-56. Hupet, Michel & Brigitte Tilmant. 1986. What are clefts good for? Some consequences for comprehension. Journal of Memory and Language 25, pp. 419-430. Iatridou, Sabine & Spyridoula Varlokosta. 1998. Pseudoclefts crosslinguistically. Natural Language Semantics 6, pp. 3-28. Israel, Michael. 2011. The grammar of polarity. Pragmatics, sensitivity, and the logic of scales. Cambridge, UK: Cambridge University Press. Jackendoff, Ray. 1972. Semantic interpretation in generative grammar. Cambridge, Mass.: The MIT Press. Jespersen, Otto. 1927. A modern English grammar on historical principles. Part III: Syntax. Hieldeberg: C. Winter. Jo, Jung-Min. 2013. Predicate contrastive topic constructions: Implications for morpho-syntax in Korean and copy theory of movement. Lingua 131, pp. 80-111. Kayne, Richard. 1983. Chains, categories external to S, and French complex inversion. Natural Language and Linguistic Theory 1 (1), pp. 107-139. Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, Mass.: The MIT Press. Kayne, Richard. 2000. Parameters and universals. Oxford: Oxford University Press. Keyser, Samuel Jay. 1968. Review of Adverbial Positions in English, by Sven Jacobson. Language 44, pp. 357-374. Klima, Edward S. 1964. Negation in English. In Fodor, J. A. & J. J. Katz (eds.). The Structure of Language. Readings in the Philosophy of Language. Englewood Cliffs, N.J.: Prentice Hall, pp. 246-323. Kobele, Gregory. 2006. Generating copies. An investigation into structural identity in language and grammar. PhD dissertation. Los Angeles: UCLA. Online: http://www.linguistics.ucla.edu/general/Dissertations/Kobele06GeneratingCopies.pdf 318 Kočman, Ana. 2011. La ironía verbal como semejanza incongruente. Doctoral thesis. Salamanca: University of Salamanca. Online: http://gredos.usal.es/jspui/bitstream/10366/110705/1/DLE_Kocman_A_LaIronia.pdf Koopman, Hilda. 1996. The Spec-Head configuration. Ms. Los Angeles: UCLA. Online: http://www.linguistics.ucla.edu/people/koopman/papers/Spec_head.PDF Koster, Jan. 2005. Structure preservingness, internal Merge, and the strict locality of triads. In Karimi, Simin et al. (eds.). Phrasal and clausal architecture: Syntactic derivation and interpretation. In honor of Joseph E. Emonds. Amsterdam/Philadelphia: John Benjamins, pp. 188–205. Kovacci, Ofelia. 1999. El adverbio. In Bosque, Ignacio & Violeta Demonte (eds.). Gramática descriptiva de la lengua española. Madrid: Espasa, pp. 705-786. Krifka, Manfred. 2001. For a structured account of questions and answers. In Féry, Caroline & Sternefeld, Wolfgang (eds). Audiatur Vox Sapientiae. A Festschrift for Achim von Stechow. Berlin: Akademie-Verlag, pp. 287-319. Krifka, Manfred. 2006. Association with focus phrases. In Molnár, Valéria & Susanne Winkler (eds.). 2006. The architecture of focus. Berlin/New York: Mouton de Gruyter, pp. 105- 136. Krifka, Manfred. 2008a. Basic notions of information structure. Acta Linguistica Hungarica. Vol. 55 (3-4), pp. 243-276. Krifka, Manfred. 2008b. The semantics of questions and the focusation of answers. In Lee, Chungmin et al. (eds.). Topic and focus. Cross-linguistics perspectives in meaning and intonation. Dordrecht: Springer, pp. 139-150. Laka, Itziar. 1990. Negation in syntax. On the nature of functional categories and projections. PhD dissertation. Cambridge, MA: MIT. Lambrecht, Knud. 2001. A framework for the analysis of cleft constructions. Linguistics 39-3, pp. 463-516. Larson, Richard & Gabriel Segal. 1995. Knowledge of meaning. Cambridge, Mass.: The MIT Press. Lorenzo, Guillermo. 1998. Curso de morfología generativa. Lima, Peru: Educare. Lutz, Uli & Jürgen Pafel (eds.). 1995. On extraction and extraposition in German. Amsterdam/ Philadelphia: John Benjamins. 319 Maienborn, Claudia. 2005. A discourse-based account of Spanish ser-estar. Linguistics 43-1, pp. 155-180. Martin, Roger et al. (eds.). 2000. Step by step. Essays on minimalist syntax in honor of Howard Lasnik. Cambridge, Mass.: The MIT Press. Martinović, Martina. 2013. Information structure and copular sentences in Wolof. Ms. Chicago: University of Chicago. Online: http://home.uchicago.edu/martinam/Selected_papers_&_presentations_files/acal2013pape r.pdf McDaniel, Dana. 1986. Condition on wh-chains. Doctoral dissertation. CUNY. McKinney-Bock, Katherine & María Luisa Zubizarreta (eds.). 2014. Primitive elements of grammatical theory. Papers by Jean-Roger Vergnaud and his collaborators. New York: Routledge. Méndez Vallejo, Dunia Catalina. 2009. Focalizing SER (‘TO BE’) in Colombian Spanish. PhD dissertation. Bloomington: Indiana University. Online: http://dcmendezvallej.people.wm.edu/index_files/FS_in_Colombian_Spanish- Mendez_Vallejo.pdf Merchant, Jason. 2001. The syntax of silence. Sluicing, islands, and the theory of ellipsis. Oxford: Oxford University Press. Mikkelsen, Line. 2005. Copular clauses: specification, predication, and equation. Amsterdam/ Philadelphia: John Benjamins. Miller, Jim. 1996. Clefts, particles and word order in languages of Europe. Language Sciences 18, 1-2, pp. 111-125. Molnár, Valéria & Susanne Winkler (eds.). 2006. The architecture of focus. Berlin/New York: Mouton de Gruyter. Molnár, Valéria & Susanne Winkler. 2010. Edges and gaps: Contrasts at the interfaces. Lingua 120, pp. 1392-1415. Moreno Cabrera, Juan Carlos. 1999. Las funciones informativas. Las perífrasis de relativo y otras construcciones perifrásticas. In Bosque, Ignacio & Violeta Demonte (eds.). Gramática descriptiva de la lengua española. Madrid: Espasa, pp. 4245-4302. 320 Müller, Gereon. 2001. Order preservation, parallel movement, and the emergence of the unmarked. In Legendre, Geraldine et al. (eds.). Optimality Theoretic Syntax. Cambridge, Mass.: The MIT Press, pp. 279-313. Müller, Gereon. 2002. Two types of remnant movement. In Alexiadou, Artemis et al. (eds.). Dimensions of movement. From features to remnants. Amsterdam/Philadelphia: John Benjamins, pp. 209-241. Nilsen, Øinstein. 2003. Eliminating Positions: Syntax and Semantics of Sentence Modification. PhD dissertation. Utrecht: Utrecht University. Online: http://igitur-archive.library.uu.nl/dissertations/2003-0715-112929/inhoud.htm Nisigauchi, Taisuke. 1990. Quantification in the theory of grammar. Dordrecht: Kluwer. Njkemnji, Michael. 1995. Heavy pied-piping in Nweh. PhD dissertation. Los Angeles: UCLA. Online: http://www.linguistics.ucla.edu/images/stories/nkemnji.1995.pdf Nunes, Jairo. 1999. Linearization of chains and phonetic realization of chain links. In Epstein, Samuel & Norbert Horstein (eds.). Working minimalism. Cambridge, Mass.: The MIT Press, pp. 217-249. Nunes, Jairo. 2001. Sideward movement. Linguistic Inquiry Vol. 32, No. 2, pp. 303-344. Nunes, Jairo. 2004. Linearization of chains and sideward movement. Cambridge, Mass.: The MIT Press. Nunes, Jairo (ed.). 2009. Minimalist essays on Brazilian Portuguese syntax. Amsterdam/ Philadelphia: John Benjamins. Nunes, Jairo & Ronice Quadros. In preparation. Focus duplication in Brazilian Sign Language. Ms. Universidade Estadual de Campinas/Universidade Federal de Santa Catarina. O’Rourke, Erin. 2012. The realization of contrastive focus in Peruvian Spanish intonation. Lingua 122, pp. 494-510. Ordóñez, Francisco. 1997. Word order and clause structure in Spanish and other Romance languages. PhD. Dissertation. New York: CUNY. Ordóñez, Francisco. 2000. The clausal structure of Spanish: a comparative perspective. New York: Garland. 321 Ortega-Santos, Iván. 2006. On Postverbal Subjects, PF and the Copy Theory: The Spanish Case. In Sagarra, Nuria & Almeida Jacqueline Toribio (eds.). Proceedings of the 9th Hispanic Linguistics Symposium, 13-23, Somerville, Ma: Cascadilla Proceedings Project. Ott, Dennis. 2009. Multiple NP split. A Distributed Deletion analysis. GAGL (Groninger Arbeiten zur germanistischen Linguistik) 48, pp. 65-80. Pafel, Jürgen. 1995. Kinds of extraction from noun phrases. In Lutz, Uli & Jürgen Pafel (eds.). On extraction and extraposition in German. Amsterdam/Philadelphia: John Benjamins, pp. 145-177. Patten, Amanda. 2012. The English It-cleft. A constructional account and a diachronic investigation. Berlin/Boston: De Gruyter Mouton. Paul, Ileana. 2001. Concealed pseudo-clefts. Lingua 111, pp. 707-727. Percus, Orin. 1997. Prying open the cleft. Proceedings of NELS 27, pp. 337-351. Pesetsky, David. 2000. Phrasal movement and its kin. Cambridge, Mass.: The MIT Press. Portner, Paul. 2009. Modality. Oxford: Oxford University Press. Potsdam, Eric. 2006. More concealed pseudoclefts in Malagasy and the Clausal Typing Hypothesis. Lingua 116, pp. 2154-2182. Progovac, Ljiljana. 1994. Negative and positive polarity. A binding approach. Cambridge, UK: Cambridge University Press. Pustet, Regina. 2003. Copulas. Universals in the categorization of the lexicon. Oxford: Oxford University Press. Radford, Andrew. 1997. Syntax. A minimalist introduction. Cambridge, UK: Cambridge University Press. Reeve, Matthew. 2011. The syntactic structure of English clefts. Lingua 121, pp. 142-171. Reeve, Matthew. 2012a. Clefts and their relatives. Amsterdam/Philadelphia: John Benjamins. Reeve, Matthew. 2012b. Against FP analyses of clefts. In Neeleman, Ad & Reiko Vermeulen (eds.). The syntax of topic, focus, and contrast. An interface-based approach. Berlin/Boston: De Gruyter Mouton, pp. 157-187. Reig Alamillo, Asela. 2011. The pragmatic meaning of the Spanish construction Lo que pasa es que. Journal of Pragmatics 43, pp. 1435-1450. Reinhart, Tanya. 1981. Pragmatics and linguistics: An analysis of sentence topics. Philosophica 27, pp. 53-94. 322 Rizzi, Luigi. 1990. Relativized minimality. Cambridge, Mass.: The MIT Press. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Haegeman, Liliane (ed.). Elements of grammar. Dordrecht: Kluwer. Rizzi, Luigi. 2004. Locality and beyond. In Belletti, Adriana (ed.). Structures and beyond. The cartography of syntactic structures. Vol III. Oxford: Oxford University Press, pp. 223- 251. Rochemont, Michael. 1986. Focus in generative grammar. Amsterdam: John Benjamins. Rodríguez Ramalle, Teresa María. 2001. Los adverbios de manera como predicados de los subeventos verbales. REL (Revista Española de Lingüística) 31, 1, pp. 107-143. Rodríguez Ramalle, Teresa María. 2009. Sobre la estructura discursiva de la oración y su proyección sintáctica: el caso de los adverbios oracionales y otros constituyentes de la periferia oracional. ELUA (Estudios de Lingüística de la Universidad de Alicante) 23, pp. 265-288. Rooth, Mats. 1985. Association with focus. PhD dissertation. Amherst: University of Massachusetts. Rudin, Catherine et al. 1999. Macedonian and Bulgarian li questions: Beyond syntax. Natural Language and Linguistic Theory 17, pp. 541-586. Saab, Andrés. 2008. Hacia una teoría de la identidad parcial en la elipsis. Doctoral thesis. Buenos Aires: University of Buenos Aires. Schlenker, Philippe. 2003. Clausal equations. (A Note on the connectivity problem). Natural Language and Linguistic Theory 21, pp.157-214. Selkirk, Elizabeth. 2005. Comment on intonational phrasing in English. In Frota, Sónia et al. (eds.). Prosodies. Berlin/New York: Mouton de Gruyter, pp. 11-58. Shell, Robert. 1972. The analysis of clefts and pseudo-clefts sentences in English. PhD dissertation. Urbana-Champaign: University of Illinois at Urbana-Champaign. Shu, Chih-hsiang. 2006. The syntax of high adverbs: overt and covert positions. Ms. New York: Stony Brook University. Shu, Chih-hsiang. 2011. Sentence adverbs in the kingdom of Agree. PhD dissertation. New York: Stony Brook University. Simpson, Andrew & Zoe Wu. 2002. IP-raising, tone sandhi and the creation of S-final particles. Evidence for cyclic Spell-Out. Journal of East Asian Linguistics 11, pp. 167-199. 323 Smit, Johan. 2013. An investigation into the adequacy of Cinque’s functional theory as a framework for the analysis of adverbs in Afrikaans. MA thesis. Stellenbosch: Stellenbosch University. Online: http://scholar.sun.ac.za/handle/10019.1/85752 Sportiche, Dominique. 1995. Sketch of a reductionist approach to syntactic variation and dependencies. In Campos, Héctor & Paula Kempchinsky (eds.). Evolution and revolution in linguistic theory. Washington D.C.: Georgetown University Press, pp. 356–398. Stabler, Edward. 1999. Remnant movement and complexity. In Bouma, Gosse et al. (eds.). Constraints and resources in natural language syntax and semantics. Stanford: CSLI, pp. 299–326. Tancredi, Christopher. 1992. Deletion, deaccenting, and presupposition. PhD dissertation. Cambridge, MA: MIT. Thornton, Rosalind. 1990. Adventures in long-distance movement: The acquisition of complex wh-questions. Doctoral dissertation. Storrs: University of Connecticut. Torner Castells, Sergi. 2005. Aspectos de la semántica de los adverbios de modo en español. Doctoral thesis. Barcelona: Universitat Pompeu Fabra. Trask, R.L. & Peter Stockwell (ed.). 2007. Language and linguistics. The key concepts. Oxon: Routledge. Truckenbrodt, Hubert. 1995. Phonological phrases: their relation to syntax, focus and prominence. PhD dissertation. Cambridge, Mass.: MIT. Vallduví, Enric. 1988. Functional load, prosody and syntax: left detachment in Catalan and Spanish. CLS (Chicago Linguistic Society) 24, pp. 391-404. van Riemsdijk, Henk.1989. Movement and regeneration. In Benincà, Paola (ed.). Dialectal variation and the theory of grammar. Dordrecht: Foris, pp. 105-136. Vanden Wyngaerd, Guido & Jan-Wouter Zwart. 1999. Antecedent-contained deletion as deletion. Linguistics in the Netherlands 16-1, pp. 203-216. Velásquez, Héctor. 2009. On the relative order of Spanish adverbs. Ms. Los Angeles: University of Southern California. Velásquez, Héctor. 2011a. Sentential adverb frankly as modifier of a deleted speech verb. Linguistic Society of America, Annual Meeting 2011 extended abstracts. Online: http://www.elanguage.net/journals/index.php/lsameeting/issue/view/269 324 Velásquez, Héctor. 2011b. Spanish modal adverb probablemente as a focus-sensitive unit. A syntactic-prosodic approach. Ms. Los Angeles: University of Southern California. Vergnaud, Jean-Roger. 2014. Some explanatory avatars of conceptual necessity: Elements of UG. In McKinney-Bock, Katherine & María Luisa Zubizarreta (eds.). 2014. Primitive elements of grammatical theory. Papers by Jean-Roger Vergnaud and his collaborators. New York: Routledge, pp. 123-206. Vergnaud, Jean-Roger & María Luisa Zubizarreta. 2001. Types of focus and types of questions. Séminaire de syntaxe, June 2001. Ms. Paris: Université Paris VIII. von Fintel, Kai. 2006. Modality and language. In Borchert, Donald M. (ed.). Encyclopedia of Philosophy, 2 nd edition. Detroit: MacMillan Reference USA. Online: http://mit.edu/fintel/www/modality.pdf Weinert, Regina & Jim Miller. 1996. Cleft constructions in spoken language. Journal of Pragmatics 25, pp. 173-206. Wilder, Chris. 1995. Rightward movement as leftward deletion. In Lutz, Uli & Jürgen Pafel (eds.). On extraction and extraposition in German. Amsterdam/Philadelphia: John Benjamins, pp. 273-309. Williams, Edwin. 1999. Economy as Shape Conservation. Ms. Princeton N.J.: Princeton University. Winkler, Sussane. 2005. Ellipsis and focus in generative grammar. Berlin: Mouton de Gruyter. Zagona, Karen. 2002. The syntax of Spanish. Cambridge, UK: Cambridge University Press. Zimmermann, Malte & Edgar Onea. 2011. Focus marking and focus interpretation. Lingua 121. Zubizarreta, María Luisa. 1982. On the relationship of the lexicon to syntax. PhD dissertation. Cambridge, Mass.: MIT. Zubizarreta, María Luisa. 1998. Prosody, focus, and word order. Cambridge, Mass.: The MIT Press. Zubizarreta, María Luisa. 1999. Las funciones informativas: tema y foco. In Bosque, Ignacio & Violeta Demonte (eds.). Gramática descriptiva de la lengua española. Madrid: Espasa, pp. 4216-4244. Zubizarreta, María Luisa. 2009. The left edge in the Spanish clausal structure. In Masullo, Pascual José et al. (eds.). Romance Linguistics 2007. Selected papers from the 37th 325 Linguistic Symposium on Romance Languages. Amsterdam/Philadelphia: John Benjamins, pp. 339–357. Zubizarreta, María Luisa. 2013. On the grammaticalization of Assertion Structure: a view from Spanish. Ms. Los Angeles: University of Southern California. Online: http://dornsife.usc.edu/assets/sites/295/docs/Zubizarreta2013.Assertion_Structure.pdf Zubizarreta, María Luisa & Jean-Roger Vergnaud. 2006. Phrasal stress and syntax. In Everaert, Martin & Henk van Riemsdijk (eds.). The Blackwell companion to syntax Vol. III, Malden, MA: Blackwell, pp. 522-568. Zyman, Erik. 2012. Two investigations of adverbs and clause structure in English. Senior thesis. Princeton: Princeton University. Online: http://people.ucsc.edu/~ezyman/adverbs_clause_structure.pdf hv.
Abstract (if available)
Abstract
The main problem that this work deals with is the question about the syntactic representation one must attribute to sentence modal adverbs in Spanish (for instance, probablemente ‘probably’) when they surface in low sentence positions. Starting by the verification of the propositional scope of this kind of adverbs, the fact that they can surface adjacent to sub‐clausal units, as vP, DP, AdjP, etc. seems puzzling. Whereas the adverb probablemente is visibly located in a syntactic position with clausal scope in the sentence Probablemente, Juan comprará un carro ‘Probably, J. will‐buy a car’, this is not the case in the sentence Juan comprará un carro probablemente rojo ‘J. will buy a car such that its color will probably be red’. ❧ First, I observe that the adjacency between the adverb and the sub‐clausal unit is triggered by focus sensitivity: the adverb appears immediately adjacent to the focus marked unit. This adjacency may be a mechanism in virtue of which the left edge of the unit in focus gets unambiguously marked. I propose that the adverb is always base generated in a high sentence position, but focus sensitivity on the adverb may trigger later reduplication whenever a unit is focus‐marked within its c‐command domain. I claim that the unit to be copied is the clause c‐ commanded by the sentence modal adverb. This clause is then copied and re‐merged in the Spec position of the maximal projection of the adverb. ❧ I anchor my proposal in the Copy Theory of Movement (Nunes 2001, 2004). This theory interprets movement as Copy+Merge. Crucially, the ‘original’ clause and its clone form a chain. Linearization then enforces a process of chain reduction, involving deletion. In this case, ellipsis applies on the lower copy, but cannot delete the unit in focus, since information relevant for interpretation would be lost (Copy cannot reproduce F‐marking). As a result, the modal adverb ends up by surfacing adjacent to the F‐marked unit. On the basis of this proposal, I extend the reduplicative structure to other cases of focus‐related words and constructions.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Narrowing the focus: experimental studies on exhaustivity and contrast
PDF
Copy theory of movement and PF conditions on spell-out
PDF
Superlative ambiguities: a comparative perspective
PDF
Silence in answers: a study of ellipsis in Hindi
PDF
Perspective in Turkish complementation
PDF
The Spanish feminine el at the syntax-phonology interface
PDF
A minimalist analysis of participial constructions: towards a phase account of non-finite structures
PDF
Agreement on the left edge: the syntax of left dislocation in Spanish
PDF
The case of a person: The person case constraint in German
PDF
The morphosyntax of states: deriving aspect and event roles from argument structure
PDF
Syntax-prosody interactions in the clausal domain: head movement and coalescence
PDF
Towards the unity of movement: implications from verb movement in Cantonese
Asset Metadata
Creator
Velásquez Chafloque, Héctor
(author)
Core Title
A reduplicative analysis of sentence modal adverbs in Spanish
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Linguistics (Hispanic Linguistics)
Publication Date
08/06/2014
Defense Date
06/16/2014
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
Copy Theory of Movement,ellipsis,informative focus,modal adverbs,OAI-PMH Harvest,reduplication,sentence adverbs
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Zubizarreta, Maria Luisa (
committee chair
), Pancheva, Roumyana (
committee member
), Saltarelli, Mario (
committee member
)
Creator Email
hvelasq@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-455060
Unique identifier
UC11287595
Identifier
etd-VelasquezC-2771.pdf (filename),usctheses-c3-455060 (legacy record id)
Legacy Identifier
etd-VelasquezC-2771.pdf
Dmrecord
455060
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Velásquez Chafloque, Héctor; Velasquez Chafloque, Hector
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
Copy Theory of Movement
ellipsis
informative focus
modal adverbs
reduplication
sentence adverbs